SRIF Formulation and UDU?

greenspun.com : LUSENET : CU Statistical Orbit Determination : One Thread

My SRIF code keeps diverging and I can't figure out why. One question I have is for the EKF, is b_bar associated with the state itself or the state error? Do I initialize b to the state values or to apriori state error values (=0)? I've tried to paste in my code to this email in case you could spot something off the bat. I havn't started the UDU' portion yet, but I'm assuming we follow the measurement update steps Dr. Born gave us in class: F=U_bar'*H' V = d_bar*F alpha = [I+V*F] . . . and use the time update you had posted. What do we use for the initial U_bar and D_bar? I think they must come from the P_0 somehow, but I'm not sure how.

-- Brian Lathrop (Brian.Lathrop@Colorado.EDU), February 08, 2000

Answers

here is the algorithm: given: W, P0, xbar0 = 0 (xbar is the a priori state deviation) phi = I R = chol(inv(P0))

b = R * xbar0

start loop, proces each obs

time update X and phi calculate measurement residual y calculate H matrix

bbar = b Rbar = R * inv(phi)

[ R b ] [ Rbar bbar ] [ 0 e ] = givens [ sqrt(W) * H sqrt(W) *y ]

end loop

xhat = backsub([R b]) As far as Ubar0, Dbar0...Pbar0 = Ubar0 * Dbar0 * Ubar0' If Pbar0 is diagonal then Ubar0 = I and Dbar0 = Pbar0 Dave G.

-- David Goldstein (david.goldstein@colorado.edu), February 08, 2000.


Moderation questions? read the FAQ