A Nonlinear Finite-Memory Filter with Application to Tracking

This paper considers the estimation based on nonlinear measurements of a state characterized by a model valid for short intervals of time when the components of the state are related by a nonlinear feasibility constraint. An estimation algorithm is formulated which takes into account the exact nonlinearity in the measurement and constraint equation without linearization. The error of this filter is shown under certain assumptions to be bounded in time, in contrast to the divergent behavior which sometimes plagues linearized estimators. The proposed method is applied to the tracking of maneuvering aircraft based on acoustic measurements and its performance evaluated relative to that of a linearized finite-memory filter by means of Monte Carlo simulations.