LiDAR-Based 3D SLAM for Indoor Mapping
Teng Hooi Chan, Henrik Hesse, Song Guang Ho
Abstract
Aiming to develop methods for real-time 3D scanning of building interiors, this work evaluates the performance of state-of-the-art LiDAR-based approaches for 3D simultaneous localisation and mapping (SLAM) in indoor environments. A simulation framework using ROS and Gazebo has been implemented to compare different methods based on LiDAR odometry and mapping (LOAM). The featureless environments typically found in interiors of commercial and industrial buildings pose significant challenges for LiDAR-based SLAM frameworks, resulting in drift or breakdown of the processes. The results from this paper provide performance criteria for indoor SLAM applications, comparing different room topologies and levels of clutter. The modular nature of the simulation environment provides a framework for future SLAM development and benchmarking specific to indoor environments.