Edge Computing Task Unloading Decision Optimization Algorithm Based on Deep Reinforcement Learning

Mobile Edge Computing (MEC) is used to enhance the data processing capability of low-power networks and has now become an efficient computing paradigm. In this paper, an edge-cloud collaborative system composed of multiple terminals (Mobile Terminal, MT) and its resource allocation strategy are considered. In order to reduce the total delay of MTs, a variety of offloading modes are adopted, and an edge-cloud collaborative task offloading algorithm based on deep reinforcement learning (DDPG) is proposed. The coordinated serial task dynamic allocation processing provides approximately optimal task allocation and offloading strategies for different user equipment applications. The simulation results show that compared with the Deep Q Neural Network (DQN) algorithm and the Deep Deterministic Policy Gradient (DDPG) algorithm, the proposed algorithm significantly improves the maximum performance gain. Experiments show that the DDPG algorithm has better system performance than the traditional scheme, and improves the service quality of the application.