Nonlinear observer with observability-based parameter adaptation for vehicle motion estimation

Abstract

In the framework of self-driving cars and driver-assistance systems the demand for reliable information about the vehicle ego-motion is increasing. This paper describes an estimation scheme, based on a nonlinear observer design, that provides velocity and attitude angle estimates. The approach relies on a state-affine representation of a kinematic model bolstered by a dynamic model-based measurement equation. By means of a thorough observability analysis, global exponential convergence is theoretically guaranteed. Additionally, in order to minimize the errors introduced by the dynamic model limitations, an observer tuning rule is proposed. The adaptation of the tuning parameters is built upon an online observability assessment of the system without the support of the dynamic model. Experimental results show that the presented approach reliably estimates the motion states.