Exploring choice set generation approaches for public transport connection choice

Abstract

This technical working paper documents the choice set generation algorithms for public transport connection choice in their state of development in August 2014. The choice set generation methods have been developed to generate the non-chosen alternatives for public transport trips that have been observed through GPS traces recorded either by dedicated GPS devices or by smart-phones. The choice set generation approaches that we have developed build on the idea of generating a routing network that contains the entire schedule and modifying that Dijkstra algorithm so that takes into account multiple start and end stops in the public transport network and the corresponding access and egress segments. The basic version is then combined with other least cost path based choice set generation procedures: First, with a procedure to randomly vary the link cost and cost function parameters to obtain a doubly stochastic choice set generation and, second, with a new approach of selecting potential transfer opportunities in the network enforcing a routing via these selected points. A first evaluation of the generated choice sets was done with public transport trips found in GPS traces originating from a person-based GPS diary survey with participants living in and around Zurich. It shows that all three approaches generally work. However, it also reveals several areas for improvement that are discussed within in this working paper.