Load-balanced multi-user mobility-aware task offloading in multi-access edge computing

In scenarios with dense user network service requests, multi-access edge computing demonstrates significant advantages in reducing user device load and decreasing service response time. However, the dynamic changes in user trajectories cause edge server load fluctuations, inevitably impacting the overall service processing performance. To tackle this problem, this paper introduces a load-balanced multi-user mobility-aware service request offloading method, achieving efficient service request offloading in mobile user scenarios. Specifically, this paper divides the service request offloading problem into two stages: dynamic edge server allocation and real-time offloading decision generation. In the first stage, users are allocated edge servers based on their location distribution, implementing an adaptive decreasing variance optimization server load balancing algorithm to achieve edge server load balancing. In the second stage, based on the edge server allocation results from the first stage, a latency performance self-optimizing task offloading decision-making algorithm is employed to minimize the processing latency of user requests, utilizing dueling double deep Q-network to generate real-time decisions on whether to offload service requests to the corresponding edge servers. According to experimental results, the proposed algorithm markedly decreases the processing latency of user network service requests in scenarios of different scales, with an average task completion rate of 99.94%. This effectively addresses the problem of inefficient processing requests caused by load fluctuations due to user movement in multi-access edge computing.