Real-time 2D localization of a car-like mobile robot using dead reckoning and GPS, with satellite masking prediction
Résumé
The paper deals with the study and the implementation of an extended Kalman filter (EKF) applied to real-time 2D vehicle positioning. This has special applications in fields where inaccurate models or disturbed sensors are involved, such as in driving assistance or mobile robotics in urban area. Particular attention has been payed to the adaptation of the algorithm to the real-time conditions of data acquisition. The contribution of this work also includes some tests performed in real conditions to check the localization algorithm efficiency. To do so, a vehicle is equipped with gyroscope and odometer sensors to be fused with GPS solutions. A Mahalanobis test is also implemented to verify the GPS consistency and reject outliers. Finally, a satellite constellation prediction tool has been used, that encompasses both the planned trajectories and the GPS ephemeris, as well as the surrounding structures.