In context
As aging populations drive demand for intelligent nursing equipment, bed-and-chair robots—extensions of smart wheelchairs—require reliable indoor localization to navigate safely. However, building environments such as corridors and garages often present low-texture surfaces and dynamic obstacles, challenging conventional point-feature SLAM systems. This research, published in the Chinese journal Robot in early 2025, addresses these hurdles with a multi-sensor fusion approach tailored to structured building scenes.
What was reported
Researchers at Beijing University of Technology proposed a SLAM method integrating inertial, encoder, visual, and LiDAR data with multi-dimensional visual constraints. The system employs a high-frequency front-end for pose estimation and a low-frequency back-end for optimization, using Manhattan World Hypothesis constraints to correct accumulated rotation errors without sacrificing real-time performance.
To mitigate the limitations of Manhattan World assumptions in cluttered indoor settings, the team introduced a dual-camera strategy: one camera captures ground texture while another observes upper-space planes that conform to the Manhattan model, filtering out non-Manhattan objects. Multi-primitive features—points, lines, and planes—are fused to enhance robustness in sparse-texture conditions. Line features are extracted using the LSD algorithm, while plane features are obtained via RGB-D depth data and a PCA-based extraction method.
Experimental results indicated that the proposed method significantly improves localization accuracy and robustness compared to existing approaches, while maintaining real-time operation, thereby supporting deployment of intelligent bed-and-chair robots in building environments.
Why it mattered
This work demonstrates a practical path to reliable autonomous navigation in challenging indoor environments, combining geometric constraints with multi-sensor fusion. For industrial automation and service robotics, such advances enable safer and more dependable mobile platforms in human-centric spaces, potentially extending to logistics and healthcare applications where dynamic and low-texture settings are common.
“Experimental results demonstrate that, compared to the existing methods, the proposed method significantly improves the localization accuracy and robustness while ensuring real-time performance.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-01-15 · “基于多维度视觉约束的床椅机器人SLAM”