Search

In unmanned aerial vehicle ad‐hoc network (UANET), the node speed of unmanned aerial vehicles (UAVs) may reach up to 400 km/h. The fast or slow movement of UAV nodes leads to different speeds of topology change of the nodes. Traditional optimized link state routing (OLSR) protocol cannot adaptively adjust the routing update period when the network topology changes, which may lead to the nodes calculating incorrect routing tables. This increases the average end‐to‐end delay and packet loss rate for packet transmission. To enhance the adaptability of OLSR routing protocol to network topology changes, this paper proposes a multi‐agent independent deep deterministic policy gradient‐OLSR (MA‐IDDPG‐OLSR) routing protocol based on distributed multi‐agent reinforcement learning. The protocol deploys DDPG algorithm on each UAV node, and each UAV node adaptively adjusts the Hello and TC message sending intervals, according to the one‐hop neighbouring nodes as well as its own state. Simulation results show that the proposed protocol is able to improve the throughput and reduce the packet loss rate as compared to traditional AODV, GRP, OLSR, and distributed multiple‐agent independent proximal policy optimization‐OLSR (MA‐IPPO‐OLSR), distributed multiple‐agent independent twin delayed deep deterministic policy gradient‐OLSR (MA‐ITD3‐OLSR) routing protocols. Since MA‐IDDPG‐OLSR relies only on local information, there is a minor performance degradation in MA‐IDDPG‐OLSR compared to centralized single‐agent DQN‐OLSR routing protocol. But it is more suitable to a completely distributed UAV network without a centralized node. [ABSTRACT FROM AUTHOR]

Nowadays, the development of Artificial Intelligence of Things (AIoT) is advancing rapidly, and intelligent devices are increasingly exposed to more security risks on the network. Deep learning‐based intrusion detection is an effective security defence approach. Federated learning (FL) is capable of enabling deep learning models to be trained on local clients without uploading their data to a central server. This paper proposes a novel federated learning aggregation algorithm called fed‐dynamic gravitational search algorithm (Fed‐DGSA), which incorporates the GSA algorithm to optimize the weight updating process of FL local models. During the updating process, the decay rate of the gravity coefficient is optimized and random perturbations and dynamic weights are introduced to ensure a more stable and efficient FL aggregation process. The experimental results show that the detection accuracy of Fed‐DGSA has reached about 97.8%, and it is demonstrated that the model trained using Fed‐DGSA achieves higher accuracy compared to Fed‐Avg. [ABSTRACT FROM AUTHOR]

Multi‐user (MU) massive multiple input multiple output (mMIMO) is considered a potential technology for fifth generation (5G) and sixth‐generation (6G) wireless systems. The presence of the antenna arrays at the base station level to communicate with the users or to serve tens of single antenna users leads to excessively high system costs and power consumption. The deployment 1‐bit digital‐to‐analogue converters (DACs) in the base station can solve these problems. This paper starts by presenting an analytical study centered on the effects of 1‐bit DACs on the system envisaged for a Rayleigh‐type fading channel. Compact‐form expressions are derived for the symbol error rate. Afterwards, an efficient end‐to‐end deep learning technique to compensate for the joint effect of 1‐bit DAC and imperfect channel state information in downlink mMIMO systems. Moreover, to improve the performance of the considered system, a DAC mixed architecture is proposed, where a number of antennas use 1 bit DACs while the others do not. The simulations results showed the improvement in transmission quality of the downlink of the MU‐mMIMO system in the presence of hardware imperfections using the considered end‐to‐end compensation technique. [ABSTRACT FROM AUTHOR]

This paper presents a multi‐input deep learning‐based joint pilot decontamination and symbol detection (SD) technique for 5G massive multiple‐input multiple‐output (MAMIMO) systems. It consists of a fully convolutional neural network (FCNN) that finds the 5G channel coefficients using pre‐known sounding reference signals (SRSs) under pilot contamination and a symbol detector derived from projected gradient descent iterations. The study considers pilot contamination caused by inter‐cell interferences between SRSs during channel estimation (CE). The proposed scheme accepts entire 5G orthogonal frequency modulated (OFDM) data and least square estimates and produces the transmitted OFDM signal. Simulation experiments demonstrated that the proposed technique has better CE and SD performance with reduced trainable parameters. Moreover, it is faster due to the lowest elapsed time during end‐to‐end OFDM symbol detection. This paper proposes a joint pilot decontamination and signal detection for 5G MAMIMO systems. It achieves better detection performance with the lowest number of trainable parameters and memory requirements. It is applicable in 5G OFDM symbol detection and channel estimation under pilot contamination. [ABSTRACT FROM AUTHOR]

The current wireless communication system has higher performance demands for receivers. This paper considers exploiting the natural redundancy (NR) that is widely existed in the transmission sources to improve the receiver's performance of sequence detection. The type of NR discussed in this paper is the inherent redundancy in data caused by the correlation between symbols in the memory sources. Since the correlation between transmitted symbols is too challenging to obtain, deep learning (DL) is introduced into sequence detection, which has a powerful ability to extract complex patterns from data. Specifically, intended for a specific memory source that is convenient for performance analysis, namely the Markov source, an iterative sequence detection algorithm based on the long short‐term memory network is proposed. Simulation results demonstrate that the receiver's performance can be improved by exploiting the NR and the proposed DL‐based sequence detection scheme can obtain optimal bit error rate performance with blind state transition probability of the transmitted Markov source. [ABSTRACT FROM AUTHOR]

Delta compression, as a complementary technique for data deduplication, has gained widespread attention in network storage systems. It can eliminate redundant data between non‐duplicate but similar chunks that cannot be identified by data deduplication. The network transmission overhead between servers and clients can be greatly reduced by using data deduplication and delta compression techniques. Resemblance detection is a technique that identifies similar chunks for post‐deduplication delta compression in network storage systems. The existing resemblance detection approaches fail to detect similar chunks with arbitrary similarity by setting a similarity threshold, which can be suboptimal. In this paper, the authors propose Chunk2vec, a resemblance detection scheme for delta compression that utilizes deep learning techniques and Approximate Nearest Neighbour Search technique to detect similar chunks with any given similarity range. Chunk2vec uses a deep neural network, Sentence‐BERT, to extract an approximate feature vector for each chunk while preserving its similarity with other chunks. The experimental results on five real‐world datasets indicate that Chunk2vec improves the accuracy of resemblance detection for delta compression and achieves higher compression ratio than the state‐of‐the‐art resemblance detection technique. [ABSTRACT FROM AUTHOR]

Gamma spectrum denoising can reduce the adverse effects of statistical fluctuations of radioactivity, gamma ray scattering, and electronic noise on the measured gamma spectrum. Traditional denoising methods are intricate and require analytical expertise in gamma spectrum analysis. This paper proposes a segmentation‐enhanced Convolutional Neural Network‐Stacked Denoising Autoencoder (CNN‐SDAE) method based on convolutional feature extraction network and stacked denoising autoencoder to achieve gamma spectrum denoising, which adopts the idea of data segmentation to enhance the learning ability of the neural network. By dividing the complete gamma spectrum into multiple segments and then using the segmentation‐enhanced CNN‐SDAE method for denoising, the method can achieve adaptive denoising without manually setting the threshold. The experimental results show that our method can effectively achieve gamma spectrum denoising while retaining the characteristics of the gamma spectrum. Compared with traditional methods, the denoising speed and effectiveness have been significantly improved, and the proposed method demonstrates an approximately 1.72‐fold enhancement in smoothing performance than the empirical mode decomposition method. Furthermore, in terms of retaining gamma spectrum characteristics, it also achieves a performance improvement of approximately three orders of magnitude than the wavelet method. [ABSTRACT FROM AUTHOR]

Internet‐of‐things is an enabling technology in the fourth‐generation industrial revolution. The freshness of the data sent by a sensor node (SN) is an important parameter in the Internet‐of‐things, unlike the throughput in cellular communications. A relatively new performance metric named age of information (AoI) is used in this paper to quantify the freshness of the data. The SN, located in the transport infrastructure, harvests energy from radio frequency signals transmitted by the FD‐UAV. This is used to transmit real‐time sensor observations to the data sink via FD‐UAV. The SN generates an update after replenishing the battery and transmits it by using the harvested energy. A closed‐form expression for average AoI is derived as a function of time allocated for energy harvesting. The optimal time allocation for energy harvesting that maximizes the freshness of data update is identified. A deep learning technique namely long‐short term memory is used to predict the average AoI. Simulation results demonstrate the usefulness of the performance bounds in terms of the freshness of data updates. [ABSTRACT FROM AUTHOR]

This paper investigates intelligent reconfigurable surface‐assisted non‐orthogonal multiple access (NOMA) networks for short packet communications where the source with multiple antennas utilizes beamforming technology to serve two end users including a near user and a far user. The analysis system, closed‐form expressions of the average block error rates (BLERs), latency and reliability at all users are derived under assuming perfect channel state information over quasi‐static Rayleigh fading channels. For evaluating the improvement of the network system, a deep neural network (DNN) framework is designed to predict the performance parameters in two cases: (i) BLERs prediction of the users, (ii) transmit power allocation under the constraints of BLERs at each user which satisfy the extremely high reliability requirements of uRLLCs. The extensive results demonstrate that first, the DNN‐based estimation results are more accurate than Monte‐Carlo simulation results with low execution time. Second, the allocated power for each user can be accurately predicted at the high layers. Finally, two methods of evaluating the effectiveness of the DNN model, i.e. root mean square error and mean absolute percentage error, reach low error which verify the power of the designed DNN model for future analysis. [ABSTRACT FROM AUTHOR]

Thanks for its powerful feature extraction capability and unprecedented success in computer vision and language processing tasks, deep learning (DL) has also been applied in recent years to wireless communications. However, DL models are proved to be inherently vulnerable to adversarial perturbations: carefully crafted perturbations that appear imperceptible but can fool the models to induce misclassification. Existing researches in wireless communications employ the gradient‐based or optimization based methods, for example, FGSM, PGD and iterative FGSM, to craft input‐dependent adversarial perturbations. However, these methods have a high computational complexity and the perturbations they craft cannot be readily applied to fool DL‐based algorithms in wireless communications. In this paper, the efficient generation of universal (input‐agnostic) adversarial perturbations (UAPs) used to attack DL‐based modulation classification algorithms is studied. Communication signals exhibit unique characteristics, for example, received modulated signals after transmitting through noisy channels are densely distributed around the modulation constellation points. In this work, the authors have proposed, by making use of this unique characteristics, a generative network to model the distribution of adversarial perturbations for high‐efficiency generation of UAPs, which can fool the DL‐based modulation classification algorithms for most inputs. Instead of attacking all regions of modulated signals, attention mechanisms are introduced in generative network design that enables the perturbations to be concentrated around the constellation points. In this way, a higher attack efficiency and a lower perceptibility are achieved by crafting UAPs with smooth variations and more focused attack. In addition, a diversity term is included in loss function to help capture a wide range distribution of adversary perturbations that will help the generation of more diverse adversary perturbations. Once trained, the generative network can construct with ease a large number of UAPs with some favorable features, which will benefit more efficient adversarial attack and faster adversarial training. Experimental results show the superiority of our proposed method in terms of attack efficiency, imperceptibility and diversity over existing methods. [ABSTRACT FROM AUTHOR]

This paper presents a path loss model based on path profile in urban propagation environments for 5G systems. Although deep learning approaches are indeed powerful in tasks involving prediction or classification, they often lack transparency and suffer from high computational complexity. The proposed model combines the log‐distance path loss model for line‐of‐sight propagation scenarios and a machine‐learning‐based model for non‐line‐of‐sight (NLOS) cases. This paper uses the principal component analysis algorithm to extract relevant features out of some selected attributes of the path profile for NLOS cases. Then, the path loss model can be constructed based on the approach of polynomial regression. Simulation results show that the proposed model outperforms the conventional models when operating in the 3.5 GHz frequency band. The standard deviation of prediction error was reduced by about 22.2–37.2% dB when compared to the conventional models. Furthermore, the prediction performance was also evaluated in a non‐standalone 5G New Radio network in the urban environment of Taipei city. The real‐world measurements show that the standard deviation of prediction error can be reduced by 3.33–6.13 dB when compared to the conventional models. [ABSTRACT FROM AUTHOR]

Aiming at the issues of low individual identification accuracy of radiation sources under the condition of little samples, this paper proposes a way of individual identification of radiation source based on strengthening global deep convolutional generative adversarial network (SGDCGAN) and deep convolutional neural network (DCNN) to achieve data augmentation. The method first performs IQ map feature splicing processing on the input signal, and then uses DCNN to automatically obtain the deep essential features of the data. Besides, an adaptive improvement is created to the deep convolutional generative adversarial network, and a self‐attention mechanism is introduced into the discriminator and the generator to reinforce the integrity and authenticity of the generated samples. For the common gradient disappearance during model training, the gradient penalty mechanism and spectral normalization are added to make the training process more stable. Through comparison experiments on the collected ADS‐B signals, the experimental results have demonstrated that when the signal‐to‐noise ratio is 0 dB and the number of original samples in each class is 40, the recognition accuracy is improved by 23.6% after doubling the data. Compared with the DCGAN‐DCNN and GAN‐DCNN methods, the recognition accuracy is improved by 3.5% and 4%. [ABSTRACT FROM AUTHOR]

In this paper, the authors study the compensation of high‐power amplifier (HPA) non‐linear distortion in the multi‐user (MU) massive multiple‐input multiple‐output (MIMO) systems and focus on uplink transmission, where the base station (BS) uses a large antenna array. First, the authors present a non‐linear distortion iterative cancellation (NDIC) algorithm‐based MMSE and approximate message passing (AMP) at the receiver level, in order to estimate and mitigate jointly a non‐linear distortion and the channel noise. Second, the authors propose a novel distortion cancellation technique based on deep learning. At this level, the authors first introduce a multilayer neural network, trained in the Levenberg–Marquardt algorithm by eliminating HPA non‐linearities on the 'Pre distortion' transmitter and 'Post distortion' receiver side. Next, the authors developed a novel end‐to‐end (E2E) learning approach for the joint transmitter and non‐coherent receiver in the Rayleigh fading channel. The basic idea lies in the use of deep neural networks (DNNs), auto encoder (AE) for unknown channels, where DNNs are applied to perform several functions and modules existing in the transmission chain. The simulation results demonstrate the strong potential of the proposed approach E2E in terms of improving the link quality and symbol error rate (SER) compared to other compensation techniques presented in this work. [ABSTRACT FROM AUTHOR]

In this paper, a deep learning integrated reinforcement learning (DLIRL) algorithm is proposed for comprehending intelligent beamsteering in Beyond Fifth Generation (B5G) networks. The smart base station in B5G networks aims to steer the beam towards appropriate user equipment based on the acquaintance of isotropic transmissions. The foremost methodology is to optimize beam direction through reinforcement learning that delivers significant improvement in signal to noise ratio (SNR). This includes alternate path finding during path obstruction and steering the beam appropriately between the smart base station and user equipment. The DLIRL is realized through supervised learning with deep neural networks and deep Q‐learning schemes. The proposed algorithm comprises of an online learning phase for training the weights and a working phase for carrying out the prediction. Results confirm that the performance of the B5G system is improved considerably as compared to its counterparts with a spectral efficiency of 11 bps/Hz at SNR = 10 dB for a bit error rate performance of 10−5. As compared to reinforced learning and deep neural network with a deviation of ±3o and ±5°, respectively, the DLIRL beamforming displays a deviation of ±2o. Moreover, the DLIRL can track the user equipment and steer the beam in its direction with an accuracy of 92%. [ABSTRACT FROM AUTHOR]

Numerous researches on communication receiver design with deep learning algorithms have been implemented effectively recently. In this paper, an innovative scheme is proposed to construct an end‐to‐end neural network (NN) receiver to directly recover information bits from unsynchronized waveform sequences. Considering the correlation between samples, multiple bidirectional long‐short term memory (BiLSTM) layers to process oversampled signals. Then, shifted binary cross‐entropy (SBCE) function is devised to tackle the overfitting issue and eliminate instances with abnormal bit error rate (BER), which originate from standard binary cross‐entropy (BCE) loss function. Substantial simulation results demonstrate that the trained NN receiver can achieve theory BER values under excellent conditions for transceivers. Compared with traditional synchronization algorithms, the proposed method can obtain significant BER performance gain under harsh conditions, such as low oversample rate, small roll‐off factor, short frame length. [ABSTRACT FROM AUTHOR]

The recent concept of contextual information in error correcting code (ECC) can significantly improve the capacity of the blind recognition of codes with deep learning (DL) approaches. However, the fundamental challenges of existing DL‐based methods are inflexible structure and limited kernel size which bring great difficulties to exploit the characteristics of contextual information in ECC. To handle this problem, in this paper, a state‐of‐the‐art framework for natural language processing (NLP), bidirectional encoder representation from transformers (BERT), is utilized in ECC classification scenarios. To strengthen the effectiveness of contextual information, the BERT model is improved by weighted relative positional encoding and error bit embedding. The proposed approach achieves higher classification accuracy than the methods based on Gauss‐Jordan elimination and traditional deep learning schemes. Further simulation results show that the classification accuracy is affected by block length and the employment of weighted relative positional encoding and error bit embedding to a large extent. [ABSTRACT FROM AUTHOR]

In this paper, massive multiple‐input‐multiple‐output (MIMO) wireless communication systems are considered to investigate joint transceiver beamforming. A base station (BS) equipped with a uniform planar array (UPA) serves several multi‐antennas users in a single cell. Based on the channel state information (CSI), the low complexity design of transceiver beamforming to minimize the transmit power subject to some quality of service (QoS) constraints is investigated. As the upper bound of the transmit power performance, the existing iteration‐based algorithms are leveraged as a reference. A general deep learning (DL)‐based framework and deep neural network (DNN) structure are proposed to reduce the complexity of the existing algorithms, where the properly trained DNN structure can learn directly from CSI. Consider the complexity of the DNN structure itself, a heuristic algorithm is proposed to replace the DNN structure, which takes the max‐eigenvalue‐eigenvector of the CSI as the direction of receive beamforming directly. The DNN structure is trained in the offline stage, therefore, only the complexity in the online stage is taken into consideration. Based on the numerical simulation, the complexity of the proposed DL‐based framework and the transceiver beamforming algorithms is reduced significantly while maintaining nearly the optimal performance compared with the existing iterative algorithms. [ABSTRACT FROM AUTHOR]

Deep learning (DL) has been applied to digital signal modulation identification (DSMI) due to its powerful feature learning ability. However, most of the existing DL‐based DSMI methods are limited to specific experimental scene relating to the additive white Gaussian noise (AWGN) channel or static multipath channel. The result is that the trained network has deteriorative identification accuracy when the channel conditions change unless retrained. To solve the problem, this paper proposes a DSMI method suitable for orthogonal frequency division multiplexing (OFDM) under different multipath channels, including the variation of delay, path number and channel coefficient. This method can accurately detect the modulation feature rather than the channel's to identify the modulation type, thus reducing the network training amount. The method is divided into two parts. Firstly, traditional signal processing methods are combined, including various channel estimators and equalisers to compensate for the channel. Then a robust DL network, RSN‐MI, is designed as a classifier. Unlike other DL‐based DSMI methods, the influence of signal processing algorithms on DSMI performance are focused on rather than model parameters. Besides, the proposed classifier is compared with the DSMI classifier in other contributions. The results show that the classifier works better in different multipath channels. [ABSTRACT FROM AUTHOR]

The electromagnetic situation, which can promote the abilities of understanding and decision‐making for the battlefield, has attracted significant interest recently in information‐based warfare. This paper investigates the deep learning‐based electromagnetic situation analysis and judgment in a complicated battlefield environment. To comprehensively simulate the two‐sided battling process, a turn‐based confrontation strategy is proposed, and an electromagnetic situation analysis and judgment model are then designed based on the AlphaGo Zero algorithm to achieve efficient situation analysis and decision‐making. In addition, an electromagnetic situation‐based attack‐defense platform is developed to realize and evaluate this designed model. Simulation results demonstrate that this designed model achieves significant performance in electromagnetic situation analysis and judgment compared with the Monte Carlo Tree Search based baseline. [ABSTRACT FROM AUTHOR]