Multivariate adaptive regression splines for two-stage design optimization of bio-inspired drilling into the lunar regolith

This paper proposes an adaptively updating two-stage multi-objective optimization design framework based on Multivariate Adaptive Regression Splines (MARS) for the design of the bio-inspired drill into the lunar regolith. In this framework, the geometry parameters and the controlling strategies are optimized in Stage 1 and Stage 2, respectively, aided by discrete element modeling of the bio-inspired drilling process. Minimizing construction cost and total power consumption are the design objectives in Stage 1, whereas maximizing drilling effectiveness and drilling efficiency are the design objectives in Stage 2. The Pareto front-based design optimization method is utilized to identify the optimal design in each stage. The optimal geometry design from Stage 1 is adopted for the design optimization of controlling strategies in Stage 2 and the optimal design in Stage 2 is determined as the final design of the lunar drill. The selected design space in each stage is first discretized with coarse intervals for MARS model construction. Then the design objectives in the global design space are predicted by MARS. The optimal design derived from MARS is iteratively added to train the MARS model until the optimal design is converged. A comparison between the optimal design with and without the MARS model validates the superiority of the adaptively updating MARS-based optimization framework. The proposed framework provides an efficient solution and valuable insights for the optimal design of a bio-inspired drill into the lunar regolith, which will set the stage for the design of bio-inspired tools and technologies applicable to other extraterrestrial environments.