Development of a High-assurance Confined-space Detection Robot Platform With Perception Techniques for Extreme Ceiling-interstitial Environments

This paper presents a high-assurance tracked robot platform and integrated environment perception pipeline for confined ceiling-interstitial spaces, where low illumination, dense structural clutter, and locomotion-induced shocks and vibrations can severely degrade sensing quality and mission reliability. The proposed system employs a LiDAR - IMU SLAM (FAST-LIO2) backbone with LiDAR camera fusion to generate accurate 3D maps and provide a colorized 3D map for improved human interpretability. To mitigate perception degradation under adverse conditions, the pipeline integrates low-light enhancement, odometry-aided image stabilization, overhead (ceiling) structure removal, and point-cloud colorization. Experiments in a ceiling-environment testbed demonstrate accurate mapping, while the proposed stabilization reduces RMS global motion in a vibration- and step-induced setting. These results indicate that the developed platform and perception stack enhance mapping consistency and visual interpretability, supporting practical applications such as safety inspection, reconnaissance, and disaster-response missions in confined environments.