Multi-agent reinforcement learning for cost-aware collaborative task execution in energy-harvesting D2D networks

In device-to-device (D2D) networks, multiple resource-limited mobile devices cooperate with one another to execute computation tasks. As the battery capacity of is limited, tasks running on will terminate once dead. order achieve sustainable computation, energy-harvesting technology has been introduced into D2D networks. At present, how make energy harvesting work collaboratively minimize long-term system cost for task execution under limited computing, network and constraint a challenging issue. To deal such challenge, in this paper, we design multi-agent deep deterministic policy gradient (MADDPG) based cost-aware collaborative task-execution (CACTE) scheme (EH-D2D) validate CACTE scheme's performance, conducted extensive experiments compare four baseline algorithms, including Local, Random, ECLB (Energy Capacity Load Balance) CCLB (Computing Balance). Experiments were accompanied by various parameters, as device's capacity, workload, bandwidth so on. The experimental results show that can effectively many more higher reward, lower latency fewer dropped