This paper presents a Physics-informed Hierarchical Controller (PHC) for adaptive locomotion of multi-legged space robots in microgravity. The controller comprises a high-level reinforcement learning policy that incorporates dynamic priors for gait decision-making, and a low-level trajectory optimization module that generates dynamically constrained foot trajectories, synergistically balancing stability and mobility. A dual-track evaluation strategy is adopted. In microgravity simulations on the Hexapod Rover, PHC achieves autonomous gait transitions across sloped terrains with effective foot slippage suppression. In physical ground experiments on the JetHexa, the policy successfully transfers to real hardware, showing improved stability over fixed-gait baselines and energy consumption comparable to the most efficient tripod gait. The simulation results provide evidence for effectiveness of the PHC framework in the target microgravity environment, while the ground experiments suggest its engineering feasibility and cross-platform generalization capability. Compared to end-to-end control methods, PHC significantly accelerates policy convergence and enables smooth terrain-adaptive gait transitions. While the current validation is limited to simulation and ground-based experiments, the proposed method offers a promising framework for autonomous motion control of space robots, with strong algorithmic compatibility and practical deployability.
Martian regolith simulants are widely used as proxy materials in laboratory studies of martian surface materials and in situ resource utilization (ISRU). Orbital and in situ observations provide important constraints on regolith mineralogy and chemistry, but preparing a simulant also requires selecting suitable terrestrial mineral feedstocks. Because minerals within the same broad group can differ substantially in composition, feedstock selection can affect the mineralogical and chemical characteristics of the resulting simulant. In this study, BH-Mars-N was developed by combining mineral-scale information from the nakhlite NWA 10153 with regolith-scale compositional constraints. Mineralogical, compositional, and crystallographic analyses were used to characterize NWA 10153 and evaluate candidate terrestrial feedstocks. NWA 10153 served as a phase-specific reference for feedstock selection, whereas the final mineral proportions were determined independently from reported martian regolith mineral-abundance ranges and bulk major-oxide optimization. BH-Mars-N consists of 40 wt% plagioclase, 36 wt% pyroxene, 15 wt% olivine, 8 wt% magnetite, and 1 wt% α-cristobalite. The selected feedstocks yielded Single-Mineral Chemical Composition Figure of Merit (FoM) values of 0.8074–0.9256, and the bulk FoM relative to the mean martian regolith composition was 0.9368. Using the same calculation basis, comparison with several representative martian regolith simulants showed relatively high bulk major-oxide correspondence for BH-Mars-N across multiple martian reference compositions. Rietveld refinement and bulk XRD showed overall correspondence of the major crystalline phases in the final mixture with the nominal formulation and the selected martian surface references. The particle-size distribution fell within the ranges represented by the selected martian surface and simulant references. BH-Mars-N provides a compositionally defined and component-traceable material basis for comparative studies of martian regolith processing.
