Datamining as a foundation for science-enabling autonomy

In situ exploration by spacecraft and planetary rovers will increasingly require knowledge "on demand" in the future as downlink constraints limit the amount of information that can be transmitted from these platforms back to earth. Several onboard processing methods have the potential to significantly enhance scientific results in these settings. They include automatic detection of natural satellites of planetary bodies, investigation of possible surface motions on planets and planetary moons, and directed acquisition of scientific data by planetary rovers. The key ingredient in all three cases is the need to process scientific data directly onboard, so that information can be rapidly provided to an automated spacecraft executive and/or to ground-based Principal Investigators (PI's). We discuss here recent developments in data mining technology that were designed initially for ground-based scientific data analysis. We then outline how these ideas can be migrated to onboard platforms to dramatically enhance the scientific capabilities of autonomous spacecraft.