Learning and Autonomy for Extraterrestrial Terrain Sampling: An Experience Report from OWLAT Deployment (2311.17405v2)

Abstract: Extraterrestrial autonomous lander missions increasingly demand adaptive capabilities to handle the unpredictable and diverse nature of the terrain. This paper discusses the deployment of a Deep Meta-Learning with Controlled Deployment Gaps (CoDeGa) trained model for terrain scooping tasks in Ocean Worlds Lander Autonomy Testbed (OWLAT) at NASA Jet Propulsion Laboratory. The CoDeGa-powered scooping strategy is designed to adapt to novel terrains, selecting scooping actions based on the available RGB-D image data and limited experience. The paper presents our experiences with transferring the scooping framework with CoDeGa-trained model from a low-fidelity testbed to the high-fidelity OWLAT testbed. Additionally, it validates the method's performance in novel, realistic environments, and shares the lessons learned from deploying learning-based autonomy algorithms for space exploration. Experimental results from OWLAT substantiate the efficacy of CoDeGa in rapidly adapting to unfamiliar terrains and effectively making autonomous decisions under considerable domain shifts, thereby endorsing its potential utility in future extraterrestrial missions.

• The paper demonstrates that the CoDeGa model effectively adapts to unpredictable extraterrestrial terrains, enabling robust autonomous scooping.
• It employs deep Gaussian processes and meta-learning techniques to select optimal scooping actions from imagery data in a realistic testbed environment.
• Deployment on NASA's OWLAT confirms the model's transferability and adaptability, offering promising insights for future space exploration missions.

Introduction

Exploration in extraterrestrial environments, such as those on ocean worlds, presents unique technical challenges that necessitate high levels of autonomy in robotic landers. This paper details an experience report on the application of an advanced machine learning approach for terrain sampling autonomy. Such an autonomous system has been developed to tackle the task of scooping materials on planetary bodies within our solar system, which is essential for the scientific analysis of these alien worlds.

Adaptation to Unpredictable Terrains

The Deep Meta-Learning with Controlled Deployment Gaps (CoDeGa) model discussed in the paper leverages an innovative training approach using deep Gaussian processes. It is tailored to handle the unpredictability of extraterrestrial terrains. The CoDeGa model focuses particularly on scooping action selection, utilizing imagery data to inform decision-making. The transferability and adaptive nature of the model are assessed through its application to a sophisticated testbed, the Ocean Worlds Lander Autonomy Testbed (OWLAT), developed by NASA's Jet Propulsion Laboratory, offering a significant step up from earlier low-fidelity testbeds in terms of realism.

Deployment Experiences and Validation

The authors describe the employment of the aforementioned model on OWLAT and outline their findings in regard to model transferability across different fidelity testbeds, adaptability to novel environments, and the salient points of deploying AI algorithms in space exploration contexts. Experiments conducted with OWLAT confirmed the model's ability to adapt to substantial environmental variations, suggesting promising implications for its use in future space missions.

Conclusion and Implications

The success of deploying CoDeGa in the OWLAT high-fidelity testbed demonstrates the practicality of utilizing AI and machine learning algorithms in physical robotic systems for space exploration. Its performance affirms that learning-based approaches can be transferred from controlled settings to real-world-like environments, providing a pathway towards enhancing autonomy and efficiency of sample collection processes in future extraterrestrial missions. The deployment also underscores the value of machine learning techniques in dynamically dealing with uncertainties and domain shifts that are inherent in outer space exploration.