Aug 2026· ACM Transactions on Internet Technology· 0 citations· 23 references
TL;DR
This work introduces KAT-MAPPO, a Kolmogorov–Arnold Transformer-enhanced Multi-Agent Proximal Policy Optimization scheme that significantly outperforms existing baselines and highlights the promise of reinforcement learning considering temporal dynamics for efficient offloading in heterogeneous CEG systems.
Abstract
In post-disaster or infrastructure-limited remote-area environments where local terrestrial networking infrastructure is unavailable or severely degraded, heterogeneous cloud–edge–ground (CEG) systems provide a viable platform for computation offloading. Ground devices such as IoT sensors and unmanned ground vehicles (UGVs) generate computation-intensive and latency-sensitive tasks, while fixed-wing unmanned aerial vehicles (UAVs) with edge servers offer intermediate processing and relay capabilities to the cloud. We aim to minimize system energy consumption and task latency by jointly considering offloading ratios, CPU frequencies and transmission power of the ground devices. However, efficiently partitioning tasks across these tiers is challenging due to dynamic wireless network conditions, limited onboard resources, and diverse task demands. We propose a temporal dynamics-aware multi-agent offloading and resource allocation framework for such CEG systems in this work. We introduce KAT-MAPPO, a Kolmogorov–Arnold Transformer-enhanced Multi-Agent Proximal Policy Optimization scheme. By embedding Kolmogorov–Arnold Network layers into the transformer encoder, the framework captures long-range temporal dependencies and improves feature representation, while MAPPO enables decentralized offloading under dynamic conditions. Resource allocation is further decoupled and solved analytically via convex optimization. Using a real-world setup, our simulation results show that KAT-MAPPO significantly outperforms existing baselines, achieving faster convergence, a 43% reduction in average task latency in the training evaluation setting, and up to 53.57% latency reduction for the case study scenarios, while also reducing energy consumption. These results highlight the promise of reinforcement learning considering temporal dynamics for efficient offloading in heterogeneous CEG systems.
This paper proposes a hierarchical computation framework that flexibly supports task execution across local vehicles, neighboring vehicles, RSUs, and cloud resources, and designs an efficient task migration and resource scheduling strategy that improves overall system performance under dynamic network conditions.
Simulation results confirm that the proposed JORC framework substantially reduces latency, energy consumption, and overall system cost, while increasing the successful task completion ratio compared to existing baseline approaches.
Tanmay Baidya, Sangman Moh· Italian National Conference...· 0 citations
: Space-Air-Ground Integrated Networks (SAGIN) provide a multi-layered, wide-coverage computing infrastructure for distributed urban sensing systems. However, their heterogeneity and dynamics pose unprecedented challenges for task offloading and resource allocation. Existing methods struggle to simultaneously address the complexity of cross-layer decision-making and reliability assurance under uncertain conditions. This paper proposes a novel framework, termed DRL-RA, which synergistically integrates Deep Reinforcement Learning (DRL) with reliability-aware optimization. The framework consists of two complementary components: (1) a Dueling Double Deep Q-Network (D3QN) module that learns adaptive policies to make offloading decisions among various options including local execution, terrestrial edge, UAVs, and satellites; (2) a Reliability-Aware Multi-Objective Optimization Framework (RA-MOOF) that introduces explicit reliability guarantees through cross-layer link reliability modeling, node availability estimation, and smooth reliability proxy functions. Addressing the heterogeneous communication characteristics of the SAGIN architecture, this paper establishes a complete cross-layer delay model and composite reliability metrics. The reliability formulation is defined under explicitly stated conditional-independence assumptions, and the proposed smooth constraint terms are treated as surrogate CMDP costs rather than exact hard chance-constraint guarantees. Extensive experiments in a SAGIN simulation environment demonstrate that the proposed method improves the task completion rate by 3.8%, reduces average latency by 11.1%, and increases system reliability by 3.9% compared to state-of-the-art benchmarks. The optimization-only RA-Opt baseline is used as a non-real-time optimization reference for assessing reliability-aware offloading decision quality, while deployment-time decision-latency comparisons are interpreted primarily among learned inference policies. Comprehensive ablation studies and statistical validation across multiple random seeds confirm the contributions of each component, while cross-layer offloading decision analysis verifies the effectiveness of the method across different network layer selections.
Fei-Yan Bu, Zheng Wang, Yong Pan et al.· Computers, Materials & C...· 0 citations
6G vehicular services, including cooperative perception, augmented reality navigation, and high-definition map updating, need computation support close to moving vehicles. Vehicular Edge Computing (VEC) is a natural solution, but the offloading decision becomes difficult when wireless channel conditions, vehicle density, and edge server loads vary simultaneously. In this paper, we study joint task offloading and resource allocation in 6G VEC with high- and low-frequency cooperation (HL-FC). We formulate the problem as a decentralized partially observable Markov decision process (Dec-POMDP). Each vehicle decides its offloading ratio, transmission power, server association, and edge CPU request from local observations. To evaluate the proposed policy, we build a lightweight equation-driven Python simulator and compare MAPPO with Local-only, Edge-only, Random, and Greedy policies. Compared with Edge-only, MAPPO reduces the average system cost by 32.15%, 23.51%, and 17.13% under 10, 15, and 20 vehicles, respectively. It also improves the task completion rate by 21.00, 20.49, and 17.65 percentage points. Additional blockage experiments show that HL-FC keeps the policy more robust than high-frequency-only transmission under severe high-frequency blockage. The results reveal that MAPPO delivers better performance when edge resources become congested than in lightly loaded scenarios.
Zi-Heng Gu· 2026 8th International Confe...· 0 citations
A task-driven offloading algorithm based on Balanced Multi-Agent Deep Deterministic Policy Gradient (BMADDPG) that reduces average task processing latency by approximately 22.67% and decreases total system cost by at least 18.32% under high-load scenarios.
The Internet of Things (IoT) has grown rapidly in recent years, enabling the interconnection of a large number of heterogeneous and distributed devices. This number is expected to exceed 70 billion according to Statista. With this massive scale, fulfilling complex IoT applications that require combinations of multiple objects remains a real challenge. Moreover, several Quality of Service (QoS) requirements must be satisfied, making the problem of selecting appropriate IoT services NP-hard. In such environments, task offloading is a key mechanism to efficiently distribute computational workloads across edge, fog, and cloud resources. However, selecting the optimal offloading decision remains a difficult NP-hard problem due to system heterogeneity and conflicting objectives. In this paper, we propose a GNN-DQN-based approach for task offloading in edge–fog–cloud environments. Unlike prior GNN-DQN approaches limited to single- or dual-tier architectures, our framework explicitly models heterogeneous node types and inter-tier communication links, enabling more balanced and scalable resource allocation. Experimental results show that GNN-DQN achieves a mean latency of 2.64 s, representing improvements of 70.2% over Random, 7.8% over DQN-only, and 3.5% over Greedy. A GNN-A2C baseline is also included to broaden the comparison with a modern DRL method. Despite sharing the same GNN encoder, it underperforms GNN-DQN across all metrics, confirming the superiority of the DQN learning backbone. These results highlight the effectiveness of integrating graph-based representation with reinforcement learning, while also revealing a trade-off between latency optimization and energy efficiency.
Sirine Hakim, Sonia Yassa· International Conference on...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
