Data-driven analyses of future electric personal mobility

Personal mobility is moving towards the era of electrification. Adopting electric vehicles (EV) is widely regarded as an effective solution to energy crisis and air pollution. Many automakers have announced their roadmap to electrification in the next 1-2 decades. At the same time, limited electric range and insufficient charging infrastructure are still obstacles to EV large-scale adoption. However, with the emerging technologies of ride-hailing, connected vehicles, and autonomous vehicles, these obstacles are being solved effectively, and the EV market penetration is expected to increase significantly. Among the many kinds of electric mobility, electric taxis and personal battery electric vehicles (BEV) especially are gaining increasing popularity and acceptance among customers. This dissertation studies the future challenges of electric taxis and personal BEVs. First, this dissertation examines the BEV feasibility from the spatial-temporal travel patterns of taxis. The BEV feasibility of a taxi is quantified as the percentage of occupied trips that can be completed by BEVs during a year. It is found that taxis with certain characteristics are more suitable for switching to BEVs, such as fewer daily shifts, shorter daily driving distance, and higher likelihood to dwell at the borough of Manhattan. Second, we model and simulate the operations of electric autonomous vehicle (EAV) taxis. EAV taxis are dispatched by the optimization-based model and the neural network-based model. The neural network dispatch model is able to learn the optimal dispatch strategies and runs much faster. The EAV taxis dispatched by the neural network-based model can improve operational efficiency in term of less empty travel distance and smaller fleet size. Third, this dissertation proposes a cumulative prospect