All righty then...

So, I coded up the consider covariance as a batch. I'm calaculating my matrices numerically, not analytically. My step size is 0.001 and in using this I get a good phi matrix (by cross peer validation). So, I know that my code works. However, I'm not getting consistent answers for my data noise covariance nor my consider covariance. I have no idea why! I've looked at my A and B matrices, and they seem to be on the money (formulation).

For my A matrix I have: [0,1;-(g/l),0] For my B matrix I have: [0,0;0,-(theta/l)]

Phidot = A*Phi Thetadot = A*Theta + B

with Phi(t0,t0) = I and Theta(t0,t0) = zeros

can anyone contribute to solving my dilemma?

-- Moriba Jah (Moriba.Jah@Colorado.EDU), March 03, 2000

Answers

one more thing...

The sequential algorithm on page 361 of the class notes has something peculiar:

Pbarc and Pbarxc have nothing to do with the measurement update of Pc and Pxc...why not? Dave, is the algorithm correct?

-- Moriba Jah (Moriba.Jah@Colorado.EDU), March 03, 2000.