With the advancement of communication technology, there is a higher demand for high-precision and high-generalization channel path loss models as it is fundamental to communication systems. For traditional stochastic and deterministic models, it is difficult to strike a balance between prediction accuracy and generalizability. This paper proposes a novel deep learning-based path loss prediction model using satellite images. In order to efficiently extract environment features from satellite images, residual structure, attention mechanism, and spatial pyramid pooling layer are developed in the network based on expert knowledge. Using a convolutional network activation visualization method, the interpretability of the proposed model is improved. Finally, the proposed model achieves a prediction accuracy with a root mean square error of 5.05 dB, demonstrating an improvement of 3.07 dB over a reference empirical propagation model.
{"title":"Channel Path Loss Prediction Using Satellite Images: A Deep Learning Approach","authors":"Chenlong Wang;Bo Ai;Ruisi He;Mi Yang;Shun Zhou;Long Yu;Yuxin Zhang;Zhicheng Qiu;Zhangdui Zhong;Jianhua Fan","doi":"10.1109/TMLCN.2024.3454019","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3454019","url":null,"abstract":"With the advancement of communication technology, there is a higher demand for high-precision and high-generalization channel path loss models as it is fundamental to communication systems. For traditional stochastic and deterministic models, it is difficult to strike a balance between prediction accuracy and generalizability. This paper proposes a novel deep learning-based path loss prediction model using satellite images. In order to efficiently extract environment features from satellite images, residual structure, attention mechanism, and spatial pyramid pooling layer are developed in the network based on expert knowledge. Using a convolutional network activation visualization method, the interpretability of the proposed model is improved. Finally, the proposed model achieves a prediction accuracy with a root mean square error of 5.05 dB, demonstrating an improvement of 3.07 dB over a reference empirical propagation model.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1357-1368"},"PeriodicalIF":0.0,"publicationDate":"2024-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10663692","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246437","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper studies an unmanned aerial vehicle (UAV)-enabled communication network, in which the UAV acts as an air relay serving multiple ground users (GUs) to jointly construct an accurate radio map or channel knowledge maps (CKM) through a federated learning (FL) algorithm. Radio map or CKM is a site-specific database that contains detailed channel-related information for specific locations. This information includes channel power gains, shadowing, interference, and angles of arrival (AoA) and departure (AoD), all of which are crucial for enabling environment-aware wireless communications. Because the wireless communication network has limited resource blocks (RBs), only a subset of users can be selected to transmit the model parameters at each iteration. Since the FL training process requires multiple transmission model parameters, the energy limitation of the wireless device will seriously affect the quality of the FL result. In this sense, the energy consumption and resource allocation have a significance to the final FL training result. We formulate an optimization problem by jointly considering user selection, wireless resource allocation, and UAV deployment, with the goal of minimizing the computation energy and wireless transmission energy. To solve the problem, we first propose a probabilistic user selection mechanism to reduce the total number of FL iterations, whereby the users who have a larger impact on the global model in each iteration are more likely to be selected. Then the convex optimization technique is utilized to optimize bandwidth allocation. Furthermore, to further save communication transmission energy, we use deep reinforcement learning (DRL) to optimize the deployment location of the UAV. The DRL-based method enables the UAV to learn from its interaction with the environment and ascertain the most energy-efficient deployment locations through an evaluation of energy consumption during the training process. Finally, the simulation results show that our proposed algorithm can reduce the total energy consumption by nearly 38%, compared to the standard FL algorithm.
{"title":"Energy Minimization for Federated Learning Based Radio Map Construction","authors":"Fahui Wu;Yunfei Gao;Lin Xiao;Dingcheng Yang;Jiangbin Lyu","doi":"10.1109/TMLCN.2024.3453212","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3453212","url":null,"abstract":"This paper studies an unmanned aerial vehicle (UAV)-enabled communication network, in which the UAV acts as an air relay serving multiple ground users (GUs) to jointly construct an accurate radio map or channel knowledge maps (CKM) through a federated learning (FL) algorithm. Radio map or CKM is a site-specific database that contains detailed channel-related information for specific locations. This information includes channel power gains, shadowing, interference, and angles of arrival (AoA) and departure (AoD), all of which are crucial for enabling environment-aware wireless communications. Because the wireless communication network has limited resource blocks (RBs), only a subset of users can be selected to transmit the model parameters at each iteration. Since the FL training process requires multiple transmission model parameters, the energy limitation of the wireless device will seriously affect the quality of the FL result. In this sense, the energy consumption and resource allocation have a significance to the final FL training result. We formulate an optimization problem by jointly considering user selection, wireless resource allocation, and UAV deployment, with the goal of minimizing the computation energy and wireless transmission energy. To solve the problem, we first propose a probabilistic user selection mechanism to reduce the total number of FL iterations, whereby the users who have a larger impact on the global model in each iteration are more likely to be selected. Then the convex optimization technique is utilized to optimize bandwidth allocation. Furthermore, to further save communication transmission energy, we use deep reinforcement learning (DRL) to optimize the deployment location of the UAV. The DRL-based method enables the UAV to learn from its interaction with the environment and ascertain the most energy-efficient deployment locations through an evaluation of energy consumption during the training process. Finally, the simulation results show that our proposed algorithm can reduce the total energy consumption by nearly 38%, compared to the standard FL algorithm.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1248-1264"},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10662910","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169609","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-29DOI: 10.1109/TMLCN.2024.3452057
Pranav S. Page;Anand S. Siyote;Vivek S. Borkar;Gaurav S. Kasbekar
The Internet of Things (IoT) is emerging as a critical technology to connect resource-constrained devices such as sensors and actuators as well as appliances to the Internet. In this paper, a novel methodology for node cardinality estimation in wireless networks such as the IoT and Radio-Frequency Identification (RFID) systems is proposed, which uses the Privileged Feature Distillation (PFD) technique and works using a neural network with a teacher-student model. This paper is the first to use the powerful PFD technique for node cardinality estimation in wireless networks. The teacher is trained using both privileged and regular features, and the student is trained with predictions from the teacher and regular features. Node cardinality estimation algorithms based on the PFD technique are proposed for homogeneous wireless networks as well as heterogeneous wireless networks with $T geq 2$ types of nodes. Extensive simulations, using a synthetic dataset as well as a real dataset, are used to show that the proposed PFD based algorithms for homogeneous as well as heterogeneous networks achieve much lower mean squared errors (MSEs) in the computed node cardinality estimates than state-of-the-art protocols proposed in prior work. In particular, our simulation results for the real dataset show that our proposed PFD based technique for homogeneous (respectively, heterogeneous) networks achieves a MSE that is 92.35% (respectively, 94.08%) lower on average than that achieved by the Simple RFID Counting (SRCs) protocol (respectively, T-SRCs protocol) proposed in prior work while taking the same number of time slots to execute.
{"title":"Node Cardinality Estimation in the Internet of Things Using Privileged Feature Distillation","authors":"Pranav S. Page;Anand S. Siyote;Vivek S. Borkar;Gaurav S. Kasbekar","doi":"10.1109/TMLCN.2024.3452057","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3452057","url":null,"abstract":"The Internet of Things (IoT) is emerging as a critical technology to connect resource-constrained devices such as sensors and actuators as well as appliances to the Internet. In this paper, a novel methodology for node cardinality estimation in wireless networks such as the IoT and Radio-Frequency Identification (RFID) systems is proposed, which uses the Privileged Feature Distillation (PFD) technique and works using a neural network with a teacher-student model. This paper is the first to use the powerful PFD technique for node cardinality estimation in wireless networks. The teacher is trained using both privileged and regular features, and the student is trained with predictions from the teacher and regular features. Node cardinality estimation algorithms based on the PFD technique are proposed for homogeneous wireless networks as well as heterogeneous wireless networks with \u0000<inline-formula> <tex-math>$T geq 2$ </tex-math></inline-formula>\u0000 types of nodes. Extensive simulations, using a synthetic dataset as well as a real dataset, are used to show that the proposed PFD based algorithms for homogeneous as well as heterogeneous networks achieve much lower mean squared errors (MSEs) in the computed node cardinality estimates than state-of-the-art protocols proposed in prior work. In particular, our simulation results for the real dataset show that our proposed PFD based technique for homogeneous (respectively, heterogeneous) networks achieves a MSE that is 92.35% (respectively, 94.08%) lower on average than that achieved by the Simple RFID Counting (SRCs) protocol (respectively, T-SRCs protocol) proposed in prior work while taking the same number of time slots to execute.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1229-1247"},"PeriodicalIF":0.0,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10659215","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142152132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-28DOI: 10.1109/TMLCN.2024.3450829
Maria Raftopoulou;José Mairton B. da Silva;Remco Litjens;H. Vincent Poor;Piet van Mieghem
Federated learning is an effective method to train a machine learning model without requiring to aggregate the potentially sensitive data of agents in a central server. However, the limited communication bandwidth, the hardware of the agents and a potential application-specific latency requirement impact how many and which agents can participate in the learning process at each communication round. In this paper, we propose a selection metric characterizing each agent’s importance with respect to both the learning process and the resource efficiency of its wireless communication channel. Leveraging this importance metric, we formulate a general agent selection optimization problem, which can be adapted to different environments with latency or resource-oriented constraints. Considering an example wireless environment with latency constraints, the agent selection problem reduces to the 0/1 Knapsack problem, which we solve with a fully polynomial approximation. We then evaluate the agent selection policy in different scenarios, using extensive simulations for an example task of object classification of European traffic signs. The results indicate that agent selection policies which consider both learning and channel aspects provide benefits in terms of the attainable global model accuracy and/or the time needed to achieve a targeted accuracy level. However, in scenarios where agents have a limited number of data samples or where the latency requirement is very stringent, a pure learning-based agent selection policy is shown to be more beneficial during the early or late stages of the learning process.
{"title":"Agent Selection Framework for Federated Learning in Resource-Constrained Wireless Networks","authors":"Maria Raftopoulou;José Mairton B. da Silva;Remco Litjens;H. Vincent Poor;Piet van Mieghem","doi":"10.1109/TMLCN.2024.3450829","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3450829","url":null,"abstract":"Federated learning is an effective method to train a machine learning model without requiring to aggregate the potentially sensitive data of agents in a central server. However, the limited communication bandwidth, the hardware of the agents and a potential application-specific latency requirement impact how many and which agents can participate in the learning process at each communication round. In this paper, we propose a selection metric characterizing each agent’s importance with respect to both the learning process and the resource efficiency of its wireless communication channel. Leveraging this importance metric, we formulate a general agent selection optimization problem, which can be adapted to different environments with latency or resource-oriented constraints. Considering an example wireless environment with latency constraints, the agent selection problem reduces to the 0/1 Knapsack problem, which we solve with a fully polynomial approximation. We then evaluate the agent selection policy in different scenarios, using extensive simulations for an example task of object classification of European traffic signs. The results indicate that agent selection policies which consider both learning and channel aspects provide benefits in terms of the attainable global model accuracy and/or the time needed to achieve a targeted accuracy level. However, in scenarios where agents have a limited number of data samples or where the latency requirement is very stringent, a pure learning-based agent selection policy is shown to be more beneficial during the early or late stages of the learning process.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1265-1282"},"PeriodicalIF":0.0,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10654373","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142173968","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-26DOI: 10.1109/TMLCN.2024.3449834
Marc Jean;Murat Yuksel;Xun Gong
The increasing demand for gigabit-per-second speeds and higher wireless node density is driving the need for spatial reuse and the utilization of higher frequencies above the legacy sub-6 GHz bands. Since these super-6 GHz bands experience high path loss, directional beamforming has been the main method of access to the large amount of bandwidth available at these higher frequencies. Hence, the programming of wireless beams with specific directions is emerging as a requirement for software-defined radio (SDR) platforms. To address this need, we introduce an affordable millimeter-wave (mmWave) testbed. Using a multi-threaded software architecture, the testbed allows for the convenient programming of mmWave beam directions using a high-level programming language, while also providing access to machine learning (ML) libraries as well as SDR methods traditionally deployed in Universal Software Radio Peripheral (USRP) devices. To showcase the potential of the testbed, we tackle the Angle-of-Arrival (AoA) detection problem using reinforcement learning (RL) methods on the receiver side. AoA detection and direction finding is a crucial need for the emerging use of super-6 GHz spectra. We design and implement Q-learning, Double Q-learning, and Deep Q-learning algorithms that passively inspect the Received Signal Strength (RSS) of the mmWave beam and autonomously determine the predicted AoA. The results indicate the feasibility of programming directionality of the wireless beams via ML-based methods as well as solving difficult problems pertaining to emerging directional wireless systems.
对每秒千兆位速度和更高无线节点密度的需求不断增长，推动了对空间重用和利用传统 6 GHz 以下频段以上更高频率的需求。由于这些超 6 GHz 频段的路径损耗较高，定向波束成形一直是利用这些较高频率的大量带宽的主要方法。因此，软件定义无线电（SDR）平台需要对特定方向的无线波束进行编程。为了满足这一需求，我们推出了一种经济实惠的毫米波（mmWave）测试平台。该测试平台采用多线程软件架构，可使用高级编程语言方便地对毫米波波束方向进行编程，同时还可访问机器学习（ML）库以及传统上部署在通用软件无线电外设（USRP）设备中的 SDR 方法。为了展示该测试平台的潜力，我们在接收端使用强化学习（RL）方法解决了到达角（AoA）检测问题。AoA检测和测向是超6 GHz频谱新兴应用的关键需求。我们设计并实施了 Q-learning、Double Q-learning 和 Deep Q-learning 算法，这些算法可被动检测毫米波波束的接收信号强度 (RSS)，并自主确定预测的 AoA。研究结果表明，通过基于 ML 的方法对无线波束的方向性进行编程以及解决与新兴定向无线系统相关的难题是可行的。
{"title":"ML-Enabled Millimeter-Wave Software-Defined Radio With Programmable Directionality","authors":"Marc Jean;Murat Yuksel;Xun Gong","doi":"10.1109/TMLCN.2024.3449834","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3449834","url":null,"abstract":"The increasing demand for gigabit-per-second speeds and higher wireless node density is driving the need for spatial reuse and the utilization of higher frequencies above the legacy sub-6 GHz bands. Since these super-6 GHz bands experience high path loss, directional beamforming has been the main method of access to the large amount of bandwidth available at these higher frequencies. Hence, the programming of wireless beams with specific directions is emerging as a requirement for software-defined radio (SDR) platforms. To address this need, we introduce an affordable millimeter-wave (mmWave) testbed. Using a multi-threaded software architecture, the testbed allows for the convenient programming of mmWave beam directions using a high-level programming language, while also providing access to machine learning (ML) libraries as well as SDR methods traditionally deployed in Universal Software Radio Peripheral (USRP) devices. To showcase the potential of the testbed, we tackle the Angle-of-Arrival (AoA) detection problem using reinforcement learning (RL) methods on the receiver side. AoA detection and direction finding is a crucial need for the emerging use of super-6 GHz spectra. We design and implement Q-learning, Double Q-learning, and Deep Q-learning algorithms that passively inspect the Received Signal Strength (RSS) of the mmWave beam and autonomously determine the predicted AoA. The results indicate the feasibility of programming directionality of the wireless beams via ML-based methods as well as solving difficult problems pertaining to emerging directional wireless systems.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1159-1177"},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10646573","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142123044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-26DOI: 10.1109/TMLCN.2024.3449831
Xinchen Lyu;Yuewei Li;Ying He;Chenshan Ren;Wei Ni;Ren Ping Liu;Pengcheng Zhu;Qimei Cui
Split AI inference partitions an artificial intelligence (AI) model into multiple parts, enabling the offloading of computation-intensive AI services. Resource allocation is critical for the performance of split AI inference. The challenge arises from the time-sensitivity of many services versus time-varying traffic arrivals and network conditions. The conventional prediction-based resource allocation frameworks have adopted separate traffic prediction and resource optimization modules, which may be inefficient due to discrepancies between the traffic prediction accuracy and resource optimization objective. This paper proposes a new, objective-driven, differentiable optimization framework that integrates traffic prediction and resource allocation for split AI inference. The resource optimization problem (aimed to maximize network revenue while adhering to service and network constraints) is designed to be embedded as the output layer following the traffic prediction module. As such, the traffic prediction module can be trained directly based on the network revenue instead of the prediction accuracy, significantly outperforming the conventional prediction-based separate design. Employing the Lagrange duality and Karush-Kuhn-Tucker (KKT) conditions, we achieve efficient forward pass (obtaining resource allocation decisions) and backpropagation (deriving the objective-driven gradients for joint model training) of the output layer. Extensive experiments on different traffic datasets validate the superiority of the proposed approach, achieving up to 38.85% higher network revenue than the conventional predictive baselines.
{"title":"Objective-Driven Differentiable Optimization of Traffic Prediction and Resource Allocation for Split AI Inference Edge Networks","authors":"Xinchen Lyu;Yuewei Li;Ying He;Chenshan Ren;Wei Ni;Ren Ping Liu;Pengcheng Zhu;Qimei Cui","doi":"10.1109/TMLCN.2024.3449831","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3449831","url":null,"abstract":"Split AI inference partitions an artificial intelligence (AI) model into multiple parts, enabling the offloading of computation-intensive AI services. Resource allocation is critical for the performance of split AI inference. The challenge arises from the time-sensitivity of many services versus time-varying traffic arrivals and network conditions. The conventional prediction-based resource allocation frameworks have adopted separate traffic prediction and resource optimization modules, which may be inefficient due to discrepancies between the traffic prediction accuracy and resource optimization objective. This paper proposes a new, objective-driven, differentiable optimization framework that integrates traffic prediction and resource allocation for split AI inference. The resource optimization problem (aimed to maximize network revenue while adhering to service and network constraints) is designed to be embedded as the output layer following the traffic prediction module. As such, the traffic prediction module can be trained directly based on the network revenue instead of the prediction accuracy, significantly outperforming the conventional prediction-based separate design. Employing the Lagrange duality and Karush-Kuhn-Tucker (KKT) conditions, we achieve efficient forward pass (obtaining resource allocation decisions) and backpropagation (deriving the objective-driven gradients for joint model training) of the output layer. Extensive experiments on different traffic datasets validate the superiority of the proposed approach, achieving up to 38.85% higher network revenue than the conventional predictive baselines.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1178-1192"},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10646623","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142123023","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-22DOI: 10.1109/TMLCN.2024.3447634
Yuanyuan Qiao;Yong Niu;Lan Su;Shiwen Mao;Ning Wang;Zhangdui Zhong;BO Ai
Millimeter wave (mmWave) communication can meet the requirements of vehicle-to-infrastructure (V2I) systems, for high throughput and ultra-low latency. However, searching for the optimal beamforming vectors in highly dynamic environments, incurs considerable training overhead. And it is a huge challenge to achieve beam alignment between receivers and transmitters. This paper proposes a beam alignment algorithm based on vehicle position information, to achieve fast beam alignment in the V2I network. In the proposed algorithm, a roadside unit (RSU) obtains a set of candidate beams by the vehicle position information and the double deep Q network (DDQN) algorithm. Then, according to the criterion of maximizing the system spectral efficiency, the optimal beam of the candidate beam set is obtained by the exhaustive search, to achieve fast beam alignment. In this paper, the DeepMIMO dataset is utilized to fully consider the actual scene of V2I, and the effect of Doppler expansion is taken into account in the mathematical model. The simulation results show that the received signal-noise ratio (SNR) of vehicle at different positions is greater than the SNR threshold, which avoids communication interruption and improves the reliability of V2I communications. Meanwhile, we also evaluates the effect of vehicle speed. Compared with other search schemes, the proposed scheme attains higher transmission rates, effectively balances the training overhead and achievable rate, and is suitable for mmWave V2I networks.
{"title":"Deep Reinforcement Learning-Based mmWave Beam Alignment for V2I Communications","authors":"Yuanyuan Qiao;Yong Niu;Lan Su;Shiwen Mao;Ning Wang;Zhangdui Zhong;BO Ai","doi":"10.1109/TMLCN.2024.3447634","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3447634","url":null,"abstract":"Millimeter wave (mmWave) communication can meet the requirements of vehicle-to-infrastructure (V2I) systems, for high throughput and ultra-low latency. However, searching for the optimal beamforming vectors in highly dynamic environments, incurs considerable training overhead. And it is a huge challenge to achieve beam alignment between receivers and transmitters. This paper proposes a beam alignment algorithm based on vehicle position information, to achieve fast beam alignment in the V2I network. In the proposed algorithm, a roadside unit (RSU) obtains a set of candidate beams by the vehicle position information and the double deep Q network (DDQN) algorithm. Then, according to the criterion of maximizing the system spectral efficiency, the optimal beam of the candidate beam set is obtained by the exhaustive search, to achieve fast beam alignment. In this paper, the DeepMIMO dataset is utilized to fully consider the actual scene of V2I, and the effect of Doppler expansion is taken into account in the mathematical model. The simulation results show that the received signal-noise ratio (SNR) of vehicle at different positions is greater than the SNR threshold, which avoids communication interruption and improves the reliability of V2I communications. Meanwhile, we also evaluates the effect of vehicle speed. Compared with other search schemes, the proposed scheme attains higher transmission rates, effectively balances the training overhead and achievable rate, and is suitable for mmWave V2I networks.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1216-1228"},"PeriodicalIF":0.0,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10643601","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142152133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Embodying the principle of simulation intelligence, digital twin (DT) systems construct and maintain a high-fidelity virtual model of a physical system. This paper focuses on ray tracing (RT), which is widely seen as an enabling technology for DTs of the radio access network (RAN) segment of next-generation disaggregated wireless systems. RT makes it possible to simulate channel conditions, enabling data augmentation and prediction-based transmission. However, the effectiveness of RT hinges on the adaptation of the electromagnetic properties assumed by the RT to actual channel conditions, a process known as calibration. The main challenge of RT calibration is the fact that small discrepancies in the geometric model fed to the RT software hinder the accuracy of the predicted phases of the simulated propagation paths. Existing solutions to this problem either rely on the channel power profile, hence disregarding phase information, or they operate on the channel responses by assuming the simulated phases to be sufficiently accurate for calibration. This paper proposes a novel channel response-based scheme that, unlike the state of the art, estimates and compensates for the phase errors in the RT-generated channel responses. The proposed approach builds on the variational expectation maximization algorithm with a flexible choice of the prior phase-error distribution that bridges between a deterministic model with no phase errors and a stochastic model with uniform phase errors. The algorithm is computationally efficient, and is demonstrated, by leveraging the open-source differentiable RT software available within the Sionna library, to outperform existing methods in terms of the accuracy of RT predictions.
{"title":"Calibrating Wireless Ray Tracing for Digital Twinning Using Local Phase Error Estimates","authors":"Clement Ruah;Osvaldo Simeone;Jakob Hoydis;Bashir Al-Hashimi","doi":"10.1109/TMLCN.2024.3448391","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3448391","url":null,"abstract":"Embodying the principle of simulation intelligence, digital twin (DT) systems construct and maintain a high-fidelity virtual model of a physical system. This paper focuses on ray tracing (RT), which is widely seen as an enabling technology for DTs of the radio access network (RAN) segment of next-generation disaggregated wireless systems. RT makes it possible to simulate channel conditions, enabling data augmentation and prediction-based transmission. However, the effectiveness of RT hinges on the adaptation of the electromagnetic properties assumed by the RT to actual channel conditions, a process known as calibration. The main challenge of RT calibration is the fact that small discrepancies in the geometric model fed to the RT software hinder the accuracy of the predicted phases of the simulated propagation paths. Existing solutions to this problem either rely on the channel power profile, hence disregarding phase information, or they operate on the channel responses by assuming the simulated phases to be sufficiently accurate for calibration. This paper proposes a novel channel response-based scheme that, unlike the state of the art, estimates and compensates for the phase errors in the RT-generated channel responses. The proposed approach builds on the variational expectation maximization algorithm with a flexible choice of the prior phase-error distribution that bridges between a deterministic model with no phase errors and a stochastic model with uniform phase errors. The algorithm is computationally efficient, and is demonstrated, by leveraging the open-source differentiable RT software available within the Sionna library, to outperform existing methods in terms of the accuracy of RT predictions.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1193-1215"},"PeriodicalIF":0.0,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10643616","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142123043","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-21DOI: 10.1109/TMLCN.2024.3439289
Sanaullah Manzoor;Muhammad Zeeshan Shakir;Mazen O. Hasna;Khalid A. Qaraqe
With the vast proliferation of smart mobile devices, there is an ever-increasing demand for higher data rates and seamless connectivity throughout. Current 5th generation and beyond (B5G) cellular networks struggle to eradicate outage zones and ensure seamless connectivity. One promising solution to this problem is the use of unmanned aerial vehicles (UAVs) to assist the traditional ground network and provide connectivity in places where there are no small base stations or faulty ones as a result of some natural disasters such as flooding. In this paper, we propose a novel users’ mobility-aware & users’ demand-aware federated learning-based proactive UAV placement (MFPUP) framework to assist the existing ground communication network and minimise overall network outages. Our MFPUP framework utilises the federated learning-based mobility prediction model that recommends the potential outage areas to deploy UAVs using user-UAV association techniques such as the optimum association approach (OAP) and the greedy association approach (GAP). In order to validate the performance of the proposed MFPUP scheme we carried out extensive simulations. The proposed LSTM-based mobility model outperforms the DNN model with 92.88% prediction accuracy. Further, our results show that the proposed MFPUP framework associates the optimal number of users to UAVs while also improving 1.25 times users’ downlink rates as compared other UAVs placement schemes.
{"title":"Mobility-Aware Federated Learning-Based Proactive UAVs Placement in Emerging Cellular Networks","authors":"Sanaullah Manzoor;Muhammad Zeeshan Shakir;Mazen O. Hasna;Khalid A. Qaraqe","doi":"10.1109/TMLCN.2024.3439289","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3439289","url":null,"abstract":"With the vast proliferation of smart mobile devices, there is an ever-increasing demand for higher data rates and seamless connectivity throughout. Current 5th generation and beyond (B5G) cellular networks struggle to eradicate outage zones and ensure seamless connectivity. One promising solution to this problem is the use of unmanned aerial vehicles (UAVs) to assist the traditional ground network and provide connectivity in places where there are no small base stations or faulty ones as a result of some natural disasters such as flooding. In this paper, we propose a novel users’ mobility-aware & users’ demand-aware federated learning-based proactive UAV placement (MFPUP) framework to assist the existing ground communication network and minimise overall network outages. Our MFPUP framework utilises the federated learning-based mobility prediction model that recommends the potential outage areas to deploy UAVs using user-UAV association techniques such as the optimum association approach (OAP) and the greedy association approach (GAP). In order to validate the performance of the proposed MFPUP scheme we carried out extensive simulations. The proposed LSTM-based mobility model outperforms the DNN model with 92.88% prediction accuracy. Further, our results show that the proposed MFPUP framework associates the optimal number of users to UAVs while also improving 1.25 times users’ downlink rates as compared other UAVs placement schemes.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1305-1318"},"PeriodicalIF":0.0,"publicationDate":"2024-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10643371","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246438","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-20DOI: 10.1109/TMLCN.2024.3446743
Abdurrahman Elmaghbub;Bechir Hamdaoui
Deep learning (DL)-based RF fingerprinting (RFFP) technology has emerged as a powerful physical-layer security mechanism, enabling device identification and authentication based on unique device-specific signatures that can be extracted from the received RF signals. However, DL-based RFFP methods face major challenges concerning their ability to adapt to domain (e.g., day/time, location, channel, etc.) changes and variability. This work proposes a novel IQ data representation and feature design, termed Double-Sided Envelope Power Spectrum or EPS , that is proven to significantly overcome the domain adaptation challenges associated with WiFi transmitter fingerprinting. By accurately capturing device hardware impairments while suppressing irrelevant domain information, EPS offers improved feature selection for DL models in RFFP. Our experimental evaluation demonstrates the effectiveness of the integration of EPS representation with a Convolution Neural Network (CNN) model, termed EPS-CNN , achieving over 99% testing accuracy in same-day/channel/location evaluations and 93% accuracy in cross-day evaluations, outperforming the traditional IQ representation. Additionally, EPS-CNN excels in cross-location evaluations, achieving a 95% accuracy. The proposed representation significantly enhances the robustness and generalizability of DL-based RFFP methods, thereby presenting a transformative solution to IQ data-based device fingerprinting.
{"title":"Distinguishable IQ Feature Representation for Domain-Adaptation Learning of WiFi Device Fingerprints","authors":"Abdurrahman Elmaghbub;Bechir Hamdaoui","doi":"10.1109/TMLCN.2024.3446743","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3446743","url":null,"abstract":"Deep learning (DL)-based RF fingerprinting (RFFP) technology has emerged as a powerful physical-layer security mechanism, enabling device identification and authentication based on unique device-specific signatures that can be extracted from the received RF signals. However, DL-based RFFP methods face major challenges concerning their ability to adapt to domain (e.g., day/time, location, channel, etc.) changes and variability. This work proposes a novel IQ data representation and feature design, termed Double-Sided Envelope Power Spectrum or \u0000<monospace>EPS</monospace>\u0000, that is proven to significantly overcome the domain adaptation challenges associated with WiFi transmitter fingerprinting. By accurately capturing device hardware impairments while suppressing irrelevant domain information, \u0000<monospace>EPS</monospace>\u0000 offers improved feature selection for DL models in RFFP. Our experimental evaluation demonstrates the effectiveness of the integration of \u0000<monospace>EPS</monospace>\u0000 representation with a Convolution Neural Network (CNN) model, termed \u0000<monospace>EPS-CNN</monospace>\u0000, achieving over 99% testing accuracy in same-day/channel/location evaluations and 93% accuracy in cross-day evaluations, outperforming the traditional IQ representation. Additionally, \u0000<monospace>EPS-CNN</monospace>\u0000 excels in cross-location evaluations, achieving a 95% accuracy. The proposed representation significantly enhances the robustness and generalizability of DL-based RFFP methods, thereby presenting a transformative solution to IQ data-based device fingerprinting.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"1404-1423"},"PeriodicalIF":0.0,"publicationDate":"2024-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10640139","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142320477","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：