Tracking Algorithms for Air Traffic Control Applications

Abstract

This article presents an analysis of algorithms for central tracking in air traffic control systems. Three algorithms were examined: an adaptive alpha/beta filter representative of systems now in use; an adaptive single model Kalman filter; and an interacting multiple model Kalman filter. The interacting multiple model filter included the turn rate as an additional filter state and estimated it as part of an extended Kalman filter. Average performance for the three filters was evaluated over a collection of scenarios simulating civilian and military targets with single-and two-sensor coverage and a number of ranges and target aspects. Existing radar systems and future Mode-S sensors were modeled. With the exception of straight-line targets, there was a dramatic performance difference between the Kalman filters and the alpha/beta algorithm. In tests that simulated tracking of high speed maneuvering targets with fused Mode-S data from two radar systems, the Kalman filters reduced the root mean square position error by more than an order of magnitude relative to the alpha/beta filter. The greatest difference between the two Kalman filters was in the magnitude of their peak errors during turns. In these cases the interacting multiple model filter typically yielded a 30% reduction in peak error relative to the adaptive single model filter. We also assessed the relative computational complexity of the algorithms. The alpha/beta filter required approximately 700 floating point operations per update. Relative to this, the single model Kalman filter required a two fold increase in computation and the multiple model filter required an eight fold increase. This is easily supported by modern personal computers. © 1996 John Wiley & Sons, Inc.