Model-based threat assessment for lane guidance systems

This paper considers a threat assessment problem in a lane guidance application for semi-autonomous vehicles. In particular, in order to issue an autonomous assisting intervention, we assess the vehicle\´s ability to safely travel along a path subject to limitations arising from the vehicle\´s dynamics and the driver\´s ability. We first introduce a set of constraints describing "safe driving". For the specific lane guidance application considered in this paper, the constraints are set by the lane boundaries and the vehicle\´s stability limits. We then formulate the threat assessment problem as a constraints satisfaction problem over a finite time horizon, solved by resorting to reachabililty analysis and invariant set theory. Validation with experimental data demonstrates the capability of the proposed threat assessment method of predicting vehicle instability or crossing of the lane boundaries.