Understanding the daily operations of electric taxis from macro-patterns to micro-behaviors

Electrifying taxi fleets represents a significant step towards a sustainable urban transportation system. However, there is a current gap in our comprehensive understanding of the day-to-day operations of electric taxi services. In this study, we utilize operational data from electric taxi fleets in Shenzhen, employing a two-stage approach to analyze the daily charging, cruising, and serving activities of drivers. Initially, we use latent profile analysis to thoroughly investigate the diverse operating patterns of 14,660 taxis. This analysis categorizes the taxis into six distinct subgroups, highlighting notable differences in aspects like charging demand, operational durations, and spatio-temporal distributions. Building on these subgroups, we further employ an inverse reinforcement learning framework to uncover various decision-making preferences across operating patterns, derived from the operational data. This in-depth analysis reveals the diverse spatio-temporal preferences of the subgroups, particularly in relation to range anxiety, charging, and cruising behaviors.