Intelligent Simulation-based Test Methods for Certification of Advanced Planning Algorithms

The robustness and operational flexibility of UAS will be significantly enhanced through the integration of autonomous path-planning algorithms that modify system behavior based on changing mission requirements and environmental factors. Many such algorithms have been developed, but the certification of these algorithms typically involves brute-force testing in high-fidelity simulation, which is a significant challenge that impedes widespread adoption in operational UAS. The overall goal of the Phase I research effort is to develop a methodology for certification of autonomous path-planning algorithms for UAS that significantly reduces the computational burden relative to brute-force testing. The proposed approach is based on the application of advanced analysis methods to computationally optimized simulation models to gain insight into combinations of inputs and test parameters that may lead to undesirable behavior of the system being evaluated. These results are used to intelligently guide simulation-based testing that uses the full high-fidelity simulation. The team will show the applicability of the test methodology to two different classes of planning algorithm: a deterministic heuristic search-based algorithm and a non-deterministic random sample-based algorithm. The results will be used to show that analysts can draw similar conclusions about the airworthiness of a planning algorithm as an exhaustive brute-force test.