Stochastic Optimization Approach for the Multi Depot Vehicle Scheduling Problem

Vehicle scheduling is one of the main planning problems for transit agencies. While it is relatively simple to solve in the deterministic and single-depot setting, these assumptions are unrealistic in real-world applications. Specifically, ignoring major sources of uncertainty (such as travel times) and making decisions over average predictions can lead to inferior schedules that incur additional costs and reduce the quality of service during execution. In this project, we consider the multi-depot vehicle scheduling problem under uncertainty. We propose a mathematical model that receives a timetable and historical data (e.g., travel times) to generate a long-term schedule while taking into account the day-to-day operational stage and potential future scenarios of various uncertain parameter realizations therein. The resulting schedule is optimal in terms of average total (scheduling and operational) cost and the rescheduling needs on a day-to-day basis due to uncertainty.

Faculty Supervisor:

Merve Bodur;Amer Shalaby

Student:

Margarita Castro

Partner:

Trapeze Group

Discipline:

Engineering - mechanical

Sector:

Transportation and warehousing

University:

University of Toronto