Joint Task Offloading and Service Caching for Vehicular Edge Computing with Energy Constraints

Vehicular edge computing (VEC) enables vehicles to offload their tasks to idle vehicles for processing. To process tasks, the needed service model should be stored ahead. Considering the storage of vehicles is limited and downloading the service model upon tasks causes repeated data transmission, it is necessary to determine which service models should be cached and which tasks should be offloaded. For vehicles, the task completion time is crucial due to their mobility nature. Therefore, balancing the completion time and energy consumption for vehicles with limited battery capacity in highly dynamic system remains a significant challenge. To address this, we jointly designed task offloading and service caching scheme to minimize the tasks completion time in a cache-assisted VEC scenario while satisfying the long-term average energy constraints. We adopt Lyapunov optimization and Deep reinforcement learning (D RL) based Task offloading and Service caching (LDTS) to solve the proposed problem. Specifically, the Lyapunov optimization is used to tackle the issues of long-term energy consumption constraints by transforming the original problem into per-slot optimization problems that can be resolved using a model-free DRL method. Simulation results demonstrate that the LDTS method can minimize the completion time while satisfying the constraints of the long-term energy consumption budget. The results indicate that the LDTS method outperforms other benchmark methods.