SafeLight: A Reinforcement Learning Method toward Collision-Free Traffic Signal Control

Traffic signal control is safety-critical for our daily life. Roughly one-quarter of road accidents in the U.S. happen at intersections due to problematic signal timing, urging the development of safety-oriented intersection control. However, existing studies on adaptive traffic signal control using reinforcement learning technologies have focused mainly on minimizing traffic delay while neglecting potential exposure to unsafe conditions. We, for the first time, incorporate road safety standards as enforcement to ensure the safety of existing reinforcement learning methods, aiming toward operating intersections with zero collisions. We propose a safety-enhanced residual reinforcement learning method (SafeLight) and employ multiple optimization techniques, such as a multi-objective loss function and reward shaping, for better knowledge integration. Extensive experiments on both synthetic and real-world benchmark datasets show that our method can significantly reduce collisions while increasing traffic mobility.

Code: gitlab.com/wenlu057/traffic-safety

Recommended citation: Du, W., Ye, J., Gu, J., Li, J., Wei, H., & Wang, G. (2023). "SafeLight: A Reinforcement Learning Method toward Collision-Free Traffic Signal Control." Proceedings of the AAAI Conference on Artificial Intelligence, 37(12), 14801–14810.
Download Paper