P084-07
Imaging and Data Processing Autonomy Techniques for Lunar and Planetary Payloads and Missions

Wednesday, 16 December 2020: 07:18
Virtual
Kaizad Raimalwala1, Michele Faragalli1, Melissa Battler1, Ewan Reid1, Evan Smal1, Edward Cloutis2, Gordon Osinski3, Ryan Ewing4, Krzysztof Skonieczny5 and Kenneth McIsaac6, (1)Mission Control Space Services, Ottawa, ON, Canada, (2)University of Winnipeg, Department of Geography, Winnipeg, MB, Canada, (3)University of Western Ontario, Earth Sciences, London, ON, Canada, (4)Texas A&M, Department of Geology and Geophysics, College Station, TX, United States, (5)Concordia University, Montreal, Canada, (6)University of Western Ontario, Electrical and Computer Engineering, London, Canada
Abstract:
Visual planetary surface characterization is a common first step in many scientific investigations. However, this can be a time-consuming activity in a constrained operations cycle. To help automate this in planetary rover missions, Mission Control is developing a compact and advanced computing payload called ASAS-CRATERS (Autonomous Soil Assessment System: Contextualizing Rocks, Anomalies and Terrains in Exploratory Robotic Science).

This system can benefit a wide range of science investigations. It comprises cutting-edge algorithms for terrain classification and novelty detection using convolutional neural networks, and for data aggregation to produce relevant data products for supporting science operations.

Automated annotation of common terrain classes can help speed up scientific terrain assessment in short-duration decision-making cycles for lunar missions. Novelty detection can aid scientists that may miss valuable and rare features like bedrock outcrop and pyroclasts. For high-priority features, it can inform algorithms onboard for autonomous instrument targeting, and prioritized downlink. It can also inform path planning algorithms to enable autonomous and intelligent navigation.

These algorithms will be embedded on a compact, low-cost, and high-performance COTS processor with flight heritage, enabling relatively rapid integration onto lunar rovers and landers. ASAS-CRATERS is currently at TRL 4 and is being prototyped with relevant lunar and lunar analogue datasets. The terrain classifier was previously developed in the context of Mars rover operations and field-tested to TRL 5 at a high-fidelity Mars analogue site in Iceland. ASAS-CRATERS is currently planned to be demonstrated on upcoming lunar rover missions in 2022 and 2023.

As an advanced image analysis system, this technology can address a wide range of challenges beyond lunar rover exploration. Close encounters with asteroids and comets can be transient events that can benefit from onboard imaging autonomy to ensure that targets are quickly identified, instruments targeted, and prioritized data is downlinked. Satellites around dynamic bodies such as Mars or Io can benefit from continuous ‘change detection’ technology which is similar to the novelty detection technology being developed in this project.

<< Previous Abstract | Next Abstract >>