AI Assisted Dynamic Bus Lane Control in Connected Urban Environments: The Case of Intermittent Dynamic Bus Lanes.

This paper introduces a novel bus prioritization approach called Intermittent Dynamic Bus Lanes (IDBL) that aims to enhance the reliability and efficiency of bus operations, while minimizing their impact on adjacent traffic. It does so by utilizing a fully decentralized intelligent control mechanism for individual buses that regulates nearby traffic by deciding whether to activate or deactivate a dynamic exclusion zone of fixed distance downstream of the bus, based on the existing traffic conditions and the delay of the bus based on the scheduled arrival times at each bus stop. The intelligent mechanism is trained based on the principles of reinforcement learning in a single-agent non-stationary episodic task, yet evaluated in multi-agent settings with continuous bus flow (agents) under different demand patterns, showing its excellent transferability. Moreover, the proposed scheme is compared to typical Dynamic Bus Lanes (DBL) schemes, Exclusive Bus Lanes (EBL) and Mixed Traffic conditions on the public transport operational stability and reliability, as well as the disturbance to general vehicular traffic. Results indicate that IDBL outperforms the other strategies, rendering IDBL as an efficient transit prioritization strategy and highlighting the potential of using intelligent mechanisms for bus operation control.