基于A2C-DDQN网络的异构任务分布式卸载优化算法OA

In mobile edge computing(MEC),heterogeneous tasks need to be processed as a whole due to their indivisibility,while the existing distributed offloading algorithms have problems such as insufficient adapta-tion to heterogeneous features and difficulty in collaborative optimization of delay and energy consumption.In re-sponse to the aforementioned issue,this paper proposes a heterogeneous task distributed offloading optimization(HDOO)algorithm.The HDOO algorithm integrates advantage actor-critic(A2C)and double deep Q-network(DDQN)technologies to construct an A2C-DDQN network so as to improve training efficiency and policy stabili-ty,and then introduces a heterogeneous task feature perception module to adapt to the requirements of tasks with different computational complexities.Simulation results show that,compared with typical deep reinforcement learning offloading algorithms,the HDOO algorithm has significant optimizations in terms of such indicators as average delay,energy consumption and task drop rate,effectively improving the overall performance of the MEC system.