Module for Event Driven Operations on Spacecraft (MEDOS) Expanded for Multi-Agent Decision Making and Teaming

Operations engineers face the task of maintaining the health and functionality of a spacecraft, based on extremely limited information. This task becomes even more complicated when considering lag time between contacts, the light-time delay of bidirectional communications, and the prospect of multiple spacecraft operating simultaneously. Simple pre-programmed time tagged commands become increasingly burdensome in multi-spacecraft missions, particularly when trying to target ephemeral events in space that are difficult to predict with any accuracy. The Module for Event Driven Operations on Spacecraft (MEDOS) puts the operations engineer on the spacecraft. MEDOS takes raw telemetry and fuses them together into physically meaningful quantities that an engineer understands turning voltages and decay times into methane concentration, or turning count rates and field vectors into plasma density. Then, MEDOS compares the physical parameters to generalized event definitions that scientists use all the time when the plasma density jumps and the magnetic field reverses direction, we crossed a boundary. MEDOS computes a mathematical distance between measured parameters and rules from engineer to provide a numerical confidence in an event occurring. MEDOS does not require any training data, discrete if/then rules, or a PhD in AI to tune the model. All that MEDOS requires is for an engineer or scientist to describe the process that they go through every day and then MEDOS encodes that process onto the spacecraft. MEDOS is able to analyze the ambient environment in real time, and respond to it in a way consistent with the response of an operations engineer. In Phase II we will expand MEDOSs capability for multi-agent teaming with a new module, MADEM (Multi Agent Decision Engine for MEDOS.) We will also port the entire MEDOS framework into NASAs core Flight System (cFS) and verify + validate MEDOS on flight qualified hardware, readying it for mission deployment. MEDOS proposes to put an operations engineer on the spacecraft. MEDOS takes raw data and fuses them together into physically meaningful quantities that an engineer understands – voltages and decay rates into “methane concentration”, count rates and field vectors into “plasma density”. Then, MEDOS compares the physical parameters to generalized event definitions that experts use in their analysis – “when the plasma density jumps and the magnetic field reverses direction, it means we crossed a boundary.” MEDOS computes a mathematical distance between the measured parameters and the fuzzy rules set out by the engineer to provide a numerical confidence in the likelihood that an event is occurring, leveraging knowledge received from other spacecraft and, with the development of MADEM, allows a heterogenous team to work together. MEDOS does not require any training data, discrete if/then rules, or a PhD in AI to tune a model. All that MEDOS requires is for an engineer to describe the process that they go through every day – and then MEDOS encodes that process onto the spacecraft. Major Technical Objectives: Improvements to MEDOS (Single Agent) Implement Improved Feature Extraction, Event Detection, and Uncertainty Quantification Introduce Multi-Spacecraft Telemetry in MEDOS Implement MEDOS in cFS Testing in Flight-Like Environment Configure Telemetry Feeder to Interface With MMS Flatsat and EM-CIDP Develop Complex Science Signature Synthetic Mission Testing on MMS Flatsat EM-CIDP Implement Multi-Agent Decision Engine for MEDOS (MADEM) Design Policy Interpreters/Problem Definitions Design Robotic Teaming Environment in CAIRO Lab Perform Software Simulations of MADEM Perform Robotic Testing Validation of MADEM Reporting / Task Closeout / Margin Deliverables: MEDOS Software Python Port cFS Port with MADEM implementation Updated Telemetry Feeder Data and Logs from Verification and Validation Testing Final Report
Benefits: MEDOS has secured a spot to fly on the SCENIC platform on the ISS (PI C. Wilson), and has been selected to fly on the NAMASTE effort, detecting methane from melting permafrost in Alaska (PI M. Sultana), pending Phase II completion. We have had communication with the MMS project scientist (G. Le) about infusing MEDOS into the MMS mission pending Phase II. We are also in contact with teams in earlier mission phases (GDC, 7 Sisters, Enceladus). We are also working to complement AutoNGC with MEDOS. Outside of NASA we aim to target other space players, such as NOAA and DoD, particularly for detection of and response to surface events (fires, missile launches, etc.) We will also approach the private space industry (e.g. Starlink and Iridium.) Beyond space, underwater submersibles (such as the Titan submarine) as well as mining and resource extraction system could greatly benefit from MEDOS.