PROJECT TITLE :

Deep Reinforcement Learning for Intelligent Transportation Systems A Survey

ABSTRACT:

The most recent technological advancements have led to an improvement in the standard of transportation. New data-driven approaches give rise to a new research direction for all control-based systems, such as those used in transportation, robotics, the internet of things, and power systems, among others. The integration of data-driven applications and transportation systems is playing an increasingly important role in modern transportation applications. In this paper, a survey of the most recent applications of deep reinforcement learning (RL), which are focused on traffic control, is presented. Specifically, applications for traffic signal control (TSC) based on (deep) RL, which have been studied extensively throughout the body of published research, are dissected and analyzed in minute detail. A comprehensive discussion is held on a variety of problem formulations, RL parameters, and simulation environments pertaining to TSC. There are also several applications of autonomous driving that have been studied using deep RL models in the published research. Our investigation provides a comprehensive summary of previous research in this area by classifying it according to application types, control models, and studied algorithmic approaches. In conclusion, we talk about the difficulties and open questions that arise when applying deep reinforcement learning to transportation applications.