Managing Exponential Decision Spaces
Combinatorial optimization problems such as vehicle routing problems, team orienteering problems, and scheduling problems are pervasive in today’s society. Logistics companies such UPS, FedEx and Amazon solve these types of problems daily in order to most efficiently deliver packages. Despite their pervasive nature in today’s society these types of problems are notoriously difficult to solve, and optimal solutions are often unattainable in a reasonable amount of time. The Managing Exponential Decision Spaces (MEDS) project evaluated how machine learning approaches could aid in improving solve time and solution quality for these difficult class of problems.
Dr. Katie McConky advised the work of mechanical and industrial engineering Ph.D. student, Prashant Sankaran who developed novel deep learning architectures to solve collaborative multi-vehicle routing problems, developed novel machine learning driven parent selection algorithms that can be incorporated into genetic algorithms, and developed a hybrid machine learning optimization solver that ties together three common solution methods into a collaborative platform. The solutions developed enable scalable models that provide quality solutions in faster amounts of time than previously possible. We are very proud that Prashant Sankaran will be joining SUNY Buffalo as an Assistant Professor in the industrial and systems engineering department in the fall of 2023.