Dear Yi Cao,
According to the paper'performance evaluation of UKF-based nonlinear filtering',choose:f=@(x)[x(1)+tao*x(2);x(2)-tao*x(1)+tao*(x(1)^2+x(2)^2-1)*x(2)];
h=@(x)x(1),with covariance of the process noise w(k)given as:Q=0.003^2I,the covaiance of the noise v(k) is given by:R=0.001^2I,initial state:x0=[2.3;2.2],P0=I,the true value of the initial state:x=[0.8;0.2]. The paper proof that when given all these,UKF tends to be divergent.However,based on this code,it seems that the estimator is stable.Does it owe to the weights chosen when doing the prediction?