טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentHadas Yuval
SubjectDeveloping a Model for Public Transportation with Flexible
Routes Based on Distributed Computing
DepartmentDepartment of Civil and Environmental Engineering
Supervisor Professor Emeritus Avishai Ceder
Full Thesis textFull thesis text - English Version


Abstract

The use of transfers in public transit has the advantages of reducing operation costs and introducing more flexible and efficient route planning. In contrast, the main drawback from the passengers’ point of view is the inconvenience of traveling multi-legged trips. In order to diminish the waiting time caused by passenger transfers, synchronized timetables were introduced. The use of synchronized (timed) transfers suffers from uncertainty about the simultaneous arrival of two (or more) vehicles at an existing stop which can lead to deterioration in system reliability. In order to alleviate the uncertainty of simultaneous arrivals, a new passenger-transfer concept was developed that extends the commonly used single-point encounter (at single transit stop) to a road-segment encounter (any point along the road segment constitutes a possible encounter point) as well as an optimization model based on dynamic programming for aiming at reducing the total travel time of the public transit system. The objectives of this work are: (a) to define vehicle-encounter probability along a road-segment, (b) to introduce a simulation model to estimate the vehicle-encounter probability, (c) to model the vehicle-encounter probability upper bound and, (d) to construct of an optimization model based on distributed dynamic programming. Emphasis and examples refer to a bus transit system. It is believed that the new proposed concept will reduce the uncertainty of meeting at a point and will enable more flexibility in deploying on-line operational tactics (hold, skip stop, short-turn, etc). A simulation model validated the benefits of the dynamic programming model results with an average of 10% reduction of total average travel time and more than 200% increase of direct transfers (transfers in which both vehicles arrive simultaneously to the transfer point). The simulation was executed on different public transit scenarios characterized by the headway, level of synchronization, forecast range and travel time parameters.