Edge-Cloud based Framework to Offload Task in Vehicular Edge Computing Using Optimized Deep Reinforcement Learning Approaches

Vehicular edge computing (VEC) is rapidly gaining traction to satisfy computationally demanding vehicular applications such as autonomous driving, monitoring traffic flow, remote vehicle diagnostics, and vehicular online gaming. Nevertheless, many existing offloading schemes fail to consider the edge computing environment’s fluctuating nature, leading to increased delays and high energy consumption. This paper introduces an optimized edge-cloud based task offloading scheme utilizing deep reinforcement learning for VEC networks. Depending on the specific computation and memory needs, this system enables tasks requiring a lot of computing to be managed locally or delegated to a cloud or edge server. By simultaneously considering i) the dynamic nature of the vehicular edge computing network and ii) the necessity for swift offloading decisions, our approach leverages deep reinforcement learning to reduce energy usage and the average processing delay of tasks through seamless integration of the cloud, edge servers, and vehicular computing resources.