M.Sc Thesis
M.Sc Student Lazarev Michael Development of a Timetable Based Transit Assignment Algorithm Department of Civil and Environmental Engineering Professor Shlomo Bekhor

Abstract

This thesis investigates existing transit assignment models and algorithms and develops a modification of an existing algorithm aiming to solve the problem of passenger transit assignment. The model allows examining the effect of various factors on the choice of the optimal route by public transport passengers. According to the literature, transit assignment algorithms can be classified in two main categories: frequency-based and timetable-based algorithms. This study focuses on timetable-based assignment algorithms. Tong and Richardson (1984) developed a timetable-based algorithm to find the optimal route.  This thesis offers a modification to this algorithm, by "splitting" a transfer node in the network into several nodes in order to find the optimal path.  The split network allows the activation of Dijkstra's algorithm (1959), thus maintaining the optimality principle. As with previous studies, this thesis assumes that the passenger weighs several travel components: walk to and from the station, waiting and riding transit lines, and sometimes transferring between lines. The main factors affecting the selection of the route depend on the components' weighting. The model assumes that the passengers have complete information on the public transport system, namely the itineraries, travel times and line schedules.  Transit passengers have different travel time perceptions. The approach of this study was to group the passengers into representative classes and assign different "penalties" for waiting and walking.  Special penalties were assigned for transfer - partly a sum of money and partly convenience, and for passing a specific station.  The optimal route in the algorithm is calculated according to the minimal weighted time, while the passenger uses an exact and correct schedule, and the weighted time is calculated according to the penalties assigned to different arcs for each group.  The algorithm gathers the optimal routes from each origin-destination pair in order to assign the passenger matrix on the public transport network.  A test case was performed for the network of Petah-Tikva.  The impact of various factors on the system results was examined for such results as average times of traveling, walking and waiting, number of ascends and number of transfers. The results show that the average times of the different passenger groups vary according to the value of the penalties on a specific component of a specific group.  It is possible to extend the proposed algorithm in several directions, such as building a stochastic assignment model, dynamic assignment or examination of additional factors on the selection of the optimal route.