计算机应用研究2026，Vol.43Issue(5)：1489-1496,8.DOI:10.19734/j.issn.1001-3695.2025.09.0305
SDN辅助与博弈激励的多跳任务卸载策略
Multi-hop task offloading strategy with SDN assistance and game-theoretic incentives
摘要
Abstract
Multi-hop task offloading expands the collaboration range and improves resource utilization and scheduling flexibi-lity in edge networks,whereas one-hop offloading remains constrained by communication distance and limited available resources,making it inadequate under high-load conditions.However,multi-hop offloading introduces strong coupling between path selection and computation allocation,and idle devices lack incentives to contribute computing or forwarding capabilities.To address these challenges,this paper developed a centralized SDN-based cooperative offloading framework that leveraged global network visibility to jointly optimize communication and computation resources,and incorporated a task-priority mecha-nism to alleviate link contention in multi-task scenarios.In addition,this paper designed a Stackelberg game-based pricing model to incentivize idle devices to participate in computation and forwarding,thereby jointly improving delay and energy per-formance.Building on this framework,this paper proposed the BGLS-PPO algorithm,which integrated a one-step Bellman-based look-ahead estimate into the action selection process to enhance exploration efficiency and convergence stability.Simula-tion results demonstrate that multi-hop offloading significantly outperforms one-hop schemes in overall system utility,and the proposed method exhibits superior adaptability and robustness compared with existing baselines.关键词
边缘计算/设备到设备/软件定义网络/斯坦克伯格博弈/多跳任务卸载/深度强化学习Key words
edge computing/D2D/software defined networking/Stackelberg game/multi-hop task offloading/deep rein-forcement learning分类
信息技术与安全科学引用本文复制引用
黎作鹏,黄时宇,贾泽桐..SDN辅助与博弈激励的多跳任务卸载策略[J].计算机应用研究,2026,43(5):1489-1496,8.基金项目
国家自然科学基金资助项目(62101174) （62101174）
河北省高等学校科学技术研究项目(BJK2022025) （BJK2022025）
河北省光纤生物传感与通信器件重点实验室资助项目(SZX2022010) （SZX2022010）
邯郸市科学技术研究与发展计划项目(24422021193ZC) （24422021193ZC）
