A PREDICTIVE APPROACH TO ROADWAY DEPARTURE PREVENTION

In this paper, we investigate predictive control approaches to the problem ofroadway departure prevention via steering and braking. We assume a sensinginfrastructure detecting road geometry and consider a two layers architectureconsisting of a threat assessment and an intervention layer. In particular, theupper threat assessment layer detects the risk of roadway departure or vehi-cle instability within a future time horizon. If a risk of roadway departure orvehicle instability is detected, the lower intervention layer is enabled. The lat-ter is designed based on Model Predictive Control (MPC) approaches, wheresteering and braking interventions are the result of an optimization problem.This is formulated on the basis of vehicle state measurements and coming roadinformation (e.g., road geometry, surface adhesion) and repeatedly solved overa moving future time horizon.Simulation and experimental results are presented, showing that the proposedapproach effectively exploits road preview capabilities in order to issue earlierand less intrusive interventions, compared to standard Electronic Stability Con-trol (ESC) systems