1

I am running experiments on the kalman filter, with a particular interest on the covariance Matrix. I am using the formulas given at http://en.wikipedia.org/wiki/Kalman_filter and something which seems weird to me is that the covariance matrix does not seem to be influenced at all by the measurements, because the formulas using the measurement $z_k$ concerns only the corrected state and not the corrected covariance.

My specific concern is that since I need to compute the "quality" of the current value (it's expected accuracy) I was thinking about using the covariance matrix to get this information. But if this value is not influenced by the measurements, I think it is not a very good idea...

I didn't write too much detailed informations in order to keep a readable question, but I can give more details if needed.

Martin Sleziak
  • 56,060
medrimonia
  • 11
  • 3

1 Answers1

2

The covariance matrix is influenced by a measurement. However you are absolutely correct in your excellent observation that the measured value does not affect the covariance matrix. Only the measurement noise affects the covariance matrix. See this question for more details.

The diagonal elements of the covariance matrix are the variances of the corresponding states and can be used as a measure of the quality of the estimate.

Community
  • 1
Pradu
  • 782
  • 5
  • 12