Revolutionizing Transportation Using Deep Reinforcement Learning: A Comprehensive Review

摘要

Deep Reinforcement Learning (DRL) is revolutionizing the way many challenging transportation decision-making problems are approached. This potent learning-based technology is used more frequently to tackle difficult challenges in the domain of transportation. Any control-based system, including those in electricity, robotics, transportation, and the Internet of Things, can benefit from new data-driven research methodologies. This paper provides a summary of the mathematical foundation of DRL, as well as several very successful DRL extensions and well-liked and prospective DRL algorithms. On the basis of above analysis the paper further explore the suitability, advantages, disadvantages, and general and particular application-related problems of DRL approaches for their use in transportation. Finally, this paper outlines the future research paths and resources that might be used to put DRL into practice.