Probabilistic representation of driver space and its inference from trajectory data

Abstract

Driver space describes the vehicle-centred area where other road users cannot intrude without causing discomfort. It is traditionally measured by a deterministic distance to the boundary within which discomfort is caused. However, driver space is better regarded as probabilistic because intrusion into the driver space may cause varying levels of discomfort. In this study, we present a probabilistic representation of driver space, which conceptually captures the proximity resistance of a vehicle to its surrounding vehicle. Specifically, we develop a method to empirically infer driver space and demonstrate the method by applying it to an urban trajectory dataset. Our results show that driver space grows quadratically with the relative speed of a vehicle to its surrounding vehicle. Furthermore, we find that the longitudinal boundaries of driver space are significantly less sharp than the lateral ones. This implies that traditional distance-based approaches are inadequate for the longitudinal measurement of driver space.