Robust Model Predictive Control based Automated Driving Control Algorithm for Improvement of Safety and Ride Comfort

Abstract

Over the last decade, traffic accidents caused by human error have been accounted for 90 percent of all traffic accidents. For this reason, various active safety systems which assist drivers to avoid risky situations or mitigate collision, e.g. lane keeping assistance system (LKAS), smart cruise control system (SCC), advanced emergency braking system (AEBS), traffic jam assistance system (TJA), and blind spot intervention (BSI) have been developed and marketed by major automakers. Furthermore, in recent years, the interest of automotive industry changes from the development of active safety systems to the development of an automated driving system capable of sensing surrounding environments and travelling without human input. In developing a highly automated driving control algorithm, one of the key issues is how to cope with probable risky situation which could be occurred in the foreseeable future for the enhancement of safety. The other issue in developing a highly automated driving control algorithm is to achieve robust control performance under model uncertainties and external disturbances. This dissertation focuses on the development of an automated driving control algorithm which handles probable risky situation due to the possible change of traffic situation surrounding the subject vehicle while satisfying a robust control performance with respect to model parameter uncertainties and exogenous disturbances. In order to enhance safety with respect to the potential behaviors of surrounding vehicles, not only current states of surrounding vehicles but also predictable behaviors of surrounding vehicles should be considered. Based on the probabilistic prediction of surrounding vehicles over a finite prediction horizon, a desired driving motion is determined. Then a safe driving envelope which indicates the safe driving condition over a finite prediction horizon is determined in consideration of probabilistic prediction of future states of surrounding vehicles. In order to obtain desired steering angle and desired longitudinal acceleration for keeping the subject vehicle in the safe driving envelope over a finite prediction horizon and reducing a computational burden, distributed control architecture is designed based on a Model Predictive Control (MPC) approach. For the restriction of the maximum deviation between the actual states and the nominal states due to model uncertainties and disturbances, linear state feedback control inputs are added to nominal control inputs based on the analysis of robust invariant sets. The performance of the proposed algorithm is verified via computer simulations and vehicle tests. From the simulation and vehicle test results, it has been shown that the proposed automated driving control algorithm enhances safety with respect to the potential risk while providing permissible ride comfort. Furthermore, it has been shown that the proposed algorithm achieves robust control performance under additional disturbances.