Using Machine Learning to Identify Hidden Constraints in Vehicle Routing Problems
Published in Computers and Operations Research, 2024
This paper develops machine learning methods to identify hidden or implicit constraints affecting vehicle routing problem solutions. The approach uses LSTM-based classification to detect and locate constraint violations that may not be explicitly specified in problem formulations.
Key Contributions
- LSTM-based system for detecting hidden constraint violations in VRP solutions
- Comprehensive classification of constraint types: time windows, capacity, precedence
- Methodology to pinpoint exact violation locations in routes
- Integration of tacit driver knowledge into optimization algorithms
Applications
The identified hidden constraints can be integrated back into optimization algorithms to improve solution quality and better reflect real-world operational constraints.
Recommended citation: Konovalenko, A., Hvattum, L.M., & Msakni, M.K. (2024). "Using Machine Learning to Identify Hidden Constraints in Vehicle Routing Problems." Computers and Operations Research. https://doi.org/10.1016/j.cor.2025.107029
Download Paper