Performance Metrics for Tactical Aircraft to Aircraft Conflicts

Abstract

An air traffic control system's main function is to separate aircraft. The computer supporting the system assists the air traffic controllers by generating a conflict alert whenever it predicts that two aircraft are about to get too close to each other. The performance of the conflict alert function is a key element to the overall functioning of the air traffic control system. A set of metrics has been designed to measure the conflict alerting performance of an aircraft traffic control system. The key factors are the missed alert rate and the false alert rate. However there are several secondary factors that are essential to measuring the performance, especially in a simulation environment. This paper describes a set of metrics that have been developed to evaluate the performance of the FAA's en route aircraft traffic control system. They have been applied to the existing system, the Host Computer System, and will be used to establish metrics for the new system now under development, the En Route Automation Modernization system. The metrics are calculated by post processing recorded data that has been produced by running a real time simulation of the air traffic system without controllers, using as input field recorded aircraft data that has been time shifted to induce aircraft-to-aircraft conflicts.

Assessing tactical alert function accuracy performance

“…This approach relies on the use of a linear predictor operating on truth data to check the validity and timeliness of an alert. A similar approach is described in [7] and [8], using however smooth and interpolated track outputs instead of truth data.…”

“…For 3 The term "valid-late" has also been applied to such alerts [8] instance, ERAM had requirements to be able to perform surveillance processing over an area as large as 2000 nmi by 2000 nmi with a specified mix of short and long range radars providing multiple radar coverage in some areas and single radar coverage in others.…”

“…6 In ERAM Release 1 formal test, a linear predictor was used as the Test Predictor. Similar approaches to validate the issuance of an alert for non-conflict near LoS situations also make use of a linear predictor [7,8,10]. In ERAM formal and regression test, the velocity is computed using the central difference method from "ground truth" position points in the past up to the evaluation time.…”

“…Such a definition avoids the need to define what constitutes a 'nuisance opportunity' or near encounter. This was presented as one of two False Alert rate definitions in references [7,8].…”

“…5 [15]. This definition has been used in analyses of conflict probe [25] and was one of two metrics presented in [7,8]. An alternative approach to defining nuisance opportunities has been used at NASA Ames in research on TSAFE [26] whereby a near encounter is defined as a situation where the 1 minute dead-reckoning (based on track data but alternatively could use truth data when available) projects the aircraft to get within 6 nmi and 1200 ft, except if the aircraft are flying level and in conformance to their cleared and separated altitudes.…”