Soft Actor-Critic-Based Service Migration in Multiuser MEC Systems

Multi-access Edge Computing (MEC) is an emerging computing paradigm that provides abundant resource support for the next generation of the Internet of Things (IoT). When users move between edge servers with limited coverage, it is necessary to determine service migration strategies to ensure service quality. However, traditional research on service migration often oversimplifies the multi-user scenario, focusing only on the service migration of individual users. This limitation hinders these methods from achieving optimality in real MEC environments. To address this issue, this paper models the multi-user service migration scenario and focuses on exploring the impact of service migration on user latency and system energy consumption. We propose a service migration method based on the discrete version Soft Actor-Critic algorithm (SACDM). Through simulation experiments to evaluate the performance of our proposed solution, the results demonstrate our method outperforms service migration based on Deep Q-learning (DQNM) by reducing latency by approximately 11.12%. Additionally, our method also achieves a reduction in energy consumption by approximately 6.66% compared to DQNM.