Sensor fusion based state estimation for localization of autonomous vehicle

Abstract

Localization is an estimate of vehicle position for a given environment. This work focuses on the state estimation of a vehicle for localization functionality using the Schmidt Kalman filter for fused sensor data. The Kalman filter provides an efficient approach in reducing the errors presented by the sensors. Further, computational complexity is reduced through pre-processed initialization in the Schmidt Kalman filter. The two sensors used are GPS (Global Positioning System) and IMU (Inertial Measurement Unit), where GPS provides the position, and IMU provides acceleration/direction. Since GPS data has a dependency on the external environmental factors resulting in discontinuities, it is augmented with similar data until the corrected GPS data is resumed. The error in the position determined by GPS can be as high as 12m. This work presents a method for fusing sensor data using the Schmidt Kalman filter in a practical scenario.