Probabilistic Threat Assessment and Driver Modeling in Collision Avoidance Systems

Paper in proceeding, 2011

This paper presents a probabilistic framework for decision-making in collision avoidance systems, targeting all types of collision scenarios with all types of single road users and objects. Decisions on when and how to assist the driver are made by taking a Bayesian approach to estimate how a collision can be avoided by an autonomous brake intervention, and the probability that the driver will consider the intervention as motivated. The driver model makes it possible to initiate earlier braking when it is estimated that the driver acceptance for interventions is high. The framework and the proposed driver model are evaluated in several scenarios, using authentic tracker data and a differential GPS. It is shown that the driver model can increase the benefit of collision avoidance systems — particularly in traffic situations where the future trajectory of another road user is hard for the driver to predict, e.g. when a playing child enters the roadway.