Vehicle trajectory data has been widely applied in research, ranging from human mobility analysis to short-term traffic prediction. Besides the capability of uncovering aggregative behavioral dynamics, the trajectory data also contain rich information about individual vehicles. In this talk, I will present two studies on learning the individual vehicle behaviors from a large dataset of taxi trajectories. Specifically, the first aims to recognize various passenger search strategies among taxi drivers, and the second tackles the problem of predicting the vehicle occupancy state given noisy observations.
Download the slides for this talk.Download ( PDF, 5699.85 MB)