Computer Communications最新文献

{"title":"Deep-reinforcement-learning–guided resource allocation and task offloading for 6G edge intelligence","authors":"Jiangbo Tang","doi":"10.1016/j.comcom.2025.108364","DOIUrl":"10.1016/j.comcom.2025.108364","url":null,"abstract":"<div><div>The proliferation of Internet of Everything (IoE) applications in 6G-enabled smart edge environments has intensified the demand for resource-efficient, latency-aware, and adaptive scheduling mechanisms. Conventional static or fairness-driven algorithms often fail to meet the scalability and responsiveness requirements of resource-constrained infrastructures. To address this challenge, we propose a novel hybrid scheduling framework, termed EOTSA (Energy-Optimized Task Scheduling and Allocation), which synergistically integrates Deep Q-Network (DQN)-based reinforcement learning with Particle Swarm Optimization (PSO). In this hybrid paradigm, DQN serves as the primary decision-maker, learning long-term optimal policies for dynamic selection among three specialized scheduling strategies Dynamic Priority Scheduling (DPS), Energy-Aware Fair Scheduling (EAFS), and Latency-Sensitive Adaptive Scheduling (LSAS) by modeling the task offloading process as a Markov Decision Process and maximizing a multi-objective reward function encompassing latency, energy, and QoS. However, DQN suffers from slow convergence and susceptibility to local optima in high-dimensional, dynamic 6G environments. PSO complements DQN by enhancing exploration efficiency: it generates a population of candidate task-device assignments, refines them iteratively using velocity-position updates, and provides high-quality initial actions to the DQN agent, accelerating convergence and escaping local optima. Individually, DQN excels in sequential decision-making under uncertainty, while PSO is superior in parallel global search for continuous optimization spaces. Together, the hybrid framework resolves the limitations of pure DRL and pure swarm intelligence, achieving robust, real-time adaptability in heterogeneous workloads. Simulation results benchmarked against RR, SC, MR, PF, PB, DRA, and ATO demonstrate that EOTSA achieves up to 30 % energy savings, a 25 % reduction in task completion time, and a 20 % improvement in Quality of Service (QoS). These results highlight EOTSA's superior adaptability across heterogeneous execution scenarios, positioning it as a scalable and sustainable solution for intelligent edge computing in forthcoming 6G-based IoE networks.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"245 ","pages":"Article 108364"},"PeriodicalIF":4.3,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145580322","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":"Analytical-based resource allocation framework for NOMA-assisted Semi-ISAC systems","authors":"Dinh Van Tung ,&nbsp;Thai-Hoc Vu ,&nbsp;Nguyen Tien Hoa","doi":"10.1016/j.comcom.2025.108354","DOIUrl":"10.1016/j.comcom.2025.108354","url":null,"abstract":"<div><div>Emerging 6G applications, such as autonomous systems and immersive extended reality, require joint communication and sensing to meet stringent performance demands. Integrated sensing and communication (ISAC) has thus emerged as a promising paradigm for supporting such dual functionality in future wireless networks. This paper proposes a novel optimization framework for joint spectrum and power allocation in semi-ISAC systems assisted by non-orthogonal multiple access (NOMA). The objective is to maximize the minimum ergodic achievable rate under statistical channel state information (CSI), thereby ensuring fairness across heterogeneous communication and sensing services. The non-convex problem is reformulated using successive convex approximation (SCA) for efficient and tractable optimization. Closed-form expressions for ergodic rates are derived under two NOMA configurations: single layer per sub-band and multiple layers per sub-band, highlighting the trade-off between decoding complexity and spectral efficiency. Numerical results highlight four key performance benefits: (i) a guaranteed minimum rate of 2 Gbps per user at 20 dBm transmit power, (ii) improved fairness based on Jain’s index, (iii) higher ergodic sum rate compared to benchmark schemes, and (iv) robustness to channel fading and target variations such as Nakagami-m parameters, sensing distance, and radar cross-section. These findings confirm the adaptability and efficiency of the proposed framework for dense deployment scenarios in semi-ISAC networks.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"245 ","pages":"Article 108354"},"PeriodicalIF":4.3,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528997","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":"Operator coexistence in IRS-assisted mmWave networks: A wideband approach","authors":"Joana Angjo,&nbsp;Anatolij Zubow,&nbsp;Falko Dressler","doi":"10.1016/j.comcom.2025.108341","DOIUrl":"10.1016/j.comcom.2025.108341","url":null,"abstract":"<div><div>In sixth generation (6G) mobile networks, the push towards high-frequency bands for ultra-fast data rates intensifies the challenges of signal attenuation and reduced coverage range. Intelligent reconfigurable surfaces (IRSs) present a promising solution to these challenges by enhancing signal coverage and directing reflections, which also contribute to minimize loss. However, there are multiple challenges associated with IRS, which have to be addressed before full incorporation of this technology into existing networks. A key issue arises from the inability of IRS to filter out non-target signals from other frequency bands due to lack of bandpass filtering. In areas where multiple wireless operators are spatially nearby, even if they use different frequency bands, this may cause unwanted reflections that may degrade their communication performances. To address this challenge, we previously proposed a solution, which relied on partitioning an IRS into sub-surfaces (<span>sub_IRS</span>) and dynamically assigning operators to these <span>sub_IRS</span>. Results have shown that a proper assignment of wireless operators to <span>sub_IRS</span> can improve the overall performance compared to a random assignment. In this paper, we introduce a wideband approach, demonstrating that the impact from unwanted reflections can be mitigated by using wideband channels, as the average signal to noise ratio (SNR) across subcarriers is less adversely affected. This approach leverages frequency diversity to reduce SNR variance, as some of the subcarriers may be negatively affected while others benefit, resulting in maintaining a more consistent and robust system performance in the presence of IRS-induced unwanted reflections. Simulations and real-world measurements confirm that the deployment of wideband IRS provides a robust strategy for combating inter-operator reflections in next generation IRS-assisted networks. Additionally, the wideband approach comes at no additional necessity for centralized resource control in future multi-operator networks. According to simulations, the SNR variance for a 1.28<!--> <!-->GHz channel is approximately 20 dB lower than that of a 10<!--> <!-->MHz channel when coexistence is considered. Similarly, measurements confirm a threefold reduction in SNR variation when transitioning from narrowband (10<!--> <!-->MHz) to wideband (320<!--> <!-->MHz) transmission. In overall, the usage of wideband channels in this context allows the system to be more stable and predictable.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108341"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145366115","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":"M2M group-based AI routing protocol for IoT networks","authors":"Jaime Lloret ,&nbsp;Jesús Tomás ,&nbsp;Pedro Luis González Ramírez ,&nbsp;Martín Jiménez-Piedrahita ,&nbsp;Milos Stojmenovic","doi":"10.1016/j.comcom.2025.108338","DOIUrl":"10.1016/j.comcom.2025.108338","url":null,"abstract":"<div><div>Internet of Things (IoT) networks are heterogeneous networks that offer different services depending on their application environment through smart devices (also called nodes). Machine-to-Machine (M2M) routing protocols provide interconnection between nodes. However, IoT and Internet of Everything (IoE) networks are limited to constraints defined by the scenario of the services they implement. For this reason, in this paper, we propose a new routing protocol that uses Artificial intelligence (AI) to interconnect the nodes of an IoT network efficiently. The AI decides the routes based on the tables created by the Publish/Subscribe (Pub/Sub) communication system, relating it to the dataset of features of each device connected in the network (content-based filtering). Consequently, the IoT gateway selects the best node that can provide it without compromising the performance of the network. We performed a routing protocol simulation between the machine that requests resources and who provides them. The proposal is based on a multi-objective optimization technique that uses evolutionary algorithms. The results show that the algorithm selects the most optimal node that can provide the requested resources without any constraint violation and evaluates the most efficient criteria. Consequently, the proposed algorithm can be applied to provide IoT services attending to several constraints derived from the scenarios where the system is implemented.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108338"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145366116","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":"Clustered federated learning with heterogeneous differential privacy on Non-IID data","authors":"Ping Guo ,&nbsp;Cheng Bai ,&nbsp;Mingxing Zhang ,&nbsp;Puwadol Oak Dusadeerungsikul","doi":"10.1016/j.comcom.2025.108339","DOIUrl":"10.1016/j.comcom.2025.108339","url":null,"abstract":"<div><div>Federated Learning (FL) has emerged as a promising technology that has garnered significant attention in the Internet of Things (IoT) domain. However, the non-independent and identically distributed (Non-IID) nature of IoT data, coupled with the vulnerability of gradient transmission in traditional federated learning frameworks, limits its broader applicability. Heterogeneous differential privacy offers tailored privacy protection for individual clients, making it particularly well-suited for the diverse functional requirements of IoT devices. This study proposes a clustered federated learning method with heterogeneous differential privacy (FedCDP) to balance model utility and privacy preservation on Non-IID data. Specifically, we employed a two-stage clustering technique to enhance clustering accuracy amidst noise perturbations, and implement a client verification procedure to mitigate the detrimental effects of erroneous clustering and malicious data injection. To solve the problem of noise accumulation in cluster models, we introduced an intra-cluster privacy budget weighting mechanism, and used model shuffling to prevent the server from obtaining the cluster identity corresponding to the local model. We conducted experimental evaluations under multiple data distribution scenarios, and these experimental results show that our method effectively improves robustness to noise and significantly improves model performance compared to the baseline methods. In addition, we perform ablation experiments on each module to further analyze the impact of each module on the method. These findings underscore the usability and robustness of the proposed method.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108339"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145366113","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":"Multi-antenna mobile charger scheduling optimization scheme for wireless rechargeable sensor networks","authors":"Jinyi Li ,&nbsp;Yong Feng ,&nbsp;Nianbo Liu ,&nbsp;Ming Liu ,&nbsp;Yingna Li","doi":"10.1016/j.comcom.2025.108343","DOIUrl":"10.1016/j.comcom.2025.108343","url":null,"abstract":"<div><div>Multi-antenna mobile chargers (MC) featuring directional multi-beam functionality present a promising solution for energy replenishment in wireless rechargeable sensor networks. However, existing multi-antenna scheduling schemes encounter challenges in jointly optimizing the coupled problem of Antenna Configuration and Path Planning (ACPP) while balancing MC’s coverage efficiency with energy consumption. To address this gap, this paper investigates the complex interdependencies and stringent constraints inherent in ACPP, and proposes a phased hybrid optimization scheme, PHMS-ACPP, integrating multi-objective optimization and deep reinforcement learning to compute approximate solutions. We first employ a modified Gaussian mixture model incorporating physical coverage constraints via the Expectation–Maximization algorithm to partition clusters, thereby reducing problem complexity. Within each cluster, the subproblem of determining optimal antenna count and orientation is solved using the Multi-objective Grey Wolf Optimizer to simultaneously optimize MC’s coverage efficiency and energy consumption. Then, we utilize Double Deep Q-Network to plan MC’s charging path across clusters, which captures long-term temporal dependencies between the evolution of nodes’ energy states and the spatial allocation of charging resources, enhancing both global scheduling efficacy and long-term charging efficiency. Extensive simulations demonstrate that PHMS-ACPP significantly outperforms state-of-the-art baselines in reducing node failure rate and minimizing average charging delay, with reductions of approximately 21.6% and 14.4%, respectively.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108343"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145366114","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":"Multi-agent deep reinforcement learning for service function chain deployment in software defined LEO satellite networks","authors":"Pingduo Xu ,&nbsp;Debin Wei ,&nbsp;Jinglong Wen ,&nbsp;Li Yang","doi":"10.1016/j.comcom.2025.108342","DOIUrl":"10.1016/j.comcom.2025.108342","url":null,"abstract":"<div><div>Large-scale low earth orbit (LEO) satellite networks constitute a core component of future sixth-generation (6G) communication systems. To address the challenges of resource scarcity and highly dynamic topologies, the integration of software-defined networking (SDN) and network function virtualization (NFV) technologies into LEO satellite networks has become imperative. We proposes a hybrid centralized-distributed software-defined LEO satellite network architecture. Within this framework, This study focuses on the service function chain (SFC) deployment problem in LEO space-ground integrated networks. time-expanded graphs (TEGs) are employed to model satellite networks with dynamic topological variations, aiming to satisfy diverse user requirements while jointly optimizing resource consumption costs and service latency. The problem is formulated as a weighted sum minimization of resource consumption costs and service latency, and this problem is proven to be NP-complete. Subsequently, we integrate the twin delayed deep deterministic policy gradient method with multi-agent techniques to design a multi-agent deep reinforcement learning SFC deployment (MADRL-D) framework for optimizing our objectives. Experimental results demonstrate that the proposed MADRL-D framework outperforms existing alternatives in terms of resource utilization efficiency, resource consumption costs, and service latency.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108342"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145340548","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":"Dynamic distance-based load balancing in mobile edge computing with deep reinforcement learning","authors":"Mohammad Esmaeil Esmaeili ,&nbsp;Ahmad Khonsari ,&nbsp;Mahdi Dolati","doi":"10.1016/j.comcom.2025.108337","DOIUrl":"10.1016/j.comcom.2025.108337","url":null,"abstract":"<div><div>Edge computing reduces latency by bringing computation closer to end devices, but the growing scale and heterogeneity of edge networks make resource management increasingly complex. Load balancing is essential for efficient resource use and low response times, yet static approaches struggle in dynamic environments. This calls for adaptable, data-driven load balancing methods that can continuously respond to changing conditions and optimize performance. This paper addresses the problem of load balancing in edge computing, where the distance between servers plays a critical role in performance. We propose two deep reinforcement learning (DRL)-based algorithms – Deep Q-Learning (DQL) and Long Short-Term Memory (LSTM) – that dynamically adjust the neighbor radius for load distribution in response to environmental changes. Unlike static approaches, our methods learn the radius online in a data-driven manner without requiring global coordination. Simulation results demonstrate that both algorithms adapt effectively to dynamic conditions. In scenarios with 80–100 edge servers and 500–1000 requests per second, DQL achieves up to 18% higher throughput, 21% lower average response time, and 23% lower blocking rate compared to recent methods, while LSTM remains competitive under stable workloads.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108337"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145419909","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":"Macroscopic diffusion prediction in social networks based on spatio-temporal and trend features","authors":"Xueqin Zhang ,&nbsp;Yisong Lu ,&nbsp;Gang Liu ,&nbsp;Xiaowei Chen","doi":"10.1016/j.comcom.2025.108352","DOIUrl":"10.1016/j.comcom.2025.108352","url":null,"abstract":"<div><div>Predicting the scale of information diffusion in social networks can sense the future propagation of information in advance, which plays a crucial role in controlling the diffusion of harmful information. We propose STTFP (Spatio-Temporal and Trend Features for Prediction), a deep learning framework that integrates temporal, spatial, and trend features to improve macroscopic diffusion prediction accuracy. This framework first utilizes graph attention networks to extract node interaction features from cascaded graphs. It captures node position features from diffusion sequences, and uses sparse matrix factorization to extract node features from social network graphs. Then it adopts bi-directional gated recurrent units and self-attention mechanisms to deeply mine spatio-temporal features. Additionally, we design an attention-based convolutional neural network to capture the short-term fluctuations in the information propagation process, while long short-term memory networks are used to uncover historical forwarding variation in diffusion scales. By fusing these features, the framework achieves incremental predictions of information diffusion. Experiments on three public datasets show that our method effectively enhances the accuracy of macroscopic diffusion predictions.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108352"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145467569","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":"FANT: Flexible active in-band network telemetry","authors":"Mingwei Cui,&nbsp;Yufan Peng,&nbsp;Ying Wang,&nbsp;Tianyuan Niu,&nbsp;Fan Yang","doi":"10.1016/j.comcom.2025.108336","DOIUrl":"10.1016/j.comcom.2025.108336","url":null,"abstract":"<div><div>With the ever-increasing complexity of networks, the need for fine-grained and comprehensive network monitoring has become crucial. In order to provide higher precision and multidimensional network status, In-band Network Telemetry (INT) has been incorporated into certain modern merchant silicon devices. INT, specifically active INT (ANT), in conjunction with SDN, enables network-wide visibility by capturing device-internal status along probe traces. However, as network management applications require more measurements from different perspectives to identify potential network issues, switch vendors are continuously expanding the scope of telemetry measurements available on their devices. The proliferation of telemetry items results in increased overhead for the network and places a heavier processing load on the INT analyzer. In this paper, we introduce a novel approach called Flexible Active in-band Network Telemetry (FANT) that enhances the measurement capabilities of the system by decoupling it into a mechanism and a probe-switching policy. FANT divides the measurement space into basic and detailed parts based on the relationship between INT metadata. The mechanism of FANT leverages a combination of ANT and Software-Defined Networking (SDN) to perform network-wide basic measurements. By centrally analyzing these measurements, FANT dynamically adjusts certain port measurements from basic to detailed. Additionally, we propose two probe-switching policies to generate <em>low-redundant</em> Detail Probes that cover all abnormal ports. In the first policy, we present an optimized <em>Iterative deepening</em> <span><math><msup><mrow><mi>A</mi></mrow><mrow><mo>∗</mo></mrow></msup></math></span> algorithm that provides near-optimal solutions in low measurement frequency scenarios. The second policy achieves improved performance and reduced overcontrol using the <em>random cover</em> algorithm in high measurement frequency scenarios. To validate the effectiveness of the proposed mechanism, we implement a FANT prototype system. Furthermore, we extensively evaluate FANT’s performance in various topologies, including wide area networks and data center networks of different scales.</div></div>","PeriodicalId":55224,"journal":{"name":"Computer Communications","volume":"244 ","pages":"Article 108336"},"PeriodicalIF":4.3,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145467570","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}