Developing a Low Altitude Manned Encounter Model Using ADS-B Observations

With the integration of unmanned aircraft systems into the U.S. National Airspace System, low altitude regions are being stressed in historically new ways. The FAA must understand and quantify the risk of UAS collision with manned aircraft during desired low altitude unmanned operations in order to produce regulations and standards. A key component of these risk assessments are statistical models of aircraft flight. Previous risk assessments used models for manned aircraft based primarily on Mode C-based secondary surveillance radar observations. However, these models have some important limitations when used at low altitude. We demonstrate a methodology for developing statistical models of low altitude manned flight or applicable at low altitudes that leverages the OpenSky Network, a crowdsourced ADS-B receiver network that provides open access to the aircraft data, and the FAA aircraft registry, an open database of registered aircraft. Unlike Mode C surveillance, a key advantage to this method is the availability of necessary metadata to distinguish between different types of low altitude aircraft. For example, previous models did not discriminate a large commercial aircraft transiting to higher altitudes from low altitude or small general aviation aircraft cruising at low altitudes. We use an aircraft's unique Mode S address to correlate ADS-B reports with aircraft type information from the FAA registry. We filter surveillance data and statistically characterize the low altitude airspace based on aircraft type, performance, and location. Lastly, we leverage the characterization and aircraft tracks to develop a Dynamic Bayesian Network that models the behavior of low altitude manned aircraft, an extension of previous aircraft modeling approaches that have employed Bayesian networks. By sampling representative trajectories from the Bayesian network, we can model encounters between manned and unmanned aircraft at low altitudes to assess collision risk, a key supporting technology to support safe integration of unmanned aircraft.