Kalman Filter

[ehofm]

In the code, covariance (P_{k-1}) is being used to calculate newcovariance (P{k}) rather than projectedcovariance (P{k|k-1}). However, in the formulation for the Kalman Filter as shown below, projected_covariance is used to calculate new_covariance:

The same is found in the calculation of Kalman gain. Can you shed light as to why this is?

[nwojke]

Sorry for the late reply. I think we are talking about this line of code. Note that the app calls predict() followed by update() at each time step.

• At the beginning of the time step, track.covariance contains the posterior covariance P_{k - 1}.
• After predict(), track.covariance contains the predicted covariance P_{k|k-1}.
• After update(), track.covariance contains the posterior covariance P_k.

If you carefully trace the code, you'll find that the Kalman filter update does indeed work as your equation suggests. I hope this helps.