Abstract: Roving exploration using mobile robots is crucial for extending the breadth and depth of deep space exploration. However, current terrestrial perception methods employed by mobile robots in orbit struggle to effectively identify non-geometric hazards(such as soft sandy areas) under extraterrestrial conditions characterized by rugged and natural terrain, high texture homogeneity, extreme illumination variations, and limited prior knowledge. This limitation consequently compromises the navigational safety of mobile robots.This study designs a soil mechanical properties inference experiment based on multispectral characteristics: spectral characteristics curves and corresponding mechanical parameter databases covering multiple simulated soils are constructed through spectral characteristic measurement experiments and in-situ Bevameter mechanical properties measurement experiments; a hierarchical mapping mechanism of "spectral characteristics-soil type-mechanical properties" is established to qualitatively infer soil mechanical properties; validation is conducted using multispectral image data captured by Zhurong rover. The validation results demonstrate that identifying dominant soil types through spectral features and correlating them with mechanical properties via database allows qualitative inference of surface mechanical parameters, which provides decision support for safe path planning of extraterrestrial robots.

Key words: extraterrestrial exploration, mobile robot, multi spectrum, mechanical properties