An improved robust filtering for GPS/IMU integration navigation

Accurate positioning information underpins the safety and efficiency operation of vehicles in general aviation. In particular, the Global Positioning System (GPS) and Inertial Measurement Units (IMU) integrated system is widely used to provide positioning information. In order to obtain the accuracy and reliable positioning information, robust filtering is often used to suppress the errors in the GPS observations and therefore improves the performance of GPS/IMU integrated navigation systems. In the loosely coupled GPS/IMU integrated navigation system, the state vector is with lower dimension than the tightly couple fusion scheme. The computed positioning results are with various positioning accuracy in different axises, i.e. the accuracy in the easting and northing directions is always higher than it is in the up direction. Therefore, the conventional robust filtering algorithm that builds standardized prediction residuals based on the median of observations may not be able to correctly and effectively detect all errors in observations. To solve this problem, we have proposed an improved robust filtering algorithm for GPS/IMU integration navigation system. The algorithm constructs the standardized prediction residuals of each observation, respectively, based on the historical information using a windowing scheme. Experimental results show that the proposed algorithm has the highest positioning accuracy. The values of root mean square error for the proposed algorithm are 0.45m, 1.75m and 2.37m in the east, north and up direction, respectively.