Graph optimization Based Long-distance GPS/IMU Integrated Navigation

Abstract

The accurate position estimation plays an critical role in the autonomous navigation for Micro Aerial Vehicles (MAV). Global positioning System (GPS) and inertial measurement unit (IMU) are two common sensors for navigation widely used on MAVs in the urban environment. Both of them have its distinct disadvantages that the GPS is susceptible to environmental interference and the IMU has accumulative errors. To overcome these problems, a GPS/IMU integrated system based on the factor graph optimization is developed in this paper. Unlike the conventional extended Kalman filter (EKF)-based method, the graph optimization method takes the whole trajectory into consideration so that it can achieve enough accuracy even after a long distance. Furthermore, the IMU preintegration method is used to avoid the repeated computation of high-rate IMU data. Compared with the EKF method, the experimental results on the Zurich urban micro aerial vehicle dataset show the superior accuracy of the proposed factor graph optimization algorithm.