A comparison of track to truth assignment methods

Evaluating the performance of a radar target tracking simulation architecture typically requires creating a mapping between between the tracks generated by that architecture and the true trajectories of the objects in the simulation. The mapping is commonly generated by the solution to a 2D assignment problem in which the costs are based on some measure of the distance between tracks and truth. In the past, distances such as the Euclidean and Mahalanobis distances have been used, and, more recently, the contribution of truth objects to tracks has been used to derive a distance from the track content, which can be more robust in the presence of unresolved targets. This paper describes and compares various methods of defining assignment costs and explores the differences in tracking effectiveness and performance metrics that arise from using them in the context of closely spaced air targets.