Deep Reinforcement Learning Based Task Offloading Strategy Under Dynamic Pricing in Edge Computing

Bing Shi, Feiyang Chen, Xing Tang

Abstract: Mobile edge computing has become a new paradigm for efficient computing, which allows users to offload computing tasks to edge servers to accomplish the tasks. However, in the real world, users usually keep moving, and the edge servers may dynamically change the offered service prices in order to maximize their own profits. At this moment, we need a highly efficient task offloading strategy for users. In this paper, we design a task offloading strategy when users are on the movement and edge servers dynamically change the service prices based on the deep reinforcement learning algorithm, which is named as DUTO. Furthermore, we run extensive experiments to evaluate our offloading strategy against four benchmark offloading strategies. The experimental results show that DUTO task offloading strategy can effectively improve the long-term profits of users in the dynamic environment with different experimental settings.