A Multi-Objective Task Migration Approach for Satellite Vehicle Mounted MEC Networks

Aiming at the problem of edge server mobility during computation offloading, a multi-objective task migration method for satellite vehicle MEC networks is proposed, solved using a deep reinforcement learning approach. Firstly, a Satellite Vehicle Mobile Edge Multi-Region Migration System (SaV-MEM) is constructed. The system aims to jointly optimize task processing delay and system energy consumption, forming a multi-objective optimization problem. By fully considering inter-task correlation, the system reduces task processing latency and energy consumption by selecting appropriate edge nodes for task migration. Through this optimization approach, the SaV-MEM system can efficiently allocate computational resources and achieve effective task processing and migration in a dynamic network environment. Secondly, the multi-objective optimization problem is modeled by Markov Decision Process (MDP).The state normalization method is used to solve the quantity difference issue, and the multi-objective migration algorithm based on Deep Deterministic Policy Gradient (DDPG) is proposed. Finally, experiments show that the proposed algorithm can converge stably and effectively reduce the task processing delay and system energy consumption, demonstrating good stability and convergence.