Lidar-inertial-wheel SLAM for ground robots in complex scenarios

Abstract

Lidar-Inertial Simultaneous Localization and Mapping (LI-SLAM) is widely used in robot navigation and 3D mapping. However, existing methods suffer from error accumulation due to IMU-based speed constraints and reduced matching accuracy in feature-degraded environments, impairing reliability in complex scenarios. To address this, we propose a tightly-coupled SLAM algorithm integrating LiDAR, IMU and a wheel odometer. Firstly, we derive a manifold-based incremental wheel odometer model to enhance its 3D measurement accuracy. Further, an adaptive strategy adjusts heterogeneous sensor fusion weights dynamically to enhance robustness against single-sensor failures. Experiments on open source datasets and our vehicle show the proposed algorithm achieves higher localization accuracy and stronger fault tolerance in abnormal situations compared to mainstream methods.