MEC下基于深度强化学习的分布式多目标任务卸载算法OA

To address the problem that offloading tasks from edge terminals in the mobile edge computing(MEC)scenario can reduce both costs and consumption,but edge nodes are prone to overload because of massive tasks,leading to delay threshold or discarding of partial tasks,this paper proposes a distributed multi-objective task offloading(DMTO)algorithm based on deep reinforcement learning.The proposed algorithm integrates the advantages of the dueling deep Q-network(dueling DQN)and the double deep Q-network(double DQN)to opti-mize the network training efficiency and the decision-making stability.Simulation results show that compared with the mainstream distributed deep reinforcement learning algorithms such as IDDPG,DDTO,DDLO and EDRL,the DMTO algorithm performs better in terms of average delay,energy consumption control and success rate of task offloading.