Personalization of future urban mobility

Abstract

In the past few years, we have been experiencing rapid growth of new mobility solutions fueled by a myriad of innovations in technologies such as automated vehicles and in business models such as shared-ride services. The emerging mobility solutions are often required to be profitable, sustainable, and efficient while serving heterogeneous needs of mobility consumers. Given high-resolution consumer mobility behavior collected from smartphones and other GPS-enabled devices, the operational management strategies for future urban mobility can be personalized and serve for various system objectives. This thesis focuses on the personalization of future urban mobility through the personalized menu optimization model. The model built upon individual consumer's choice behavior generates a personalized menu for app-based mobility solutions. It integrates behavioral modeling of consumer mobility choice with optimization objectives. Individual choice behavior is modeled through logit mixture and the parameters are estimated with a hierarchical Bayes (HB) procedure. In this thesis, we first present an enhancement to HB procedure with alternative priors for covariance matrix estimation in order to improve the estimation performance. We also evaluate the benefits of personalization through a Boston case study based on real travel survey data. In addition, we present a sequential personalized menu optimization algorithm that addresses trade-off between exploration (learn uncertain demand of menus) and exploitation (offer the best menu based on current knowledge). We illustrate the benefits of exploration under different conditions including different types of heterogeneity.