Algorithm Optimization of Deep Reinforcement Learning for Traffic Signal Control of Municipal Road Engineering

Abstract

Urban road traffic signal control is the key factor affecting the road capacity, but due to the strong randomness of traffic flow, the traditional inductive control method only considers the current phase of vehicle arrival information, which has great limitations. In the stage of intelligent traffic signal control and regulation of traffic engineering, we should do a good job in traffic flow prediction, intelligent traffic signal regulation and control, as well as hardware and software control, so as to improve the quality of intelligent traffic signal control and regulation of traffic engineering. Taking the municipal road at a single intersection as an example, this paper applies deep learning to predict the municipal traffic flow, and then optimizes the deep reinforcement learning algorithm for its traffic signal control and carries out information modeling. The conclusion is that the traffic signal control effect based on algorithm optimization is significantly better than the traditional timing control method, which can effectively reduce the probability of traffic accidents and realize the intellectualization of municipal traffic engineering.