Task Offloading Strategy Based on TD3 Algorithm in Cloud-Edge Collaborative MEC
Haiwen Gong, Haibo Ge, Shixiong Ma, Ao Sun, Xutao Chen, Linhuan Liu · 2022 4th International Conference on Natural Language Processing (ICNLP) · 2022
Aiming at the high cost of the existing TD3 algorithm for processing mobile edge computing (MEC) task offloading systems, an offloading strategy based on the improved TD3 algorithm in mobile edge computing is proposed. This paper constructs a cloud-edge collaborative MEC system and studies the computational offloading problem to minimize the total system cost (the weighted sum of delay and energy consumption). Through Markov Decision Process (MDP) the problem is transformed into an offloading scheme based on deep reinforcement learning (DRL), and solved by the improved TD3 algorithm HERTD3, which uses Hindsight Experience Replay (HER) to improve the experience replay unit of the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. It solves the sparse reward problem in reinforcement learning and improves the convergence speed of the algorithm. The simulation results show that the proposed algorithm has good convergence and significantly reduces the total cost of the system.