StylePredict: Machine Theory of Mind for Human Driver Behavior From\n Trajectories

Abstract

Studies have shown that autonomous vehicles (AVs) behave conservatively in a\ntraffic environment composed of human drivers and do not adapt to local\nconditions and socio-cultural norms. It is known that socially aware AVs can be\ndesigned if there exist a mechanism to understand the behaviors of human\ndrivers. We present a notion of Machine Theory of Mind (M-ToM) to infer the\nbehaviors of human drivers by observing the trajectory of their vehicles. Our\nM-ToM approach, called StylePredict, is based on trajectory analysis of\nvehicles, which has been investigated in robotics and computer vision.\nStylePredict mimics human ToM to infer driver behaviors, or styles, using a\ncomputational mapping between the extracted trajectory of a vehicle in traffic\nand the driver behaviors using graph-theoretic techniques, including spectral\nanalysis and centrality functions. We use StylePredict to analyze driver\nbehavior in different cultures in the USA, China, India, and Singapore, based\non traffic density, heterogeneity, and conformity to traffic rules and observe\nan inverse correlation between longitudinal (overspeeding) and lateral\n(overtaking, lane-changes) driving styles.\n