Development of key performance indicators for trajectory prediction accuracy

Key Performance Indicators (KPI) for trajectory prediction accuracy were developed by applying factor analysis to a wider set of accuracy metrics obtained from a literature search. A Monte-Carlo simulation was conducted under operationally-representative conditions to provide a data set for the analysis. It is shown that the derived KPI can be linearly combined to estimate the larger set of metrics. These estimates provide good rank correlation with the actual metrics computed. KPIs can describe both the accuracy of trajectory prediction in addition to the quality of the input data supplied to a trajectory predictor. Various applications of these KPI are discussed including the specification of requirements on prediction performance. While certain KPI are described in this study, various values could have been selected. It is argued that TP KPIs should be made consistent with measurements used to express vertical and longitudinal RNP as those get defined.