Higher-order Kalman filter to support fast target tracking in a multi-function radar system

Abstract

State-of-the-art multi-function radars to support area defence systems are currently under development-these allow steered searchlight-type looks to be performed in 3D and will provide the capability to track targets via both surveillance (search) looks and, additionally, looks dedicated to a specific track. Central to the derivation of this tracking capability is the development of a fast target track smoothing algorithm within these systems-this smoother will utilise associating plot outputs to update existing fast target tracks. As well as supporting weapon systems functions, the track data is also used as the basis for scheduling future looks, thus closing the loop in the radar system. A conventional approach to the tracking of fast targets is to utilise a Kalman filter using a system model based on the assumption of constant target velocity. However, the increased manoeuvrability of projected threats compromises the fundamental assumption of constant velocity, thereby limiting the potential performance of such an approach. A candidate solution is to utilise a higher order filter (HOF) which specifically embraces the possibility of target manoeuvre. While this addresses the problem, the formulation of the filter must necessarily include increased uncertainty in the assumed target behaviour. This results in larger covariances being attributed to the estimated position and velocity of tracked objects, resulting in reduced accuracy in the derived estimates of target heading in the important case of straight line target motion. A hybrid solution is possible, where a constant velocity filter is adopted as the baseline solution and a switching regime selects the HOF only when necessary. Such an approach retains all the benefits of the constant velocity approach in the regimes where its performance advantages occur. This paper describes a candidate switching approach, together with the structure of the selected second order filter, and illustrates the comparative performance against a conventional constant velocity filter approach.