Abstract
Traffic congestion continues to pose serious challenges in metropolitan areas, including travel delays, air pollution, noise, and increased accident risks. These impacts translate to substantial societal costs, with major U.S. cities experiencing billions of dollars in annual congestion-related losses. A prominent feature of congestion is its temporal concentration during peak periods, driven largely by travelers’ similar trip timing decisions, such as arriving at work by 9:00 AM. Shifting demand away from peak periods is thus a critical strategy for mitigating congestion. This dissertation addresses this issue by developing and analyzing stable day-to-day dynamical models of travelers’ departure time choices and examining how optimal pricing influences these decisions to improve system efficiency.While traditional transportation analyses focus on equilibrium outcomes, they often overlook the dynamic process by which such equilibria are achieved. This dissertation focuses on how travelers adjust their departure times from day to day and collectively converge to a departure time user equilibrium (DTUE). By capturing these dynamics, we can better design pricing policies that can drive the system toward a system optimal (SO) state. A major emphasis is on the theoretical stability of the dynamical models, which is essential to ensure that they reflect real-world travel behavior that is empirically observed to be convergent.This dissertation advances the field in four key areas. First, it extends an existing stable single-class day-to-day departure time dynamical model at the corridor level to a multi-class setting, with traveler heterogeneity. When desired arrival times are identical, with different queuing costs relative to unpunctuality costs, the proposed multi-class model is proven to be asymptotically stable, and its stationary state is equivalent to a multi-class DTUE. Second, it applies various pricing schemes to both single- and multi-class dynamical models. It is demonstrated, both theoretically and numerically, that appropriate pricing, such as optimal fine tolling, can drive the system from a DTUE to a stable, stationary SO state. Third, the dissertation extends departure time modeling from a single corridor to the network level using Vickrey’s cordon pricing framework with 48-hour entry data from Manhattan, New York. Results show that Vickrey’s marginal cost pricing — combined with day-to-day adjustments — can drive the system to a practically stable optimal state. Finally, the dissertation integrates the network-level departure time dynamics with dynamic traffic assignment (DTA) modeling. Results show that Vickrey’s marginal cost pricing significantly improves peak-period congestion, average speeds, vehicle miles traveled (VMT), and vehicle hours traveled (VHT). The work also explores the generalized bathtub model and develops insights on effectively reproducing DTA results at much less computational cost, offering a unique sketch-level dynamic planning method for large networks.In summary, the dissertation: (1) introduces the first provably stable multi-class day-to-day departure time dynamics; (2) shows that optimal pricing can drive both single- and multi-class systems to a stable SO state; (3) empirically validates Vickrey’s pricing in a real-world urban context; and (4) compares the effects of departure-time versus route-choice optimization using DTA. Collectively, these contributions provide a stable modeling framework for analyzing and managing departure time choices at both corridor and network levels, which offers new theoretical insights for designing congestion pricing policies, and provides practical insights on a notably more efficient process for transportation planning at large.