MaaS Personalised Multi-Modal Multi-Objective Journey Planning with Machine Learning Guided Shortest Path Algorithms

This work presents a methodology for rapidly generating many multi-modal journey alternatives including cheap, fast, green, convenient and low effort alternatives to private car journeys. The proposed methodology firstly generates a Pareto set of journey profiles based on static inter-transfer zone objective criteria contribution estimates. Secondly, integrated and extended versions of existing shortest path algorithms for open and public transport networks are used to optimise paths and transfer points in a procedure guided by neural networks while using real-time transport network information. A novel hybrid k-means and Dijkstra’s algorithm is introduced for generating transfer zone samples, in a way that accounts for transport network connectivity to ensure transfer feasibility. The end result is a modularised algorithm that knits together and extends the most effective existing shortest path algorithms for open and public transport networks using neural networks as a look ahead mechanism for generating many efficient multi-modal journey alternatives. In experiments based on a large scale transport network, query response times are shown to be suitable for real-time applications, and are found to be independent of transfer zone sample size, despite larger transfer zone samples leading to higher quality and more diverse Pareto sets of journeys, a win-win scenario.