Research on Offloading Strategy Based on Deep Reinforcement Learning in Edge Computing

Abstract:
The traditional edge computing offloading strategy is difficult to achieve the long-term efficiency of the system, and the offload strategy based on deep reinforcement learning (DRL) is easy to converge to the local optimal solution. In order to solve the above problems, an edge computing offloading strategy based on DRL with entropy is designed. The information entropy is added to the value and strategy network update function. So, the agent needs to maximize the entropy While ensuring the maximum reward, which will improve the exploration ability of the agent. The results of the experiment show that the proposed offloading strategy TO-DRL has increased comprehensive efficiency by 12.09% compared with the EOC and increased by 3.93% compared with the TO-DRL without entropy, which effectively balances the delay and energy consumption of the system.