To address multipath interference, sparse echoes, and motion distortion of millimetre-wave radar point clouds in underground unstructured environments, this paper proposes a tightly coupled millimetre-wave radar-inertial measurement unit (IMU) simultaneous localisation and mapping (SLAM) system for quadrotor unmanned aerial vehicles (UAVs), enabling real-time 3D reconstruction and high-precision pose estimation in complex roadways. Three-level point cloud pre-processing (statistical outlier removal SOR, voxel grid downsampling VGD, and density-based spatial clustering of applications with noise DBSCAN) is adopted to effectively suppress radar clutter and multipath interference. Within a sliding window optimisation framework, we build a tightly coupled graph-based SLAM system integrating radar point cloud registration error and IMU pre-integration constraints, and design an IMU pre-integration-based undistortion algorithm to correct radar point cloud motion distortion. Simulations show that in strong multipath environments, the proposed method reduces ATE RMSE by 35%–55% and RPE by 38% compared with LC-SLAM and IMU-only methods, with terminal drift below 0.3%. Undistortion processing cuts peak RMSE by 45% and lowers local median error to 0.016 m. The SLAM system costs 70–100 ms per frame, with the undistortion module taking 21–25 ms per frame, satisfying the 15 Hz radar real-time requirement. This method provides algorithmic support and engineering reference for positioning and 3D modelling in underground unstructured spaces.
Zheng et al. (Fri,) studied this question.