Scalable physics-aware runtime assurance for autonomy

Runtime Assurance (RTA) is a  technology for enabling rapid experimentation and  prototyping of advanced control algorithms in safety-critical autonomous systems. The idea, is to keep  a verified controller as a backup option, while an unverified experimental controller is tested, refined, and developed. Feasibility of RTA systems has been established in a number of recent experiments with aircraft engine control, unmanned aerial vehicles (UAV), and multi-agent systems. The key challenge, as identified in the solicitation, is to develop the switching logic that avoids false alarms. That is, RTA system should not switch from the experimental to the verified controller, unnecessarily, as such switching limits the benefits of experimentation, and the potential performance benefits of the unverified controller. In this Phase 1 project, we will develop and evaluate a scalable, physics-aware software tool for generating the switching logic that resolves this issue.