Safety ALgorithm Test System (SALTS)

Unmanned Autonomous Systems are required to operate in cluttered, unknown, and unstructured environments. Guaranteeing the safety of these systems is critical for their successful deployment. Autonomous vehicles often need to operate near their limits to effectively complete their tasks. Safety of these vehicles needs to be imposed while respecting the constraints imposed by sensory and dynamic limitations of the vehicle. Current state-of-the-art certification approaches for evaluating the route planners in the presence of uncertainty typically involve large Monte Carlo simulations, and the use of subject matter experts to tailor the scope of unbounded testing (in order to fit into a given computing footprint or project timeline). We propose to develop Safety ALgorithm Test System (SALTS) that verifies safety and evaluates the robustness of in-flight algorithms (e.g. learning-based trajectory generation algorithms) in the presence of uncertainty without the need of exhaustive testing. The notion of MTL robustness is a valuable tool for testing and certification since it allows us to convert a decision problem into an optimization problem. In other words, an optimization algorithm can now be utilized in order to search for system trajectories that minimize the robustness of the system instead of exhaustive and time consuming Monte Carlo simulations.