Discovering Probabilistic Models of Pilot Behavior from Aircraft Telemetry Data

摘要: A mobile ground-based sense-and-avoid system for Unmanned Aircraft System (UAS) operations was developed by the University of North Dakota. This detected proximate aircraft with various sensor systems, including a 2D radar and an Automatic Dependent Surveillance – Broadcast (ADS-B) receiver. Information about those then displayed to UAS operators customized visualization software. Its risk mitigation subsystem designed estimate current midair collision below 18,000 feet MSL. However, accurate probabilistic models behavior pilots manned flying at these altitudes were needed before this could be implemented. In paper authors present results data mining Flight Data Monitoring (FDM) set from consecutive 9 month period in 2011. Arbitrarily complex subpaths discovered using ant colony algorithm. Then, mined extensions Genetic K-Means Expectation-Maximization algorithms.