3D LiDAR SLAM Integration with GPS/INS for UAVs in Urban GPS-Degraded Environments

This paper presents a data fusion algorithm, using an Adaptive Extended Kalman filter (AFK) for estimation of velocity and position of a UAV. A LIDAR sensor provides local position updates using a SLAM technique, a GPS provides corrections when available and an Inertial Navigation System (INS) is used as an additional input to the Extended Kalman filter. We adapt the measurement noise covariance (R) of the AKF based on both the Global Positioning System (GPS) receiver error as well as on the LiDAR point cloud point-to-point match error. A simulation environment was developed to test the proposed SLAM as well as navigation (e.g., autopilot) algorithms in a virtual, but accurate environment. We show that by adapting the measurement noise covariance (R) of the AKF we improve both the accuracy and reliability of the position estimate, specially in areas with GPS signal drop outs such as urban canyon environments.