Terry <terryzz24@gmail.com> writes:> the measurement equation is: > y(t) = a1(t)+a2(t)*exp(-a3(t)*y(t-1)^2)*y(t-1)+a4(t)*exp(- > a5(t)*y(t-1)^2)*y(t-2); > > and the state equations are > a1(t) = a1(t-1) + v1; > a2(t) = a2(t-1) + v2; > a3(t) = a3(t-1) + v3; > a4(t) = a4(t-1) + v4; > a5(t) = a5(t-1) + v5; > > there is no measurement noise.Then you still haven't read enough about Kalman filters. Ciao, Peter K. -- "And he sees the vision splendid of the sunlit plains extended And at night the wondrous glory of the everlasting stars."
can anyone help design a state space model applying into Extended Kalman Filter
Started by ●January 21, 2008
Reply by ●January 22, 20082008-01-22
Reply by ●January 22, 20082008-01-22
sorry i may not clearly represent my question. In my point of view, KF is a sort of estimator. y(t) = a1(t)+a2(t)*exp(-a3(t)*y(t-1)^2)*y(t-1)+a4(t)*exp(- a5(t)*y(t-1)^2)*y(t-2) + e(t); during the estimation process y_hat(t) = a1(t)+a2(t)*exp(-a3(t)*y(t-1)^2)*y(t-1)+a4(t)*exp(- a5(t)*y(t-1)^2)*y(t-2); the measurement error e(t) = y(t) - y_hat(t); the measurement covariance Q(t) = cov(e(t)); then predicted covariance P(t|t-1) = F*P(t|t)*F+Q(t) where F is the state process model. If i want to apply EKF to y(t) = a1(t)+a2(t)*exp(- a3(t)*y(t-1)^2)*y(t-1)+a4(t)*exp(-a5(t)*y(t-1)^2)*y(t-2) + e(t), how can I the State-space mode?