Dynamic edge orchestration with cross-domain actions for edge offloading in 5G networks

As smart cities’ infrastructure advances, the emergence of Cooperative, Connected, and Automated Mobility (CCAM) introduces a paradigm shift that requires real-time environmental perception and collaborative decision-making. These safety-critical applications demand ultra-low latency and uninterrupted service continuity. Although the convergence of Fifth Generation (5G) and Multi-Access Edge Computing (MEC) provides the necessary foundational infrastructure, vehicular mobility often causes application instances to remain anchored to distant edge nodes, leading to severe Quality of Service (QoS) degradation. This article addresses this orchestration gap by proposing a holistic, resource and mobility aware beyond 5G-Edge architecture. This work aggregates, processes and normalizes real-time telemetry from the Radio Access Network (RAN), edge nodes, and User Equipment (UE) to actively anticipate service degradation. By employing a proactive service replication strategy tailored for workloads, the system dynamically positions application instances along a vehicle’s trajectory and performs inter-gNB handovers, to preserve service continuity. The proposed solution is experimentally validated on a real beyond 5G-Edge testbed using an autonomous vehicle for cooperative perception offloading. Experimental results under realistic mobility and resource-constrained conditions demonstrate that the proposed multi-domain orchestration effectively mitigates mobility- and stress-induced disruptions, preventing Quality of Experience (QoE) degradation by maintaining more than 20 processed frames per second during stress situations, compared with 15 frames per second in the baseline orchestration, and reducing pipeline stalls in 87%, during network and edge transitions. The proposed approach proactively triggers network handovers to ensure that the Received Signal Strength Indicator (RSSI) does not drop below the critical threshold of -85 dBm, while successfully managing edge resources by reducing RAM imbalance by approximately 10 pp and CPU usage by 20%. Overall, the proposed solution preserves the stringent QoS required for safety-critical CCAM operations.