Distributed reinforcement learning-based optimization of resource scheduling for telematics

Abstract

As a reliable computational model in telematics, Mobile Edge Computing (MEC) is utilized to lighten the load on individual autonomous vehicles. It enables computationally expensive and latency-sensitive jobs to be offloaded from cars with limited processing power to servers at edge nodes. This paper presents the joint optimization problem as a mixed nonlinear integer programming problem. The scheduling of vehicle tasks and the distribution of communication resources are then rationalized using a resource allocation algorithm based on Deep Deterministic Policy Gradient (DDPG) reinforcement learning to arrive at a suboptimal solution based on the resource allocation policies of single-intelligent DDPG and multi-intelligent DDPG. The numerical simulation demonstrates the superior service experience of the proposed joint optimization strategy of distributed task offloading and resource allocation based on multi-intelligent DDPG over other association and channel selection strategies and joint resource allocation based on single-intelligent DDPG.