Real-time ego-motion estimation using Lidar and a vehicle model based Extended Kalman Filter

Automated driving maneuvers enable a highly reproducible validation of preventive vehicle safety systems. However, the automation of vehicle guidance requires an exact and reliable knowledge of current vehicle position and motion. This paper presents a new method for the real-time estimation of the vehicle position and of further longitudinal and lateral dynamic state variables. Fundamental idea is the fusion of the Lidar-based range and bearing measurements of landmarks with the information of various vehicle sensors by means of an advanced vehicle model based Extended Kalman Filter. It takes into account the nonlinear tire characteristics at the limits of driving physics when estimating the variables. Moreover, the proposed ego-localization and ego-motion estimation scheme incorporates an approach for the automated association of Lidar-detected objects to predefined landmarks. Using the experimental results of a highly dynamic driving maneuver the accuracy and robustness of the proposed method is demonstrated.