A data-driven machine learning model for predicting forces required for lunar regolith excavations
Year: 2026
Type: article
Abstract: Accurate prediction of excavation forces in lunar regolith (i.e., Moon’s surface) is critical for designing efficient excavation systems for future lunar surface operations. Traditional analytical models often fail to capture the complex interactions among excavation parameters. While machine learning (ML) approaches have proven effective in terrestrial excavation as well as geotechnical applications, their use in predicting excavation forces in lunar regolith, especially involving percussive tools, remains unexplored. To address this gap, four ML regression models, multiple linear regression (MLR), support vector regression (SVR), random forest regression (RFR), and extreme gradient boosting regression (XGBR) were developed in this study to predict horizontal and vertical excavation forces using experimental data from tests on lunar regolith simulants. Each model’s performance was assessed with standard regression metrics, and feature importance analysis was conducted for the model that achieved the best results. Among the ML models, XGBR recorded the lowest prediction errors, with RMSE values of 11.305 N and 6.824 N for horizontal and vertical excavation forces, respectively. Notably, XGBR model surpassed the performance of the best analytical model, Lockheed-Martin model by reducing the RMSE in horizontal and vertical force predictions by 64% and 85.5% respectively. A graphical user interface (GUI) was ultimately designed as a tool for generating accessible force predictions from user-provided input conditions. This study contributes a novel ML based approach for predicting excavation forces in lunar regolith, offering an alternative to traditional analytical methods to support lunar excavation planning. Furthermore, SHAP-based interpretability analysis identifies the key parameters governing excavation forces, enabling informed decision-making for the design and optimization of lunar excavation tools. The developed ML framework and GUI also provide practical engineering tools that can guide future lunar excavation experiments and support the design of optimized excavation systems for space missions. By enabling engineers to evaluate suitable excavation tool configurations prior to mission deployment, the proposed framework has the potential to reduce design uncertainties, improve operational planning, and contribute to lowering the cost and risk associated with expensive lunar exploration missions.
Source: Acta Astronautica
Institution: University of Alberta
Language: English
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,350
Funder: University of Alberta
OpenAlex ID: W7203707249