Vehicle Dead-Reckoning Autonomous Algorithm Based on Turn Velocity Updates in Kalman Filter

Abstract

The paper presents a Kalman-based dead-reckoning algorithm for a vehicle. The algorithm uses inertial data only. No data from other sources of information are utilized. The proposed technique relies on two aspects: pseudo-acceleration removal procedure and novel turn velocity update (TVU) correction technique applied when the vehicle performs a maneuver. The dynamic algorithm performance was evaluated using real data obtained from the forklift and compared with the output of the ultra-wideband positioning system. To achieve this, the synchronization between UWB and inertial data was employed by aligning velocities in acceleration and deceleration moments. Seven experiments were carried out to test the method. The duration of each experiment varied from 4 to 24 minutes. The travelled distance corresponded to the range from 118 to 380 meters. The algorithm shows fair position estimation results with the data obtained from commercial off-the-shelf MEMS sensors on a long-term run. The median position error did not exceed 1.2 meters for all performed tests. The end position estimation error, respectively, was not worse than 1.2% of total travelled distance.