A Service Consolidation Approach for Edge-Vehicle Network using Multi-agent Decision-making Method

Roadside Unit (RSU) plays a significant role in Vehicle-to-Everything (V2X) service to accomplish adequate performance for the autonomous driving system. Haphazard RSU deployment and selection cause to diminish the frame-work efficiency and fail to meet the service capability of most vehicles. Reducing latency by timely delivering service is a phenomenal concern of edge orchestration to enhance the quality of service (QoS) with high reliability. With this motivation, this study focuses on maintaining a trade-off between RSU deployment and selection to achieve a high convergence rate of service offloading policy for effective service consolidation. An RSU-Assisted Service Consolidation (RASC) approach is proposed considering probabilities of three service demand rates, Computation-intensive service rates, and resource prediction rates based on Deep Reinforcement Learning. A resource-weight factor is designed to classify and map the arrived services to the potential RSU based on a discrete concave-convex procedure to avoid the execution hiccups for achieving a high convergence rate of service offloading policy. The average probability rate helps to select the potential RSU to fulfill the requirements of delay-sensitive service requests. The simulation results shows the RASC method improves the service reliability rate by at least 69.5%, the energy preservation rate by 27.1%, and the service offloading reduction rate by 73.08% compared to state-of-art schemes.