Midterm colloquium Olaf de Vries

 

Solving Dynamic Vehicle Routing Problems Using Inverse Optimization

Supervisors: Dr. Peyman Mohajerin Esfahani and Pedro Zattoni Scroccaro

Abstract:
The Vehicle Routing Problem is an important optimization problem in operations research. It is widely studied and efficient heuristic algorithms exist. In the dynamic variant however, these heuristics are not well equipped to find efficient solutions, since a different decision problem arises: when to dispatch to which requests. Machine Learning algorithms do show promise with this decision problem, and, given the recent rise of routing problem research in the Machine Learning literature, the 'EURO meets Neurips 2022 Vehicle Routing Competition' was created. We will participate retroactively with a machine learning method called Inverse Optimisation. The preliminary results rank high in the competition, showing that Inverse Optimisation is suited for such routing problems.