Characterization of Tracker Performance Evaluation Algorithms for Radar Model Validation and Accreditation Using Monte Carlo Data Set Ensembles

摘要

Validation and accreditation of tracker algorithm performance in radar models is complicated by the fact that tracker output data are intrinsically correlated. Recently tracker validation algorithms have been developed by MIT Lincoln Laboratory (MIT/LL) and the Ballistic Missile Defense System(BMDS) OTA . These algorithms are tested for self consistency by verifying their probability characteristics against statistics derived from ensembles of Monte Carlo generated data. These studies show that the MIT/LL technique provides an estimate of the tracker correlation time t with a 90% confidence level error of +/-50%. The self consistency tests show that these algorithms can provide results with quantitatively meaningful confidence levels. The sensitivity of the tests to systematic errors is discussed in the results presented below.