research paper on wireless network communication

Advances, Systems and Applications

• Open access
• Published: 10 April 2020

Deep learning-driven wireless communication for edge-cloud computing: opportunities and challenges

• Huaming Wu 1 ,
• Xiangyi Li 1 &
• Yingjun Deng 1  

Journal of Cloud Computing volume  9 , Article number:  21 ( 2020 ) Cite this article

9268 Accesses

32 Citations

4 Altmetric

Metrics details

Future wireless communications are becoming increasingly complex with different radio access technologies, transmission backhauls, and network slices, and they play an important role in the emerging edge computing paradigm, which aims to reduce the wireless transmission latency between end-users and edge clouds. Deep learning techniques, which have already demonstrated overwhelming advantages in a wide range of internet of things (IoT) applications, show significant promise for solving such complicated real-world scenarios. Although the convergence of radio access networks and deep learning is still in the preliminary exploration stage, it has already attracted tremendous concern from both academia and industry. To address emerging theoretical and practical issues, ranging from basic concepts to research directions in future wireless networking applications and architectures, this paper mainly reviews the latest research progress and major technological deployment of deep learning in the development of wireless communications. We highlight the intuitions and key technologies of deep learning-driven wireless communication from the aspects of end-to-end communication, signal detection, channel estimation and compression sensing, encoding and decoding, and security and privacy. Main challenges, potential opportunities and future trends in incorporating deep learning schemes in wireless communications environments are further illustrated.

Introduction

Along with the incredible growth of mobile data generated in internet of things (IoT) and the explosion of complicated wireless applications, e.g., virtual reality (VR) and augmented reality (AR), the fifth-generation (5G) technology demonstrates high-dimensional, high-capacity and high-density characteristics [ 1 , 2 ]. Moreover, future wireless communication systems will become ever-more demanding for edge-cloud computing since the edge servers are in proximity of the IoT devices and communicate with them via different wireless communication technologies [ 3 , 4 ]. The requirements of high bandwidth and low latency for wireless communications have posed enormous challenges to the design, configuration, and optimization of next-generation networks (NGN) [ 5 , 6 ]. In the meantime, massive multiple-input multiple-output (MIMO) is widely regarded as a major technology for future wireless communication systems. In order to improve the quality of wireless signal transmission, the system uses multiple antennas as multiple transmitters at the base station (BS) and receivers at a user equipment (UE) to realize the multipath transmitting, which can double the channel capacity without increasing spectrum resources or antenna transmit power. However, conventional communication systems and theories exhibit inherent limitations in the utilization of system structure information and the processing of big data. Therefore, it is urgent to establish new communication models, develop more effective solutions to address such complicated scenarios and further fulfill the requirements of future wireless communication systems, e.g., beyond the fifth-generation (B5G) networks.

Along with the fast convergence of communication and computing in popular paradigms of edge computing and cloud computing [ 7 , 8 ], intelligent communication is considered to be one of the mainstream directions for the extensive development of future 5G and beyond wireless networks, since it can optimize wireless communication systems performance. In addition, with tremendous progress in artificial intelligence (AI) technology, it offers alternative options for addressing these challenges and replacing the design concepts of conventional wireless communications. Deep learning (DL) is playing an increasingly crucial role in the field of wireless communications due to its high efficiency in dealing with tremendous complex calculations, and is regarded as one of the effective tools for dealing with communication issues. Although deep learning has performed well in some IoT applications, “no free lunch” theorem [ 9 ] shows that a model cannot solve all problems once and for all, and we cannot learn a general model for a wide range of communication scenarios. This means that for any particular mobile and wireless network issue, we still need to adopt different deep learning architectures such as convolutional neural networks (CNN), deep neural networks (DNN) and recurrent neural networks (RNN), in order to achieve better performance of the communication systems.

As a classic model of deep learning, autoencoder is widely used in the design paradigms of communication system models. Autoencoder-based wireless communication models are drawing more and more attention [ 10 , 11 , 12 ]. Generative adversarial network (GAN) [ 13 ] is a promising technique, which has attracted great attention in the field of mobile and wireless networking. The architecture of GAN is composed of two networks, i.e., a discriminative model and a generative model, in which a discriminator D is trained to distinguish the real and fake samples, while the generator G is trained to fool the discriminator D with generated samples. The feature of GAN is very appropriate for training. GAN-driven models and algorithms can facilitate the development of next-generation wireless networks, especially coping with the growth in volumes of communication and computation for emerging IoT applications. However, the incorporation of AI technology in the field of wireless communications is still in its early stages, and learning-driven algorithms in mobile wireless systems are immature and inefficient. More endeavors are required to bridge the gap between deep learning and wireless communication research, e.g., customize GAN techniques for network analytics and diagnosis and wireless resource management in heterogeneous mobile environments [ 14 ].

This survey explores the crossovers and the integration of wireless communication and AI technology, aims at solving specific issues in the mobile networking domain, and greatly improve the performance of wireless communication systems. We gather, investigate and analyze latest research works in emerging deep learning methods for processing and transferring data in the field of wireless communications or related scenarios, including strengths and weaknesses. The main focus is on how to customize deep learning for mobile network applications from three perspectives: mobile data generation, end-to-end wireless communications and network traffic control that adapts to dynamic mobile network environments. Several potential deep learning-driven underlying communication technologies are described, which will promote the further development of future wireless communications.

The rest of this paper is organized as follows: we first draw an overall picture of the latest literature on deep learning technologies in the field of wireless communications. Then, we present important open issues and main challenges faced by researchers for intelligent communications. After that, several potential techniques and research topics in deep learning-driven wireless communications are pointed out. Finally, the paper is concluded.

Emerging deep learning Technologies in Wireless Communications

A list of emerging technology initiatives of incorporating AI schemes for communication research is provided by IEEE Communications Society. Footnote 1 This section selects and introduces the latest research progress of deep learning-driven wireless communication from the aspects of end-to-end communication, signal detection, channel estimation, channel estimation and compression sensing, encoding and decoding, and security and privacy.

End-to-end communications

The guiding principle in communication system design is to decompose signal processing into chains with multiple independent blocks. Each independent block performs a well-defined and isolated function, such as source coding/decoding, channel coding/decoding, modulation, channel estimation and equalization [ 15 ]. This kind of approach yields today’s efficient, versatile, and controllable wireless communication systems. However, it is unclear whether the optimization of individual processing blocks can achieve optimal end-to-end performance, while deep learning can realize theoretically global optimal performance. Thus, deep learning has produced far-reaching significance for wireless communication systems and has shown promising performance improvements.

As shown in Fig.  1 , an autoencoder consists of an encoder and a decoder, where the input data is first processed by the encoder at the transmitter, and then it is decoded at the receiver in order to get the output. The transmitter encodes the input s as a one-hot vector, and a conditional probability density function p ( y|x ) is applied to indicate the wireless channel. After receiving the message, the receiver selects the one with the maximum probability over all possible messages as the output \( \hat{s} \) [ 10 ]. Autoencoder is mainly constructed by neural networks, i.e., an encoding network and a decoding network, the wireless communication system is divided into multiple physical layers to facilitate the propagation of information via the neural network thereon.

Autoencoder-based communication systems

In addition, the idea of end-to-end learning in communication systems has also attracted widespread attention in the wireless communications community [ 16 ]. Several emerging trends for deep learning in communication physical layer were elaborated in [ 10 ]. By treating the wireless communication system as an autoencoder, redefining it as the transmitter and receiver, a local optimum of the end-to-end refactoring process can be achieved. Moreover, different conditions were set in the physical layer to simulate different transmission environments in reality.

The design paradigms of conventional wireless communication systems have to consider the influence of various uncertain factors in hardware implementation, and compensate for delay and phase, which is not efficient and scalable. In contrast, model-free training of end-to-end communication systems based on autoencoder was built by hardware implementations on software-defined radios (SDRs) [ 17 , 18 ], which was simpler, faster, and more efficient. Furthermore, the first entire neural network-based communication system using SDRs was implemented in [ 19 ], where an entire communication system was solely composed of neural networks for training and running. Since such a system fully considered time-varying in the actual channel, its performance was comparable to that of existing wireless communication systems.

A conditional generative adversarial network (CGAN) was applied in [ 20 ] to construct an end-to-end wireless communication system with unknown channel conditions. The encoded signal for transmitting was treated as condition information, and the transmitter and receiver of the wireless communication system were each replaced by a DNN. CGAN acted as a bridge between the transmitter and the receiver, allowing backpropagation to proceed smoothly, thereby jointly training and optimizing both the transmitter and receiver DNNs. This approach makes a significant breakthrough in the modeling mode of conventional wireless communications and opens up a new way for the design of future wireless communication systems.

Signal detection

Deep learning-based signal detection is getting more and more popular. Unlike the conventional model-based detection algorithms that rely on the estimation of the instantaneous channel state information (CSI) for detection, the deep learning-based detection method does not require to know the underlying channel model or the knowledge of the CSI when the channel model is known [ 21 ]. A sliding bidirectional recurrent neural network (SBRNN) was proposed in [ 22 ] for signal detection, where the trained detector was robust to changing channel conditions, eliminating the requirement for instantaneous CSI estimation.

Unlike traditional orthogonal frequency-division multiplexing (OFDM) receivers that first estimate the CSI explicitly, and then the estimated CSI is used to detect or restore the transmitted symbols, the deep learning-based method in [ 23 ] estimated the CSI implicitly and then recovered the transmitted signals directly. The estimated CSI was to solve the problem that a large amount of training data and high training cost were required due to a large increase in the number of parameters caused by DNNs.

Some recent works have suggested the use of DNNs in the context of MIMO detection and have developed model-driven deep learning networks for MIMO detection. For example, a network specifically designed for MIMO communication [ 24 ] can cope with time-varying channel in only one training phase. Instead of addressing a single fixed channel, a network obtained by unfolding the iterations of a projected gradient descent algorithm can handle multiple time-invariant and time-varying channels simultaneously in a single training phase [ 25 ]. Deep learning-based networks as demonstrated in [ 26 ] can reach near-optimal detection performance, guaranteed accuracy and robustness with low and flexible computational complexity.

Channel estimation and compression sensing

Channel estimation and compression sensing are key technologies for the real-time implementation of wireless communication systems. Channel estimation is the process of estimating the parameters of a certain channel model from the received data, while compression sensing is a technique to acquire and reconstruct sparse or compressible signals. Deep learning-based channel estimation and compression sensing methods have been suggested in several recent works [ 27 , 28 , 29 , 30 ].

To tackle the challenge of channel estimation when the receiver is equipped with a limited number of radio frequency (RF) chains in massive MIMO systems, a learned denoising-based approximate message passing (LDAMP) network was exploited in [ 27 ], where the channel structure can be learned and estimated from a large amount of training data. Experiment results demonstrated that the LDAMP network significantly outperforms state-of-the-art compressed sensing-based algorithms.

Motivated by the covariance matrix structure, a deep learning-based channel estimator was proposed in [ 28 ], where the estimated channel vector was a conditional Gaussian random variable, and the covariance matrix was random. Assisted by CNN and the minimum mean squared error (MMSE) estimator, the proposed channel estimator can ensure the state-of-the-art accuracy of channel estimation at a very lower computational complexity.

The basic architecture of deep learning-based CSI feedback is as shown in Fig.  2 . Recently, more and more researchers have focused on the benefits of CSI feedback that the transmitter can utilize it to precode the signals before the transmission, thus we can gain the improvement of MIMO systems. The precoding technique can help to realize the high quality of restoring signals and are widely adopted in wireless communication systems. By exploiting CSI, the MIMO system can substantially reduce multi-user (MU) interference and provide a multifold increase in cell throughput. In the network of frequency division duplex (FDD) or time division duplex (TDD), the receiver UE can estimate the downlink CSI and transmit it back to the BS once they obtain it and help BS to perform precoding for the next signal. BS can also obtain the uplink CSI to help rectify the transmission at UE. The procedure of CSI feedback transmitting has drawn much attention, since high quality reconstructed CSI received by BS guarantees a good precoding, improving the stability and efficiency of the MIMO system.

Deep learning-based CSI feedback

Inspired by traditional compressed sensing technologies, a new CNN-based CSI sensing and recovery mechanism called CsiNet was proposed in [ 29 ], which effectively used the feedback information of training samples to sense and recover CSI, and achieved the potential benefits of a massive MIMO. The encoder of CsiNet converted the original CSI matrix into a codebook using CNN, and then the decoder restored the received codebook to the original CSI signal using the fully-connected network and refine networks.

To further improve the correctness of CSI feedback, a real-time long short-term memory (LSTM) based CSI feedback architecture named CsiNet-LSTM was proposed in [ 31 ], where CNN and RNN are applied to extract the spatial and temporal correlation features of CSI, respectively. Using time-varying MIMO channel time correlation and structural features, CsiNet-LSTM can achieve a tradeoff between compression ratio, CSI reconstruction quality, and complexity. Compared to CsiNet, the CsiNet-LSTM network can trade time efficiency for CSI reconstruction quality. Further, the deep autoencoder-based CSI feedback in the frequency division duplex (FDD) massive MIMO system was modelled in [ 30 ], which involved feedback transmission errors and delays.

As shown in Fig.  3 , a novel effective CSI sensing and recovery mechanism in the FDD MIMO system was proposed in our previous work [ 32 ], referred to as ConvlstmCsiNet, which takes advantage of the memory characteristic of RNN in modules of feature extraction, compression and decompression, respectively. Moreover, we adopt depthwise separable convolutions in feature recovery to reduce the size of the model and interact information between channels. The feature extraction module is also elaborately devised by studying decoupled spatio-temporal feature representations in different structures.

The architecture of ConvlstmCsiNet with P3D block [ 32 ]

Encoding and decoding

In digital communications, source coding and channel coding are typically required in data transmission. Deep learning methods have been suggested in some recent works [ 33 , 34 , 35 , 36 , 37 , 38 ] that can be used to improve standard source decoding and solve the problem of high computational complexity in channel decoding.

A DNN-based channel decoding method applied in [ 33 ] can directly realize the conversion from receiving codewords to information bits when considering the decoding part as a black box. Although this method shows advantages in performance improvement, learning is constrained with exponential complexity as the length of codewords increases. Therefore, it is neither fit for random codes, nor for codewords with long code lengths.

The issue of joint source encoding and channel encoding of structured data over a noisy channel was addressed in [ 38 ], a lower word error rate (WER) was achieved by developing deep learning-based encoders and decoders. This approach was optimal in minimizing end-to-end distortion where both the source and channel codes have arbitrarily large block lengths, however, it is limited in using a fixed length of information bits to encode sentences of different lengths.

Belief propagation (BP) algorithm can be combined with deep learning networks for channel decoding. Novel deep learning methods were proposed in [ 36 , 37 ] to improve the performance of the BP algorithm. It demonstrated that the neural BP decoder can offer a tradeoff between error-correction performance and implementation complexity, but can only learn a single codeword instead of an exponential number of codewords. Neural network decoding was only feasible for very short block lengths, since the training complexity of deep learning-based channel decoders scaled exponentially with the number of information bits and. A deep learning polarization code decoding network with partitioned sub-blocks was proposed in [ 34 ] to improve its decoding performance for high-density parity check (HDPC) codes. By dividing the original codec into smaller sub-blocks, each of which can be independently encoded/decoded, it provided a promising solution to the dimensional problem. Furthermore, Liang et al. [ 35 ] proposed an iterative channel decoding algorithm BP-CNN, which combined CNN with a standard BP decoder to estimate information bits in a noisy environment.

Security and privacy

Due to the shared and broadcast nature of wireless medium, wireless communication systems are extremely vulnerable to attacks, counterfeiting and eavesdropping, and the security and privacy of wireless communications have received much attention [ 39 , 40 ]. Moreover, wireless communication systems are becoming increasingly complex, and there is a close relationship between various modules of the system. Once a module is attacked, it will affect the operation of the entire wireless communication system.

Running AI functions on nearby edge servers or remote cloud servers is very vulnerable to security and AI data privacy issues. Thus, offloading AI learning models and collected data to external cloud servers for training and further processing may result in data loss due to the user's reluctancy of providing sensitive data such as location information. Many research efforts have focused on bridging DL and wireless security, including adversarial DL techniques, privacy Issues of DL solutions and DL hardening solutions [ 41 , 42 ], to meet critical privacy and security requirements in wireless communications.

Conventional wireless communication systems generally suffer from jamming attacks, while autoencoder-based end-to-end communication systems are extremely susceptible to physical adversarial attacks. Small disturbances can be easily designed and generated by attackers. New algorithms for making effective white-box and black-box attacks on a classifier (or transmitter) were designed in [ 43 , 44 ]. They demonstrated that physical adversarial attacks were more destructive in reducing the transmitter’s throughput and success ratio when compared to jamming attacks. In addition, how to keep security and enhance the robustness of intelligent communication systems is still under discussion. Defense strategies in future communication systems are still immature and inefficient. Therefore, further research on the defense mechanisms of adversarial attacks and the security and robustness of deep learning-based wireless systems is very necessary.

One possible defense mechanism is to train the autoencoder to have an antagonistic perturbation, which is a technique that enhances robustness, known as the adversarial training [ 45 ]. Adversarial deep learning is applied in [ 46 ] to launch an exploratory attack on cognitive radio transmissions. In a canonical wireless communication scenario with one transmitter, one receiver, one attacker, and some background traffic, even the transmitter’s algorithm is unknown to the attacker, it can still sense a channel, detect transmission feedback, apply a deep learning algorithm to build a reliable classifier, and effectively jam such transmissions. A defense strategy against an intelligent jamming attack on wireless communications was designed in [ 47 ] to successfully fool the attacker into making wrong predicts. To avoid the inaccurate learned model due to interference of the adversary, one possible way is to use DNNs in conjunction with GANs for learning in adversarial radio frequency (RF) environments, which are capable of distinguishing between adversarial and trusted signals and sources [ 48 ].

Open challenges

This section discusses several open challenges of deep learning-driven wireless communications from the aspects of baseline and dataset, model compression and acceleration, CSI feedback and reconstruction, complex neural networks, training at different SNRs and fast learning.

Baseline and dataset

The rapid development of computer vision, speech recognition, and natural language processing have benefited most from the existence of many well-known and effective datasets in computer science, such as ImageNet [ 49 ] and MNIST [ 50 ]. For fairness, performance comparisons between different approaches should be performed under the same experimental environment by using common datasets. In order to compare the performance of newly proposed deep learning models and algorithms, it is critical to have some well-developed algorithms serving as benchmarks. Experiment results based on these benchmarks are usually called baselines, which are very important to show the development of a research field [ 51 ]. The quality and quantity of open datasets will have a huge impact on the performance of deep learning-based communication systems.

Wireless communication systems involve inherently artificial signals that can be synthesized and generated accurately, the local bispectrum, envelope, instantaneous frequency, and symbol rate of the signal can be used as input features. Therefore, in some cases, we should pay more attention to the standardization of data generation rules rather than the data itself.

In the field of intelligent wireless communications, however, there are few existing and public datasets that can be directly applied for training. It is necessary to either create generic and reliable datasets for different communication problems or develop new simulation software to generate datasets in various communication scenarios. On the basis of such dataset or data generation software, widely used datasets similar to ImageNet and MNIST can be created. Then, we can treat them as baselines or benchmarks for further comparison and research.

Model compression and acceleration

Deep neural networks (DNN) have achieved significant success in computer vision and speech recognition, in the meanwhile, their depth and width are still boosting, which lead to a sharp increase in the computational complexity of networks. At present, the number of parameters in DNN models is very huge (parameters are generally tens of millions to hundreds of millions) and thus the amount of calculation is extremely large. Current deep learning models either rely on mobile terminals or edge-cloud server to run AI functions and are under tremendous pressure in terms of high data storage and processing demands [ 41 ]. Offloading complex compute tasks from mobile terminals to a central cloud with AI functions can alleviate the limitation of computation capacity, but also results in high latency for AI processing due to long-distance transmissions. Therefore, it is not appropriate to offload AI learning model to the central cloud server, especially for data-intensive and delay-sensitive tasks.

Some deep learning algorithms deployed on mobile terminals can only rely on cloud graphic processing units (GPUs) to accelerate computing, however, the wireless bandwidth, the communication delay, and the security of cloud computing will incur enormous obstacles. The large memory and high computational consumption required by the DNN greatly restricts the use of deep learning on mobile terminals with limited resources. Deep learning-based communication systems are also difficult to deploy on small mobile devices such as smartphones, smartwatches and tablets.

Due to the huge redundancy of the parameters in DNN models, these models can be compressed and accelerated to build a lightweight network, which is an inevitable trend in the development of related technologies in the future. Methods like low-rank factorization, parameter pruning and sharing, quantization, and knowledge distillation can be applied in DNN models. Specifically, on the one hand, it is possible to consider quantifying the parameters of DNN models to further compress the network model; on the other hand, channel pruning and structured sparse constraints can be applied to eliminate part of the redundant structure and accelerate the calculation speed [ 52 ].

Lightweight AI engines at the mobile terminals are required to perform real-time mobile computing and decision making without the reliance of edge-cloud servers, where the centralized model is stored in the cloud server while all training data is stored on the mobile terminals. In addition, learning parameter settings or updates are implemented by local mobile devices. In some cases, if the floating-point calculation or storage capacity of the network model is greatly reduced, but the performance of the existing DNNs remains essentially unchanged, such a network model can run efficiently on resource-constrained mobile devices.

CSI feedback and reconstruction

The massive multiple-input multiple-output (MIMO) system is usually operated in OFDM over a large number of subcarriers, leading to a problem of channel state information (CSI) feedback overload. Moreover, in order to substantially provide a multifold increase in cell throughput, each base station is equipped with thousands of antennas in a centralized or distributed manner [ 29 ]. Therefore, it is crucial to utilize the available CSI at the transmitter for precoding to improve the performance of FDD networks [ 32 ]. However, compressing a large amount of CSI feedback overload in massive MIMO systems is very challenging. Traditional estimation approaches like compressive sensing (CS) can only achieve poor performance on CSI compression in real MIMO system due to the harsh preconditions.

Although DL-based CSI methods outperform much than the CS ones, the price of training cost remains high, which requires large quantities of channel estimates. Once the wireless environment changes significantly, a trained model still has to be retrained [ 53 ]. In addition, a more able and efficient structure of DNN is needed. The design of CSI feedback link and precoding mode still remains an open issue that different MIMO systems should adopt their own appropriate designed CSI feedback link and precoding manner. Furthermore, DL-based CSI feedback models are still immature when adopted in real massive MIMO systems and suffer constraints of realistic factors, e.g., time-varying channel with fading, SRS measurement period, channel capacity limitation, hardware or device configuration, channel estimation and signal interference in MU systems. These challenges may hinder the general applications temporarily and will be addressed by future DL-based models with a more exquisite and advanced architecture.

Complex neural networks

Due to the widely used baseband representations in wireless communication systems, data is generally processed in complex numbers, and most of the associated signal processing algorithms rely on phase rotation, complex conjugate, absolute values, and so on [ 10 ].

Therefore, neural networks have to run on complex values rather than real numbers. However, current deep learning libraries usually do not support complex processing. While complex neural networks may be easier to train and consume less memory, they do not provide any significant advantages in terms of performance. At present, we can only think of a complex number as a real number and an imaginary number. Complex neural networks that are suitable for wireless communication models should be developed.

Training at different SNRs

Up to now, it is still not clear which signal-to-noise (SNR) ratio the deep learning model should be trained on. The ideal deep learning model should be applied to any SNR regardless of the SNR used for training or the range of SNR it is in. In fact, however, this is not the case. The results of training deep learning models under certain SNR conditions are often not suitable for other SNR ranges [ 10 ].

For example, training at lower SNRs does not reveal important structural features of wireless communication systems at higher SNRs, and similarly, training at higher SNRs can not reveal important structural features of wireless communication systems at lower SNRs. Training the deep learning model across different SNRs can also seriously affect the training time. In addition, how to construct an appropriate loss function, how to adjust parameters and data representation for wireless communication systems are still big problems that must be solved.

Fast learning

For end-to-end training of wireless communication systems including encoders, channels, and decoders, a specific channel model is usually required. The trained model needs to be applied to its corresponding channel model, otherwise, mismatch problems will occur, which will cause severe degradation of system performance.

In real-world scenarios, however, due to many environmental factors, the channel environment often changes at any time and place, e.g., the change of the movement speed and direction of user terminals, the change of the propagation medium, the change of the refractive scattering environment. Once the channel environment changes, a large amount of training data is needed to retrain, which means that for different channel environments at each moment, such repeated training tasks need to be performed, which consumes resources and weakens the performance of the system.

Retraining is required when the system configuration changes because the system model does not have a good generalization ability. Adaptation is done on a per-task basis and is specific to the channel model [ 54 ]. Some changes in the channel environment may lead to a sharp decline in system performance. Therefore, we need to seek systems with stronger generalization ability, in order to adapt to the changing channel environment.

Potential opportunities

This section mainly describes the profound potential opportunities and the promising research directions in wireless communications assisted by the rapid development of deep learning.

Deep learning-driven CSI feedback in massive MIMO system

Recent researches indicate that applying deep learning (DL) in MIMO systems to address the nonlinear problems or challenges can indeed boost the quality of CSI feedback compression. Different from the traditional CS-based approaches, DL-based CSI methods adopt several neural network (NN) layers as an encoder replacing the CS model to compress CSI as well as a decoder to recover the original CSI, which can speed up the transmitting runtime nearly 100 times of CS ones.

The structure of autoencoder-based MIMO systems is depicted in Fig.  4 , which only considers the downlink CSI feedback process, assuming that the feedback channel is perfect enough to transmit CSI with no impairments. In fact, a large part of the overload CSI serves redundant and the CSI matrix falls to sparse in the delay domain. In order to remove the information redundancy, CNNs are applied here, which has the ability to eliminate the threshold of domain expertise since CNNs use hierarchical feature extraction, which can effectively extract information and obtain increasingly abstract correlations from the data while minimizing data preprocessing workload.

The structure of autoencoder-based MIMO systems with downlink CSI feedback

We can consider both the issues of feedback delay and feedback errors. Assume that one signal is transmitted into n time slots due to the restriction of downlink bandwidth resource, thus demanding a n -length time series of CSI feedback estimation within a signal transmitting period and the SRS measurement period. The time-varying channel is also under the condition of known overdue CSI or partial CSI characteristics, such as Doppler or beam-delay information. Furthermore, the feedback errors from MU interference brought by multiple UE at middle or high moving speed are also taken into account. When transmitting the compressed CSI feedback, the imperfections, e.g. additive white Gaussian noise (AWGN), in uplink CSI feedback channel would also bring feedback errors. The model is trained to minimize the feedback errors via the minimum mean square error (MMSE) detector.

The architecture of DL-based autoencoder in CSI feedback compression is also advanced via taking the advantages of RNN’s memory characteristic to deal with the feature extraction in time-varying channel, which can have an active effect on time correlation exploring and better performance on CSI recovery [ 30 ]. Similarly, a DL-based autoencoder of CSI estimation method can be applied in this MIMO system, which is exposed to more practical restrictions.

In the future, we can use DL methods of CSI feedback with time-varying channel in massive MU-MIMO system to improve the compression efficiency and speed up the transmitting process, as well as develop novel theoretical contributions and practical research related to the new technologies, analysis and applications with the help of CNN and RNN.

GAN-based Mobile data augmentation

Mobile data typically comes from a variety of sources with various formats and exhibits complex correlations and heterogeneity. According to the mobile data, conventional machine learning tools require cumbersome feature engineering to make accurate inferences and decisions. Deep learning has eliminated the threshold of domain expertise because it uses hierarchical feature extraction, which can effectively extract information and obtain increasingly abstract correlations from the data while minimizing data pre-processing workload [ 55 ]. However, inefficiency in training time is an enormous challenge when applying learning algorithms in wireless systems. Traditional supervised learning methods, which learn a function that maps the input data to some desired output class label, is only effective when sufficient labeled data is available. On the contrary, generative models, e.g., GAN and variational autoencoder (VAE), can learn the joint probability of the input data and labels simultaneously via Bayes rule [ 56 ]. Therefore, GANs and VAEs are well suitable for learning in wireless environments since most current mobile systems generate unlabeled or semi-labeled data.

GANs can be used to enhance the configuration of mobile and wireless networks and help address the growth of data volumes and algorithm-driven applications to satisfy the large data needs of DL algorithms. GAN is a method that allows exploiting unlabeled data to learn useful patterns in an unsupervised manner. GANs can be further applied in B5G mobile and wireless networks, especially in dealing with heterogeneous data generated by mobile environments.

As shown in Fig.  5 , the GAN model consists of two neural networks that compete against each other. The generator network tries to generate samples that resemble the real data such that the discriminator cannot tell whether it is real or fake. After training the GAN, the output of the generator is fed to a classifier network during the inference phase. We can use GAN to generate real data according to previously collected real-world data. Furthermore, it can be used for path planning, trajectories analysis and mobility analysis.

GAN-based mobile data generation

Monitoring large-scale mobile traffic is, however, a complex and costly process that relies on dedicated probes, which have limited precision or coverage and gather tens of gigabytes of logs daily [ 57 ]. Heterogeneous network traffic control is an enormous obstacle due to the highly dynamic nature of large-scale heterogeneous networks. As for a deep learning system, it has difficulty in characterizing the appropriate input and output patterns [ 58 ].

GANs can be applied in resource management and parameter optimization to adapt to the changes in the wireless environment. To make this happen, intelligent control of network traffic can be applied to infer fine-grained mobile traffic patterns, from aggregate measurements collected by network probes. New loss functions are required to stabilize the adversarial model training process, and prevent model collapse or non-convergence problems. Further, data processing and augmentation procedure are required to handle the insufficiency of training data and prevent the neural network model from over-fitted.

Deep learning-driven end-to-end communication

The purpose of autoencoder is to make the input and the output as similar as possible, which is achieved by performing backpropagation of the error and continuing optimization after each output. Similarly, a simple wireless communication system consists of a transmitter (encoder), a receiver (decoder) through a channel, and an abundant of physical layer transmission technologies can be adopted in the wireless communication process. A communication system over an additive white gaussian noise (AWGN) or Rayleigh fading channel can be represented as a particular type of autoencoder. The purpose of wireless communication is to make the output signal and the input signal as similar as possible. However, how to adapt an end-to-end communications system trained on a statistical model to a real-world implementation remains an open question.

As shown in Fig.  6 , we can extend the above single channel model to two or more channels, where multiple transmitter and multiple receivers are competing for the channel capacity. As soon as some of the transmitters and receivers are non-cooperative, adversarial training strategies such as GANs could be adopted. We can perform joint optimization for common or individual performance metrics such as block error rate (BLER). However, how to train two mutually coupled autoencoders is still a challenge. One suggestion is to assign dynamic weights to different autoencoders and minimize the weighted sum of the two loss functions.

Autoencoder-based MIMO System

The diagram of the energy-based generative adversarial network (EBGAN) [ 59 ] in wireless communications is depicted in Fig.  7 . We use an encoder instead of a transmitter, and a decoder instead of a receiver for intelligent communications. The generative network is applied to generate the canonicalized signal, and then fed into the discriminative network for further classification. Inverse filtering can be applied to simplify the task of the learned discriminative network. Similarly, the purpose of EBGAN-based end-to-end communication is to make the output signal and the input signal as close as possible.

EBGAN with an autoencoder discriminator in wireless communications

The discriminator D is structured as an autoencoder:

where Enc (·) and Dec (·) denote the encoder function and decoder function, respectively.

Given a positive margin m, a data sample x and a generated sample G ( z ), the discriminator loss LD and the generator loss LG are formally defined by:

where {·} + is the hinge loss function, the generator is trained to produce contrastive samples with minimal energies, while the discriminator is trained to assign high energies to generated samples [ 59 ].

Most mathematical models in wireless communication systems are static, linear, and Gaussian-compliant optimization models. However, a realistic communication system has many imperfect and non-linear problems, e.g., nonlinear power amplifiers, which can only be captured by these models. The EBGAN-based wireless communication system no longer requires a mathematical linear processing model that can be optimized for specific hardware configurations or spatially correlated channels. With the help of EBGAN, we can learn about the implementation details of the transmitter and receiver and even the information coding without any prior knowledge.

Meta-learning to wireless communication

In real-world scenarios, it is not worthwhile to perform multi-tasks training from scratch just because of different channel models, because these tasks are closely related, they share the same encoder and decoder network structure, and their parameter changes are only affected by the channel model. Training from scratch is under the assumption that such tasks are completely independent and cannot make full use of the connections, resulting in many repetitive and redundant training steps, however, it is not true.

Meta-learning, or learning to learn [ 60 ], that is, to make the model a learner. It learns a priori knowledge in multi-tasking and then quickly applies it to the learning of new tasks, so that fast learning and few-shot learning can be realized. Meta-learning provides a way to perform multi-task learning and optimizes the system parameters toward a common gradient descent direction during training, thereby achieving the optimal generalization ability and reduced training data and/or time complexity. In the meantime, when a new task arrives, the system can train on a few rounds of iterative (or even one round of iterative) with very little training data, so that the parameters can be dynamically fine-tuned on the basis of the original learning model to adapt to the new channel model, where the dynamic parameter tuning is possible. Thus, meta-learning can be implemented for end-to-end learning of encoder and decoder with unknown or changing wireless channels, and it outperforms conventional training and joint training in wireless communication systems [ 54 ].

A specific example of meta-training methods known as model agnostic meta-learning (MAML) [ 61 ]. Its core idea is to find a common initialization point that allows for a quick adaptation towards the optimal performance on the new task. MAML updates parameters through one or more stochastic gradient descent (SGD) steps, which are calculated using only a small amount of data from the new task. Therefore, instead of training a common system model for all channel models, we can apply MAML to find a common initialization vector so that it supports fast training on any channel [ 54 ].

Several recent efforts have focused on intelligent communications to harvest remarkable potential benefits. We have mainly discussed the potential applicability of deep learning in the field of wireless communications for edge-cloud computing, such as model-free training method for end-to-end wireless communications, and further demonstrated their superior performance over conventional wireless communications. Implementation of many emerging deep learning technologies are still in the preliminary stage, and profound potential solutions to solving wireless communication problems have to be further studied. This survey attempts to summarize the research progress in deep learning-driven wireless communications and point out existing bottlenecks, future opportunities and trends.

In the research of B5G wireless networks and communication systems, the low efficiency of training time is a bottleneck when applying learning algorithms in wireless systems. Although deep learning is not mature in wireless communications, it is regarded as a powerful tool and hot research topic in many potential application areas, e.g., channel estimation, wireless data analysis, mobility analysis, complicated decision-making, network management, and resource optimization. It is worthwhile to investigate the use of deep learning techniques in wireless communication systems to speed up the training process and develop novel theoretical contributions and practical research related to the new technologies, analysis and applications for edge-cloud computing.

Availability of data and materials

Machine Learning For Communications Emerging Technologies Initiative https://mlc.committees.comsoc . org/research-library.

Abbreviations

Artificial intelligence

Augmented reality

Additive white gaussian noise

Block error rate

Belief propagation

Base station

Beyond the fifth-generation

Conditional generative adversarial network

Convolutional neural network

Channel state information

Deep neural network

Energy-based generative adversarial network

Frequency division duplex

Generative adversarial network

Graphics processing unit

High-density parity check

Learned denoising-based approximate message passing

Long short-term memory

Agnostic meta-learning

Minimum mean squared error

Multiple-input multiple-output

Next-generation network

Orthogonal frequency-division multiplexing

Radio frequency

Recurrent neural network

Software-defined radio

Sliding bidirectional recurrent neural network

Stochastic gradient descent

Time division duplex

• Internet of things

User Equipment

Virtual reality

Word error rate

Fifth-Generation

Ma Z, Xiao M, Xiao Y, Pang Z, Poor HV, Vucetic B (2019) High-reliability and low-latency wireless communication for internet of things: challenges, fundamentals, and enabling technologies. IEEE Internet Things J 6(5):7946–7970

Article   Google Scholar  

Liu G, Wang Z, Hu J, Ding Z, Fan P (2019) Cooperative NOMA broadcasting/multicasting for low-latency and high-reliability 5g cellular v2x communications. IEEE Internet Things J 6(5):7828–7838

Xu X, Liu X, Xu Z, Dai F, Zhang X, Qi L (2019) Trust-oriented IoT service placement for smart cities in edge computing. IEEE Internet Things J

Lai P, He Q, Cui G, Xia X, Abdelrazek M, Chen F, Hosking J, Grundy J, Yang Y (2019) Edge user allocation with dynamic quality of service. In: International Conference on Service-Oriented Computing. Springer, Cham, pp 86–101

Qi L, Chen Y, Yuan Y, Fu S, Zhang X, Xu X (2020) A QoS-aware virtual machine scheduling method for energy conservation in cloud-based cyber-physical systems. World Wide Web 23, pp 1275-1297

Xu X, Chen Y, Zhang X, Liu Q, Liu X, Qi L (2019) A blockchain-based computation offloading method for edge computing in 5G networks. Software: Practice and Experience. Wiley, pp 1–18

Xu X, Zhang X, Gao H, Xue Y, Qi L, Dou W (2020) Become: Blockchain-enabled computation offloading for IoT in mobile edge computing. IEEE Trans Ind Inform 16(6):4187-4195

Wu H, Wolter K (2018) Stochastic analysis of delayed mobile offloading in heterogeneous networks. IEEE Trans Mob Comput 17(2):461–474

Wolpert DH, Macready WG (1997) No free lunch theorems for optimization. IEEE Trans Evol Comput 1(1):67–82

O’Shea T, Hoydis J (2017) An introduction to deep learning for the physical layer. IEEE Trans Cogn Commun Netw 3(4):563–575

Felix A, Cammerer S, Dörner S, Hoydis J, Ten Brink S (2018) OFDM-autoencoder for end-to-end learning of communications systems. In: 2018 IEEE 19th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC). IEEE, pp 1–5

Jang Y, Kong G, Jung M, Choi S, Kim I-M (2019) Deep autoencoder based CSI feedback with feedback errors and feedback delay in FDD massive MIMO systems. IEEE Wireless Commun Lett 8(3):833–836

Goodfellow IJ, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, Courville A, Bengio Y (2014) Generative adversarial nets. In: Proceedings of the 27th International Conference on Neural Information Processing Systems-volume 2. MIT Press, pp 2672–2680

Wu H, Han Z, Wolter K, Zhao Y, Ko H (2019) Deep learning driven wireless communications and mobile computing. Wirel Commun Mob Comput 2019:1–2

Google Scholar  

O’Shea TJ, Erpek T, Clancy TC (2017) Deep learning based MIMO communications. arXiv preprint arXiv 1707:07980

Qin Z, Ye H, Li GY, Juang B-HF (2019) Deep learning in physical layer communications. IEEE Wirel Commun 26(2):93–99

Aoudia FA, Hoydis J (2018) End-to-end learning of communications systems without a channel model. In: 2018 52nd Asilomar Conference on Signals, Systems, and Computers. IEEE, pp 298–303

Aoudia FA, Hoydis J (2019) Model-free training of end-to-end communication systems. IEEE J Selected Areas Commun 37(11):2503–2516

Dörner S, Cammerer S, Hoydis J, ten Brink S (2018) Deep learning based communication over the air. IEEE J Selected Top Signal Process 12(1):132–143

Ye H, Li GY, Juang B-HF, Sivanesan K (2018) Channel agnostic end-to-end learning based communication systems with conditional GAN. In: 2018 IEEE Globecom Workshops (GC Wkshps). IEEE, pp 1–5

Wang T, Wen C-K, Wang H, Gao F, Jiang T, Jin S (2017) Deep learning for wireless physical layer: opportunities and challenges. China Communications 14(11):92–111

Farsad N, Goldsmith A (2018) Neural network detection of data sequences in communication systems. IEEE Trans Signal Process 66(21):5663–5678

Article   MathSciNet   Google Scholar  

Ye H, Li GY, Juang B-H (2018) Power of deep learning for channel estimation and signal detection in OFDM systems. IEEE Wireless Commun Lett 7(1):114–117

He H, Wen C-K, Jin S, Li GY (2018) A model-driven deep learning network for MIMO detection. In: 2018 IEEE Global Conference on Signal and Information Processing (GlobalSIP). IEEE, pp 584–588

Samuel N, Diskin T, Wiesel A (2017) Deep MIMO detection. In: 18th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), IEEE, pp 1–5

Samuel N, Diskin T, Wiesel A (2019) Learning to detect. IEEE Trans Signal Process 67(10):2554–2564

He, H., Jin, S., Wen, C., Gao, F., Ye Li, G., Xu, Z.: Model-driven deep learning for physical layer communications. IEEE Wireless Commun, 1–7 (2019)

Neumann D, Wiese T, Utschick W (2018) Learning the MMSE channel estimator. IEEE Trans Signal Process 66(11):2905–2917

MathSciNet   MATH   Google Scholar  

Wen C-K, Shih W-T, Jin S (2018) Deep learning for massive MIMO CSI feedback. IEEE Wireless Commun Lett 7(5):748–751

Lu C, Xu W, Shen H, Zhu J, Wang K (2019) MIMO channel information feedback using deep recurrent network. IEEE Commun Lett 23(1):188–191

Wang T, Wen C-K, Jin S, Li GY (2019) Deep learning-based CSI feedback approach for time-varying massive MIMO channels. IEEE Wireless Commun Lett 8(2):416–419

Li X, Wu H (2020) Spatio-temporal representation with deep neural recurrent network in MIMO CSI feedback. IEEE Wireless Communications Letters

Gruber, T., Cammerer, S., Hoydis, J., Ten Brink, S.: On deep learning-based channel decoding. In: 2017 51st Annual Conference on Information Sciences and Systems (CISS) , pp. 1–6 (2017). IEEE

Cammerer S, Gruber T, Hoydis J, ten Brink S (2017) Scaling deep learning-based decoding of polar codes via partitioning. In: GLOBECOM 2017-2017 IEEE Global Communications Conference. IEEE, pp 1–6

Liang F, Shen C, Wu F (2018) An iterative BP-CNN architecture for channel decoding. IEEE J Selected Top Signal Process 12(1):144–159

Nachmani E, Be’ery Y, Burshtein D (2016) Learning to decode linear codes using deep learning. In: 2016 54th Annual Allerton Conference on Communication, Control, and Computing (Allerton). IEEE, pp 341–346

Nachmani E, Marciano E, Lugosch L, Gross WJ, Burshtein D, Be’ery Y (2018) Deep learning methods for improved decoding of linear codes. IEEE J Selected Top Signal Process 12(1):119–131

Farsad N, Rao M, Goldsmith A (2018) Deep learning for joint source-channel coding of text. In: 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, pp 2326–2330

Meng T, Wolter K, Wu H, Wang Q (2018) A secure and cost-efficient offloading policy for mobile cloud computing against timing attacks. Pervasive Mobile Comput 45:4–18

Tian Q, Lin Y, Guo X, Wen J, Fang Y, Rodriguez J, Mumtaz S (2019) New security mechanisms of high-reliability IoT communication based on radio frequency fingerprint. IEEE Internet Things J 6(5):7980–7987

Nguyen, D.C., Cheng, P., Ding, M., Lopez-Perez, D., Pathirana, P.N., Li, J., Seneviratne, A.: Wireless AI: enabling an AI-governed data life cycle (2020)

Sagduyu YE, Shi Y, Erpek T, Headley W, Flowers B, Stantchev G, Lu Z (2020) When wireless security meets machine learning: Motivation, challenges, and research directions. arXiv preprint arXiv 2001:08883

Sadeghi M, Larsson EG (2019) Physical adversarial attacks against end-to-end autoencoder communication systems. IEEE Commun Lett 23(5):847–850

Sadeghi M, Larsson EG (2019) Adversarial attacks on deep-learning based radio signal classification. IEEE Wireless Commun Lett 8(1):213–216

Goodfellow IJ, Shlens J, Szegedy C (2015) Explaining and harnessing adversarial examples. Stat 1050:20

Shi Y, Sagduyu YE, Erpek T, Davaslioglu K, Lu Z, Li JH (2018) Adversarial deep learning for cognitive radio security: jamming attack and defense strategies. In: 2018 IEEE International Conference on Communications Workshops (ICC Workshops). IEEE, pp 1–6

Erpek T, Sagduyu YE, Shi Y (2019) Deep learning for launching and mitigating wireless jamming attacks. IEEE Trans Cognitive Commun Netw 5(1):2–14

Roy D, Mukherjee T, Chatterjee M (2019) Machine learning in adversarial RF environments. IEEE Commun Mag 57(5):82–87

Deng J, Dong W, Socher R, Li L-J, Li K, Fei-Fei L (2009) Imagenet: a large-scale hierarchical image database. In: IEEE Conference on Computer Vision and Pattern Recognition. IEEE, pp 248–255

Deng L (2012) The MNIST database of handwritten digit images for machine learning research [best of the web]. IEEE Signal Process Mag 29(6):141–142

Zhang MM, Shang K, Wu H (2019) Learning deep discriminative face features by customized weighted constraint. Neurocomputing 332:71–79

Liu C, Wu H (2019) Channel pruning based on mean gradient for accelerating convolutional neural networks. Signal Process 156:84–91

Qing C, Cai B, Yang Q, Wang J, Huang C (2019) Deep learning for CSI feedback based on superimposed coding. IEEE Access 7:93723–93733

Simeone O, Park S, Kang J (2020) From learning to meta-learning: Reduced training overhead and complexity for communication systems. arXiv preprint arXiv:2001–01227

Zhang C, Patras P, Haddadi H (2019) Deep learning in mobile and wireless networking: a survey. IEEE Commun Surv Tutorials 21(3):2224–2287

Jagannath J, Polosky N, Jagannath A, Restuccia F, Melodia T (2019) Machine learning for wireless communications in the internet of things: a comprehensive survey. Ad Hoc Netw 93:101913

Mohammadi, M., Al-Fuqaha, A., Oh, J.-S.: Path planning in support of smart mobility applications using generative adversarial networks. In: 2018 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData), pp. 878–885 (2018). IEEE

Wang M, Cui Y, Wang X, Xiao S, Jiang J (2018) Machine learning for networking: workflow, advances and opportunities. IEEE Netw 32(2):92–99

Zhao J, Mathieu M, LeCun Y (2017) Energy-based generative adversarial network. In: 5th International Conference on Learning Representations, ICLR 2017, Toulon

Andrychowicz M, Denil M, Gomez S, Hoffman MW, Pfau D, Schaul T, Shillingford B, De Freitas N (2016) Learning to learn by gradient descent by gradient descent. In: Advances in Neural Information Processing Systems. Curran Associates, Inc., pp. 3981–3989

Finn C, Abbeel P, Levine S (2017) Model-agnostic meta-learning for fast adaptation of deep networks. In: Proceedings of the 34th International Conference on Machine Learning-Volume 70. Sydney, JMLR.org, pp 1126–1135

Download references

Acknowledgements

The authors thank the editor and anonymous reviewers for their helpful comments and valuable suggestions.

Authors’ informations

Huaming Wu received the B.E. and M.S. degrees from Harbin Institute of Technology, China in 2009 and 2011, respectively, both in electrical engineering. He received the Ph.D. degree with the highest honor in computer science at Free University of Berlin, Germany in 2015. He is currently an associate professor in the Center for Applied Mathematics, Tianjin University. His research interests include mobile cloud computing, edge computing, fog computing, internet of things (IoTs), and deep learning.

Xiangyi Li received the B.S. in Applied Mathematics from Tianjin University, China. She is currently a M.S. student majoring in applied mathematics at Tianjin University, China. Her research interests include deep learning, wireless communications and generative models.

Yingjun Deng received the B.S. in Applied Mathematics (2009) and M.S. in Computational Mathematics (2011) from Harbin Institute of Technology, China. He got his Ph.D. in Systems Optimization and Dependability from Troyes University of Technology, France in 2015. He worked as a postdoctoral fellow, respectively at University of Waterloo in Canada (2015–2016), and Eindhoven University of Technology in Netherlands (2018–2019). He became a lecturer since 2016 in the Center for Applied Mathematics, Tianjin University, China. His research interests include applied statistics, deep learning, prognostic and health management, and predictive maintenance.

This work is partially supported by the National Key R & D Program of China (2018YFC0809800), the National Natural Science Foundation of China (61801325), the Huawei Innovation Research Program (HO2018085138), the Natural Science Foundation of Tianjin City (18JCQNJC00600), and the Major Science and Technology Project of Tianjin (18ZXRHSY00160).

Author information

Authors and affiliations.

Center of Applied Mathematics, Tianjin University, Weijin Road, Tianjin, China

Huaming Wu, Xiangyi Li & Yingjun Deng

You can also search for this author in PubMed   Google Scholar

Contributions

HW designed the survey and led the write up of the manuscript. XL contributed part of the writing of the manuscript. YD took part in the discussion of the work described in this paper. All authors have read and approved the manuscript.

Corresponding author

Correspondence to Huaming Wu .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Wu, H., Li, X. & Deng, Y. Deep learning-driven wireless communication for edge-cloud computing: opportunities and challenges. J Cloud Comp 9 , 21 (2020). https://doi.org/10.1186/s13677-020-00168-9

Download citation

Received : 11 February 2020

Accepted : 30 March 2020

Published : 10 April 2020

DOI : https://doi.org/10.1186/s13677-020-00168-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Wireless communication
• Future network
• Edge-cloud computing

Help | Advanced Search

Computer Science > Networking and Internet Architecture

Title: empowering wireless networks with artificial intelligence generated graph.

Abstract: In wireless communications, transforming network into graphs and processing them using deep learning models, such as Graph Neural Networks (GNNs), is one of the mainstream network optimization approaches. While effective, the generative AI (GAI) shows stronger capabilities in graph analysis, processing, and generation, than conventional methods such as GNN, offering a broader exploration space for graph-based network optimization. Therefore, this article proposes to use GAI-based graph generation to support wireless networks. Specifically, we first explore applications of graphs in wireless networks. Then, we introduce and analyze common GAI models from the perspective of graph generation. On this basis, we propose a framework that incorporates the conditional diffusion model and an evaluation network, which can be trained with reward functions and conditions customized by network designers and users. Once trained, the proposed framework can create graphs based on new conditions, helping to tackle problems specified by the user in wireless networks. Finally, using the link selection in integrated sensing and communication (ISAC) as an example, the effectiveness of the proposed framework is validated.

Submission history

Access paper:.

• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

Journals By Subject

• Proceedings

Information

An Overview Research on Wireless Communication Network

Mohaiminul Islam

School of Intelligent Technology and Engineering, Chongqing University of Science and Technology, Chongqing, China

Shangzhu Jin

Add to Mendeley

Communication Systems can be Wired or Wireless and the medium used for communication can be Guided or Unguided. In Wired Communication, the medium is a physical path like Co-axial Cables, Twisted Pair Cables and Optical Fiber Links etc. which guides the signal to propagate from one point to other. Such type of medium is called Guided Medium. The term wireless refers to the communication or transmission of information over a distance without requiring wires, cables or any other electrical conductors. Wireless communication is one of the important mediums of transmission of data or information to other devices. The Communication is set and the information is transmitted through the air, without requiring any cables, by using electromagnetic waves like radio frequencies, infrared, satellite, etc., in a wireless communication technology network. At the end of the 19th century, the first wireless communication systems were introduced and the technology has significantly been developed over the intervening and subsequent years. Today, the term wireless refers to a variety of devices and technologies ranging from smart phones to laptops, tabs, computers, printers, Bluetooth, etc. On the other hand, Wireless Communication doesn’t require any physical medium but propagates the signal through space. Since, space only allows for signal transmission without any guidance, the medium used in Wireless Communication is called Unguided Medium. In the present days, wireless communication system has become an essential part of various types of wireless communication devices, that permits user to communicate even from remote operated areas. There are many devices used for wireless communication like mobiles. Cordless telephones, GPS, Wi-Fi, satellite television and wireless computer parts. Current wireless phones include 3 and 4G networks, Bluetooth and Wi-Fi technologies. This paper is focused on elements of Wireless Communication system, Types of Wireless Communication, Advantage & Disadvantage of it, Smart city, wireless network security.

Wireless Communication, Bluetooth, WALNs, Infrared Communication, Transmission, Network

Mohaiminul Islam, Shangzhu Jin. (2019). An Overview Research on Wireless Communication Network. Advances in Wireless Communications and Networks , 5 (1), 19-28. https://doi.org/10.11648/j.awcn.20190501.13

Mohaiminul Islam; Shangzhu Jin. An Overview Research on Wireless Communication Network. Adv. Wirel. Commun. Netw. 2019 , 5 (1), 19-28. doi: 10.11648/j.awcn.20190501.13

Mohaiminul Islam, Shangzhu Jin. An Overview Research on Wireless Communication Network. Adv Wirel Commun Netw . 2019;5(1):19-28. doi: 10.11648/j.awcn.20190501.13

Cite This Article

• Author Information

Verification Code/

The verification code is required.

Verification code is not valid.

Science Publishing Group (SciencePG) is an Open Access publisher, with more than 300 online, peer-reviewed journals covering a wide range of academic disciplines.

Learn More About SciencePG

• Special Issues
• AcademicEvents
• ScholarProfiles
• For Authors
• For Reviewers
• For Editors
• For Conference Organizers
• For Librarians
• Article Processing Charges
• Special Issues Guidelines
• Editorial Process
• Peer Review at SciencePG
• Open Access
• Ethical Guidelines

Important Link

• Manuscript Submission
• Propose a Special Issue
• Join the Editorial Board
• Become a Reviewer

• Open access
• Published: 28 September 2009

Optimization Techniques in Wireless Communications

• Sergiy A. Vorobyov 1 ,
• Shuguang Cui 2 ,
• Yonina C. Eldar 3 ,
• Wing-Kin Ma 4 &
• Wolfgang Utschick 5  

EURASIP Journal on Wireless Communications and Networking volume  2009 , Article number:  567416 ( 2009 ) Cite this article

5951 Accesses

7 Citations

Metrics details

Welcome to this Special Issue of the EURASIP Journal on Wireless Communications and Networking (JWCN). This issue collects several research results on the use of optimization techniques in wireless communications. Recent advances in linear and nonlinear optimization facilitate progress in many areas of communications. In wireless and mobile communications this progress provides opportunities for introducing new standards and improving existing services. Supporting multimedia traffic with end-to-end quality-of-service (QoS) guarantee over multi-hop wireless networks (e.g., wireless sensor networks, mobile ad hoc networks, wireless mesh networks) is a challenging technical problem due to various factors and constraints: limited bandwidth and battery power, channel variability and user mobility, protocol and standard compatibility, fairness consideration, higher data rates, system robustness, and seamless service, to name a few. In addition, several wireless networks may be allowed to co-exist and share the same spectrum, which leads to the requirement of minimal (acceptable) interference between different networks.

Optimization methods have been recognized as extremely useful techniques in helping with addressing the aforementioned challenges. Although optimization techniques are not limited by the convex optimization category, the convex optimization framework has been most successfully applied to a number of problems in wireless communications and signal processing. Over the last few years, convex optimization has found a place among the most useful techniques for algorithm design and analysis of wireless communication systems, and has become a standard engineering tool shared by a large number of researchers worldwide.

The success of convex optimization techniques is largely attributed to several of their unique features. First, very efficient and fast algorithms for solving convex problems have been developed and implemented, which makes convex optimization easy to use in practical wireless communication systems. Second, convex optimization often helps with gaining insight into the optimal solution structures that reveal the very nature of the problems in wireless communications. It makes the convex optimization framework a useful research tool. Third, the general theory of convex optimization is already relatively well developed which makes it very appealing for engineering applications. However, as time has shown, there is still a lot of room for research. This special issue is specifically devoted to such kind of studies with a main focus on the physical layer of wireless communication systems.

We have received about 30 paper submissions for this Special Issue by the deadline in December 2008. After extensive and careful reviews followed by the Editorial Board discussions, we accepted 7 papers that bear the highest quality and the best fit with the topic of this Special Issue. The accepted papers are categorized into 3 categories: Optimization Techniques for Resource Allocation in Wireless Systems, Optimization Techniques for Beamforming and Precoding in Wireless Systems, and Optimization Techniques for Scheduling in Wireless Systems. Three papers are included in each of the first and the second categories, while one paper fits under the third category.

Optimization Techniques for Resource Allocation in Wireless Systems . This part describes the recent advances on resource allocation for energy-constrained systems, broadcasting systems, and multi-user relay systems.

In the first paper, "On power allocation for parallel Gaussian broadcast channels with common information," Gohary and Davidson consider a broadcast system in which a single transmitter sends messages to a number of receivers over multiple unmatched parallel scalar Gaussian channels. The set of all rate tuples, for such systems, is parameterized by a set of power loads and partitions, and the problem of finding the boundaries of such sets is formulated as an optimization problem. Although this problem is non-convex, the tight inner and outer bounds can be efficiently computed. These bounds are computed using (convex) geometrical programming.

In "Power allocation and admission control in multiuser relay networks via convex programming: centralized and distributed schemes," Phan et al. address the power allocation problem for multiuser amplify-and-forward relay networks, in which multiple users share the same set of relay nodes. The problems of minimum rate and sum-rate maximization are shown to be convex. However, the joint power allocation and admission control problem is not convex that necessitates the development of approximate algorithms. Two configurations: centralized and decentralized are considered, while in the latter one the Lagrange decomposition method is applied.

In the third paper, "Stochastic resource allocation for energy-constrained systems," Sachs and Jones consider the battery-powered wireless systems with energy constraint. In the traditional resource allocation problem setup, allocation is done by assuming that the same tasks will run from the start-up until a specific future time. In this case, the energy and runtime constraints can be converted into a single power constraint. More general energy and runtime constraints are considered in this paper for the case when these constraints are not convertible into a single power constraint. The problem considered is NP-hard, where efficient stochastic recourse allocation method is developed based on the Lagrange optimization approach.

Optimization Techniques for Precoding and Beamforming in Wireless Systems . In this part, it is demonstrated how the optimization techniques can be used for developing precoding and beamforming methods in multiple-input multiple-output (MIMO) ad hoc networks, MIMO relay networks, and seamless ad hoc networks.

In the first paper of this category, "Transmission strategies in MIMO ad hoc networks," Fakih et al. address the precoding problem in MIMO ad hoc networks via maximizing the system mutual information under power constraints. A fast and distributed algorithm based on the quasi-Newton method is developed to solve the aforementioned problem.

In the paper, "Joint linear filter design in multiuser cooperative non-regenerative MIMO relay systems," Li et al. develop a new relay communication protocol in which linear filters are employed at both the transmitter and the relays. The joint design and optimization of transmitter and relay filters via the minimization of the mean squared error is considered. The work can be viewed as an extension of the traditional amplify-and-forward relay protocol.

In the last paper of this category, "On connectivity limits in ad hoc networks with beamforming antennas," Kiese et al. investigate the fundamental limits on the seamlessness/connectivity in multi-hop wireless networks with beamforming antennas. Authors use the popular "keyhole" antenna model, and formulate a mixed integer program for finding the optimal antenna configurations under various setups of path probability with various auxiliary constraints, node degree, and k-connectivity. A problem-specific large-scale optimization approaches are used to find the optimal antenna configurations efficiently.

Optimization Techniques for Scheduling in Wireless Systems consists on a single paper, "A scheduling algorithm for minimizing the packet error probability in clusterized TDMA networks," in which Toyserkani et al. consider a clustered wireless network, in which transceivers in a cluster use a time-slotted mechanism to access a wireless channel that is shared among several clusters. A scheduling algorithm which minimizes the derived average packet-loss probability is developed and tested.

We are excited to edit this high quality special issue within 8 months since the submission deadline. This would have been impossible without all those who contributed their research papers, numerous patient and diligent reviewers, and the EURASIP Journal on Wireless Communications and Networking Editorial Board and the Editor-in-Chief, Dr. Luc Vandendorpe. Our thanks go to all of them. We hope you will enjoy reading the carefully selected papers on the exciting research area of Optimization Techniques in Wireless Communications.

Sergiy A. VorobyovShuguang CuiYonina C. EldarWing-Kin MaWolfgang Utschick

Author information

Authors and affiliations.

Department of Electrical and Computer Engineering, University of Alberta, AB, Canada, T6G 2V4

Sergiy A. Vorobyov

Department of Electrical and Computer Engineering, Texas A&M University, College Station, TX, 77843, USA

Shuguang Cui

Department of Electrical Engineering, Technion - Israel Institute of Technology, Haifa, 32000, Israel

Yonina C. Eldar

Department of Electronic Engineering, Chinese University of Hong Kong, Hong Kong

Wing-Kin Ma

Institute for Signal Processing, Munich University of Technology, Munich, 80290, Germany

Wolfgang Utschick

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Sergiy A. Vorobyov .

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 2.0 International License ( https://creativecommons.org/licenses/by/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article.

Vorobyov, S.A., Cui, S., Eldar, Y.C. et al. Optimization Techniques in Wireless Communications. J Wireless Com Network 2009 , 567416 (2009). https://doi.org/10.1155/2009/567416

Download citation

Received : 06 September 2009

Accepted : 06 September 2009

Published : 28 September 2009

DOI : https://doi.org/10.1155/2009/567416

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Conductor-aware Wireless Underground Sensor Networks Based on Magnetic Induction: Research Challenges

Ieee account.

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

• IEEE Xplore Digital Library
• IEEE Standards
• IEEE Spectrum Online
• More IEEE Sites

Symposium on Terahertz Communications for Future Networks

Scope and motivation.

Wireless communications in the sub-terahertz and terahertz (THz) bands (or broadly speaking, from 100 GHz up to 10 THz) have been envisioned by both academia and industry as a key enabler of future sixth-generation (6G) wireless networks. The very large available bandwidth at THz frequencies offers enormous potential to alleviate the spectrum scarcity problem and break the capacity limitation of existing wireless communication systems.

Hence, THz communications are expected to support epoch-making wireless applications that demand reliable multi-terabits per second data rates, ranging from holographic communications, extended reality, and ultra-high-definition content streaming among mobile devices, to wireless backhaul connectivity and even satellite communication networks that can help us bridge the digital divide. Moreover, beyond communications, the THz band opens the door to new forms of wireless sensing beyond radar and localization, including air quality monitoring, climate change study and even nano-bio sensing for transformative healthcare applications.

Importantly, many prospective use cases for THz communication and networking systems will have to operate with large antennas, especially for mobile setups. Hence, the systems will have to often operate in the THz near field. Consequently, contributions on the performance boundaries in the THz near field, as well as possible solutions for near-field THz communication and sensing systems are of particular interests.

TOPICS OF INTEREST

The aim of this symposium is to collect the contributions presenting the latest updates on the THz wireless technologies development for communication and sensing purposes. We also welcome submissions presenting novel insights on the design, analysis, development, and deployment of THz communications, sensing, and networking solutions, related but not restricted to the following topics:

• Information-theoretic analysis of Terahertz communications
• Channel models for Terahertz communications
• Channel estimation techniques for Terahertz communications
• System-level modeling and experimental demonstrations for Terahertz communications
• Coexistence of Terahertz with millimeter wave and sub-6GHz transmissions
• Transceivers for Terahertz communications
• Antenna and massive antenna arrays for Terahertz communications
• Ultra-broadband modulation and waveform design for Terahertz communications
• Beamforming, precoding, and space-time coding schemes for Terahertz communications
• Near-field Terahertz communications
• MAC layer design for Terahertz communications
• Interference management for Terahertz communications
• Relaying and routing in Terahertz communications
• Mobility management (including near-field to far-field mobility)
• Spectrum and power allocation algorithms
• Terahertz for space communications
• Terahertz for nano-networks
• Terahertz for industrial IoT
• Terahertz for network sensing
• Terahertz for vehicular networks
• Terahertz for eXtended Reality (XR) services

SYMPOSIUM PAPER SUBMISSION

All papers should be submitted via  EDAS , Symposium on Terahertz Communications for Future Networks track. Full instructions on how to submit papers are provided on the IEEE FNWF 2024 website:  https://fnwf2024.ieee.org/authors/information-authors

Financial Co-Sponsors

UAV-enabled software defined data collection from an adaptive WSN

• Original Paper
• Open access
• Published: 29 April 2024

Cite this article

You have full access to this open access article

• Pejman A. Karegar   ORCID: orcid.org/0000-0003-2179-4199 1 ,
• Duaa Zuhair Al-Hamid 1 &
• Peter Han Joo Chong 1  

143 Accesses

Explore all metrics

Unmanned aerial vehicle (UAV)-based data gathering from wireless sensor networks is one of the recent research topics that has currently attracted research interest. One of the challenges for the UAV-aided WSN data collection efforts is to design an energy-efficient UAV/drone communication with arbitrarily dispersed ground sensors by improving the ground network structure. This paper aims to develop a technique titled UAV Fuzzy Travel Path' that supports UAV smooth path design and enables ground network topology shifting. A comprehensive UAV-based data collection model is proposed to enable dynamic orchestration/re-orchestration of wireless ground sensors to jointly improve network performance and UAV path fluidity. This provides a more flexible ground network framework that can be restructured based on network demands and UAV optimal paths, effectively allowing for a software-defined network concept. The main contribution of this work is the implementation of the software-defined wireless sensor network on the ground network that adaptably supports the movement of the UAV and enhances the communication network’s energy efficiency with a proposed latency analytical analysis via network orchestration/re-orchestration phases. The main significance of this research is in offering a flexible span for UAV path design than being fixed in one strict route for data gathering purposes. Four various simulation tools are employed for modelling and performance evaluation, namely MATLAB, CupCarbon, Contiki-Cooja and Mission Planner. The proposed software-defined ground network system demonstrates encouraging results in terms of network performance metrics including energy consumption of UAV versus ground sensor nodes energy usage, packet delivery rate, and the communication time of the ground orchestrated or/and re-orchestrated network.

Similar content being viewed by others

Unmanned aerial vehicles (UAVs): practical aspects, applications, open challenges, security issues, and future trends

Swarm intelligence algorithms for multiple unmanned aerial vehicles collaboration: a comprehensive review

UAV Path Planning Using Optimization Approaches: A Survey

Avoid common mistakes on your manuscript.

1 Introduction

WSNs have rapidly increased in popularity during recent years among research key subjects. WSNs typically include low-cost, battery-powered, and energy-constrained dispersed sensor nodes (SNs), each of which is restrained to limited energy resources, which makes it challenging to replenish after use. Therefore, to extend the lifespan of the entire network, energy-efficient communication strategies with less packet loss amongst SNs are essential. There are several approaches to mitigate energy usage and packet loss amongst ground networks. One of the solutions is the use of UAV, which offers more flexibility and manoeuvrability in data capturing use cases. The UAV can also be employed to offer a reasonably less polluted solution for data gathering from the dispersed ground SNs. It allows the WSN to pass their data through SNs’ representatives on the ground and then forward the buffered data to the UAV with higher percentage of Line of Sight (LoS) connectivity over vertical air-to-Ground communication rather than having plain horizontal on-the-Ground communication with Non-Line of Sight (NLoS) connectivity. Utilising the UAV for data gathering purposes from ground network could also facilitate the data gathering from the distant and isolated areas where the horizontal communication is not viable owing to uneven terrain shape or dense plantations.

UAV movement energy efficiency is also critical in WSN data gathering effort, as improving UAV energy efficiency can directly extend the flight time of the UAV to connect to more WSNs before it needs to be called back to recharge. Therefore, creating an energy-aware UAV path plan that takes into account the communication factors with SNs is essential to enhance the network performance cost.

Three key areas should be taken into account whereas designing a cost-effective UAV-based WSN data gathering model: an energy-efficient ground network architecture, a reliable UAV-Ground communication approach, and a power-efficient UAV path planning model. Each area could be represented by the main factors that can influence the structure of the proposed system.

Based on the conventional UAV-based WSN data gathering arrangement, the UAV path is restrained to limited options for collecting the data once physical topology of each group of ground sensor nodes is set. Setting the gateways to be selected dynamically offers adaptability to the ground network global topology taking the service demands into account. This allows for the groups topologies to be flexibly re-organise through software. The proposed concept of creating what we refer to as a ‘Fuzzy flight Path’ utilizes the capability to software redefine the groups representatives or network topology to align with the network requirements. The concept of adaptability of the UAV path with the ground SNs offers a flexibility for the entire network that ensures an improvement in the energy efficiency of either the UAV or the SNs. Therefore, the timely distribution of the workload of the collected data to a large number of gateways across multiple ground network structures can offer a host of advantages for the entire system. This includes improving the energy efficiency of the ground sensor network and allowing the UAV path design to be selected from multiple routes instead of being limited to one route. This ideology can eliminate the use of a single fixed architecture by distributing the workload of the data across multiple ground network entities, gateways, rather than a single gateway node.

With regard to wireless network energy efficiency and workload fairness in UAV-enabled data gathering use case, employing the recent technologies such as the software-defined network (SDN) and integrating the WSN three core functions, namely ‘leaf function’, ‘router function’, ‘Gateway function’ in the ground network through the softwarization approach to bring the concept of software-defined wireless sensor network (SDWSN) forward can offer a promising solution [ 1 ]. Herein, the control plane is separated from the data plane using a central controller such as a cloud computation processor [ 2 ]. The control plane transmits the logical operations and all relevant decisions to orchestrate the network structure, whereas the data plane forwards the data packets to an appropriate interface such as a cloud processing centre for computation. The separation of these two planes enables an intelligent routing mechanism and orchestration and re-orchestration (if necessary) of the network topology to support the network energy consumption and enhance UAV path through ground network flexibility.

This paper intends to leverage the SDWSN concept to facilitate a span for UAV path design definition. This necessitates flexible orchestration/re-orchestration of the topology providing flexibility for the network formations. It is worth mentioning that the main evaluation and performance metrics for the proposed approach focus on testing the energy consumption within the ground network and UAV, the packet delivery rate, and the communication time during the orchestration/re-orchestration process.

In the previous work [ 3 ], the UAV fuzzy path concept was briefly introduced and a preliminary point-to-point air-to-Ground communication between the UAV and the sensor nodes was initiated. Other previous work [ 4 ] focused on the UAV path relaxation concept within the fuzzy route in terms of UAV propulsion energy consumption, and preliminary in-Ground and air-to-Ground communication among the network components were established considering various ground SNs’ distributions and densities. Whereas within [ 5 ], the concept of SDN was aligned with the UAV path design to improve the ground network formation by proposing diverse packet frame designs for control and data packets signalling on the UAV-Ground communication. This paper aims to identify a solution for the UAV-aided WSN data gathering model, which considers three crucial areas jointly: ground network structure and re-structure, UAV-Ground communication, and UAV path planning model with utilization of the SDN functionality that can support the orchestration and re-orchestration of the ground network and thus offer the flexibility of the optimal UAV path. The main contributions of this work are summarized as follows:

The ‘Fuzzy Travel Route’ concept is defined as the UAV path span that enables the UAV flight path to be elected from a wider range of alternatives rather than being fixed in one defined path. This organisation allows the UAV path to be dynamically adjusted according to the updated ground network topology.

Proposing an effective solution for the integration between UAV path design, air-to-Ground connectivity, and ground network communication in a large field.

Applying the software-defined wireless sensor network (SDWSN) on the ground network that flexibly supports the movement of the UAV and enhances the energy efficiency of the communication network with latency analytical analysis through network orchestration/re-orchestration.

Obtaining an optimal UAV path design within the UAV fuzzy domain based on the updated network formations by defining an optimization problem considering jointly minimizing the UAV propulsion energy usage and ground SNs energy consumption while maximizing the packet delivery to the UAV.

The proposed model enhances the UAV path versus ground network energy efficiency with a higher percentage of served sensor nodes and an improved packet delivery rate to the UAV buffer. The remainder of this paper is structured as follows: Section 2 discusses the state-of-the-art and related work. Section 3 presents the proposed algorithms. Section 4 evaluates the proposed algorithm through extensive simulations. Finally, in Section 5 , conclusions of the work and future work suggestions are provided.

2 Related work

Wireless communications via flying unmanned vehicles such as drones has recently gained popularity due to its numerous advantages, including rapid assembling, controllable moving, and offering line-of-sight (LoS) communication with the ground base stations [ 6 ]. There are three various use cases for UAV- aided wireless communications including UAV-enabled mobile relaying [ 7 ], UAV-enabled base station [ 8 ], UAV-enabled data acquisition [ 9 ]. One special application for the UAV is through the use of that as a mobile relay. Zeng et al. [ 10 ] have proposed UAV-enabled multicasting systems in which a UAV is used to disseminate a common file to a set of ground terminals. The main goal of the research is to minimize the UAV mission completion time, whereas ensuring that each ground terminal can successfully recover the file with a targeting probability. In this study, a set of optimal waypoints for the UAV trajectory is found, and then the instantaneous UAV speed is optimised along with the path connecting these waypoints.

The UAV serves as a flying base station in UAV-based aerial base stations use case, offering a reliable communication particularly in emergency situations to the ground users [ 11 ]. Another use case of UAV-based wireless communication is reflected in data acquisition leveraging the UAV from the ground sensor networks. Numerous studies have addressed UAV data collection use cases from a distributed sensor network, where the UAV visits the sensor nodes individually or through the SNs representatives. The aim is to maximize network performance by mitigating the risk of high network latency and packet loss rate to identify the shortest UAV path that can serve the maximum data points within the permissible operational link during a limited flight time. To this end, Karunanithy et al. [ 12 ] have utilized UAV as an intelligent data collector for water irrigation applications, where a number of randomly distributed sensor nodes disseminate their data to the UAV based on a suggested UAV-Ground communication structure. To minimize signal attenuation, they propose a communication transaction diagram in which the UAV initiates communication to the corresponding SNs. The key issues in creating a scalable, energy-efficient and delay tolerant UAV-capable WSN data gathering model are the high mobility, frequent movements of the UAV over the ground network and disruptions in the communication network. To overcome this issue, UAV trajectory designing domain has been suggested based on a wide variety of ever-changing methods such as the geometric based path planning [ 13 , 14 , 15 ] and heuristically trajectory planning [ 16 , 17 , 18 ].

In keeping with the context of UAV path planning design, optimization solutions are proposed to overcome the mobility and frequent disruptions in the communication network. This includes the optimization of the UAV’s trajectory parameters such as altitude, velocity, and energy usage of the UAV whereas ensuring the reliable communication network. Zeng et al. [ 18 ] have suggested a theoretical model for energy efficiency of a fixed wing UAV that relates its propulsion energy consumption with the flying velocity and acceleration. The aim of the paper is to propose an efficient design for enhancing the UAV’s energy efficiency considering general constraints on the UAV movement ensuring maximal communication bit rate. However, integrating the energy efficiency of the ground network with the air-to-Ground scheduling and UAV path planning can offer more realistic outcomes. Herein, it is worth mentioning that other aspects of energy efficiency and scalability of the UAV-aided WSN should be considered to simultaneously curtail the energy consumption of the UAV and WSN in the design. In this regard, Zhan et al. in [ 9 ] have focused on the energy efficiency of the ground SNs by developing an optimization problem for SNs’ wake-up schedule and UAV’s trajectory to minimize the energy consumption of all SNs whereas ensuring that a target amount of data is forwarded from each SN to the UAV. The proposed scheme achieved significant energy savings for the SNs as compared with static data collector. Ebrahimi et al. [ 19 ] have also outlined the issue of energy-efficient data collection in dense WSNs using minimising the length of UAV flight path and SNs’ transmission power. A data gathering model that takes into account the energy efficiency of UAV and SNs whereas ensuring the maximum transmission rate of communication packets in either ground or air-to-ground communications is missing from all the above references. Moreover, all such heuristic methods share similar characteristics and shortcomings. They can get stuck in local minima, although they offer fast and near-optimal solutions.

One of the key operational processes in WSN is data dissemination and management between ground nodes. Herein, an efficient data routing solution supported by the ground network organization should be presented to manage the resource constraints of the nodes such as computational capabilities and nodes distribution. This highlights the need for an adaptive network structure approach that could follow the demands of UAV-Ground interaction. Due to the considerable flight altitude of the UAV, a direct connection of the UAV to all nodes is not an energy-efficient way of data collection. This problem can be solved by dividing the nodes into clusters/groups so that only the Cluster Head (CH) nodes can communicate with the drone. The clustering process involves two steps: selecting the CHs and constructing the cluster. Several protocols and algorithms have been designed for clustering, serving different purposes based on the network applications requirements, such as reduction the energy consumption in WSN. Sengaliappan et al. [ 20 ] have proposed an improved version of the existing general self-organized tree-based energy balance routing protocol [ 21 ]. The authors [ 20 ] based their work on clustering, assuming that the nodes are randomly distributed in a square field and that only one base station is located far from the area for the root node assignment. The cluster tree network is built, with the base station allocating a root node to each level and broadcasting this selection to all nodes. Although the simulation results show that the protocol performs better than the existing protocol [ 21 ] in terms of packet loss, involving the base station in the allocation of the root node in each round could introduce delays in the process.

The clustering criteria for selecting a CH has been of significant importance in formulating a stable and adaptive cluster structure. Herein, the selection criteria for the CHs in UAV-enabled data gathering applications can be based on parameters related to the remaining energy of the nodes, the position density of the nodes, and their distance from the drone path [ 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 ]. In addition, a clustering model should be based on improving various performance metrics such as increasing cluster size, enhancing WSN functioning time, minimizing data collection time and reducing latency in network-based clustering to deliver real-time data to the drone with less delay. In this regard, Bagga et al. [ 30 ] have proposed a cluster-tree based routing protocol for data dissemination to save energy. The criteria for selecting a CH is based on a cluster’s load that can be supported by its CH. In addition, the node energy, the distance to the sink and the degree of the neighbouring nodes are three parameters for cluster selection process in their work. However, selecting a CH from nodes placed around the centre of a grid could limit the role of the head to a small group of nodes. To further emphasize the role of clustering for the energy efficiency of the network, different types of clustering structures and cluster sizes can impact the energy efficiency of the network and thus the communication between the clusters and the UAV. Alagirisamy et al. [ 22 ] have utilized unequal cluster sizes in their WSN design to reduce the additional power consumed by the CHs once the data is routed to the sink. The results based on static and mobile sink nodes have indicated that considering unequal cluster sizes in the WSN design significantly extended the network lifetime of SNs. However, this method of clustering SNs can be efficient for the uniform distribution of sensor nodes, making it inapplicable for the random distribution of SNs which is close to an actual application.

As the number of distributed sensor nodes in the field increases, the complexity of the system becomes excessive, reducing data collection fairness between the network components. Herein, ground network management is required wherein the nodes need to be flexible in approaching the various tasks as well as processing the data. Hence, the nodes could be re-programmable when other tasks need to be prioritized during the network operation [ 31 ]. The Software-defined networking (SDN) concept has been proposed as another major trend in networking due to its potential in facilitating the network management, increasing network capability, easing virtualization within the network and allowing for innovation through network programmability [ 32 , 33 ]. SDN has been reflected in several applications such as offloading network computation from UAV-assisted vehicles to perform computationally complex and time-sensitive tasks whereas reducing the risk of higher network latency and packet loss rate [ 34 ]. In this regard, Zhang et al. [ 35 ] have utilized the SDN concept on the internet of UAVs, where the entire network can “forward looking” the uploaded information to potentially idle nodes to achieve the optimized system performance [ 35 ]. In a bid to enable adaptability and flexibility in WSN, the integration of SDN in WSN has been referred to as Software defined wireless sensor network (SDWSN) [ 31 , 36 , 37 , 38 ]. This integration can enhance the management and control of sensor networks taking into account the frequent changes to the network state and functions wherein the sensor functionalities can be adjusted by invoking various programs [ 39 ]. Furthermore, the network will be controlled and maintained easily in case of network failure. To examine the realistic communication link prior to the real implementation test, there are several methods to virtualize the network behaviour within the cloud [ 40 ]. Herein, the flexible orchestration of such ground WSN plays an important role in monitoring the operation of the physical environment. To a considerable extent, a cloud-based architecture tends to act as a viable solution since it encompasses a multitude of software-based computational capabilities, including virtualization and data management. The key components of cloud architecture are virtualization and softwarization, which contribute to network flexibility and can tackle the issues related to the response to any event. Software-driven virtualization offers a testing ground for conducting and analysing soft-trials of dynamic network scenarios. Such parallel co-simulation running in the cloud can significantly aid in leaning out the network configuration process by means of obviating the hardware requirements (during the testing process). For example, in our previous work [ 41 ], the Contiki-Cooja simulator is adopted as a virtualization platform for certain target hardware (Motes such as Texas Instruments CC2538 Evaluation Module). Karegar et al. [ 42 ] emphasize the importance of real world implementation for environmental monitoring using Raspberry Pi and Arduino in driving towards a flexible IoT-based sensor network organization. Cloud-based virtualization is adopted to plan and test various WSN orchestration scenarios when an event occurs. Network performance (i.e. packet loss, consumed energy, network downtime, etc.) can be analysed so that the most appropriate orchestration structure can be applied to the physical network [ 43 , 44 ]. This can support flexible network operation and lessen the impact of network failure [ 45 ].

The previous research works mostly focused on the fixed data gathering arrangement in which the UAV path is restrained to limited options for collecting the data once physical topology of each group of ground sensor nodes is set. This restriction results in the fixed UAV path design following the ground network topology set up. Hence, this shortcoming prompted us to look into new path designing algorithms which could facilitate the UAV path design flexibility by adapting to the changeable ground network topologies straight away.

To employ SDWSN concept on the UAV-enabled data gathering, the core idea of the fuzzy path approach is examined in our earlier works [ 3 , 4 , 5 ]. The proposed approach in this paper is based on a geographic grouping structure with one or more representative nodes that can act as ground data collection points (s). The active collection points are then assigned based on the proposed design of the fitness model. This not only leads to more efficient and smoother UAV travel planning, but it can also aid in the fair routeing of network traffic. The optimal and smooth UAV flight path should be aligned with the updated network structure analysis, which provides fairness for ground network power consumption with minimal communication latency per round.

3 System model

3.1 uav-based data collection of a distributed sdwsn: system organization and key assumptions.

Conventionally, UAV-assisted data collection is based on either direct connection with individual nodes or indirect communication with group representative nodes (are referred to as cluster head). Whereas grouping in this formation allows for more efficient data transmission, it still limits the UAV path for data gathering to the limited options once the actual topology of each group is structured. The main question is, can we relax the UAV route options by allowing the cluster heads to be elected to match the data collection path?

Flight path relaxation allows the UAV to lower its energy expenditure in turning points once the UAV is within the sharp edges. This facilitates the minimum variations in speed and acceleration required at turning points reducing the UAV flight propulsion energy consumption [ 4 ]. This method also allows for the dynamic redefining of group topologies via softwarization. In this case, the CHs are software redefined to meet the requirements of the flight route.

The term ‘Fuzzy Travel Route’ is used in this paper to describe the ability to software redefine the network organization to align with the UAV flying optimal route. It is necessary here to clarify exactly what is meant by UAV fuzzy path. Fuzzy travel route concept is defined as a UAV path span that enables the UAV flight path to be chosen from a wider range of alternatives rather than being fixed in one defined path. The choice of the UAV flight’s dynamic parameters such as path shape, flight speed and acceleration variations could provide more options and be designed to maximize the efficiency of travel whereas reducing data loss. This organization allows the UAV path to be dynamically adjusted in accordance with the updated ground network topology. The main contribution of this research is on employing the software defined network orchestration concept applied on ground network and exploring its impact on data collection process’s robustness and evaluating its effect on the UAV’s movement and ground network’s energy efficiency and jointly analysing the UAV path energy consumption and ground network energy expenditure.

The fuzzy route approach is based on the ability of modifying a given node functionality from leaf node role to a gateway one or vice versa. These nodes are known as gateway-capable nodes. Due to the UAV path relaxation arising from a larger fuzzy range, this method can enhance the overall system robustness and efficiency. Herein, the proposed topology structure eases the UAV path by incorporating efficient and optimal route designs within the UAV route fuzzy region and providing the resilience for adaptive gateway election closer to the UAV path. For generality, the proposed model considers the random distribution of ground network, considering both dense and sparse nodes’ distributions over a given testing area. The goal is to identify the model by taking the distribution and density of ground network into account as two key factors (discussed in the following section). According to the proposed model, neighbouring nodes that are within LoS of each other form a WSN group that is controlled by one or more ground representatives acting as cluster head(s).

The design of fuzzy path following with the inserted smooth and optimal paths for each distribution and density scenarios are shown in Fig.  1 a–d. The non-uniform path planning method via Bezier curves in sharp edges is one solution for path alignment within fuzzy route to overcome higher energy expenditure in sharp edges (See Fig.  1 a–c) [ 4 ]. This UAV path design, on the other hand, has variable speed, which causes inflexibility in real-world path planning and leads to UAV energy efficiency degradation. Hence, an example of a relaxed UAV path fitted within a fuzzy flight route using circular arcs/lines geometry is proposed as shown in Fig.  1 d [ 4 ]. This is demonstrated using two semi-circular routes with constant velocity and two linear and level flight paths with variable velocities. The proposed geometric semi-circular routes provide a better performance in terms of energy efficiency allowing for constant velocity travels for the UAV throughout the curves [ 4 ]. The velocity variation relaxation improves the UAV energy consumption for the entire tour and at the same time lower the complexity of the flying path. The detailed explanation on the designed methods of smooth paths is discussed further in our previous work [ 4 ]. In this paper, an optimal path design within the fuzzy range is obtained based on solving a proposed cost function for optimization problem.

The design of fuzzy route and smooth path with various distribution of gateway-capable nodes (from a to d from top left to down right, the distribution of gateway-capable nodes increased)

According to the simulation outcomes, using the fuzzy smooth and optimal path not only improves the energy efficiency of the entire tour but also provides the adaptability with the updated ground network formation. This paper also focuses on software defined enabled communication network formation via ground and air-to-Ground connectivity. The proposed model is highly dependent on the percentage of gateway-capable nodes distribution and density. As shown in these figures, the nominated gateways are represented with solid-line circles for each group whereas the remining potential gateways are depicted with dashed-line circles in the network structure. In short, the fuzzy flight route approach presents the potential for relaxing the specified options for the flight tour over a given group without jeopardizing the communication LoS. Depending on the capability of the WSN group members to presume the role of gateway, the more spread out the gateway-capable members, the better the smooth flight space.

3.2 The proposed network communication system

The proposed network communication architecture is based on splitting the entire system into multiple network components: Ground Network, Drone, Cloud stations. Leaf, Router and Gateway roles are three major roles in ground network formation as shown in Fig.  2 . A node may be assigned to one or more of these roles. The drone terminal is also a mobile vehicle that collects data from the ground and sends that to the cloud via the Internet for instantaneous orchestration/re-orchestration evaluations and virtualization purposes.

Topological arrangement for WSN data gathering

Mainly, the communication among the UAV and ground in the proposed SDWSN is based on a set of control/sensing data information messages that requires to be passed over the cloud to either get the network configured during a phase called ‘topological pre-orchestration scanning phase’ or gather the sensing data through another phase called ‘data collection post-orchestration phase’. The proposed system model requires another phase called ‘orchestration phase’ to notify the ground network entities about their new roles through the backward communication from the cloud. The communication among the three acting sub-units which involves in these three phases aims at:

Collecting the control data by the UAV on available gateways and the relevant nodes that can access them and passing the control data to the Cloud.

Cloud level analysis of the collected control data and identifying the elected gateways and related network structure and operational parameters.

Passing the outcomes of the analysis from cloud to the ground networks within control data information offering the updated ground networks set up.

Collecting the ground sensing data through the data collection post-orchestration phase as a subsequent phase running the UAV movement within the designed updated fuzzy route.

The sequence diagram of the signalling within the control and data flow collection phases is expressed in our previous work [ 5 ].

3.3 Various roles in the network system

The designed communication network framework is based on the definition of multiple ground/aerial network roles. This includes leaf nodes, router-capable nodes, gateway-capable nodes, drone level and remote cloud server. Leaf nodes are the lowest functional roles in the network, predefined as prior to the UAV traveling; Leaf nodes are responsible for the data acquisition from connected sensors and passing sensing data and the critical control information to the higher-layer routers or gateways. They are not in direct contact with the UAV during the SDWSN phases. Leaf nodes can also be software redefined to act as fully functioned router nodes or to enable or disable a given query or data processing function.

The router-capable nodes also have the capability to be software defined as the high-layer fully functioned router nodes or reduced-function leaf nodes via the control packets received through the gateway. Once elected as routers, they can transfer information from other leaf nodes to upper-layer routers or active gateways using multi-hop communication protocol. They are presumed not to have direct communication with the UAV. Their roles as routers/leaf nodes are assigned by the UAV in the orchestration phase.

The gateway-capable nodes can be software defined to operate as gateways or dropped to lower-level routers and leaf nodes. Depending on their updated softwarised roles and their distributions in random positions within the ground network, they can present a variety of network topologies such as tree or star ones. They also determine the UAV fuzzy path. Their roles have been notified by the drone during the orchestration phase. Once selected as gateways for a given network of sensors, they are responsible for collecting the data coming from the lower-layer nodes and instantaneously forwarding that to the drone during the data collection phase. The spatial arrangement in a large-scale network distribution may include more than one node defined as a gateway. The presence of multiple gateways on a large-scale network facilitates the prevention of rapid energy draining of gateway-configured nodes and improves reliability by mitigating the risk of the failure of a single gateway. Herein, the flexibility of the network to handle any changes, such as on-going healing in response to a rupture caused by the failure of network components/gateways, can be achieved through the backup/multiple nodes. Also, it can provide a broader range of UAV fuzzy path where the route is dynamically adjusted to benefit either the drone energy efficiency or the ground network’s energy consumption balancing.

The UAV acts as an upper-layer router between the gateways and the cloud server providing access to the remote cloud server to pass the data through upstream control data flow within the pre-orchestration phase or the operational sensing data flow within data collection post-orchestration phase and downstream flow of reconfiguration of ground stations during the orchestration phase. The cloud platform is the central station for processing real-world data passed by drone and virtualising network stations. The decisions for gateway and router elections have also been made in cloud servers using a proposed fitness model. Following the decision, the UAV fuzzy path along with the elected gateways and routers are assigned to each node’s ID in the remote cloud station. Next, through the notification messages, the decisions are returned to the selected routers and gateways as well as their assigned leaf-nodes to structure the ground network. The proposed topological arrangement for a given WSN has been broadly depicted in Fig.  2 .

In the proposed topological set-up, we assume that the gateway-capable nodes are spread such that one or more of them are not isolated from the remaining gateway capable nodes (there should not be any isolated gateway-capable node in the ground network).

The distribution of gateway-capable nodes factor \(D\) specifies the percentage of the gateway-capable nodes population out of the total sensor node population. Also, the gateway election factor \({\xi }_{i}\) represents the percentage of the gateway capable nodes population that have been reconfigured as gateways in each configuration election phase. The gateway election factor for each reconfiguration election process is obtained from:

where in \({n}_{i}\) is the number of gateways elected in \(i\) th reconfiguration election process, \(m\) stands for the number of predefined gateway-capable nodes.

The other factor is spread factor which reflects the dispersing aspect of elected gateway nodes within the ground network after each reconfiguration election process. That is obtained from:

wherein \({\sigma \_gc}_{i}\) is the spread factor after the election process of \(i\) th, \({q}_{0}^{i}\) is the coordinates of elected gateway nodes, \(\mu\) is the average coordinates of elected gateway nodes, \(N\) is the number of gateway nodes.

These three factors are the main parameters in defining the distribution and density of gateway-capable and elected gateway nodes among the total ground network entities.

As shown in Fig.  1 a–d, the effect of increasing the percentage of the gateway-capable nodes and spread factor ratio whereas maintaining a constant amount of gateway election factor on the fuzzy path region and hence UAV flying route is demonstrated. Once the percentage of the gateway-capable nodes and the spread factor ratio are limited to the specified threshold of \({D}_{gc}\left(th\right)\) and \({\sigma }_{gc}\left(th\right)\) , the UAV path shape loses its fluidity. On the other hand, this can be improved to a more relaxed shape with less sharp edges by utilizing the Bezier curve enhancement concept (according to Fig.  1 a–c) [ 4 ].

Conversely, once the percentage of the gateway-capable nodes and the spread factor ratio are higher than the given threshold, the UAV fuzzy region is expanded, facilitating more flexibilities for the UAV path definition, and hence the UAV path gets closer to the smooth/optimal path with minimal sharp edges (See Fig.  1 d). Hence, the proposed topology organization allows for the UAV’s smooth route capability by minimizing the variations in velocity, acceleration, and length of the path. This approach also provides the flexibility of gateways election closer to the UAV route design [ 4 ].

3.4 Energy consumption model for communication network

The proposed cost-effective energy model for evaluating the communication cost in both SDWSN topological scanning pre-orchestration and process data collection post-orchestration phases is represented as below:

where in the overall communication cost equals with the aggregation of communication costs of both sensors on the ground and access points ground communication with UAV. The amount of each network components’ power consumption for enabling the ground network communication equals with:

that means the average power consumption of each node is the summation of the average power consumption of node in four modes: Idle mode, Low power mode (LPM), receiving and transmitting modes. The Idle mode ( \({P}_{Idle}\) ) is activated whenever the node is listening (the time interval that the CPU is non-active prior to the radio transmitter or receiver gets active). LPM mode ( \({P}_{LPM}\) ) is activated when the sensor node goes to low power mode. Rx mode ( \({P}_{Rx}\) ) is activated in the radio receive mode and finally Tx mode ( \({P}_{Tx}\) ) is activated in transmission mode.

Sensor nodes operate in either active mode or sleep mode. The ratio of the time spent in active mode to a total data period is defined as duty cycle. In general, sensors consume energy mainly in data receiving and transmitting, and idle listening when they are in active mode.

The proposed energy model uses the Contiki powertracker to measure the time intervals that each node spends in these four modes [ 46 ]. Hence, the overall energy consumption per node can be calculated considering the equivalent consumed energy for these intervals as following:

where in \({P}_{i}\) represents the value of consumed power within each power mode, \({T}_{i}\) is the time spent during a specific mode i. The power calculation analysis has been conducted based on the information explored from CC2538 datasheet for values of current usages, once the module is in active receiving, transmitting, idle and low power modes as shown in Table  1 .

For enabling the air-to-Ground communication among the UAV and gateway nodes, we use the radio model [ 48 ], for modelling the energy consumptions of transmitting and receiving of data for \(b\) bits as shown in below:

\({E}_{elec}\) stands for transmitting circuit loss and \({d}_{0}\) is the threshold distance. \({\varepsilon }_{fs}\) and \({\varepsilon }_{amp}\) are the energy for power amplification in the free space channel model and multipath fading channel model respectively. The number of transmitted/received bits for each node is denoted by \(b\) . It is clear from ( 6 ) and ( 7 ) that the consumed energies for transmitter \({E}_{t}\) and receiver \({E}_{r}\) are highly dependent on received/disseminated bits as \(b\) and the instantaneous distance among the transmitter and receiver as \(d\) . The same energy model is implemented in CupCarbon for calculating the energy consumption of air-to-Ground communication. We presume that the transmission distance \(d\) during the communication among the UAV and each gateway is less than the threshold distance \({d}_{0}\) , and the free space channel model is adopted accordingly. As discussed in section 4 , the overall communication cost in both SDWSN topological pre-orchestration scanning and data collection post-orchestration phases are analyzed based on various distribution and density factors of ground network.

3.5 Fitness election model

The fitness model computation is presumed to be executed in the cloud environment enabling the updated structure of a ground network organization. The suggested parameters in the fitness model election process for the involved potential gateways are briefly expressed as:

3.5.1 Link quality based on radio signal strength intensity (RSSI)

The link quality between a UAV and its neighbour gateways is obtained by using the information of received signal strength indication (RSSI) of received packets. The link quality between UAV and each network component assuming free space path loss, LQ can be expressed as [ 34 ]:

Herein, \({N}_{rssi}\) is the total number of RSSI samples received on the UAV from each gateway-capable and \({R}_{k}\) is the \(RSSI\) value of the \(k\) -th sample.

3.5.2 Energy consumption factor per node

The overall energy consumption per node \({cost}_{i}\) is another key factor in election process calculated from ( 3 ).

3.5.3 Capacity factor per node

The capacity factor per node \({H}_{i}\) is another parameter in the proposed fitness model definition which is expressed as:

wherein \({M}_{i}\) is the current number of connected nodes to the gateway node \(i\) th and \({{Q}_{max}}_{i}\) is the maximum capacity of the gateway node \(i\) th. Note that once the number of connected nodes to the gateway equals with the defined maximum capacity, the capacity factor equals with 0. This means that the gateway is connected to its neighbors’ routers and leaf nodes with its full capacity, and it consumes higher energy than other gateways. This implies that this gateway should have a lower preference over the other gateways with higher capacity factor. Thus, the fitness model is defined as following:

wherein \({LQ}_{GWi-UAV}\) is the link quality factor of the gateway-capable node \(i\) th received on the UAV, \({Cost}_{i}\) is the accumulative energy cost of gateway-capable node \(i\) th, Lower the \(LQ\) and \(Cost\) parameters are, the higher likelihood of electing as gateway node, \({H}_{i}\) is the capacity factor of the gateway node \(i\) th. According to the proposed fitness model, the ground network can be structured in favor of the gateway-capable nodes election with better link quality, lower energy consumption and with higher capacities. \(\alpha , \beta\) and ζ denote three weights assigned to the three parameters based on their priorities in the updated network structure. Following the execution of election process computation in the cloud, the gateway-capable node with the higher value of \({W\_election}_{i}\) will be elected as the updated gateway node and those that are not elected as gateways will be dropped down to leaf node functionality. The updated ground network architecture will be organized based on the locations of the current gateway nodes.

3.6 An analytical formulation for UAV path optimization problem

Following with multiple scenarios assumed in the UAV fuzzy range, the output performance of the proposed approach includes the percentage of served sensor nodes, ground network energy consumption, and average UAV energy consumption. It is obvious that there is a trade-off between UAV propulsion energy consumption and ground network energy cost. While a heuristic UAV smooth path design [ 4 ] can offer an energy efficient path from the UAV cost point of view, it suggests an energy inefficient data gathering model from ground SNs perspective. The Bezier curve UAV path design [ 4 ] although results in UAV energy consumption degradation per mission and lowers the percentage of served sensor nodes, it benefits the ground network formation in terms of energy efficiency. Hence, it is required to define an optimization problem to solve the optimal solution considering jointly minimizing the UAV propulsion energy usage and ground SNs energy consumption while maximizing the packet delivery to the UAV.

To this end, to enhance the packet delivery in the air-to-Ground communication, a statistical model for modelling the communication throughput amongst the UAV and SNs considering LoS communications needs to be developed. The air-to-Ground connectivity model for occasional link blockage due to NLoS links is not presumed in this paper. Hence, the total amount of information bits transferred to the UAV over the duration \(T\) is a function of UAV trajectory expressed as [ 47 ]:

where \(B\) stands for the channel bandwidth, \({\gamma }_{0}\) is the reference received signal-to-noise ratio (SNR) at \({d}_{0} = 1\mathrm{ m}\) . \(H\) is the altitude of the UAV while flying over the ground SNs. Also, the UAV energy consumption for a fixed-wing UAV considering variable velocity \(v\left(t\right)\) and acceleration vectors \(a\left(t\right)\) is expressed in ( 12 ) [ 18 ]:

in which \({c}_{1}\) and \({c}_{2}\) are two parameters related to the aircraft’s weight, wing, air density, etc., \(g\) is the gravitational acceleration with nominal value \(9.8\,\mathrm{ m}/{{\text{s}}}^{2}\) , \(m\) is the mass of the UAV including all its payload. The speed of wind is considered zero.

To define the approximation optimization problem, a cost function of multiple parameters is required to be specified as:

where \(q\left(t\right)\) is the UAV path design. \({\overline{E }(q\left(t\right))}_{UA{V}_{-propulsion}}\) is the UAV propulsion power consumption obtained from ( 12 ), \({\overline{R }(q(t))}_{air-to-Ground}\) is the communication throughput obtained from ( 11 ) and \({Cost}_{Total}\) is the ground energy cost emerged from ( 3 ). According to our assumptions, \({(Cost}_{Ground-network})\) within ( 3 ) is static and not dependant on the location of the UAV, while \({(Cost}_{air-to-Ground})\) is highly dependent on the distance between the UAV and SNs and as a result, the locations of the UAV based on ( 6 ). Hence, the impact of the ground SNs energy consumption while communicating with each other \({(Cost}_{Ground-network})\) is disregarded in our optimisation problem formulation by only taking the UAV-Ground communication energy usage \({(Cost}_{air-to-Ground})\) into account.

Identifying the constraints for the optimization problem is dependent on the definition of the approximation for the proposed scenario. The outcome of optimization problem should probably be an optimal and energy efficient UAV path design that adapts to the updated ground network formations supporting the requirement of maximized communication throughput and mitigating the high values of ground energy cost. By discretizing the time horizon \(T\) into \(N + 2\) slots with step size \({\delta }_{t}\) , i.e., \(t = n{\delta }_{t}n = 0, 1, \cdot \cdot \cdot , N + 1\) , the UAV’s trajectory \(q(t)\) can be well characterized by the discrete-time UAV location \(q\left[n\right]=q(n{\delta }_{t}),\) the velocity \(v\left[n\right]= v(n{\delta }_{t})\) , as well as the acceleration \(a\left[n\right]=a(n{\delta }_{t})\) .

The constraints that should be satisfied on the UAV path optimization problem are UAV initial/final location and velocity, the minimum/maximum speed and acceleration and minimum throughput requirement on the UAV-Ground connectivity. The list of the constraints in the UAV path optimization problem is expressed in Table  2 .

Since some of the constraints in this table such as the minimum UAV speed and communication throughput constraints are not convex, solving approximation optimization problem in ( 13 ) is a challenging task due to the non-convex problem formulation. There are various mathematical algorithms to search the optimal results for non-convex problems. Herein, a specific mathematical solution for finding the optimal UAV path, velocity and acceleration with maximized communication throughput and minimized ground energy consumption considering the updated network formation is required. Sequential convex approximation (SCA) technique is chosen to solve the optimization problem using slack variables to convert the problem into linear programming which is solvable by CVX MATLAB.

3.7 Ground network latency analytical model

As the ground network can be flexibly structured based on distributed groups/clusters, with each CH communicating with the higher level of the system, i.e., UAV, latency analytical modelling can provide better analysis for network simulation. Herein, the topological pre-orchestration scanning phase entails latency within the pre-configuration round, whereas the orchestration phase entails latency within the election outcome notification round. The exchange of data between the leaf nodes and the gateway nodes can entail propagation and transmission latencies that need to be taken into consideration [ 49 ]. When the gateway node forwards the data to the UAV, the latency occurred by the data transmission from the gateway and the propagation latency between the gateway and UAV are part of the latency of the pre-orchestration phase. Finally, the data forwarded from the UAV to the cloud may also entail latency in data transmission of the UAV and propagation latency between the cloud and the UAV. The resulting transmission latency \({L}_{TR}\) and propagation latency \({L}_{Prop}\) are each expressed as:

where in rein \({P}_{length}\) is the length of a packet/message transmitted by a node. Herein, \({P}_{length}\) of a router or gateway node can vary depending on the number of connections. \({S}_{rate}\) is the communication message transmission rate.

where in \({L}_{Prop}\) is the difference between time stamps of the message receipt ( \({T}_{received}\) ) of the destination and transmission ( \({T}_{transmission}\) ) of the source node.

The latency experienced during the pre-orchestration phase is expressed as:

where in \(N\) is the total number of leaf nodes, whereas \(M\) is the total number of gateway nodes. The latency experienced during the notification round for the orchestration phase is expressed as:

where in \({T}_{Process(Fitness)}\) is considered as a configuration parameter in this work which defines the processing latency caused by running the fitness model for orchestrating the ground network. As the notification of the ground network orchestration takes place, the system can enable the data gathering post-orchestration phase.

It is worth mentioning that the communication latency is supported by the designed packet structure for sensing/control data packets based on our previous work [ 5 ]. The latency for data gathering post-orchestration phase is calculated the same as ( 16 ) taking the packet length and the number of hops into consideration. Herein, as the structure of the ground network post-orchestration can be based on multi-hop approach, \({L}_{Prop}\) may vary depending on the number of hops that the packet travels from source to destination.

4 Model testing and evaluation

To conduct testing and evaluation for the proposed model, several sequential processes utilizing multiple simulation tools must be followed. As shown in Fig.  3 , simulation through each software tool has composed of multiple computations that might get triggered in the event of either a defined component within the same simulation tool or an outputted parameter from other simulation tool. For the proposed model performance analysis in this thesis, four different software tools are used: MATLAB, Contiki-Cooja, CupCarbon, and Mission Planner. For instance, MATLAB output parameters can be applied as inputs for Cooja, CupCarbon and Mission Planner facilitating the network orchestration and smooth path design models at the same time. The outcomes of the proposed ground network softwarization and UAV path design are to enhance ground sensor nodes’ energy consumption, the communication latency, overall packet loss and the UAV energy expenditure during the smooth path design. According to Fig.  3 , each software module simulation is based on a sequence of multiple stages per module. The following section contains a detailed description of each task on each sub-module.

Development of software modules for the proposed model

4.1 Development of fuzzy path design

The simulation software used for UAV path design is MATLAB 2020b in which the UAV fuzzy route, UAV relaxed path and the air-to-Ground connectivity window of time for the spatially dispersed sensor nodes are defined [ 4 ]. The UAV fuzzy path concept is initially implemented in MATLAB to understand and investigate the impact of data-capturing dependent parameters on the proposed model. MATLAB also generates visual outputs that can be fed into SITL Mission Planner in order to validate the proposed paths. The performances including length of the path, mission time, instantaneous velocity and acceleration of the UAV and UAV energy efficiency are assessed in MATLAB. Within MATLAB, the first simulation module (See Fig.  4 ) is through the spatial distribution of ground sensor nodes considering two given variants: the distribution of gateway capable nodes and their density spread factor. The outputs of this module can be employed for ground network communication analysis in Contiki-Cooja as the locations of dispersed sensor nodes are transferred one by one from this module of MATLAB to the Cooja network simulator. Then, within the second component of testbed simulation in MATLAB, the UAV fuzzy route and UAV flight relaxed path are designed. Following that, the communication window of connectivity enabling interaction among the UAV and the elected gateways is sketched as part of the third process. The outputs of these two modules (UAV smooth path design and communication window of connectivity design) can support the air-to-Ground communication simulation in CupCarbon. The UAV smooth path design can be validated in Mission Planner based on real world scenario visualization. Finally, the UAV performance evaluation mainly on the UAV energy expenditure, length of the path and velocity of UAV for the proposed smooth path is worked out in [ 4 ].

UAV fuzzy range and smooth flight path design on MATLAB

According to Fig.  4 , various scenarios of SNs deployment are simulated in MATLAB based on given densities and distributions of sensor nodes and then fuzzy route is aligned based on the predefined locations of gateway-capable and router nodes. As the proposed window of connectivity is expressed on [ 4 ], the connectivity window of time is plotted over the UAV relaxed path within the fuzzy route (pink circle). The fuzzy route is identified (the hatched region in Fig.  4 ) accounting for the average percentage of Gateway-capable nodes distribution and spread factor equal to 20% and between (700 and 850), respectively. The optimal path is also represented in this figure as the pink UAV flying route within the fuzzy hatched range. The connectivity window is identified as the time once the UAV is within the entry and departed points calculated based on a range of UAV’s speeds [ 4 ]. Herein, with raising the average UAV’s speed, the average connectivity time is dropped which results in network performance degradation.

4.2 Ground network simulation based on SDWSN strategy

As discussed in the previous section, the network communication model in SDWSN is divided into three main phases: topological scanning pre-orchestration phase via control information messages, orchestration phase via notification messages and data gathering post-orchestration phase via sensing data information messages. To begin with, a test is performed for a network with multiple functionalities (Leaf, router, and gateway) in which the topological structure mode can be dynamically re-configured based on single hop, multi-hop network structures. The ground network is generated in Contiki-Cooja shown in Fig.  5 with all gateway-capable nodes involved in data gathering model (This is represented as the time preceding the network topological orchestration). In this network structure, the ground network consists of star network arrangements prior to orchestration phase. The UAV path design in this phase is obtained from solving an optimization problem for this specified network formation. Following the orchestration phase and passing the control data to the drone, the most appropriate gateways for each group are elected based on fitness model computation and various diverse network architectures such as one/multi-hop data transmission groups are formed (See Fig.  4 ). This phase can be represented as the data collection phase. The sensing data upstream flow takes place during this phase and the UAV traverses over each elected gateway to gather the accumulated data. Herein, the UAV path in this phase is considered as the proposed optimal paths from solving optimisation problem.

The distribution of network components within the UAV scanning pre- orchestration phase

The simulation goal is to evaluate the residual energy, packet delivery and communication latency performances either during scanning pre-orchestration phase or data gathering post-orchestration phase. A testbed is designed for ground network for various density spread factors \({\sigma }_{gc}\) with expanding the transmission message rate from 1 message per second to \(100\) messages per second. The ground network transmission ranges for all ground network components are set to \(50\)  meters. The simulation time is set to \(60\,\mathrm{ s}\) for data transferring and the designed packets frames for the scanning and data gathering phases are presumed based on the packet frame designs suggested in [ 5 ]. The simulation parameters are shown in Table  3 .

The ground network is structured in Contiki-Cooja considering various network formation scenarios depending on the election of gateway nodes out of the predefined gateway-capable nodes process. Herein, as shown in Figs.  5 and 6 , two various network formations for pre-orchestration and post-orchestration ground network topologies are assumed to observe the impact of network orchestration on the shape of the UAV path, ground SNs’ energy consumption, communication latency and the percentage of served SNs on the ground. Both architecture designs have two communication transaction message phases within them, which are the data flow orchestration/re-orchestration notification phase and the scanning or data gathering phase. During the data flow orchestration/re-orchestration notification phase, as discussed in our previous work [ 5 ], the updated functionality of each node emerging from fitness model decisions are returned to the elected routers and gateways as well as their specified leaf-nodes via the notification messages to orchestrate/re-orchestrate the ground network. In the scanning phase data flow, the packet frames are relayed to the potential gateways and then uploaded to the UAV to provide inputs for fitness model in order to re-define the functionality of the potential nodes, whereas, in the data gathering phase, the sensing packet frames are uploaded to the UAV.

The distribution of network components in Contiki-Cooja network simulator for post-orchestration data gathering phase

In the scanning pre-orchestration process, as shown in Fig.  5 , all gateway-capable nodes are involved in data gathering model with the gateway election factor of \(\xi\) =1 which creates the star network structure. Figure  7 upper plot depicts the data flow among the ground network components for a sample network group within this architecture. Whereas based on Fig.  6 , only a percentage of gateway-capable nodes are elected as gateways and involved in data gathering model with \(\xi =2/3\) ; hence, the multi hop based communication is generated to transfer the data from leaf nodes to the elected gateways. The data flow amongst the ground network components for a sample network group in this network architecture is represented in Fig.  7 lower figure. The UAV path designs for both scanning pre-orchestration and post-orchestration phases are drawn based on optimal paths enabling traveling over elected gateways. For each network structure, an optimal relaxed path is obtained based on the proposed cost function for the UAV to pass over the elected gateways to gather the data. Simulation outcomes including the values of received packets per gateway, the communication latency and ground energy consumption are recorded in Contiki-Cooja for each scenario. The ground network energy consumption model, as discussed in previous section, is based on the definition of each SN's cycling time for each state, including TX, RX, Idle, and Low Power. In an attempt to study the current consumption of the Cooja motes, the current profiles of two hops communication among the leaf nodes, routers and gateways are represented in Fig.  8 and 9 . The time of active TX, RX, Idle and Low Power states are obtained per cycling time based on the energytrace tool in Cooja network simulator. Also, the values of current consumption per each idle, active, and low power modes are explored from the datasheet of Texas Instrument CC2538 and presumed based on the Table  1 . The amount of voltage is considered as \(V=3\,\mathrm{ v}\) for the entire experiment. Figure  8 shows the current profiles for the Leaf and router-L1 nodes taken over \(1\,\mathrm{ s}\) . From the plot, it is easy to identify the packet interval of \(1\,\mathrm{ s}\) and to check that the device enters Power-mode in-between packets. First, the leaf node is on low power mode during the specified time cycle, then it changes its mode from low power mode to idle and later to transmission mode in order to initiate transmission (see Fig.  8 a). Then, once the whole packet is disseminated during the dissemination time, the module returns to low power mode to save power. Note that the duration in which the leaf node is in dissemination period is highly dependent on the size of designed packets, which equals 830 μs. Figure  8 b has also represented the value of current profile for the router-L1 once it allocates \(16320\) μs of its cycling time for receiving and 4530 μs for dissemination active modes respectively. There is a meaningful time in between receiving and transmitting modes for the sake of switching the transceiver from receiving mode to transmitting one. Note that the receiving time for router L-1 is higher than transmission time based on Cooja simulator due to the responsibility of receiving and relaying the data of several leaf nodes to the router-L2 at the same time.

Data flow amongst the ground network components for scanning pre-orchestration phase (upper figure) and subsequent post-orchestration data gathering phases (lower figure)

Current Consumptions for packet dissemination of each Leaf Node ( a ) and packet dissemination and receiving of each Router-L1 Node ( b ) per time cycle

Current Consumption for Packet transmission and receiving of router-L2 node ( a ) and packet receiving of gateway ( b ) per time cycle

Figure  9 a has also highlighted the current profile for the router-L2 taken over \(1\,\mathrm{ s}\) , once it allocates \(5360\) μs of cycling time for receiving and 13,840 μs for dissemination active modes respectively. The dissemination state time in router-L2 is longer than in router-L1 due to additional leaf node connection to the router-L2 in addition to other data relayed to router-L2 to be transferred to the coordinator. Finally, the current profile for the gateway node has been outlined in Fig.  9 b, in which the receiver mode is active for 13,840 μs during each cycling time.

Note that the aforementioned outcomes are part of an experiment to measure current profile for a specified two-hop network using the Contiki-Cooja simulator and CC2538 modules. The same experiment is carried out to examine the current consumption for alternative network structures such as star, single-hop network etc. The aggregated power usage is measured for a network of multiple diverse structures based on various distribution of network components such as those in Fig.  6 .

The simulation model is defined and analysed based on the obtained optimal UAV path and updated network architecture following the network orchestration phase in Fig.  6 , and the simulation outcomes are provided in Figs.  10 and 11 .

Packet received in Gateways from Leaf nodes in Ground network communication via Contiki-Cooja network simulator

Energy consumption of ground network in terms of message rate and gateway nodes’ spread factor via Contiki-Cooja network simulator

The simulation time is set to \(60\,\mathrm{ s}\) in Contiki-Cooja, since the average window of connectivity among the UAV and gateways in the air-to-Ground communication is \(60\,\mathrm{ s}\) based on the defined velocity of UAV, and the UAV-Gateways window of communication is presumed to be \(60\,\mathrm{ s}\) for all gateways. Hence, the simulation time for each group is considered \(60\,\mathrm{ s}\) in Contiki-Cooja, which is equivalent to the same period once the UAV is within the communication window of time of a specified gateway. Herein, the ground network communication performances such as accumulated received packets in gateways and the energy consumption of the entire network based on a range of various network sparsity \({\sigma }_{gc}\) and communication message rates are calculated and shown in Figs.  10 and 11 . As illustrated in these two figures, whenever network density increases, the ground network packet delivery and energy cost for message transmission from leaf to gateway are degraded due to effect of interference from other group members. Furthermore, it is obvious from these figures that as the message rate increases, the number of received packets in the gateways’ buffers decreases and the energy consumption of ground network increases. The reason for this is that as the message rate increases, so does the cycling time for the ground network components, causing the ground network components to consume more energy whereas delivering fewer packets to the gateways.

4.3 Air-to-ground communication simulation based on SDWSN strategy

CupCarbon network simulator is used for the air-to-Ground communication between elected Gateways and the UAV for the proposed SDWSN communication model. Numerous air-to-Ground data gathering scenarios can be developed in CupCarbon to assess the network performance of SDWSN communication model. The testbed design is based on defining multiple rounds for the UAV-Ground communication phases including scanning pre-orchestration or data gathering post-orchestration phases. Figures  12 and 13 depict these two simulation phases based on the distribution of selected gateways in CupCarbon.

The air-to-Ground Communication among the elected gateways and UAV for scanning pre-orchestration phase in CupCarbon

The air-to-Ground communication simulation for post-orchestration data gathering phase on CupCarbon

In both scenarios, all parameters are assumed to be fixed. The location of gateway nodes in CupCarbon is assumed to be the same as the testbed in the Contiki-Cooja ground network model, facilitating air-to-Ground communication with the drone. Also, the size of data buffered in each elected gateway is determined by the computed communication overhead per gateway in Contiki-Cooja. UAV velocity, length and shape of the path are based on the optimized path designs within the fuzzy route. The air-to-Ground transmission range, as shown in Table  3 , is considered up to \(550\,\mathrm{ m}\) in CupCarbon, once the average RSSI is below \(-\hspace{0.17em}80\,\mathrm{ dBm}\) . The reason for this is that the communication rate for this transmission range is usually fair [ 50 ].

The air-to-Ground connectivity is tested taking the movement of the UAV along the smooth and optimal paths over the defined communication window of time for each elected gateway into account. The UAV has the responsibility to pick the accumulated data from each gateway whereas passing over the communication range of each gateway. The path length equals with the length required by the UAV to pass through all gateways.

The simulation model is defined and analyzed based on the obtained optimal UAV path and updated network architecture for post-orchestration data gathering phase with the same fixed network parameters such as the gateway election factor of \(\xi =2/3\) and the distribution of gateway-capable nodes of \(D=20\mathrm{\%}\) in order to observe the impact of varying the density spread factor parameter on the network performance. The simulation outcomes are provided in Figs.  14 , 15 and 16 . The performances metrics including the percentage of overall packet delivery from leaf to the UAV, the energy consumption of UAV receiver, and the overall network cost for message transmission from leaf to the UAV are calculated in this section.

The overall packet delivery from leaf to the UAV based on message rate & gateway nodes’ spread factor

The energy consumption of UAV receiver in terms of message rate and gateway nodes’ density based on CupCarbon network simulator

The overall network cost for message transmission from leaf to the UAV based on message rate & gateway nodes’ density achieved from CupCarbon and Contiki-Cooja

Herein, the percentage of overall packets uploaded to the UAV buffer based on various communication message rates and network densities is represented in Fig.  14 . As shown in this figure, as the message rate increases from 1 to 5  \({\text{msg}}/{\text{sec}},\) the packet delivery rate on the UAV buffer decreases slightly, whereas once the message rate reaches in between 5 and 10  \({\text{msg}}/{\text{sec}}\) , the percentage of received packets on the UAV buffer stabilizes around a fixed value for a specific network density. The reason for this is that at these message rates, the UAV is capable to gather the higher amount of data stored in the gateways’ buffer during each gateway visit at its communication window of time, whereas with increasing the transferring message ratio from 20  \({\text{msg}}/{\text{sec}}\) , the ground network performance degrades moderately resulting in a decrease in the overall packet delivery to the UAV. In other words, once the message rate is bounded to lower values, the ground network outperforms the air-to-Ground one in terms of the percentage of uploaded packets via gateways and the UAV, since the air-to-Ground communication has a lower packet delivery rate due to the mobility of the UAV.

With increasing the message rate, the situation is reversed, the percentage of forwarded packets from gateways to the UAV rises whereas the ground network performance gets degraded. Hence, one solution to improve the end-to-end packet delivery performance is to adjust the speed of the drone considering the message rate, whereas the reason why the packet delivery is not raising to the values greater than 62% in this case is highly dependent on this issue. For velocities lower than \(20\,\mathrm{ m}/{\text{s}}\) , the UAV is capable to receive the entire buffered data that disseminated by the gateways. Furthermore, as depicted in this figure, whenever network density increases, the ground network packet delivery for the end-to-end message transmission from leaf to the UAV degrades due to effect of interference from other gateways on the UAV receiver.

Other energy-dependent performances including the amount of energy consumed by the UAV receiver and the overall energy consumption for message transmission from the leaf to the UAV are depicted in Figs.  15 and 16 . As shown in these two figures, as network density and message rate increase, the UAV energy cost of receiving data packets from elected gateways and consequently the overall network cost of message delivery from leaf to drone increases.

Also, the packet delivery for Ground network (from leaf nodes to gateways), the air-to-Ground network (from gateways to the UAV) and end-to-end network for two different phases (scanning pre-orchestration phase and data gathering post-orchestration phase) based on various network densities (from dense to sparse network) is represented in Fig.  17 . Note that for both scanning pre-orchestration and data gathering post-orchestration phases the optimal paths have been obtained. As shown in this figure, the packet delivery for scanning phase (either through ground, air-to-Ground and end-to-end communication) is higher than data gathering phase for various network densities. Moreover, the gateway election factor is set to \(\xi =1\) for scanning phase implying that all gateway-capable nodes contribute to passing the control data to the UAV, whereas the gateway election factor of the data gathering phase is elected as \(\xi =2/3\) , meaning that only a portion of gateway-capable nodes contribute to the data gathering process. As a result, packet delivery rate for scanning outperforms data gathering phase.

The packet delivery for Ground network (from leaf nodes to gateways), the air-to-Ground network (from gateways to the UAV) and end-to-end network for two different phases

Finally, the energy consumption for Ground network, the air-to-Ground network and end-to-end network for both the scanning phase prior to network orchestration and the data gathering phase following network orchestration based on various network densities are represented in Fig.  18 . It is obvious from this figure that data gathering phase consumes more energy than the scanning phase. Also, during both phases of network scanning and data gathering, the ground network consumes higher energy than the air-to-Ground network communication. One reason for this is that the number of network entities involved in the communication process in the ground network is higher than in the air-to-Ground one. Finally, by looking into the Figs.  17 and 18 at the same time, there is a meaningful energy gap between the overall energy consumption for data gathering and the network scanning phases, and the same gap is noticeable in overall packet delivery rate performance.

The energy consumption for Ground network (from leaf nodes to gateways), the air-to-Ground network (from gateways to the UAV) and end-to-end network for two different phases

The communication latency associated with the phases of pre-orchestration and post-orchestration based on various network densities is represented in Fig.  19 . Herein, the output of Contiki Cooja scenarios represented by the time stamps is supported by the proposed latency analytical model wherein the pre-orchestration phase experiences less communication latency due to the packet structure design [ 5 ] as well as the type of topology. In summary, the orchestration process through configuring the network roles is an efficient way to improve the network functionality as this process costs less with lower energy consumption and communication latency and at the same time higher packet delivery rates.

The communication latency for message transmission from leaf to the cloud via the UAV based on various network densities

To compare the suggested method with the method in [ 51 ], the proposed SDWSN network communication method shows better performance in terms of the rate of successful served sensor nodes following the configurability process. This comparison is based on various network scalability metric in which the proposed SDWSN network communication is compared to successive convex approximation (SCA) method in [ 51 ].

The network parameters are set to \(D=20\%\) for the distribution of gateway-capable nodes, \(\xi =2/3\) for the gateway election factor of the data gathering phase following the configurability process, and \({\sigma }_{gc}=350\) for the density spread factor which is in between the dense and sparse networks. According to Fig.  20 , with rising the number of dispersed nodes in the field, the number of served sensor nodes in the proposed method gains a larger gap from SCA algorithm. In addition, the complexity of the proposed communication network is exceedingly lower than the SCA algorithm [ 51 ] that utilizes complex non-convex/convex optimization algorithms to solve the connection-based UAV trajectory planning problems.

The comparison between the proposed SDWSN data gathering method and SCA algorithm [ 51 ] on the percentage of served sensor nodes versus scalability

4.4 Overall modelling discussion

Following the testing of the proposed SDWSN data gathering model in multiple scenarios using the suggested simulation tools, the output performance of the simulation tools, including served sensor nodes, ground network energy consumption, air-to-Ground packet delivery, average UAV energy consumption for scanning pre-orchestration and post-orchestration data gathering phases, and communication latency are recorded to evaluate the proposed model.

There is a trade-off between UAV propulsion energy consumption and ground network energy cost, as shown in Fig.  21 . Whereas post-orchestration process design can offer a cost-efficient path for the UAV and serve a higher percentage of ground sensor nodes for data gathering purposes, it suggests an inefficient model for ground sensor nodes, as expressed in Fig.  6 , due to multi-hop communication protocols in the ground network structure. In contrast, whereas the pre-orchestration process design consumes a higher amount of UAV battery per mission with lower percentage of served sensor nodes, it offers network orchestration/re-orchestration functionality to the ground network with lower amount of ground energy usage. To justify the existence of multiple gateway-capable nodes in the ground network, whereas deploying multiple gateway-capable nodes may deplete the sensor batteries and reduce the survivability of the ground network, this does not imply that a high percentage of node population should act as gateways at the same time. Based on the proposed model, the distribution of the gateway-capable nodes is bounded to 30% out of the entire network. Moreover, during the softwarization phase, the most effective gateway-capable nodes are elected as gateways which shows that the remaining gateways that are not nominated as gateways could be dropped to lower functionalities (such as leaf nodes) with lower energy expenditure.

The outcomes of multiple scenarios of data gathering model outputted from Cooja, CupCarbon and Mission Planner

The simulation outcomes of both scanning pre-orchestration and post-orchestration data gathering phases mentioned above revealed that orchestration plays an important role in serving more SNs on the ground with higher air-to-Ground packet delivery rates and improving UAV energy efficiency when compared to the situation prior to orchestration.

5 Conclusion and future work

This paper provided the motivational background behind the concept of UAV-based WSN data gathering and UAV’s feasibility to dynamically react to updated ground network. The SDWSN-enabled ground network communication was split into three main phases: the scanning topological pre-orchestration, the orchestration notification phase and sensing data collection post-orchestration phase. During the scanning and notification phases, control information messages were transferred, whereas during the data collection phase, the sensing data information was transferred to the cloud server. As the primary goal of data collection effort is on passing the sensing data information messages, the control information messages have the responsibility to pass the ground network topological orchestration data such as election data from the cloud-level to the ground network and assign the optimal functionalities to each component of the network prior to data gathering. Furthermore, the proposed model offers a flexible span for optimal UAV design paths dependent on the ground network structure.

Employing an efficient orchestration round to the UAV mission can offer benefits for better organization of the ground network. The simulation outcomes indicated that the added round is an efficient round with less consuming energy and high efficiency in data delivery and communication latency that can offer orchestration/re-orchestration functionality to the ground network. For future work, adding the UAV fixed wing to the scenarios of physical WSN ground network would provide the real time communication. Also, securing the data gathering stage by the UAV could benefit from the development of WSN secured clustering approach for data reliability prior to the stage of data gathering. In addition, using artificial intelligence AI-enabled models to predict the re-orchestration of the ground nodes once one or a group of nodes needs to be replaced due to any faulty situation would be useful to investigate as future work.

Al-Hamid, D. Z., Al-Anbuky, A. Vehicular intelligence: towards vehicular network digital-twin, In: 2022 27th Asia Pacific Conference on Communications (APCC). IEEE (2022).

Assefa, B. G., & Özkasap, Ö. (2019). A survey of energy efficiency in SDN: Software-based methods and optimization models. Journal of Network and Computer Applications, 137 , 127–143.

Article   Google Scholar  

Karegar, P. A., Al-Anbuky, A. Travel path planning for UAV as a data collector for a sparse WSN, In: 2021 17th International Conference on Distributed Computing in Sensor Systems (DCOSS). IEEE (2021).

Karegar, P. A., & Al-Anbuky, A. (2022). UAV-assisted data gathering from a sparse wireless sensor adaptive networks. Wireless Networks, 29 , 1–18.

Google Scholar  

Karegar, P. A., Al-Anbuky, A. UAV as a data ferry for a sparse adaptive WSN, In: 2022 27th Asia Pacific Conference on Communications (APCC). IEEE (2022).

Zeng, Y., Zhang, R., & Lim, T. J. (2016). Wireless communications with unmanned aerial vehicles: Opportunities and challenges. IEEE Communications Magazine, 54 (5), 36–42.

Zeng, Y., Zhang, R., & Lim, T. J. (2016). Throughput maximization for UAV-enabled mobile relaying systems. IEEE Transactions on Communications, 64 (12), 4983–4996.

Mozaffari, M., et al. (2016). Unmanned aerial vehicle with underlaid device-to-device communications: Performance and tradeoffs. IEEE Transactions on Wireless Communications, 15 (6), 3949–3963.

Zhan, C., Zeng, Y., & Zhang, R. (2017). Energy-efficient data collection in UAV enabled wireless sensor network. IEEE Wireless Communications Letters, 7 (3), 328–331.

Zeng, Y., Xu, X., & Zhang, R. (2018). Trajectory design for completion time minimization in UAV-enabled multicasting. IEEE Transactions on Wireless Communications, 17 (4), 2233–2246.

Hayajneh, K. F., et al. (2021). 3d deployment of unmanned aerial vehicle-base station assisting ground-base station. Wireless Communications and Mobile Computing, 2021 , 1–11.

Karunanithy, K., & Velusamy, B. (2020). Energy efficient cluster and travelling salesman problem based data collection using WSNs for Intelligent water irrigation and fertigation. Measurement, 161 , 107835.

Wu, Q., Zeng, Y., & Zhang, R. (2018). Joint trajectory and communication design for multi-UAV enabled wireless networks. IEEE Transactions on Wireless Communications, 17 (3), 2109–2121.

Yue, W., & Jiang, Z. (2018). Path planning for UAV to collect sensors data based on spiral decomposition. Procedia computer science, 131 , 873–879.

Zhan, C., Zeng, Y., & Zhang, R. (2018). Trajectory design for distributed estimation in UAV-enabled wireless sensor network. IEEE Transactions on Vehicular Technology, 67 (10), 10155–10159.

Ghorbel, M. B., et al. (2019). Joint position and travel path optimization for energy efficient wireless data gathering using unmanned aerial vehicles. IEEE Transactions on Vehicular Technology, 68 (3), 2165–2175.

Xu, Y., et al. (2018). Energy-efficient UAV communication with multiple GTs based on trajectory optimization. Mobile Information Systems, 2018 , 1–10.

Zeng, Y., & Zhang, R. (2017). Energy-efficient UAV communication with trajectory optimization. IEEE Transactions on Wireless Communications, 16 (6), 3747–3760.

Ebrahimi, D., et al. (2018). UAV-aided projection-based compressive data gathering in wireless sensor networks. IEEE Internet of Things Journal, 6 (2), 1893–1905.

Sengaliappan, M., Marimuthu, A. Improved general self-organized tree-based routing protocol for wireless sensor networK, Journal of Theoretical & Applied Information Technology, 68(1), (2014)

Han, Z., et al. (2014). A general self-organized tree-based energy-balance routing protocol for wireless sensor network. IEEE Transactions on Nuclear Science, 61 (2), 732–740.

Alagirisamy, M., & Chow, C.-O. (2018). An energy based cluster head selection unequal clustering algorithm with dual sink (ECH-DUAL) for continuous monitoring applications in wireless sensor networks. Cluster Computing, 21 (1), 91–103.

Kalaivanan, K., & Bhanumathi, V. (2018). Reliable location aware and cluster-tap root based data collection protocol for large scale wireless sensor networks. Journal of Network and Computer Applications, 118 , 83–101.

Velmani, R., & Kaarthick, B. (2014). An efficient cluster-tree based data collection scheme for large mobile wireless sensor networks. IEEE sensors journal, 15 (4), 2377–2390.

Tunca, C., et al. (2014). Ring routing: An energy-efficient routing protocol for wireless sensor networks with a mobile sink. IEEE Transactions on Mobile Computing, 14 (9), 1947–1960.

Hasheminejad, E., & Barati, H. (2021). A reliable tree-based data aggregation method in wireless sensor networks. Peer-to-Peer Networking and Applications, 14 (2), 873–887.

Kiamansouri, E., Barati, H., & Barati, A. (2022). A two-level clustering based on fuzzy logic and content-based routing method in the internet of things. Peer-to-Peer Networking and Applications, 15 (4), 2142–2159.

Ataei Nezhad, M., Barati, H., & Barati, A. (2022). An authentication-based secure data aggregation method in Internet of Things. Journal of Grid Computing, 20 (3), 29.

Ghorbani Dehkordi, E., & Barati, H. (2023). Cluster based routing method using mobile sinks in wireless sensor network. International Journal of Electronics, 110 (2), 360–372.

Bagga, N., et al. (2015). A cluster-tree based data dissemination routing protocol. Procedia Computer Science, 54 , 7–13.

Ndiaye, M., Hancke, G. P., & Abu-Mahfouz, A. M. (2017). Software defined networking for improved wireless sensor network management: A survey. Sensors, 17 (5), 1031.

Zilberman, N., et al. (2015). Reconfigurable network systems and software-defined networking. Proceedings of the IEEE, 103 (7), 1102–1124.

Baktir, A. C., Ozgovde, A., & Ersoy, C. (2017). How can edge computing benefit from software-defined networking: A survey, use cases, and future directions. IEEE Communications Surveys & Tutorials, 19 (4), 2359–2391.

Pu, C., Link-quality and traffic-load aware routing for UAV ad hoc networks, In: 2018 IEEE 4th International conference on collaboration and internet computing (CIC). IEEE, (2018).

Zhang, C., Dong, M., & Ota, K. (2021). Deploying SDN control in Internet of UAVs: Q-learning-based edge scheduling. IEEE Transactions on Network and Service Management, 18 (1), 526–537.

Kobo, H. I., Abu-Mahfouz, A. M., & Hancke, G. P. (2017). A survey on software-defined wireless sensor networks: Challenges and design requirements. IEEE Access, 5 , 1872–1899.

Al-Hamid, D. Z., Al-Anbuky, A., Vehicular network dynamic grouping scheme, In: 2021 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion (ACSOS-C). IEEE (2021).

Al-Hamid, D. Z., Al-Anbuky, A., Vehicular grouping protocol: towards cyber physical network intelligence, In: 2021 IEEE International Conferences on Internet of Things (iThings). IEEE (2021).

Zeng, D., et al., Evolution of software-defined sensor networks, In: Mobile Ad-hoc and Sensor Networks (MSN), 2013 IEEE Ninth International Conference on. IEEE (2013).

Khan, I., et al. (2015). Wireless sensor network virtualization: A survey. IEEE Communications Surveys & Tutorials, 18 (1), 553–576.

Al-Hamid, D. Z., & KaregarChong, P. A. P. H. J. (2023). A novel SDWSN-based testbed for IoT smart applications. Future Internet, 15 (9), 291.

Karegar, M. A., Kusche, J., Geremia-Nievinski, F., & Larson, K. M. (2022). Raspberry Pi reflector (RPR): A low-cost water-level monitoring system based on GNSS interferometric reflectometry. Water Resources Research . https://doi.org/10.1029/2021WR031713

Acharyya, I. S., Al-Anbuky, A., Sivaramakrishnan, S. Software-defined sensor networks: towards flexible architecture supported by virtualization, In: 2019 Global IoT Summit (GIoTS). IEEE (2019).

Ezdiani, S., et al., An IoT environment for WSN adaptive QoS, In: 2015 IEEE International Conference on Data Science and Data Intensive Systems. IEEE (2015).

Al-Hamid, D. Z., Al-Anbuky, A. Vehicular grouping and network formation: virtualization of network self-healing, In: International Conference on Internet of Vehicles. Springer (2018).

Amirinasab Nasab, M., et al. (2020). Energy-efficient method for wireless sensor networks low-power radio operation in internet of things. Electronics, 9 (2), 320.

Instruments, T., CC2538 Powerful Wireless Microcontroller System-On-Chip for 2.4-GHz IEEE 802.15.4, 6LoWPAN, and ZigBee® Applications, (2015).

Ren, J., et al. (2015). Lifetime and energy hole evolution analysis in data-gathering wireless sensor networks. IEEE transactions on industrial informatics, 12 (2), 788–800.

Al-Hamid, D. Z., & Al-Anbuky, A. (2023). Vehicular networks dynamic grouping and re-orchestration scenarios. Information, 14 (1), 32.

Yanmaz, E., et al., Experimental performance analysis of two-hop aerial 802.11 networks, In: 2014 IEEE Wireless Communications and Networking Conference (WCNC). IEEE (2014).

Samir, M., et al. (2019). UAV trajectory planning for data collection from time-constrained IoT devices. IEEE Transactions on Wireless Communications, 19 (1), 34–46.

Download references

Acknowledgements

The authors would like to thank the Department of Electrical and Electronic Engineering in the School of Engineering, Computer and Mathematical Sciences at Auckland University of Technology for providing advice for this research project.

Open Access funding enabled and organized by CAUL and its Member Institutions.

Author information

Authors and affiliations.

Electrical and Electronic Engineering, Auckland University of Technology, Auckland, New Zealand

Pejman A. Karegar, Duaa Zuhair Al-Hamid & Peter Han Joo Chong

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Pejman A. Karegar .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Karegar, P.A., Al-Hamid, D.Z. & Chong, P.H.J. UAV-enabled software defined data collection from an adaptive WSN. Wireless Netw (2024). https://doi.org/10.1007/s11276-024-03744-y

Download citation

Accepted : 02 April 2024

Published : 29 April 2024

DOI : https://doi.org/10.1007/s11276-024-03744-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Unmanned aerial vehicle (UAV)
• UAV fuzzy travel path
• Software-defined wireless sensor network (SDWSN)
• Find a journal
• Publish with us
• Track your research