Vehicular edge computing for 6G must meet millisecond scale latency and stringent energy budgets over dynamic, multi tier networks. We introduce an ML driven orchestrator that uses a Naive Bayes classifier for edge and cloud tier selection with a regression model for service time prediction on multidimensional features (task attributes, network metrics, energy profiles, CPU load), and embeds a multipath feasibility module, augmented by transmission power control and dynamic CPU frequency scaling, to jointly optimize latency, reliability, and energy consumption. In comprehensive EdgeCloudSim SUMO experiments, our framework achieves a latency of up to 35% lower end- to-end, 30% fewer task failures, and keeps energy use within 10% of optimal compared to a randomized baseline. These results demonstrate millisecond scale decision capability and robust performance under realistic VEC conditions
Energy-Efficient Workload orchestration for 6G Vehicular Edge Computing
Ali N.;Siyal F.;Aloi G.;Gravina R.;
2025-01-01
Abstract
Vehicular edge computing for 6G must meet millisecond scale latency and stringent energy budgets over dynamic, multi tier networks. We introduce an ML driven orchestrator that uses a Naive Bayes classifier for edge and cloud tier selection with a regression model for service time prediction on multidimensional features (task attributes, network metrics, energy profiles, CPU load), and embeds a multipath feasibility module, augmented by transmission power control and dynamic CPU frequency scaling, to jointly optimize latency, reliability, and energy consumption. In comprehensive EdgeCloudSim SUMO experiments, our framework achieves a latency of up to 35% lower end- to-end, 30% fewer task failures, and keeps energy use within 10% of optimal compared to a randomized baseline. These results demonstrate millisecond scale decision capability and robust performance under realistic VEC conditionsI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
