UKF data fusion of odometry and magnetic sensor for a precise indoor localization system of an autonomous vehicle

The precise localization system of autonomous vehicles is one of the important techniques in the research of autonomous driving. In the case of outdoor localization, the Global Positioning System (GPS) sensor is commonly used to attenuate the cumulative error of odometry data occurred from undesirable effects of slippage. However, considering GPS outages, its positioning inaccuracy would not be used as reference information for an accurate indoor localization. Therefore, in this paper, we describe a sensor fusion system using odometry and magnetic sensor data for the precise indoor localization system. The odometry model is based on the Ackerman steering geometry, and directly calculated from encoder information on two rear wheels and one steering. A magnetic sensor is developed to measure magnetic field produced from circular-type magnets located in the ground. Considering the sensor data fusion, Unscented Kalman Filter (UKF) is utilized so that the mean and covariance of the nonlinear system function can be estimated by the Unscented Transformation (UT) with sigma points. The simulation results are presented that demonstrate the validity of the UKF based localization system.