Mobility-on-Demand (MoD) systems are typically designed and analyzed assuming fixed, exogenous demand. This approach fails to capture the systems' impact on urban transportation, such as induced demand and its effects on transit ridership. This study introduces a unified framework for designing, optimizing, and analyzing MoD operations within a multimodal system where mode demand is a function of service level. Bayesian optimization (BO) is used to determine optimal MoD supply-side parameters (fleet size, fare). The framework is calibrated using Manhattan taxi data, with demand served by public transit and MoD services of varying capacities (1, 4, and 10 passengers). Mode choice is predicted using a model estimated from stated preference data. Numerical experiments demonstrate the convergence of the multimodal supply-demand system and BO's superiority over previous methods. A policy intervention (tax on ride-hailing) illustrates the framework's ability to quantify policy impacts for different stakeholders.

Yang Liu, Prateek Bansal, Ricardo Daziano, Samitha Samaranayake