Enhancing feeder bus service coverage with Multi-Agent Reinforcement Learning: A case study in Hong Kong

Abstract

Public transport is a vital component of modern urban mobility, playing a significant role in reducing congestion and promoting environmental sustainability. Feeder bus services are essential for connecting residents to major public transport hubs, such as metro or rail stations. In this paper, a novel framework that enhances service coverage of the feeder bus while maintaining network efficiency is proposed. The framework integrates Multi-Agent Reinforcement Learning (MARL) to simulate and optimize route designs and frequency settings. Additionally, we introduce a Cost-based Competitive Coverage (CCC) Model to evaluate the performance of the feeder bus services by considering competition with other public transport modes. A case study conducted in two new towns in Hong Kong demonstrates the effectiveness and robustness of the proposed framework, highlighting its adaptability and potential to improve public transport accessibility.