Research on active LiDAR-IMU SLAM for wheeled mobile robots in indoor environments

Abstract

Simultaneous Localization and Mapping (SLAM) is critical for enabling autonomous mobile robots to navigate and map unknown indoor environments, where Global Navigation Satellite Systems (GNSS) are not applicable due to signal obstructions and multipath effects. Traditional passive SLAM systems, reliant on fixed sensor configurations, often struggle in feature-degraded indoor settings, such as long corridors or glass-heavy offices, due to limited fields of view (FOV), sensor noise, and insufficient feature detection. Besides, with traditional LiDAR system, it is non-trivial to generate high fidelity indoor panoramic map. This thesis proposes innovative active LiDAR-IMU SLAM frameworks for wheeled mobile robots, leveraging actively actuated sensor suites to enhance localization robustness, mapping accuracy, and noise reduction in diverse indoor environments.

The research develops three integrated systems: (1) an active solid-state LiDAR SLAM system with limited FOV, employing gaze control to prioritize feature-rich regions, achieving reduction in feature detection failures in feature-degraded settings; (2) an active rotating mechanical LiDAR-IMU system with dynamic rotation speed adjustment via Bayesian optimization (BO), improving point cloud homogeneity for high-fidelity panoramic mapping; and (3) a comprehensive active LiDAR-IMU SLAM framework integrating rotation-modulated IMU pre-integration pose-graph optimization (PGO) with an Iterative Error State Kalman Filter (IESKF), reducing IMU pitch angle errors by 30% and enhancing overall system accuracy. These systems were validated through simulations (e.g., MATLAB, Isaac Sim, Gazebo), real-world experiments in varied indoor environments (e.g., foyers, corridors, 5000 m^2 parking lots), and quantitative evaluations.

The contributions of this work include robust localization in challenging environments, high-fidelity mapping for applications like Building Information Modeling (BIM), and reduced sensor noise for improved SLAM accuracy. Open-source code are provided to facilitate reproducibility and further development. The applications of the framework span assistive navigation for visually impaired people, autonomous robotic inspection, and warehouse logistics, aligned with the growing indoor positioning market. This research advances the field of indoor SLAM, offering a scalable and robust solution for autonomous navigation in complex indoor settings.

Date of Award	15 Oct 2025
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Xu Sun (Supervisor), Adam Rushworth (Supervisor), Guilin Yang (Supervisor) & Xin Dong (Supervisor)