Robust Estimation for Target Tracking

In this SBIR proposal, a variety of new and novel innovations are proposed for the target-tracking estimation problem involving large and rapidly changing target accelerations. An essential feature of the tracking problem is that the radar tracker measurements are best modeled in spherical coordinates and the target dynamics are best described in rectangular coordinates. A class of filters that produce significant improvements are tracking filters using Extended Kalman Filters (EKF) includes the Modified Gain EKF (MGEKF) and the pseudo-measurement filter. For both these filters, the estimation error propagates almost linearly. Besides proposing the MGEKF, new target models are suggested based on the assumption that certain targets execute evasive maneuvers orthogonal to their velocity vector. Furthermore, this orthogonality is also enforced by the addition of fictitious pseudo auxiliary measurements. Finally, a robust form of the MGEKF is developed based upon the dissipative inequality. The robust game-theoretic MGEKF requires adjustment by use of tuning parameters that win be adjusted during the simulation phase. Another serious problem in radar target tracking is false alarms due to clutter. Recent work by Scientific Systems has shown that the false alarms can be reduced and target resolution enhanced by using a Stochastic Realization Algorithm (SRA) instead of FFT to process the radar I&Q data. Significant improvements in tracking accuracy are expected by the use of SRA in conjunction with robust MGEKF tracking algorithms. Dr Jason Speyer, a leading expert on robust target estimation, will provide consulting support.