Estimation of kinematic attributes using passive sensors is a key problem in air and missile defense as well as anti-submarine warfare and airborne surveillance. Traditional deterministic observability criteria are based on engagement and observation kinematics, and imply infinite tracking times and noise-free measurements. Real systems have noise corrupted measurements and finite tracking times or finite numbers of measurements. These "real world" constraints may be accommodated using information-theoretic concepts. Thus criteria may be formulated to ensure the necessary and sufficient conditions for unique estimates. These criteria are based on metrics which take into account noisy and finite numbers of measurements, and provide a basis for a definition of "stochastic observability". The intent of this paper is to introduce the deterministic criteria, develop a modification using information-theoretic distance measures to form a stochastic observability criteria, and finally to illustrate this development via simple examples.
IntroductionEstimation of a target trajectory using angle-only measurements is a critical problem in satellite tracking, ballistic missile defense, homing missile guidance, antisubmarine warfare, and airborne surveillance. The key to solving this problem, in general, and for solutions to these diverse applications, in particular, is defining target observability criteria: the necessary and sufficient conditions to ensure a unique estimate of the target trajectory given a finite set of noise-corrupted observations. The intent of this paper is to apply the information-theoretic observability measure previously developed in [1].The earliest examples of trajectory estimation from angle-only measurements are the efforts of the Greeks (e.g., Ptolemy) to develop their models of planetary motion; these efforts demonstrate the impossibility of long-term trajectory prediction and estimation when the underlying dynamics models are wrong. Probably the modern era of mathematically-sound trajectory estimation based on a limited set of