The Recursive Constraint Least Square Based on UT

Estimation in nonlinear systems is extremely important because almost all practical systems involve nonlinearities of one kind or another. The recursive least square (RLS) and Kalman filter (KF) is well known for their good convergence property and small mean square error for estimation of the linear system, and the extended Kalman filter (EKF) has been widely used for 30 years for estimation of the nonlinear system. But these methods all have drawbacks. The unscented transformation (UT) is a new method presented for its ease of implementation and more accuracy for nonlinear system. In this paper, the authors bring forward a novel method-the recursive CLS based on the UT. This method yields high accuracy to nonlinear systems without the linearization steps and great efficiency with recursive algorithm. These performances can be seen from the experimental results.