Transportation and Logistics
Transportation and logistics problems provide context for much of my research. Practically, these problems are key drivers of the global supply chain. Theoretically, they are very challenging. Two things make them difficult. First, the size of the decision space is combinatorial. For example, the number of ways to route a fleet of vehicles across a supply network is considerable. Second, uncertainty is often multidimensional and unfolds across time. For instance, the number of possible demand realizations across SKUs and over an operating horizon can be substantial.
How to obtain good policies for dynamic and stochastic logistics problems is often an open question. My research approaches the issue by bringing together a deterministic legacy and recent theoretical advances. Modern transportation research leans on decades of deterministic analyses where demand, travel conditions, customer behavior, and various other factors are assumed to be known. When uncertainty enters the mix, deterministic methods generally break down. However, deterministic procedures can often be adapted to stochastic environments, leading to feasible policies, to dual bounds on the value of an optimal policy, and to insights that improve decision making.
This approach has led to problem-specific advances and to general methodological innovations. For example, see my paper titled "Electric Vehicle Routing with Public Charging Stations." For an example that also leverages developments in deep neural networks, see my paper titled "Dynamic Ridehailing with Electric Vehicles."
