Reinforcement Learning for Efficient Bus Priority Systems: Balancing Transit Efficiency and Traffic Impact

Abstract

Abstract In this research, two innovative and adaptive bus prioritization strategies are presented; Adaptive Bus Lanes with Intermittent Priority (A-BLIP) and Bus Lane Density Control (BLDC). Both strategies harness the potential of connected vehicles and Vehicle-to-Everything (V2X) communication and aim to enhance the reliability and efficiency of bus operations. Trained through reinforcement learning, they consider buses as intelligent local regulators of traffic, making informed decisions to strike a balance between optimizing bus operations and ensuring the smooth flow of vehicular traffic, based on the specific priorities set by the urban transportation authorities. An intelligent A-BLIP bus agent is trained to dynamically adjust the length of “exclusion zones” based on real-time information on bus delays and traffic conditions, while a BLDC agent controls the density of vehicles within bus lane sections. The strategies are evaluated in a multi-agent setting under various traffic and public transport demand scenarios and compared against the dedicated bus lanes strategy and mixed traffic conditions. The results demonstrate that both A-BLIP and BLDC closely approximate mixed traffic conditions' (i.e., when no restrictions are applied) impact on vehicular traffic while significantly improving bus operations. The strategies' adaptability and performance in extreme scenarios make them promising solutions for enhancing urban transportation management and promoting more sustainable mobility choices.