SLAM and Navigation for A Humanoid Robot Based on Multi-Source Sensor Information Fusion*

High-precision environmental mapping and accurate localization constitute critical prerequisites for mobile robot navigation. However, humanoid robots frequently exhibit motion blur during locomotion, which significantly degrades the localization accuracy of visual SLAM. Addressing the locomotion characteristics and capabilities of the humanoid robot GTX-III, this study proposes a multi-layer mapping method incorporating traversability maps and a localization solution based on multi-source sensor fusion. Building upon this foundation, an autonomous navigation platform for the humanoid robot GTX-III is developed. Experimental results demonstrate that the proposed approach effectively constructs hierarchical maps, achieves superior localization accuracy compared to conventional visual-inertial schemes, and successfully accomplishes autonomous navigation in custom-built complex environment.