This corrects the mean "mu" based on the new observation "z_n".
The paper then uses this simple expression to motivate the Kalman filter update equation, in which a new observation z_n is used to compute an innovation (or error signal) y:
y = z_n - H x_n-1
The innovation is then used to update the state-estimate x,
x_n = x_n-1 + K y
The two updates are formally identical, with the Kalman state "x" being analogous to the running-mean "mu". I had never noticed the connection. There is a similar connection to the on-line variance.
It starts with the familiar formula for on-line computation of the sample mean (e.g., https://en.wikipedia.org/wiki/Algorithms_for_calculating_var...) --
This corrects the mean "mu" based on the new observation "z_n".The paper then uses this simple expression to motivate the Kalman filter update equation, in which a new observation z_n is used to compute an innovation (or error signal) y:
The innovation is then used to update the state-estimate x, The two updates are formally identical, with the Kalman state "x" being analogous to the running-mean "mu". I had never noticed the connection. There is a similar connection to the on-line variance.