A tightly-coupled LIDAR-IMU SLAM method for quadruped robots

Aiming to address the issue of mapping failure resulting from unsmooth motion during SLAM (Simultaneous Localization and Mapping) performed by a quadruped robot, a tightly coupled SLAM algorithm that integrates LIDAR and IMU sensors is proposed in this paper. Firstly, the IMU information, after undergoing deviation correction, is utilized to remove point cloud distortion and serve as the initial value for point cloud registration. Subsequently, a registration algorithm based on Normal Distribution Transform (NDT) and sliding window is presented to ensure real-time positioning and accuracy. Then, an error function combining IMU and LIDAR is formulated using a factor graph, which iteratively optimizes position, attitude, and IMU deviation. Finally, loop closure detection based on Scan Context is introduced, and loop closure factors are incorporated into the factor graph to achieve effective mapping. An experimental platform is established to conduct accuracy and robustness comparison experiments. Results showed that the proposed algorithm significantly outperforms the LOAM algorithm, the NDT-based SLAM algorithm and the LeGO-LOAM algorithm in terms of positioning accuracy, with a reduction of 65.08%, 22.81%, and 37.14% in root mean square error, respectively. Moreover, the proposed algorithm exhibits superior robustness compared to LOAM, NDT-based SLAM and LeGO-LOAM.