Learning Insertion Patterns to Enhance Operational Efficiency in Large-Scale Dial-a-Ride Systems
Abstract
We study a large-scale dial-a-ride system considering around 300, 000 dynamic requests. An effi cient routing algorithm is crucial to guaranteeing the viability of the system. Large-scale requests are as- sumed to be dominated by daily commuting needs and thus should ex- hibit similar mobility patterns from one day to another. Consequently, daily vehicle trajectories should also be recurring if similar requests can be served in the same manner. We introduce a greedy insertion algo- rithm integrating a Guided Insertion Mechanism that learns insertion patterns from reference resolution to enhance the operational effi ciency of the underlined systems while maintaining high-quality solutions.