Rainbow Deep Reinforcement Learning Agent for Improved Solution of the Traffic Congestion

Abstract

While traffic congestion hits severely the world economy, adaptive traffic signal systems would efficiently provide potential solutions. In this paper, we propose a deep reinforcement learning system to control the signal lights in an isolated intersection. The proposed system uses a deep convolutional neural network to extract the crucial features from the environment state that is described by raw traffic information; i.e., vehicles positions, speeds, and waiting times. Besides, the system utilizes a multi-objective reward and the Rainbow agent which provides further space of enhancements to the conventional Deep Q-Networks agent. Extensive experiments illustrate that our proposed deep framework outperforms the baseline under a number of settings and traffic measures, including trip time, waiting time, fuel consumption, and stability.