Robust Land Vehicle Positioning with Invariant Kalman Filters and Map-Based Corrections

This paper presents a system that integrates Invariant Extended Kalman Filters with onboard sensors and map-based corrections. The system combines multiple sensors, such as an Inertial Measurement Unit and a speedometer, to improve accuracy and robustness during gaps in the map. The proposed architecture takes into consideration the different nature of each measurement used during the update step, such as map-based measurements in the world-frame and speedometer measurements in the body-frame. The experimental evaluations were conducted in diverse driving conditions. The proposed system consistently outperformed traditional solutions based on extended Kalman filters, sustaining low relative translational errors during map outages.