Safe Reinforcement Learning-Based Eco-Driving Control for Mixed Traffic Flows With Disturbances

Feb 1, 2025·
Ke Lu
,
Dongjun Li
,
Qun Wang
,
Kaidi Yang
,
Zhao Lin
,
Ziyou Song
· 0 min read
Abstract
We develop a safe reinforcement learning (Safe-RL) framework for eco-driving of connected and automated vehicles in mixed traffic flows subject to external disturbances. Safety constraints are encoded via a control barrier function, while the learned policy optimizes energy consumption and string-stability. Simulation studies show that the framework maintains safety guarantees under stochastic lead-vehicle behavior while substantially reducing energy use.
Type
Publication
IEEE Transactions on Intelligent Transportation Systems
publications
Authors
Postdoctoral Fellow
Qun Wang is a Postdoctoral Fellow at the AIMS Lab, Hong Kong Polytechnic University, working with Assistant Professor Hailong Huang. His research focuses on energy-saving optimization and safety-guaranteed control of autonomous electrified vehicles, with emphasis on reinforcement learning for eco-driving, human driving behavior modeling, and multi-agent car-following control. He has published over 10 papers in leading journals including IEEE Transactions on ITS, TTE, TCST and Energy.