The objective of this effort is to develop a real-time seismic detection technology using machine learning signal change detection for both lander-side decision-making for telemetry prioritization and rapid assessment of incoming streamed data on Earth. The technology (CORGI, short for COntinous Realtime monitorinG Infrastructure) will be modular, able to integrate diverse detection algorithms, instrument types, and data resolution. This technology is most applicable to upcoming or proposed lunar seismic missions, specifically Farside Seismic Suite (FSS, for CLPS), Lunar Environment Monitoring Station (LEMS, for Artemis 3), and Lunar Geophysical Network, but has significant potential for all scientific investigations which require automated detection and extraction of rare events within continuous data.

This project is a software development task to build code that detects rare signals in noisy continuous data. Although we will be approaching this task from the domain of seismology for this proposal, CORGI will be applicable to other data domains. This code will enable two new capabilities: (1) instantaneous assessment of rare events on streamed data back to Earth and (2) on-spacecraft decision-making for telemetry prioritization. Two versions of the code will be produced, an “unlocked” mode for Earth-side assessment of received telemetry data, and a “restricted” mode operating under power and processing limitations for spacecraft implementation.

Capability 1 (short-term benefit): Earth-side continuous data assessment

Earth seismic networks use real-time algorithms to assess streamed data. The benefit of this system is the immediate notification of interesting data, which can include the warning of hazardous seismic events. Such a system could be just as valuable for planetary missions, where change detection can provide a time-sensitive notification on an interesting scientific occurrence or a warning on an error within the instrument or spacecraft. Our proposed technology will be able to assess changes within continuous telemetry and alert for both scientific value and hazard occurrences. This capability would benefit in the short-term (1-2 years after development) by immediately assessing data from FSS and LEMS for scientific publications and ROSES proposals.

Capability 2 (long-term benefit): On-spacecraft decision-making

Research in planetary seismology is fundamentally limited by the power required to transfer high-resolution data back to Earth (5 mJ/bit/AU). Quakes are rare events, meaning that most of the continuous data sent back to Earth does not contain useful signals. By enabling triggered detection for segmented telemetry, spacecraft could record higher resolution data or even run more sensors and experiments, a capability that NASA does not currently have. Recent work by PI Civilini has shown that machine learning seismic detection techniques are able to be lightweight and flexible, but their applicability within a realistic framework has not been explored. The long-term benefit of this work would be an increase in mission capability.