Multi-Agent Anomaly Resolution System (MaARS)

We propose the Multi-Agent Anomaly Resolution System (MaARS) for Automated Fault Detection, Diagnostics and Recovery (AFDD-R). Of specific interest is how MaARS can use robotic agents to support the operation amp; maintenance of habitats on lunar or planetary surfaces or in orbital stations such as Gateway when no human crew is present. Managing such remote robotic agents and spacecraft in the presence of severe communication constraints and limited situational awareness presents a number of challenges---particularly when unexpected faults may require immediate detection and response. To address these challenges, MaARS combines TRACLabs#39; PHARAOH system for integrating autonomous and semi-autonomous robotic assets into remote mission processes with the Texas Aamp;M University DAPHNE-Anomaly Treatment system. The resulting MaARS system provides an end-to-end solution that can monitor telemetry and perform trend analysis to identify anomalies, provide a recommendation on a resolution based on a knowledge graph relating likely root causes to anomalies, and leverage robotic assets to remedy the likely faults. In our Phase I effort, we determined the requirements for MaARS and devised a suitable Concept of Operations. nbsp;We developed a candidate proof-of-concept implementation that allowed Daphne to invoke and monitor PHARAOH procedures to support diagnostics---which in turn coordinated the activity of a simulated Astrobee robotic agent to perform a number of information gathering and inspection activities. nbsp;We also identified a number of dimensions in which the Daphne tool-chain could be extended to engage robotic assets to perform autonomous iterative diagnoses until a confident root cause analysis is made, and to perform repair and recovery tasks using a library of PHARAOH procedure-based skills. nbsp;In the Phase II work, we will develop the full MaARS prototype and demonstrate the system#39;s capabilities in a number of NASA-relevant contexts.