Mobility-aware service migration and two-timescale task offloading for distributed edge computing in open RAN

Abstract

This paper proposes a mobility-aware two-timescale orchestration framework for distributed edge computing in Open Radio Access Networks, where user mobility, handover events, service migration, task offloading, and radio–compute allocation are jointly considered. The framework exploits the functional separation between the Non-Real-Time RAN Intelligent Controller (Non-RT RIC) and Near-Real-Time RIC to coordinate decisions over different temporal scales. At the slower timescale, the Non-RT RIC analyzes historical traffic, regional load, and predicted mobility flows to generate service-placement guidance, resource envelopes, and migration-aware control parameters. At the faster timescale, a Near-RT RIC xApp performs handover-coupled task offloading, service migration, bandwidth allocation, and edge-computing assignment using instantaneous channel, queue, mobility, and resource states. A graph-contextual representation is further employed to capture neighboring-region congestion, available capacity, and mobility transitions, enabling scalable inter-region coordination. The resulting stochastic optimization formulation incorporates communication, computation, and migration delays together with resource, queue-stability, and service-continuity constraints. A hybrid actor–critic learning mechanism with centralized training and decentralized execution is adopted to obtain feasible mobility-aware orchestration policies for dynamic multi-service Open RAN edge environments.