Risk assessment for Collision Avoidance Systems

Collision Avoidance Systems need to perform scene analysis and risk assessment in order to react conveniently. Based on the information provided by the perception system, scene analysis has to predict the evolution of the current driving situation for the near future. Thanks to the predicted trajectories of the relevant traffic participants, the risk of collision on the ego vehicle can be calculated. In many cases, a predicted trajectory is not defined with explicit equations but is given as a set of sampled poses, each one corresponding to a different future time instant. A predicted trajectory being always uncertain, confidence has to be estimated on the so predicted poses. We present a method that propagates the known error covariance matrix of the current pose of the ego vehicle by considering local approximations of the predicted trajectory. This allows to estimate the risk of collision of the ego vehicle with a considered target object. The proposed approach uses a Monte Carlo simulation to approximate the probability that the ego vehicle and the object come into collision at a given future time instant. Each sample time of the whole prediction horizon is considered as a potential collision time so that a curve describing the variation of the risk of collision is obtained. This allows the system to have a better comprehension of the scene and to react proportionally to the threat. The overall approach has been tested with simulated data and the consistency of results is shown.