Multimodal transit scheduler: An actor-based concurrent approach

Interactive journey planning applications are a necessity to take advantage of redundancy in multimodal public transit networks. As the scale of the transit network grows in terms of the number of stations, number of connections / services, and the number of transit operators, manual itinerary search may fail to detect best possible itineraries. Another challenging factor is the need to include the constraints imposed by intermediate stations on the creation of itineraries. One such constraint is the transfer time needed to alight one vehicle, and board another vehicle at a station. Other constraints such as non-availability of transport vehicles, and congestion have local origins, but affect the set of feasible itineraries. We introduce a generic model for a dynamic network. In our proposed dynamic network model, we incorporate both the local constraints (of a station), and the global constraints (user preferences). In this work, we use an Actor-based concurrent approach to design a transit scheduler. We model each transit station as an Actor, thereby gaining the advantage of accounting for dynamism in transit networks. Our system performs concurrent itinerary search over transit schedules. Our approach allows real-time updates to timetables, and status of transit stations. Our approach also solves the centralized schedule management problem by incorporating local update capability into the transit scheduler. Thus we enable local network operators to manage the vehicular traffic at their respective stations. We are also able to filter the itinerary search results based on user preferences. We have implemented a multimodal transit scheduler that has a median query completion time of ~0.9 seconds on the public transit networks of India.