Research On LiDAR-aided Inertial Navigation System For Urban And Indoor Environments

Due to the complementary characteristics of Global Positioning System(GPS)and Inertial Navigation System(INS),GPS is commonly integrated with INS to constrain the error accumulation in INS.However,in urban and indoor environments where GPS is susceptible to signal jamming or blockage,the positioning solutions become inaccurate or unavailable.Therefore,alternative sources of aiding information and techniques are required to provide periodic corrections to INS.Recently,with the advantages of high accuracy,high sampling rate and less computation load,Light Detection and Ranging(LiDAR)has been widely used in applications like pose estimation,mobile mapping and Simultaneous Localization and Mapping(SLAM).Meanwhile,GPS and LiDAR work in complementary environments.Therefore,the LiDAR-aided INS for both urban and indoor environments is studied in this work to achieve continuous accurate navigation performance.Two types of inertial sensor system are introduced,namely 2D Reduced Inertial Sensor System(RISS)and 3D RISS.Compared with full Inertial Measurement Unit(IMU)containing three gyroscopes and three accelerometers,RISS uses odometer to calculate velocity,and two accelerometers replace the gyroscopes to derive pitch and roll.The system cost and complexity are reduced while the accuracy is improved.For indoor office environments,the commonly existed lines are extracted as features for pose estimation.An accurate and fast line feature extraction algorithm is proposed.The residuals from line feature parameters estimation in the line feature extraction process are used as weights in the calculation of the LiDAR observations and the associated covariance.Thus,the estimated covariance can reveal the error level of the LiDAR observations.Line feature-based scan matching method and scanned point-based scan matching method are implemented respectively and compared in terms of accuracy and computational efficiency.The pose estimated from the two scan matching methods are fused with the pose predicted by 2D RISS through Extended Kalman Filter(EKF).When there are less than two nonlinear line features are matched between two scans,the line feature-based scan matching method fails to accurately estimate the pose.To address this limitation,the tight coupling of 3D RISS and LiDAR is proposed.The tight coupling is defined as the integration of line feature parameters changes over two scans estimated from LiDAR with the predictions from 3D RISS.The tight coupling of 3D RISS and LiDAR can output accurate 3D navigation solutions.To solve the initial alignment issue in the absence of absolute position information in indoor environments,an initial position and azimuth estimation algorithm using Wireless Local Area Network(WLAN)to aid the cross correlation of the environment map and LiDAR scan image is proposed.Due to the complementary properties of line feature-based scan matching method and scanned point-based scan matching method,the hybrid scan matching algorithm is proposed to combine the two scan matching methods to benefit from the advantages of each method.The integration system of INS,GPS and LiDAR for both urban and indoor environments is proposed.GPS and LiDAR are used as aiding systems to alternatively provide periodic corrections to INS in different environments.In GPS unreliable and denied environments,two integration schemes of INS and LiDAR with hybrid scan matching algorithm are implemented and compared.The experimental platform adopted in this work is Husky A200 from Clearpath Robotics.Two experiments including outdoor and indoor environments are performed to evaluate the proposed different INS/LiDAR integration schemes.Experimental results show that with the aid of LiDAR,the inertial sensors biases can be compensated for and the integrated navigation system can achieve sub-meter position accuracy during the whole trajectory.