Appendix F (CISTO): Fault Diagnosis for Safety-Critical Autonomous Systems using Reinforcement Learning

The overarching objective of the proposed research is to create and validate a principled and general framework for fault diagnosis of an autonomous system operating in a high-dimensional state/action space. The proposed research will frame the fault diagnosis problem of an autonomous system as a sequential decision-making problem in a reinforcement learning setting where the ultimate goal is to find a policy that maps states to those disturbances causing the system to fail. To identify the most likely failure paths in a simulated environment, the fundamental method developed in this research will be applied to a small satellite in a simulated suite developed by NASA. If successful, the proposed research will accelerate the safety verification of complex autonomous systems that minimizes the number of required computationally expensive validation experiments, while reliably reveals the most likely failure events.

The proposed research integrates with educational and training activities with a goal of exposing underrepresented communities in the mid-Atlantic region to the problems important to NASA and facilitate future collaborations with NASA scientists. The participants will learn about the proposed solution of reinforcement learning and explore novel solutions to the problem-space. We will use webinars and Hackweek events to introduce and engage the underrepresented communities to the problem of Fault Diagnosis in Autonomous system, thereby increasing their competitiveness for future NASA opportunities.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.