Open Collections

Abstract

Public transit systems rely on efficient scheduling to keep operating costs in check. To this end, the Vehicle Scheduling Problem (VSP) assigns trips to a fleet of vehicles at minimal operational costs. However, minimizing operating costs can result in schedules that have little buffer time between trips; these schedules are then vulnerable to disruption, as delay can propagate through the network. This thesis investigates a variant of the VSP that minimizes both operational cost and delay cost. We formulate a network flow-based stochastic optimization model that jointly sets planned trip travel times and vehicle schedules to reduce both primary and secondary delay. We evaluate our model with computational experiments on real-world transit data from Victoria, BC. We compare against practical schedules that minimize operating cost alone, or enforce fixed buffer time between trips, or rely on infrastructure improvements to reduce delay; our schedules substantially reduce delay cost with modest increases in operational cost. These improvements are achieved by adjusting travel times to balance operating and delay cost, targeting travel time adjustments for peak trips in peak directions, and pairing trips that naturally allow for buffer time and exploit anti-correlated primary delay structure.