Computing Offloading Decision Based on DDPG Algorithm in Mobile Edge Computing

Abstract

In view of the high cost of deep reinforcement learning algorithms when deployed on edge devices, user waiting delay and high energy consumption problems. In order to improve the quality of user service, this paper proposes the FPTT-DDPG algorithm based on the DDPG algorithm. First, filter pruning and tensor decomposition are used to trim, compress and retrain the deep neural network, and apply the trained neural network model to the DDPG algorithm to solve the optimal offloading decision that minimizes user energy consumption and total delay cost. Maximize the long-term utility of the offloading decision model. The simulation results show that the FPTT-DDPG algorithm proposed in this paper can effectively reduce the deployment cost of the depth enhancement algorithm in the edge device and the total cost of user delay and energy consumption. The memory size of the neural network of the FPTT-DDPG algorithm is reduced by 81%, and the total system delay cost is reduced by about 26.3% compared with the DDPG algorithm.