Airspace regulatory agencies are currently focusing on risk assessment frameworks for integrating the operation of Unmanned Aerial Systems (UAS) into National Air Space (NAS). Multiple frameworks, such as the Specific Operations Risk Assessment (SORA) framework for the European Union and similar frameworks for the US, provide defined pathways to evaluate the risk and seek approval for UAS operations. These frameworks are primarily qualitative and are sufficiently flexible to incorporate quantitative approaches, many of which have been proposed and tested in literature. Most proposed quantitative methods are still under development. Likewise, real-time analysis methods, designed to provide decision-making to unmanned systems during operations, have been proposed. Current real-time analysis methods still suffer from limitations, such as only applying to specific operations. This research applies Dempster-Shafer theory and valuation networks, a framework for reasoning with uncertainty used extensively for risk analysis, to UAS risk analysis by creating extensions which allow this framework to learn risk relationships in the UAS ecosystem based on operational results and enable this framework to be used in real-time analysis onboard small UAS. These extensions are applied to an autonomous car scenario for testing the capabilities against known baselines, then applied to the UAS scenario for testing in simulation against a previously implemented real-time health monitoring system. Finally, these extensions are demonstrated in flight on a small UAS. Application to the UAS ecosystem and conclusions are addressed based on the results of these tests.
PDF
download
878987797
028-85220240
