Engineering Notes Probabilistic Safety Assessment of Unmanned Aerial System Operations

NMANNED aircraft systems (UASs) are becoming increas-inglypopular,encompassingawidevarietyofmissionsrangingfrommilitaryreconnaissancetowildfiremonitoring.However,thereareinherentsafety concernswithUASduetothelackof anonboardhuman pilot. Currently, to operate UAS in the National AirspaceSystem, the operators must obtain either a Certificate of Authoriza-tion or Waiver or a special airworthiness certificate from the FederalAviation Administration (FAA) [1]. One of the important steps inobtaining the approval is a proof that the UAS operation can beconducted at an acceptable level of safety [1].Many past studies assessing the safety of UAS operations useduniform traffic densities. Anno [2] investigated midair collision riskusing a random collision theory and compared the results withhistoric collision data from 1969 to 1978. McGeer et al. [3,4]performedhazardestimationstudiesoftheAerosondeUAS.Inthesestudies, two different constant densities were used for the UAS andthe background traffic. A comprehensive system-wide study per-formedbyWeibelandHansman[5]usedaratioofthevolumesweptby the background aircraft to the total airspace volume. Lum andWaggoner[6]conductedastudyonbothmidaircollisionandgroundimpact based on the collision model of gas molecules. Theseapproaches are adequate for obtaining a general idea of the riskaround a given region but do not consider traffic patterns that arespecific to the region of interest.Lum et al. [7] used actual UAS trajectories for a ground impactanalysis. In this study, a realistic distribution of average glide anglewasusedtocalculatetheexpectedvalueofgroundfatalities.Sheridan[8] proposed a model to estimate the relative collision probabilitybetween two aircraft at the closest point of approach based onGaussian density functions. Maki et al. [9] created a method toefficiently estimate the probability of near midair collision usingGaussian probability distributions of proposed UAS trajectories andhistorictrackdata.Inthiswork,thenearmidaircollisionprobabilitiesare expressed as confidence intervals.Some of the work is related to quantitatively establishing theboundary of “well clear” for sense-and-avoid systems. Weibel et al.[10] used conditional probability to develop a separation standardmodelbasedonuncorrelatedencountermodel[11].Asmatetal.[12]developed a UAS-specific collision-avoidance system that cancommunicate with the existing traffic alert collision and avoidancesystem.Inthiswork,adistributedtrafficmodelsimilartoMakietal.[9]isconstructedusingactualtrafficdatacollectedoveraone-yearperiodto enable a probabilistic approach to risk assessment. The radar dataprovided by the U.S. Air Force contains not only the cooperativetraffic data but also the noncooperative traffic data with altitudeinformation. Inclusion of noncooperative traffic, mostly generalaviation (GA) traffic, is important because they tend to fly at loweraltitudes where the UAS are likely to operate, and it is harder toimplement collision mitigation measures with them. The currentstudy computes the collision rates, which are defined by the numberof collisions per unit time of UAS operation, based on UAS tracksflying through the continuous background traffic model. Theprocedures and results are explained in detail throughout thefollowing sections.Following the introduction, the area around the Grand Forks AirForce Base where the U.S. Air Force is planning to operate UAS isdescribed in Sec. II. Then, the description of the continuous-trafficmodelispresentedinSec.III.InSec.IV,mathematicalformulationsfor the continuous-traffic model and for the computation of conflictand collision probabilities are presented. Section V reviews the airtraffic characteristics of the given area in terms of average aircraftcounts and their spatial distributions, and Sec. VI presents thecollision risk computed for a potential mission scenario. Finally, theresults and recommendations are summarized in Sec. VII.