Applications on Deep Reinforcement Learning in Traffic Signal Control

Under the transportation resources and infrastructures that exist today, artificial intelligence is used as an efficient and intelligent solution to traffic congestion in strategic decision-making and trajectory planning control. This paper will introduce one of the methods - Deep Reinforcement Learning (DRL), and the application of DRL in intelligent traffic signal control. This paper would introduce the basic knowledge of DRL, explain a traffic signal control model application based on Markov Decision Process and present the relating knowledge of MDP, introduce the signal decision-making model based on Q reinforcement learning and compare the pros and cons of both models. This paper compares different solutions of traffic signal controls of varying learning modes with the standard of efficiency, implementation difficulty, advantages, and disadvantages.