Mars Exploration Rovers automatic onboard detection of dust devils.

Onboard Automated Science Investigation System (OASIS) enables prioritization of rocks or surface features according to their scientific importance.

HARVIST (Heterogeneous Agricultural Research via Interactive, Scalable Technology) is a toolkit to enable interactive analysis of relationships between multiple, potentially global, science data products, and efficient testing of competing scientific hypotheses.

The mission of Machine Learning and Instrument Autonomy is to conduct research in knowledge discovery, data mining, and machine learning, developing innovative information technology and machine intelligence to enable and facilitate ground-based and onboard data analysis. In ground based data mining we develop methods for extracting information from large, diverse data sets for which manual analysis is infeasible. We also develop methods for onboard data analysis in constrained computing environments with limited downlink bandwidth, providing methods for detecting rare or dynamic events and prioritizing and summarizing data for downlink. The group has an emphasis on developing and infusing practical solutions to challenging problems. Examples of recent successful infusion activities include a cloud mask product for the Multi-angle Imaging SpectroRadiomter (MISR), dust devil and cloud detectors on the Mars Exploration Rovers, and a land-cover classifier and sulfur detector onboard the EO-1 spacecraft.

Our core technology capabilities include feature selection, feature extraction, event detection, classification, clustering, regression, transfer learning, active learning, data mining, and data fusion.