Vehicle State Estimation for Rollover Avoidance

Abstract

To enhance vehicle/road safety, rollover warning and control systems have received considerable research interest in recent years, especially for vehicles with high center of gravity (CG). Accurate and reliable estimates of the relevant vehicle states facilitate the design of such systems. This paper investigates the state estimation for rollover avoidance, in which the relevant states include vehicle roll velocity and roll angle, as well as sideslip velocity and yaw velocity. The main challenge of the design comes from the fact that, under near-rollover situations, vehicle dynamics is complex and nonlinear. Not only vehicle suspension and tires are in their nonlinear region, but also vehicle yaw, sideslip and roll motions are highly coupled. In addition, the estimation needs to deal with sensor biases and sensor nonlinearity under this extreme condition. To address those issues, this paper proposes a vehicle state estimation design that consists of three parts: a sensor pre-filter, an Extended Kalman filter (EKF), and a sideslip velocity estimator. The sensor pre-processor removes sensor biases by utilizing the Recursive Least Square technique with a varying forgetting factor. The EKF is designed based on a linear yaw/sideslip/roll model, and its feedback gains are further scheduled based on vehicle lateral acceleration in order to reduce the effects of increased model inaccuracy as vehicle roll motion becomes more severe. The sideslip velocity estimator adjusts the sideslip velocity estimated by the EKF to extend the estimation to the nonlinear region. Both simulation and vehicle fishhook testing have been used to verify the effectiveness of the design.