Flexible bus service design under spatial and temporal demand uncertainty

Abstract

This thesis studies a zonal-based flexible bus service (ZBFBS), which is a class of demand responsive transit (DRT) service considering passenger demands’ spatial (origin-destination or OD) and volume stochastic variations. It is then extended to incorporate price and service elasticity and passenger travel time. Finally, time-dependent demand and time-dependent travel time are taken into account while a precise schedule for the ZBFBS is constructed. Dividing the service area into zones, service requests are aggregated by zonal OD pairs, time window constraints, and the number of passengers per request. Three two-phase stochastic problems with recourse are formulated, including a frequency-based model minimizing the cost, a frequency-based model maximizing the profit minus detour cost, and a schedule-based model maximizing the profit. In phase-1, given the reliability requirements, the operating profit of flexible bus services is maximized for frequency-based and schedule-based models by vehicle-to-route assignment before demand realization. In phase-2, passengers are assigned to vehicles to minimize the expected ad hoc service cost. Demand volume reliability and detour time reliability are introduced to ensure the quality of the services and separate the problem into two phases for efficient solutions. The reliabilities are then optimized by a gradient-based approach to optimize for the objective. Relaxations, heuristics, and alternative solution algorithms are proposed to enhance the performance of planning ZBFBS. Finally, a bi-level framework is proposed to jointly optimize the zonal design and flexible bus service strategy. Numerical examples are provided to illustrate the relationship between vehicle capacity, profit, demand volume, and zonal design, and that the optimal zone dimension is positively correlated with demand volume. Also, the improvements by using heuristics, relaxations, and schedule-based planning are elaborated. Moreover, the zonal design and ZBFBS strategy are validated in instances constructed based on the real data in Chengdu, China.

Date of Award	2021
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Hong Kam LO (Supervisor)