Virtual Verification Test Bed for Robust Autonomous Software Operation in Complex, Unknown Environments

Unmanned Autonomous Systems (UAS) are complex systems that exhibit a broad array of behaviors in a wide range of missions and environments. They can change their behaviors over time so the UAS rigorously tested at the range is not the same as the one operating in the field. Current testing methods are inadequate to verify that they are safe to operate in close proximity with humans. Testing Robustness of UAS Technology (TRUST) is an intelligent testing system that rapidly probes, characterizes, and assesses the safety boundaries of the complex autonomous controllers for UAS. TRUST couples the knowledge of the human test expert with an optimizing test search algorithm in a constructive simulation environment to rapidly characterize the failure modes of a safety-bounded complex controller. TRUST co-evolves with the system under test: as it improves its performance, TRUST finds more challenging tests to probe its evolving safety boundaries. TRUST incorporates three innovations: an intelligent test agent based on a Soar cognitive model, a variety of state space management techniques, and an intelligent search agent optimized for finding safety violations. Together, these technologies ensure that UAS can be thoroughly tested and verified for safe operations under a wide range of conditions.