Surfel-based lightweight LiDAR-inertial SLAM system for UAV pose estimation

Abstract

The easily occluded nature of satellite signals poses significant challenges to the pose estimation of small unmanned aerial vehicles. In this paper, a lightweight LiDAR (Light Detection and Ranging) -inertial SLAM (Simultaneous Localization and Mapping) system specifically tailored for UAV pose estimation is proposed. The proposed surfel-based LiDAR point cloud registration algorithm achieves point cloud registration and pose estimation by minimizing the distance between points and surfels, while reducing the algorithm′s computational complexity and ensuring lightweight operation by discarding un-stable surface elements. The framework of integrating this algorithm into the error-state Kalman filter based LiDAR-inertial SLAM system is also designed at the same time. The proposed SLAM system is evaluated through experiments conducted on experimental datasets. The results demonstrate superior pose estimation accuracy compared to existing LiDAR-Inertial navigation systems. Furthermore,while maintaining the performance in terms of runtime, the proposed technique reduces the average position deviation by 37.63% and the average attitude deviation by 33.94% in outdoor satellite signal denied environments.

Keywords

• uav pose estimation
• lightweight lidar-inertial slam
• multi-sensor fusion
• point cloud registration