2818. A decomposed MIP approach for a large size workforce scheduling problem in public transport sector
Invited abstract in session WC-59: Urban Mobility I, stream Transportation.
Wednesday, 12:30-14:00Room: Liberty 1.14
Authors (first author is the speaker)
| 1. | Nicolas Blandamour
|
| Sia AI |
Abstract
Workforce scheduling in public transport involves large-scale optimization with intricate constraints, including regulations, fairness, and operational rules across heterogeneous service lines. Such a problem was tackled for a leading European public transport actor. The schedules are constructed independently for each transport line, with each line involving over 100 agents whose schedules need to be defined for a one-year period.
It was solved using a three-step decomposition: (1) assigning work/rest days, (2) selecting service types, and (3) assigning specific shifts. Rest day assignment is solved using a large-scale binary MIP, adaptable to eight lines with distinct constraints. Due to project deadlines, service assignment was solved via another MIP instead of heuristics or column generation, which remain future improvements to fasten the resolution process that lasts several hours.
While such a decomposition may induce suboptimality, it allows iterative validation with business stakeholders, a key factor in change management. It also allows an alignment with existing business methodologies for better adoption. Given the substantial gains observed -up to 50% improvement on some KPIs- pursuing exact optimality was unnecessary in this preliminary phase.
Key innovations include fairness-driven optimization, workload balancing, and service hardship scores. The project is now entering industrialization for large-scale deployment and UX/UI integration.
Keywords
- Human Resources Management
- Practice of OR
- Railway Applications
Status: accepted
Back to the list of papers