Computer Communications最新文献

{"title":"Digital twin-assisted multi-layer networks for low-latency and energy-efficient communication","authors":"Muhammad Adnan Qadir ,&nbsp;Muhammad Naeem ,&nbsp;Waleed Ejaz","doi":"10.1016/j.comcom.2025.108219","DOIUrl":"10.1016/j.comcom.2025.108219","url":null,"abstract":"<div><div>The sixth-generation (6G) wireless networks are expected to provide ubiquitous connectivity, high data rate, low latency, energy efficiency, and edge intelligence for Internet of Things (IoT) applications. Digital twin technology is a promising solution to enable multi-layer wireless networks that incorporate IoT devices on the ground, unmanned aerial vehicles (UAVs) as mobile edge computing (MEC) servers, and cloud servers. Multi-layer processing can handle time-sensitive and computationally intensive tasks from IoT devices. This paper proposes a digital twin-assisted multi-layer network for low-latency and energy-efficient communication and computation. We mathematically formulate an optimization problem to minimize the latency and energy consumption of IoT devices by optimizing their association with the UAV-MECs, computation resources, communication resources, and offloading portions of tasks. We propose a two-stage scheme based on the K-means method and the deep neural network approach to solve the above optimization problem. We compare the proposed two-stage scheme with existing schemes to highlight the scalability of the proposed solution. Simulation results demonstrate that the proposed multi-layer network achieved optimization results comparable to existing schemes with less computational cost, highlighting its usefulness in achieving low latency and energy-efficient computation and communication.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108219"},"PeriodicalIF":4.5,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance of UAV-assisted C-V2X communications with 3D antenna beam-width fluctuations","authors":"Mohammad Arif ,&nbsp;Wooseong Kim ,&nbsp;Asif Mehmood","doi":"10.1016/j.comcom.2025.108267","DOIUrl":"10.1016/j.comcom.2025.108267","url":null,"abstract":"<div><div>The antenna’s three-dimensional (3D) beam-width orientation is crucial in assessing the effectiveness of vehicular communications. This paper investigates the influence of variations of millimeter waveband antenna 3D beam-width on the performance of un-crewed aerial vehicle (UAV)-assisted cellular vehicle-to-everything (C-V2X) communications. The cellular base-stations are represented using a two-dimensional Poisson point process (PPP), while vehicular nodes (V-Ns) are represented using a Poisson line process, and UAVs are represented using a 3D PPP. The typical transmitting V-N can connect with the nearest V-N in direct mode transmission or with the (macro base-station) MBS, line-of-sight (LOS) UAV, or non-LOS (NLOS) UAV in shared mode transmission. The efficiency of the system is measured by using the antenna’s 3D beam-width relative to coverage and spectrum efficiency. To that aim, analytical equations for the association and coverage probability of vehicular-to-vehicular, vehicular-to-MBS, vehicular-to-LOS UAV, and vehicular-to-NLOS UAV connections are obtained in the setting of variation in beam-width. The efficiency is also measured in terms of V-Ns, MBS, and UAVs. The findings revealed that our system, considering millimeter waveband-based UAV-assisted C-V2X network leveraging the benefits of MBSs and UAVs, performs better than the conventional V2X network. The findings reveal that the efficiency of the UAV-assisted C-V2X networks is affected by the variable 3D beam-width, hence, it needs to be thoroughly specified. Furthermore, the network’s performance degrades when the UAV’s beam-width variations grow.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"242 ","pages":"Article 108267"},"PeriodicalIF":4.5,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144687018","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Altitude control and power allocation with reinforcement learning for sum capacity maximization in high-data-rate swarm drone interventions","authors":"Oluwatosin Ahmed Amodu ,&nbsp;Rosdiadee Nordin ,&nbsp;Nor Fadzilah Abdullah ,&nbsp;Asma’ Abu-Samah","doi":"10.1016/j.comcom.2025.108257","DOIUrl":"10.1016/j.comcom.2025.108257","url":null,"abstract":"<div><div>Several efforts have been made to study the achievable performance gains of UAV-assisted communications, especially in emergency and disaster scenarios, as an alternative to traditional cellular communication, which is prone to damage. In such situations, first-hand responders require an aerial view of the terrain for situational awareness to assess the extent of damage and plan the nature and level of intervention required. Such applications require high data rates and proper UAV altitude control. Existing work on the 3D location of aerial base stations (ABS) does not study high-data-rate applications involving UAV interventions, which are critical and sometimes life-saving. Furthermore, existing work does not provide sufficient details on the mapping of state representations. It uses a fixed representation of the state space, which limits the exploration capabilities of the agent in the optimization process. Also, mutual distances affecting interference were not considered in the state space, whereas interference severely affects the sum capacity performance. In this paper, a reinforcement learning-based altitude control framework for sum capacity maximization using UAVs as ABSs is also proposed. In the scenario studied, portable ground stations, such as those that can be set up in an ad hoc manner, are assumed to assess the situation via the aerial footage of the ABS for situational awareness. In this case, the ABS should ascend whenever its location in 2D space is very close to those of the ground stations to provide extended coverage of their footage. A control mechanism is developed for UAV altitude. A new reward function, as well as, a fine-grained representation of the state transition and a more precise design of the state space, are introduced. ABS are positioned via the <span><math><mi>k</mi></math></span>-means algorithm with a large discrete search space. These ABS learn to select the transmit power that maximizes the sum capacity during training while also deploying the proposed altitude control mechanism. The proposed optimization framework improves the state-of-the-art Q-learning-based scheme without altitude control and equal power allocation algorithms, by over 20% and 55% for 8 and 16 UAVs, respectively.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108257"},"PeriodicalIF":4.5,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Pareto-based genetic algorithm for online task allocation in mobile crowdsensing","authors":"Xiangling Wu,&nbsp;Wenming Ma,&nbsp;Xiao Zhu,&nbsp;Shengyang Sun,&nbsp;Xiaoang Zhu","doi":"10.1016/j.comcom.2025.108269","DOIUrl":"10.1016/j.comcom.2025.108269","url":null,"abstract":"<div><div>Mobile CrowdSensing (MCS) is a widely adopted sensing paradigm that utilizes the strength of multiple mobile users to carry out various location-based sensing tasks. The task allocation problem, which involves assigning tasks to suitable mobile users, is a critical issue in the design of MCS systems. Many existing studies neglect the time constraints associated with both participants and tasks, often ignoring task execution times and assuming that a task is considered completed once the participant arrives at the task location. In order to take execution time into consideration, this study suggests an online heterogeneous task allocation problem with time constraints. The objective is to minimize the reward while simultaneously maximizing sensing quality. To solve this problem, we offer an enhanced approach that is based on a genetic algorithm and uses a bipartite network to account for the assignability link between participants and tasks. Additionally, we employ Pareto optimization to balance the dual objectives of minimizing reward and maximizing task quality. Results from experiments show that the suggested algorithm works better than baseline methods in many experimental configurations.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108269"},"PeriodicalIF":4.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid mobility in opportunistic networks: Insights into enhanced PIPeR variants for subway settings","authors":"Sara ElSingergy,&nbsp;Soumaia Al Ayyat,&nbsp;Sherif G. Aly","doi":"10.1016/j.comcom.2025.108277","DOIUrl":"10.1016/j.comcom.2025.108277","url":null,"abstract":"<div><div>This paper explores the performance of the Power and Interest Aware PeopleRank (PIPeR) algorithm, a prominent opportunistic network forwarding algorithm, within a new context namely, subway mobility environments. While PIPeR has demonstrated strong performance in pedestrian mobility models by accounting for both interest in disseminated content and power conservation, its application in subway settings—characterized by hybrid mobility and unique challenges—has yet to be explored. Subway mobility scenarios are particularly relevant for contexts such as emergency response for civilians use during crises, where fixed network infrastructure is limited or unavailable. In this study, PIPeR is implemented and evaluated using the AnyLogic simulator, which accurately models subway passenger flows under hybrid mobility. Additionally, five enhanced variants of the original PIPeR algorithm are proposed, designed to address the unique challenges of subway environments and any other environments of similar mobility patterns, aiming to enhance the algorithm's overall efficiency. The best-performing variant is then identified and rigorously tested through experiments to evaluate its robustness under varying conditions, including various interest distributions, battery distributions, user density, and message volume per user. The results reveal that the PIPeR algorithm in the subway environment achieves a notable 64 % increase in the F-measure and a 63 % reduction in delay compared to the pedestrian mobility environment, but at the cost of increased power consumption and cost. The proposed variants mitigate these challenges, achieving an impressive 83 % reduction in power consumption and a 38 % decrease in cost, with a trade-off of a 20 % reduction in F-measure. These findings highlight a significant step towards green computing and sustainability in opportunistic networks. Moreover, the best-performing variant, when tested in a challenging scenario with a majority of uninterested users and a lack of intermediate forwarders, demonstrates excellent performance, further underscoring its adaptability and robustness.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108277"},"PeriodicalIF":4.5,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144656399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Autonomous cyber defense for AIoT using Graph Attention Network-Enhanced reinforcement learning","authors":"Yihang Shi ,&nbsp;Huajun Zhang ,&nbsp;Lin Shi,&nbsp;Shoukun Xu","doi":"10.1016/j.comcom.2025.108265","DOIUrl":"10.1016/j.comcom.2025.108265","url":null,"abstract":"<div><div>The rapid development of the Artificial Intelligence of Things (AIoT) has established a highly interconnected device ecosystem, enhancing user convenience while exposing it to increasingly severe Advanced Persistent Threats (APTs). To address these challenges, defenders have increasingly adopted Deep Reinforcement Learning (DRL) methods, leveraging their self-learning and adaptability to strengthen network security. However, experiments reveal that existing DRL algorithms, which rely on one-dimensional observational data, often face challenges in adequately interpreting network topologies and inter-device interactions, thus impacting their defensive performance. To overcome this limitation, we propose integrating the weighting mechanism of Graph Attention Networks (GAT) to process complex network connectivity data, enabling intelligent defense agents to dynamically capture dependencies and interactions within intricate AIoT architectures and execute precise, rapid defensive actions. Accordingly, we combine GAT with DRL, integrating it with value-based Deep Q-Network (DQN) and policy-based Proximal Policy Optimization (PPO) algorithms to develop GAT-DQN and GAT-PPO. Extensive experiments on the Yawning Titan network security simulation platform demonstrate that GAT-DQN and GAT-PPO significantly outperform traditional methods, particularly as the complexity of scenarios increases, graph attention-enhanced DRL algorithms demonstrate greater stability and robustness in performance. These results highlight the effectiveness of graph-enhanced DRL in mitigating APTs, underscoring its potential for achieving autonomous network defense in future AIoT communication networks.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108265"},"PeriodicalIF":4.5,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144597226","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Novel caching framework for information-centric IoT using deep reinforcement Proximal Policy Optimization","authors":"Hamid Asmat ,&nbsp;Fasee Ullah ,&nbsp;Arfat Ahmad Khan ,&nbsp;Farman Ali ,&nbsp;Muhammad Ismail Mohmand","doi":"10.1016/j.comcom.2025.108261","DOIUrl":"10.1016/j.comcom.2025.108261","url":null,"abstract":"<div><div>The Internet of Things (IoT) continues to evolve rapidly, necessitating innovative approaches to content delivery as the number of connected devices increases. Integrating Information-Centric Networking (ICN) within IoT environments offers a transformative solution, shifting from host-centric to content-centric architectures. This shift is particularly suitable for the distributed nature of IoT applications, which enhances content retrieval and distribution efficiency. However, the dynamic and diverse patterns of IoT networks require intelligent and adaptive caching mechanisms. This paper proposes an enhanced centrally controlled cache (ECCC) scheme that integrates the Proximal Policy Optimization (PPO) algorithm to optimize caching decisions in ICN-IoT environments. The ECCC scheme adapts in real time, adjusting caching strategies based on network conditions, resulting in improved network performance, higher energy efficiency, and reduced server load. The ECCC demonstrated an average energy savings of 15% and a cache-hit ratio improvement of 10% compared to traditional schemes. Extensive simulations demonstrate that ECCC outperforms traditional caching schemes, significantly improving the efficiency of the IoT application network and resource management. Furthermore, this work opens up new opportunities for smart cities, autonomous systems, and edge computing applications, where real-time data access and efficient resource management are critical.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108261"},"PeriodicalIF":4.5,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144562914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Load-balanced scheduling optimization strategy for high-communication tasks in Kubernetes with RDMA","authors":"Donglei Xiao ,&nbsp;Wenhui Shen ,&nbsp;Huiyue Yi ,&nbsp;Wuxiong Zhang","doi":"10.1016/j.comcom.2025.108260","DOIUrl":"10.1016/j.comcom.2025.108260","url":null,"abstract":"<div><div>Remote Direct Memory Access (RDMA) is a low-latency, high-bandwidth communication technology. Efficient and balanced RDMA load balancing and bandwidth utilization are critical for optimizing the performance of high-communication tasks in heterogeneous resource environments. However, the existing Kubernetes scheduling strategies fall short in balancing the RDMA resource loads and optimizing the distribution of TCP-Pods and RDMA-Pods, thereby hindering the performance of high-communication tasks that rely on RDMA bandwidth. To address these challenges, this paper proposes a load-balanced scheduling optimization strategy that integrates RDMA bandwidth, CPU, memory utilization, and node load balancing metrics for high-communication tasks in Kubernetes with RDMA. Specifically, the proposed strategy extends Kubernetes’ resource management capabilities to support dynamic monitoring of RDMA node bandwidth states. During the filtering phase, the nodes are categorized into RDMA nodes and TCP nodes, while Pods are classified as RDMA-Pods or TCP-Pods based on their communication requirements, corresponding to high-communication and ordinary tasks, respectively. In the scoring phase, a comprehensive scoring mechanism incorporates fairness factors and resource utilization metrics to ensure that high-communication tasks are preferentially allocated to appropriate RDMA nodes, thereby avoiding resource contention and suboptimal distribution. Additionally, the distribution of ordinary tasks is optimized to reduce interference with RDMA node resources. Experimental results show that the proposed scheduling strategy significantly improves the RDMA Load Balancing Rate, meets the performance demands of high-communication tasks, and enhances the overall resource utilization and scheduling efficiency within the cluster.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108260"},"PeriodicalIF":4.5,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144605307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Latency-optimized multi-task collaborative computing mechanism based on NOMA-D2D for AIoT","authors":"Sujie Shao,&nbsp;Lili Su,&nbsp;Shaoyong Guo,&nbsp;Siya Xu,&nbsp;Xuesong Qiu","doi":"10.1016/j.comcom.2025.108266","DOIUrl":"10.1016/j.comcom.2025.108266","url":null,"abstract":"<div><div>With the development of communication technologies such as NOMA-D2D, computing power and network resource sharing in the task collaboration process have shown great potential in AIoT. However, in the face of concurrent and low-latency task requirements, improper scheduling of computing and communication resources will lead to task timeout problems. This study proposes a latency-optimized multi-task collaborative computing mechanism, focusing on task offloading and resource optimization. First, a multi-task collaborative computing architecture based on a multi-hop NOMA-D2D cellular network is constructed, and a multi-task collaborative computing delay model is designed by analyzing co-frequency reuse interference. Secondly, the HGCG algorithm is designed to achieve the optimal alliance grouping based on hedonic game by comprehensively considering factors such as distance and resource supply and demand under dynamic network state changes. In addition, the HGCG-MAPPO algorithm is designed to achieve joint optimization of task offloading, transmission power and channel resource allocation decisions under complex environments. Simulation results show that the proposed method can dynamically adapt to network state changes, effectively reduce task completion delay by at least 14.97% compared with other methods, and show strong robustness under high task loads.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108266"},"PeriodicalIF":4.5,"publicationDate":"2025-07-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144581005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Maritime communication networks: A survey on architecture, key technologies, and challenges","authors":"Ziqi Shang,&nbsp;Xian Zhang,&nbsp;Xuehua Li","doi":"10.1016/j.comcom.2025.108255","DOIUrl":"10.1016/j.comcom.2025.108255","url":null,"abstract":"<div><div>Maritime Communication Networks (MCNs) are crucial for ensuring marine security, facilitating global trade, and advancing maritime scientific research. However, establishing a stable and efficient communication network is challenging due to the complex marine environment, such as tides, wind, waves, sea surface scattering, refraction, and evaporation ducts. Additionally, the non-uniform distribution of marine terminals and the differentiated service requirements further exacerbate its complexity. Recent advancements in multi-domain technologies (shipborne/airborne/satellite systems) have driven MCNs toward cognitive and adaptive architectures. This survey begins by outlining the current applications, demands and challenges of maritime communications. Next, we sequentially overview five MCN architectures, including the maritime mobile ad-hoc network (M-MANET), onshore BSs-assisted, satellite-assisted, UAV-assisted, and space–air–ground–sea integrated network (SAGSIN) architectures. Then, we elaborate on recent advancements in key generic technologies under different network architectures, covering spectrum strategies, beam management, multi-access edge computing, and routing. Finally, we discuss and analyze the open issues and challenges in MCNs, such as cross-domain resources collaborative optimization. This survey is intended to serve as a reference for researchers interested in MCNs.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"241 ","pages":"Article 108255"},"PeriodicalIF":4.5,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144518410","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}