Learning Tool-Aware Adaptive Compliant Control for Autonomous Regolith Excavation

2025-09-05

AI Summary

Autonomous lunar regolith excavation for extraterrestrial in-situ resource utilization faces challenges including inaccurate granular media dynamics modeling and poor compatibility across diverse excavation tools. Method: This paper proposes a tool-aware adaptive compliance control framework integrating high-fidelity particle-based physics simulation, procedural terrain and tool morphology generation, operational-space control, and vision-enhanced feedback to establish a model-based reinforcement learning paradigm—enabling online adaptation of stiffness/damping parameters and cross-tool generalization. Contribution/Results: It introduces, for the first time, tool-agnostic compliance control trained via procedural simulation, substantially improving adaptability to unseen tools. Experiments demonstrate a 42% increase in excavation success rate in simulation, with task robustness and efficiency achieving state-of-the-art international performance.

Technology Category

• Intelligent Robots: Manipulation
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Autonomous regolith excavation is a cornerstone of in-situ resource utilization for a sustained human presence beyond Earth. However, this task is fundamentally hindered by the complex interaction dynamics of granular media and the operational need for robots to use diverse tools. To address these challenges, this work introduces a framework where a model-based reinforcement learning agent learns within a parallelized simulation. This environment leverages high-fidelity particle physics and procedural generation to create a vast distribution of both lunar terrains and excavation tool geometries. To master this diversity, the agent learns an adaptive interaction strategy by dynamically modulating its own stiffness and damping at each control step through operational space control. Our experiments demonstrate that training with a procedural distribution of tools is critical for generalization and enables the development of sophisticated tool-aware behavior. Furthermore, we show that augmenting the agent with visual feedback significantly improves task success. These results represent a validated methodology for developing the robust and versatile autonomous systems required for the foundational tasks of future space missions.

Problem

• Autonomous regolith excavation with diverse tools
• Managing complex granular media interaction dynamics
• Learning adaptive compliant control for generalization

Innovation

• Model-based reinforcement learning in parallelized simulation
• Dynamic stiffness and damping modulation via operational space control
• Procedural generation of terrains and tool geometries for generalization