Seminar: Closing the Loop: Human Choice and Optimization in Urban Transportation

Traveler choices shape network flows, while network conditions and mobility service operations influence those choices. Capturing this interaction is essential for effective transportation planning and operations, yet integrating traveler behavior model into system-level optimization remains under-explored. Logit-based choice models are often embedded in optimization through nonlinear constraints or bilevel formulations, creating computational difficulties for both behavioral characterization and system design. This talk presents the Primal-Dual Maximum-Utility Structured Estimator (PD-MUSE), a convex programming framework for representing and estimating logit-based discrete choice models. The convex structure enables computationally tractable characterization of traveler behavior, while duality allows behavioral parameters to be recovered from a single optimization solve. Our framework provides a modular representation of traveler behavior that can be embedded in broader network optimization problems. By representing travel patterns as flows, it models behavioral responses at the network level and supports behavior-aware optimization when service levels and traveler choices are coupled. We explore various applications in travel demand modeling, transit operations, mobility-on-demand dispatching, vertiport location planning for air mobility, and curbside pricing. The talk concludes with a research agenda for closing the loop through partnerships between public agencies and private mobility operators to share operational data, test proposed policies in the field, and use observed traveler responses to inform subsequent planning and operations.

Presented by Youngseo Kim, Assistant Professor, Civil and Environmental Engineering, UCLA