ScholarWorks@숙명여자대학교

초록

Traffic congestion is one of the common urban problems caused by increased traffic. Traffic congestion accelerates environmental pollution by wasting drivers' time and fuel and generating more fumes. Therefore, traffic congestion is an important issue to be solved. Currently, as technologies develop, a smart city that efficiently manages data information collected is in the spotlight. The smart transportation system utilizes the infrastructure and network built in the smart city to analyze traffic flow and control traffic in real-time. Accordingly, traffic congestion can be effectively alleviated. This paper proposes a smart traffic signal control system using a Deep Q-network (DQN), a type of reinforcement learning. The proposed algorithm distributes the optimal green signal time by collecting and learning information about the intersection situation. The proposed algorithm is designed to improve the performance in terms of throughput. As a result, the number of waiting vehicles also decreased. To validate the algorithm, we evaluate the performance in various traffic scenarios. © 2021 IEEE.

키워드