Efficient Algorithms for Intermodal Routing and Monitoring in Travel Information Systems

Abstract

Millions of people use online journey planning systems. However, most of the currently available systems only support finding optimal journeys for one mode of transportation (i.e. only public transportation, driving by car, etc.). This makes it hard to plan intermodal journeys combining different modes of transportation to reach a destination. In this thesis we present a real-time multi-criteria intermodal travel information system supporting various transportation modes as well as different special use cases such as intermodal routing for people with disabilities and tourist tour planning. Choosing the perfect parking to switch from private transportation (e.g. bicycle or car) to public transport (e.g. buses, trams, trains, etc.) is basically trial and error when using unimodal planning systems. This problem becomes even harder when the user wants to return to the starting point which is a common use case of intermodal travel. The optimal route for the outward trip may yield a suboptimal or even infeasible return trip and vice versa (due to the choice of the parking place). In this work, we present a novel and integrated multi-criteria approach to computing optimal journeys for both trips (outward and return trip) combined. Previous routing approaches only provide very limited functionality for people with disabilities. Many elderly people as well as persons with heavy luggage or a baby buggy would like to avoid obstacles like stairs -- however not at all costs (depending on the detour length). Our new multi-criteria approach computes all optimal trade-offs between the difficulty of the route and other optimization criteria (like travel time and the number of transfers). Additionally, we support restrictions that forbid the usage of certain obstacle types completely (like a person in a wheelchair cannot use stairs at all). The restrictions as well as the difficulty of each obstacle need to be adaptable to the profile of the person using the routing service. Our approach is customizable and computes optimal intermodal journeys in a fully integrated manner. Another use case of intermodal mobility is the planning of a tourist trip in a foreign city. Here, several constraints such as opening hours of attractions need to be considered. If planning a tour for multiple days, we want to avoid redundancy. We present a novel combination of the Time Dependent Team Orienteering Problem with Time Windows (TDTOPTW) with the Orienteering Problem with Variable Profits (OPVP). Additionally, our modeling is the first to support several entries and exits per point of interest (PoI) which is relevant in practice because for large area sites like zoos or boardwalks the public transport stop at each entry/exit may be serviced by different lines. In case of delays, cancellations, reroutings, or track changes, a journey may become infeasible. In this case, information is key to finding a solution to this problem. Informing travelers as soon as possible gives them the most options. We present an efficient approach to monitor millions of journeys in parallel. The selection of change notices to be communicated to a traveler may be flexibly adapted to the travelers individual needs. Additionally, the system is capable of providing intermodal real-time alternatives in case of a broken connection. To make the functionality described above accessible to the end-user, we have built mobile (Android) as well as web-based user interfaces. We describe the distributed modular software architecture which can resemble micro-services as well as a monolithic setup enables us to provide the approaches in a scalable and efficient way.