RTAEval: A framework for evaluating runtime assurance logic

Abstract: Runtime assurance (RTA) addresses the problem of keeping an autonomous system safe while using an untrusted (or experimental) controller. This can be done via logic that explicitly switches between the untrusted controller and a safety controller, or logic that filters the input provided by the untrusted controller. While several tools implement specific instances of RTAs, there is currently no framework for evaluating different approaches. Given the importance of the RTA problem in building safe autonomous systems, an evaluation tool is needed. In this paper, we present the RTAEval framework as a low code framework that can be used to quickly evaluate different RTA logics for different types of agents in a variety of scenarios. RTAEval is designed to quickly create scenarios, run different RTA logics, and collect data that can be used to evaluate and visualize performance. In this paper, we describe different components of RTAEval and show how it can be used to create and evaluate scenarios involving multiple aircraft models.