Seonglae ChoThe process of correcting the prediction result using the actual observation.
- If the prediction is more reliable → incorporate the observation only a little.
- If the observation is more reliable → incorporate the observation a lot.
This weighting is determined by the Kalman gain using prediction Covariance
The Kalman filter update is exactly a Bayesian update—specifically, the special case that admits a closed-form solution under linear and Gaussian assumptions. In other words, the update is a Bayesian posterior update (prior → likelihood → posterior). Under the Gaussian assumption, this becomes a closed-form optimal solution. So, the Kalman update = a Gaussian Bayesian update.