Search NASA⌕ Search
Home/Archive/NASA NTRS

NASA NTRS · 20210022218

Towards Autonomous Lunar Resource Excavation via Deep Reinforcement Learning

Joseph M. Cloud·Rolando J. Nieves·Adam K. Duke·Thomas J. Muller·Nashir A. Janmohamed·Bradley C. Buckles·Michael A. Dupuis
View original source ↗Export metadata ↓

Abstract

To support sustainable infrastructure on the Moon, NASA needs to leverage lunar resources for in-situ processing and construction. NASA’s Regolith Advanced Surface Systems Operations Robot (RASSOR) is principally designed to mine and deliver regolith for these tasks. To reliably perform these operations on the lunar surface, RASSOR's sensors and control systems need to be robust and maximize information extracted from a reduced sensor payload. Herein, we present our findings from the Intelligent Capabilities Enhanced RASSOR project. We created reduced-order simulation environments in which we applied reinforcement learning algorithms to learn autonomous trenching controllers and produced state estimation architectures. We developed two simulations: a 2D excavation simulation used to facilitate parameter selection, and a 3D simulation developed using a game physics engine to simulate simplified soil interactions and incorporate robotic agents parameterized by dynamic models. Within these simulations, we learned autonomous excavation routines that exceed excavation efficiency measures as compared against RASSOR's existing control and teleoperation-based methods.

Explore related subjects

RASSORREINFORCEMENT LEARNING

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Joseph M. Cloud, Rolando J. Nieves, Adam K. Duke, Thomas J. Muller, Nashir A. Janmohamed, Bradley C. Buckles, Michael A. Dupuis. Towards Autonomous Lunar Resource Excavation via Deep Reinforcement Learning. https://ntrs.nasa.gov/citations/20210022218

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

NASA NTRS2016

Design of an Excavation Robot: Regolith Advanced Surface Systems Operations Robot (RASSOR) 2.0

To continue on a sustainable and flexible path, NASA needs to address the challenge of collecting and moving large amounts of regolith at the destination. Acquiring the water resources on Mars will require mining significant quantities of regolith and this is not possible with the state-of-the-art low mass excavation systems. Low gravity environments (Mars = 3/8 G) and launch mass restrictions limit the traction and the resulting reaction force of the vehicle, making current terrestrial techniques impractical. This project addressed this challenge by developing a completely new technology that can mine large quantities of regolith on Mars. Recent measurements by the “Curiosity” rover on Mars have found that the regolith contains ~ 2% water by weight globally, ~4% in Jezero Crater (Human Architecture Team’s reference landing site), and much more at the poles(Leshin et al, 2013). RASSOR 2.0 is a planetary excavator, which has a mass of 66 kg, with a 0.38 kg vehicle mass per kilogram, per hour of excavation rate and power usage of 4 W per kg of regolith excavation rate. A single RASSOR 2.0 can excavate a minimum of 2.7 metric tons of regolith per day.This is accomplished by using counteracting excavation forces on two opposing digging implements called bucket drums and an autonomous mining control system. This work has addressed several major research areas outlined in the NASA Technology Area (TA) 04 Robotics & Autonomous Systems and TA 07 Human Destination Systems roadmaps. This project started at Technology Readiness Level (TRL) 4 as a low fidelity “proof of concept” prototype which has successfully demonstrated basic regolith simulant excavation functionality in a lab-scale gravity off load test. The foundational technology described here was awarded US patent number: US 9027265 for a “Zero horizontal reaction force excavator” on May 12, 2015.

RASSOR↗
NASA NTRSDate not supplied

Towards Autonomous Lunar Resource Excavation via Deep Reinforcement Learning

To support sustainable infrastructure on the Moon, NASA needs to leverage lunar resources for in-situ processing and construction. NASA’s Regolith Advanced Surface Systems Operations Robot (RASSOR) is principally designed to mine and deliver regolith for these tasks. To reliably perform these operations on the lunar surface, RASSOR's sensors and control systems need to be robust and maximize information extracted from a reduced sensor payload. Herein, we present our findings from the Intelligent Capabilities Enhanced RASSOR project. We created reduced-order simulation environments in which we applied reinforcement learning algorithms to learn autonomous trenching controllers and produced state estimation architectures. We developed two simulations: a 2D excavation simulation used to facilitate parameter selection, and a 3D simulation developed using a game physics engine to simulate simplified soil interactions and incorporate robotic agents parameterized by dynamic models. Within these simulations, we learned autonomous excavation routines that exceed excavation efficiency measures as compared against RASSOR's existing control and teleoperation-based methods.

RASSOR↗
NASA NTRSDate not supplied

ISRU Pilot Excavator: Bucket Drum Scaling Experimental Results

NASA’s Space Technology Mission Directorate (STMD) is funding the development of a robotic excavator called the “ISRU Pilot Excavator” (IPEx) which will be a technology demonstration of excavating and transporting 10 metric tons of lunar regolith on the surface of the moon with a 30kg-class robotic excavator. IPEx will be the next generation of robotic excavators to use bucket drums as excavation tools. This is an evolution of the Regolith Advanced Surface Systems Operations Robot (RASSOR) developed at NASA’s Kennedy Space Center (KSC). Bucket drums are hollow cylinders with regularly spaced scoops around the perimeter. The drums rotate in one direction to collect regolith with the scoops. The regolith slides down an internal baffling system inside the drum which prevents the regolith from falling back out of the scoops (see Figure 1). The captured regolith can then be transported while held in the drum and then deposited by rotating the drum in the opposite direction allowing the regolith to slide back down the baffling and out of the excavation scoops. Bucket drums were developed by Lockheed Martin in 2008 and used on multiple robotic excavator prototypes ever since. However, the forces on a bucket drum and considerations for scaling have not been measured in detail. Bucket drums are challenging to model using classical blade\bucket equations because of their unique geometry. Therefore, this experiment was performed to measure the forces on three bucket drums of the same geometry at different scales. Small: 9.4” (239mm) dia. x 8.1” (206mm) width, Medium: 11.6” (294mm) dia. x 10” (254mm) width, and Large: 17” (432mm) dia. x 14.1” (358mm) width. The test stand consisted of an actuated gantry with controlled motion in the vertical (Z) and horizontal (X) axes and a single rotation axis (R). The bucket drums were individually mounted to the rotary axis of the test stand and translated across a prepared bed of BP-1 lunar regolith simulant at a specified linear speed and cutting depth. The test stand was outfitted with a torque sensor in line with the rotation of the drum (R) and a 3 axis (X, Y, and Z) load cell. In addition to the three sizes of bucket drums the linear excavation speed and cutting depth were test variables. The results of these experiments show the relationship between the three scales of bucket drums for factors such as: excavation force, torque due to regolith rotation inside the drum, excavation energy, time to fill, etc. and will be discussed in detail in this paper. This fundamental data will be used in the design of IPEx and can inform the design of future bucket drum excavators.

RASSOR↗