A redundant sensor system with driving dynamic models for automated driving

Abstract

The knowledge of the current motion of the vehicle plays a central role for driving dynamics control. The over ground velocity and side slip angle for stability systems and the lane accurate position for automated driving must be available even when sensors fail or they must be accurate enough until the driver undertakes the control of the car again. This task is called dead reckoning in the navigation community. To estimate position, velocities and parameters of driving dynamics an Extended Kalman Filter (EKF), that includes a nonlinear two track model, is used. Inputs are steering wheel angle and wheel speeds. The states of the horizontal driving dynamics are position, yaw angle, longitudinal and lateral velocity and yaw rate. To make the filter adaptive to changing road conditions or tire wear, parameters of a modified Pacejka tire model are modeled as states too. To get realistic uncertainties in the model, some nuisance factors are modeled as nuisance states, which is a novelty in driving dynamic state estimation. The measurements contain the ESC sensors (yaw velocity, long. and lat. accelerations) and one sensor that can measure the absolute position and velocity in a fixed coordinate system, e.g. by a GPS.