I have a simple first-order reaction batch system for which I have some discrete measurements ($0<t_{k}\le t_{endbatchsample}$).

I have an initial guess for $x_0$ and $P_0$ and from here I perform the time update step (in a continuous-discrete EKF fashion) until $t_1$, where the measurement update kicks in. This process is repeated until $t_{endbatchsample}$ In this particular application it is useful for me two things:

  1. Prediction from $t^+_k$ (posterior) until the $t_{endbatchsample}$ with covariance (in this case I just have one state, variance) propagation through the continuous Riccati equation. This is implemented.

  2. Correction of $x_0$ and $P_0$ after the batch. For this, I have thought of the (Extended)RTS smoother, but I am not sure how to perform it since there was no measurement at $t_0$, hence no EKF estimate. Is there a conventional way to predict (extrapolate) the new $x_0$ and $P_0$ (at $t_0$) after a series of measurements are available?

Sorry for any confusion, this isn't really my field.

Best regards


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.