ELEMENT: Energy and Latency Management with Reinforcement Learning-Based Offloading in Vehicular Edge Computing

With the recent advancements in smart vehicles and ever-increasing tasks that need to be processed, Vehicular Edge Computing (VEC) emerged as a promising solution. Smart vehicles have numerous tasks and their local processing power is insufficient to process all of them. Offloading the tasks to edge servers and utilizing the processing power of those servers is a decent solution to solving this problem. Still, the decision-making for offloading is challenging due to the delay and energy con-sumption caused by transmitting data which significantly impacts two crucial parameters of a VEC environment: latency and the energy consumption of the smart vehicle. Energy and latency are closely related, and a low-latency system consumes more power, while a low-energy system has higher latencies. The balancing challenge of these two has become a popular research area with many contributions. In this paper, we propose a Reinforcement Learning (RL) approach for taking the offloading decision to find a balance between latency and the consumed energy. Utilizing RL, allowed our method to be flexible and adaptive to diverse scenarios, especially in large-scale and dynamic environments. By conducting realistic simulations using SUMO standard we gained up to 18.35% improvement in average response time while reducing energy consumption up to 5 % compared to other baselines.