Pub Date : 2024-09-18DOI: 10.1109/TMC.2024.3460403
Jing Xue;Die Wu;Jian Peng;Wenzheng Xu;Tang Liu
To guarantee the reliability for WRSNs, placing sufficient static chargers effectively ensures charging coverage for the entire network. However, this approach leads to a considerable number of sensors located within charging overlaps. The destructive wave interference caused by concurrent charging in these overlaps may weaken sensors received power, thereby negatively impacting charging performance. This work addresses a CHArging utIlity maximizatioN (CHAIN) problem, which aims to maximize the overall charging utility while considering wave interference among multiple chargers. Specifically, given a set of stationary sensors, we investigate how to determine optimal positions for a fixed number of chargers. To tackle this problem, we first develop a charging model with wave interference, then propose a two-step charger placement scheme to identify the optimal charger positions. In the first step, we maximize the overall additive power of the waves involved in interference by selecting an appropriate initial position for each charger. Then, in the second step, we maximize the overall charging utility by finding the optimal final position for each charger around its initial position. Finally, to evaluate the performance of our scheme, we conduct extensive simulations and field experiments and the results suggest that CHAIN performs better than the existing algorithms.
{"title":"Charger Placement With Wave Interference","authors":"Jing Xue;Die Wu;Jian Peng;Wenzheng Xu;Tang Liu","doi":"10.1109/TMC.2024.3460403","DOIUrl":"10.1109/TMC.2024.3460403","url":null,"abstract":"To guarantee the reliability for WRSNs, placing sufficient static chargers effectively ensures charging coverage for the entire network. However, this approach leads to a considerable number of sensors located within charging overlaps. The destructive wave interference caused by concurrent charging in these overlaps may weaken sensors received power, thereby negatively impacting charging performance. This work addresses a CHArging utIlity maximizatioN (CHAIN) problem, which aims to maximize the overall charging utility while considering wave interference among multiple chargers. Specifically, given a set of stationary sensors, we investigate how to determine optimal positions for a fixed number of chargers. To tackle this problem, we first develop a charging model with wave interference, then propose a two-step charger placement scheme to identify the optimal charger positions. In the first step, we maximize the overall additive power of the waves involved in interference by selecting an appropriate initial position for each charger. Then, in the second step, we maximize the overall charging utility by finding the optimal final position for each charger around its initial position. Finally, to evaluate the performance of our scheme, we conduct extensive simulations and field experiments and the results suggest that CHAIN performs better than the existing algorithms.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 1","pages":"261-275"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257207","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-18DOI: 10.1109/TMC.2024.3459409
Juheon Yi;Utku Günay Acer;Fahim Kawsar;Chulhong Min
Overlapping cameras offer exciting opportunities to view a scene from different angles, allowing for more advanced, comprehensive and robust analysis. However, existing video analytics systems for multi-camera streams are mostly limited to (i) per-camera processing and aggregation and (ii) workload-agnostic centralized processing architectures. In this paper, we present Argus, a distributed video analytics system with cross-camera collaboration on smart cameras. We identify multi-camera, multi-target tracking as the primary task of multi-camera video analytics and develop a novel technique that avoids redundant, processing-heavy identification tasks by leveraging object-wise spatio-temporal association in the overlapping fields of view across multiple cameras. We further develop a set of techniques to perform these operations across distributed cameras without cloud support at low latency by (i) dynamically ordering the camera and object inspection sequence and (ii) flexibly distributing the workload across smart cameras, taking into account network transmission and heterogeneous computational capacities. Evaluation of three real-world overlapping camera datasets with two Nvidia Jetson devices shows that Argus reduces the number of object identifications and end-to-end latency by up to 7.13× and 2.19× (4.86× and 1.60× compared to the state-of-the-art), while achieving comparable tracking quality.
{"title":"Argus: Enabling Cross-Camera Collaboration for Video Analytics on Distributed Smart Cameras","authors":"Juheon Yi;Utku Günay Acer;Fahim Kawsar;Chulhong Min","doi":"10.1109/TMC.2024.3459409","DOIUrl":"10.1109/TMC.2024.3459409","url":null,"abstract":"Overlapping cameras offer exciting opportunities to view a scene from different angles, allowing for more advanced, comprehensive and robust analysis. However, existing video analytics systems for multi-camera streams are mostly limited to (i) per-camera processing and aggregation and (ii) workload-agnostic centralized processing architectures. In this paper, we present Argus, a distributed video analytics system with \u0000<italic>cross-camera collaboration</i>\u0000 on smart cameras. We identify multi-camera, multi-target tracking as the primary task of multi-camera video analytics and develop a novel technique that avoids redundant, processing-heavy identification tasks by leveraging object-wise spatio-temporal association in the overlapping fields of view across multiple cameras. We further develop a set of techniques to perform these operations across distributed cameras without cloud support at low latency by (i) dynamically ordering the camera and object inspection sequence and (ii) flexibly distributing the workload across smart cameras, taking into account network transmission and heterogeneous computational capacities. Evaluation of three real-world overlapping camera datasets with two Nvidia Jetson devices shows that Argus reduces the number of object identifications and end-to-end latency by up to 7.13× and 2.19× (4.86× and 1.60× compared to the state-of-the-art), while achieving comparable tracking quality.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 1","pages":"117-134"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142269216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-18DOI: 10.1109/TMC.2024.3462437
Pengfei Hu;Yuhang Qian;Tianyue Zheng;Ang Li;Zhe Chen;Yue Gao;Xiuzhen Cheng;Jun Luo
Given the wide adoption of multimodal sensors (e.g., camera, lidar, radar) by autonomous vehicle s (AVs), deep analytics to fuse their outputs for a robust perception become imperative. However, existing fusion methods often make two assumptions rarely holding in practice: i) similar data distributions for all inputs and ii) constant availability for all sensors. Because, for example, lidars have various resolutions and failures of radars may occur, such variability often results in significant performance degradation in fusion. To this end, we present t-READi, an adaptive inference system that accommodates the variability of multimodal sensory data and thus enables robust and efficient perception. t-READi identifies variation-sensitive yet structure-specific model parameters; it then adapts only these parameters while keeping the rest intact. t-READi also leverages a cross-modality contrastive learning method to compensate for the loss from missing modalities. Both functions are implemented to maintain compatibility with existing multimodal deep fusion methods. The extensive experiments evidently demonstrate that compared with the status quo approaches, t-READi not only improves the average inference accuracy by more than 6% but also reduces the inference latency by almost 15× with the cost of only 5% extra memory overhead in the worst case under realistic data and modal variations.
{"title":"t-READi: Transformer-Powered Robust and Efficient Multimodal Inference for Autonomous Driving","authors":"Pengfei Hu;Yuhang Qian;Tianyue Zheng;Ang Li;Zhe Chen;Yue Gao;Xiuzhen Cheng;Jun Luo","doi":"10.1109/TMC.2024.3462437","DOIUrl":"10.1109/TMC.2024.3462437","url":null,"abstract":"Given the wide adoption of multimodal sensors (e.g., camera, lidar, radar) by \u0000<italic>autonomous vehicle</i>\u0000s (AVs), deep analytics to fuse their outputs for a robust perception become imperative. However, existing fusion methods often make two assumptions rarely holding in practice: i) similar data distributions for all inputs and ii) constant availability for all sensors. Because, for example, lidars have various resolutions and failures of radars may occur, such variability often results in significant performance degradation in fusion. To this end, we present t-READi, an adaptive inference system that accommodates the variability of multimodal sensory data and thus enables robust and efficient perception. t-READi identifies variation-sensitive yet \u0000<italic>structure-specific</i>\u0000 model parameters; it then adapts only these parameters while keeping the rest intact. t-READi also leverages a cross-modality contrastive learning method to compensate for the loss from missing modalities. Both functions are implemented to maintain compatibility with existing multimodal deep fusion methods. The extensive experiments evidently demonstrate that compared with the status quo approaches, t-READi not only improves the average inference accuracy by more than 6% but also reduces the inference latency by almost 15× with the cost of only 5% extra memory overhead in the worst case under realistic data and modal variations.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 1","pages":"135-149"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257208","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-18DOI: 10.1109/TMC.2024.3463417
Hang Zhao;Shengling Wang;Hongwei Shi;Jianhui Huang;Yu Guo;Xiuzhen Cheng
Game theory is an effective analytical tool for crowdsourcing. Existing studies based on it share a commonality: the influence of players’ decisions is bilateral . However, the status is broken by the zero-determinant (ZD) strategy, where the ZD player can unilaterally control the opponent's expected payoff. Thereby, crowdsourcing games trigger conclusions that differ from traditional ones. By addressing three questions, this paper is the first work to analyze the turbulence in crowdsourcing caused by the inequality between the requestor and the worker in the ZD game. The first question reveals the potential for the requestor to exploit the worker; the second question quantifies the worker's tolerance towards exploitation, providing a basis for confrontation; the third question serves as the cornerstone for maintaining the crowdsourcing, regulating the requestor's exploitative behavior. To answer these questions, we extend ZD strategies from binary games to continuous ones, not only revealing the requestor's dominance but also enriching the theoretical system of ZD strategies and broadening their application. Furthermore, we introduce the worker's dissatisfaction degree, identifying the exponential trend and decay rate, revealing optimal timing and speed for the worker's effective confrontation and maximum exploitation for the requestor. Numerical simulations have validated the effectiveness of our analyses.
{"title":"Exploitation and Confrontation: Sustainability Analysis of Crowdsourcing","authors":"Hang Zhao;Shengling Wang;Hongwei Shi;Jianhui Huang;Yu Guo;Xiuzhen Cheng","doi":"10.1109/TMC.2024.3463417","DOIUrl":"10.1109/TMC.2024.3463417","url":null,"abstract":"Game theory is an effective analytical tool for crowdsourcing. Existing studies based on it share a commonality: the influence of players’ decisions is \u0000<italic>bilateral</i>\u0000. However, the status is broken by the zero-determinant (ZD) strategy, where the ZD player can \u0000<italic>unilaterally</i>\u0000 control the opponent's expected payoff. Thereby, crowdsourcing games trigger conclusions that differ from traditional ones. By addressing three questions, this paper is the first work to analyze the turbulence in crowdsourcing caused by the inequality between the requestor and the worker in the ZD game. The first question reveals the potential for the requestor to exploit the worker; the second question quantifies the worker's tolerance towards exploitation, providing a basis for confrontation; the third question serves as the cornerstone for maintaining the crowdsourcing, regulating the requestor's exploitative behavior. To answer these questions, we extend ZD strategies from binary games to continuous ones, not only revealing the requestor's dominance but also enriching the theoretical system of ZD strategies and broadening their application. Furthermore, we introduce the worker's dissatisfaction degree, identifying the exponential trend and decay rate, revealing optimal timing and speed for the worker's effective confrontation and maximum exploitation for the requestor. Numerical simulations have validated the effectiveness of our analyses.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 2","pages":"614-626"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257209","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-18DOI: 10.1109/TMC.2024.3463735
Cheng Peng;Jun Yin;Lei Wang;Fu Xiao
Mobile edge caching network (MEN), which enables popular or reusable content caching and sharing among adjacent mobile edge devices, has become a promising solution to reduce the traffic and burden over backhaul links. Network coding (NC), represented by classical random linear network coding (RLNC), is utilized to facilitate content delivery and increase throughput in MEN. However, as the harsh decoding condition results in unacceptable time and storage overhead, classical RLNC schemes struggle to be widely deployed in practice. In this work, we propose a cost-effective NC-based content-sharing scheme based on binary sparse network coding (BSNC), called Bison, for D2D-enabled MEN. Based on the shared relationship between the binary sparse coded block (BSCB), Bison first designs a caching maintenance module to characterize the sharing progress and maintain the caching state of each edge node. Then, Bison defines a matching metric named neighbor utility to evaluate neighbors’ matching values by considering nodes’ demand and content decodability. Guiding by the metric, Bison achieves the most beneficial matching relationship among edge nodes through a proposed online matching policy. Finally, Bison devises a coded block delivery strategy to enable the sharing of valuable content between two matched edge nodes. Extensive experiments in simulations and real-world Android testbeds demonstrate its effectiveness and efficiency, wherein Bison is at least 30% less than the RLNC-based scheme on time consumption and at least 10% less than the classical BSNC-based scheme on storage overhead. The results also show that our matching policy and coded block delivery strategy can perform with a low response latency on edge and mobile devices.
{"title":"Bison: A Binary Sparse Network Coding Based Contents Sharing Scheme for D2D-Enabled Mobile Edge Caching Network","authors":"Cheng Peng;Jun Yin;Lei Wang;Fu Xiao","doi":"10.1109/TMC.2024.3463735","DOIUrl":"10.1109/TMC.2024.3463735","url":null,"abstract":"Mobile edge caching network (MEN), which enables popular or reusable content caching and sharing among adjacent mobile edge devices, has become a promising solution to reduce the traffic and burden over backhaul links. Network coding (NC), represented by classical random linear network coding (RLNC), is utilized to facilitate content delivery and increase throughput in MEN. However, as the harsh decoding condition results in unacceptable time and storage overhead, classical RLNC schemes struggle to be widely deployed in practice. In this work, we propose a cost-effective NC-based content-sharing scheme based on binary sparse network coding (BSNC), called Bison, for D2D-enabled MEN. Based on the shared relationship between the binary sparse coded block (BSCB), Bison first designs a caching maintenance module to characterize the sharing progress and maintain the caching state of each edge node. Then, Bison defines a matching metric named neighbor utility to evaluate neighbors’ matching values by considering nodes’ demand and content decodability. Guiding by the metric, Bison achieves the most beneficial matching relationship among edge nodes through a proposed online matching policy. Finally, Bison devises a coded block delivery strategy to enable the sharing of valuable content between two matched edge nodes. Extensive experiments in simulations and real-world Android testbeds demonstrate its effectiveness and efficiency, wherein Bison is at least 30% less than the RLNC-based scheme on time consumption and at least 10% less than the classical BSNC-based scheme on storage overhead. The results also show that our matching policy and coded block delivery strategy can perform with a low response latency on edge and mobile devices.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 2","pages":"677-695"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257219","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Real-time semantic segmentation (SS) is a major task for various vision-based applications such as self-driving. Due to the limited computing resources and stringent performance requirements, streaming videos from camera-embedded mobile devices to edge servers for SS is a promising approach. While there are increasing efforts on task-oriented video compression, most SS-applicable algorithms apply more uniform compression, as the sensitive regions are less obvious and concentrated. Such processing results in low compression performance and significantly limits the capacity of edge servers supporting real-time SS. In this paper, we propose STAC, a novel task-oriented DNN-driven video compressive streaming algorithm tailed for SS, to strike accuracy-bitrate balance and adapt to time-varying bandwidth. It exploits DNN's gradients as sensitivity metrics for fine-grained spatial adaptive compression and includes a temporal adaptive scheme that integrates spatial adaptation with predictive coding. Furthermore, we design a new bandwidth-aware neural network, serving as a compatible configuration tuner to fit time-varying bandwidth and content. STAC is evaluated in a system with a commodity mobile device and an edge server with real-world network traces. Experiments show that STAC can save up to 63.7–75.2% of bandwidth or improve accuracy by 3.1–9.5% compared to state-of-the-art algorithms, while capable of adapting to time-varying bandwidth.
实时语义分割（SS）是自动驾驶等各种基于视觉的应用的一项重要任务。由于有限的计算资源和严格的性能要求，从嵌入摄像头的移动设备流式传输视频到边缘服务器进行语义分割是一种很有前景的方法。虽然人们在面向任务的视频压缩方面做出了越来越多的努力，但大多数适用于 SS 的算法都采用了更均匀的压缩方式，因为敏感区域不太明显和集中。这种处理方式导致压缩性能低下，极大地限制了支持实时 SS 的边缘服务器的能力。在本文中，我们提出了 STAC，这是一种新颖的面向任务的 DNN 驱动视频压缩流算法，专为 SS 量身定制，以实现精度与比特率之间的平衡，并适应随时间变化的带宽。该算法利用 DNN 的梯度作为灵敏度度量，实现细粒度空间自适应压缩，并包含一个将空间自适应与预测编码相结合的时间自适应方案。此外，我们还设计了一种新的带宽感知神经网络，作为兼容配置调节器，以适应时变带宽和内容。STAC 在一个使用商品移动设备和边缘服务器的系统中通过真实的网络跟踪进行了评估。实验表明，与最先进的算法相比，STAC 能节省高达 63.7-75.2% 的带宽，或将准确率提高 3.1-9.5%，同时还能适应随时间变化的带宽。
{"title":"Task-Oriented Video Compressive Streaming for Real-Time Semantic Segmentation","authors":"Xuedou Xiao;Yingying Zuo;Mingxuan Yan;Wei Wang;Jianhua He;Qian Zhang","doi":"10.1109/TMC.2024.3446185","DOIUrl":"10.1109/TMC.2024.3446185","url":null,"abstract":"Real-time semantic segmentation (SS) is a major task for various vision-based applications such as self-driving. Due to the limited computing resources and stringent performance requirements, streaming videos from camera-embedded mobile devices to edge servers for SS is a promising approach. While there are increasing efforts on task-oriented video compression, most SS-applicable algorithms apply more uniform compression, as the sensitive regions are less obvious and concentrated. Such processing results in low compression performance and significantly limits the capacity of edge servers supporting real-time SS. In this paper, we propose STAC, a novel task-oriented DNN-driven video compressive streaming algorithm tailed for SS, to strike accuracy-bitrate balance and adapt to time-varying bandwidth. It exploits DNN's gradients as sensitivity metrics for fine-grained spatial adaptive compression and includes a temporal adaptive scheme that integrates spatial adaptation with predictive coding. Furthermore, we design a new bandwidth-aware neural network, serving as a compatible configuration tuner to fit time-varying bandwidth and content. STAC is evaluated in a system with a commodity mobile device and an edge server with real-world network traces. Experiments show that STAC can save up to 63.7–75.2% of bandwidth or improve accuracy by 3.1–9.5% compared to state-of-the-art algorithms, while capable of adapting to time-varying bandwidth.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"23 12","pages":"14396-14413"},"PeriodicalIF":7.7,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142269378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The ever-increasing quality of experience (QoE) demand for video streaming has prompted the integration of video super-resolution and multi-access edge computing networks (MEC). With super-resolution, the low-resolution frames can be reconstructed into high-resolution ones by edge node and end device collaboratively, which is beneficial in improving QoE. However, the existing works focus on designing video streaming strategies in single-user scenarios, which cannot be applied to multi-user scenarios due to the resource contention among users, as well as the huge solution space of coupled bitrate selection and workload share between edge-end. To fill this gap, we propose a collaborative video streaming strategy with super-resolution in multi-user MEC networks, named Co-Video, to maximize the average QoE by making optimal bitrate selection and workload share. We first formulate the problem as an optimization problem towards maximum average QoE, where the QoE incorporates playback delay, video quality, and smoothness. Then, we transform the optimization problem into a partially observable Markov decision process (POMDP) and exploit the Co-Video strategy based on the multi-agent soft actor-critic (MASAC) algorithm. Specifically, Co-Video utilizes the branching actor network to converge to good policy stably. Finally, trace-driven simulations on real-world bandwidth traces demonstrate that Co-Video outperforms the state-of-the-art baselines.
{"title":"Collaborative Video Streaming With Super-Resolution in Multi-User MEC Networks","authors":"Xiaobo Zhou;Jiaxin Zeng;Shuxin Ge;Xilai Liu;Tie Qiu","doi":"10.1109/TMC.2024.3461685","DOIUrl":"10.1109/TMC.2024.3461685","url":null,"abstract":"The ever-increasing quality of experience (QoE) demand for video streaming has prompted the integration of video super-resolution and multi-access edge computing networks (MEC). With super-resolution, the low-resolution frames can be reconstructed into high-resolution ones by edge node and end device collaboratively, which is beneficial in improving QoE. However, the existing works focus on designing video streaming strategies in single-user scenarios, which cannot be applied to multi-user scenarios due to the resource contention among users, as well as the huge solution space of coupled bitrate selection and workload share between edge-end. To fill this gap, we propose a collaborative video streaming strategy with super-resolution in multi-user MEC networks, named Co-Video, to maximize the average QoE by making optimal bitrate selection and workload share. We first formulate the problem as an optimization problem towards maximum average QoE, where the QoE incorporates playback delay, video quality, and smoothness. Then, we transform the optimization problem into a partially observable Markov decision process (POMDP) and exploit the Co-Video strategy based on the multi-agent soft actor-critic (MASAC) algorithm. Specifically, Co-Video utilizes the branching actor network to converge to good policy stably. Finally, trace-driven simulations on real-world bandwidth traces demonstrate that Co-Video outperforms the state-of-the-art baselines.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 2","pages":"571-584"},"PeriodicalIF":7.7,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257418","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-17DOI: 10.1109/TMC.2024.3461852
Xiaolu Wang;Zijian Li;Shi Jin;Jun Zhang
Federated learning (FL) is an emerging distributed training paradigm that aims to learn a common global model without exchanging or transferring the data that are stored locally at different clients. The Federated Averaging (FedAvg)-based algorithms have gained substantial popularity in FL to reduce the communication overhead, where each client conducts multiple localized iterations before communicating with a central server. In this paper, we focus on FL where the clients have diverse computation and/or communication capabilities. Under this circumstance, FedAvg can be less efficient since it requires all clients that participate in the global aggregation in a round to initiate iterations from the latest global model, and thus the synchronization among fast clients and straggler clients can severely slow down the overall training process. To address this issue, we propose an efficient asynchronous federated learning (AFL) framework called Delayed Federated Averaging (DeFedAvg) . In DeFedAvg, the clients are allowed to perform local training with different stale global models at their own paces. Theoretical analyses demonstrate that DeFedAvg achieves asymptotic convergence rates that are on par with the results of FedAvg for solving nonconvex problems. More importantly, DeFedAvg is the first AFL algorithm that provably achieves the desirable linear speedup property, which indicates its high scalability. Additionally, we carry out extensive numerical experiments using real datasets to validate the efficiency and scalability of our approach when training deep neural networks.
{"title":"Achieving Linear Speedup in Asynchronous Federated Learning With Heterogeneous Clients","authors":"Xiaolu Wang;Zijian Li;Shi Jin;Jun Zhang","doi":"10.1109/TMC.2024.3461852","DOIUrl":"10.1109/TMC.2024.3461852","url":null,"abstract":"Federated learning (FL) is an emerging distributed training paradigm that aims to learn a common global model without exchanging or transferring the data that are stored locally at different clients. The Federated Averaging (FedAvg)-based algorithms have gained substantial popularity in FL to reduce the communication overhead, where each client conducts multiple localized iterations before communicating with a central server. In this paper, we focus on FL where the clients have diverse computation and/or communication capabilities. Under this circumstance, FedAvg can be less efficient since it requires all clients that participate in the global aggregation in a round to initiate iterations from the \u0000<italic>latest</i>\u0000 global model, and thus the synchronization among fast clients and \u0000<italic>straggler clients</i>\u0000 can severely slow down the overall training process. To address this issue, we propose an efficient asynchronous federated learning (AFL) framework called \u0000<italic>Delayed Federated Averaging (DeFedAvg)</i>\u0000. In DeFedAvg, the clients are allowed to perform local training with different stale global models at their own paces. Theoretical analyses demonstrate that DeFedAvg achieves asymptotic convergence rates that are on par with the results of FedAvg for solving nonconvex problems. More importantly, DeFedAvg is the first AFL algorithm that provably achieves the desirable \u0000<italic>linear speedup</i>\u0000 property, which indicates its high scalability. Additionally, we carry out extensive numerical experiments using real datasets to validate the efficiency and scalability of our approach when training deep neural networks.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 1","pages":"435-448"},"PeriodicalIF":7.7,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142269295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Extremely Large-scale Array (ELAA) is considered a frontier technology for future communication systems, playing a crucial role in enhancing the rate and spectral efficiency of wireless networks. As ELAA employs a multitude of antennas operating at higher frequencies, users are typically situated in the near-field region where the spherical wavefront propagates. Near-field beam training requires information on both angle and distance, which inevitably leads to a significant increase in the beam training overhead. To address this challenge, we propose a near-field beam training method based on deep learning. Specifically, we employ a convolutional neural network (CNN) to efficiently extract channel characteristics from historical data by strategically selecting padding and kernel sizes. The negative value of the user average achievable rate is utilized as the loss function to optimize the beamformer, maximizing the achievable rate in multi-user networks without relying on predefined beam codebooks. Once deployed, the model requires only pre-estimated channel state information (CSI) to compute the optimal beamforming vector. Simulation results demonstrate that the proposed scheme achieves more stable beamforming gains and substantially outperforms traditional beam training approaches. Furthermore, owing to the inherent traits of deep learning methodologies, this approach substantially diminishes the near-field beam training overhead.
{"title":"Near-Field Beam Training for Extremely Large-Scale MIMO Based on Deep Learning","authors":"Jiali Nie;Yuanhao Cui;Zhaohui Yang;Weijie Yuan;Xiaojun Jing","doi":"10.1109/TMC.2024.3462960","DOIUrl":"10.1109/TMC.2024.3462960","url":null,"abstract":"Extremely Large-scale Array (ELAA) is considered a frontier technology for future communication systems, playing a crucial role in enhancing the rate and spectral efficiency of wireless networks. As ELAA employs a multitude of antennas operating at higher frequencies, users are typically situated in the near-field region where the spherical wavefront propagates. Near-field beam training requires information on both angle and distance, which inevitably leads to a significant increase in the beam training overhead. To address this challenge, we propose a near-field beam training method based on deep learning. Specifically, we employ a convolutional neural network (CNN) to efficiently extract channel characteristics from historical data by strategically selecting padding and kernel sizes. The negative value of the user average achievable rate is utilized as the loss function to optimize the beamformer, maximizing the achievable rate in multi-user networks without relying on predefined beam codebooks. Once deployed, the model requires only pre-estimated channel state information (CSI) to compute the optimal beamforming vector. Simulation results demonstrate that the proposed scheme achieves more stable beamforming gains and substantially outperforms traditional beam training approaches. Furthermore, owing to the inherent traits of deep learning methodologies, this approach substantially diminishes the near-field beam training overhead.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 1","pages":"352-362"},"PeriodicalIF":7.7,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257212","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-17DOI: 10.1109/TMC.2024.3462466
Xiaotong Zhang;Qingqiao Hu;Zhen Xiao;Tao Sun;Jiaxi Zhang;Jin Zhang;Zhenjiang Li
Wireless-based human activity recognition (WHAR) enables various promising applications. However, since WHAR is sensitive to changes in sensing conditions (e.g., different environments, users, and new activities), trained models often do not work well under new conditions. Recent research uses meta-learning to adapt models. However, they must fine-tune the model, which greatly hinders the widespread adoption of WHAR in practice because model fine-tuning is difficult to automate and requires deep-learning expertise. The fundamental reason for model fine-tuning in existing works is because their goal is to find the mapping relationship between data samples and corresponding activity labels. Since this mapping reflects the intrinsic properties of data in the perceptual scene, it is naturally related to the conditions under which the activity is sensed. To address this problem, we exploit the principle that under the same sensing condition, data of the same activity class are more similar (in a certain latent space) than data of other classes, and this property holds invariant across different conditions. Our main observation is that meta-learning can actually also transform WHAR design into a learning problem that is always under similar conditions, thus decoupling the dependence on sensing conditions. With this capability, general and accurate WHAR can be achieved, avoiding model fine-tuning. In this paper, we implement this idea through two innovative designs in a system called RoMF. Extensive experiments using FMCW, Wi-Fi and acoustic three sensing signals show that it can achieve up to 95.3% accuracy in unseen conditions, including new environments, users and activity classes.
{"title":"Few-Shot Adaptation to Unseen Conditions for Wireless-Based Human Activity Recognition Without Fine-Tuning","authors":"Xiaotong Zhang;Qingqiao Hu;Zhen Xiao;Tao Sun;Jiaxi Zhang;Jin Zhang;Zhenjiang Li","doi":"10.1109/TMC.2024.3462466","DOIUrl":"10.1109/TMC.2024.3462466","url":null,"abstract":"Wireless-based human activity recognition (WHAR) enables various promising applications. However, since WHAR is sensitive to changes in sensing conditions (e.g., different environments, users, and new activities), trained models often do not work well under new conditions. Recent research uses meta-learning to adapt models. However, they must fine-tune the model, which greatly hinders the widespread adoption of WHAR in practice because model fine-tuning is difficult to automate and requires deep-learning expertise. The fundamental reason for model fine-tuning in existing works is because their goal is to find the mapping relationship between data samples and corresponding activity labels. Since this mapping reflects the intrinsic properties of data in the perceptual scene, it is naturally related to the conditions under which the activity is sensed. To address this problem, we exploit the principle that under the same sensing condition, data of the same activity class are more similar (in a certain latent space) than data of other classes, and this property holds invariant across different conditions. Our main observation is that meta-learning can actually also transform WHAR design into a learning problem that is always under similar conditions, thus decoupling the dependence on sensing conditions. With this capability, general and accurate WHAR can be achieved, avoiding model fine-tuning. In this paper, we implement this idea through two innovative designs in a system called RoMF. Extensive experiments using FMCW, Wi-Fi and acoustic three sensing signals show that it can achieve up to 95.3% accuracy in unseen conditions, including new environments, users and activity classes.","PeriodicalId":50389,"journal":{"name":"IEEE Transactions on Mobile Computing","volume":"24 2","pages":"585-599"},"PeriodicalIF":7.7,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142257215","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","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：