Conceptually they are very similar; as you mention, Kalman is generally online, and trying to estimate the current state, whereas Viterbi can factor 'future' in

Conceptually they are very similar; as you mention, Kalman is generally online, and trying to estimate the current state, whereas Viterbi can factor 'future' information into its estimate as well. Another big difference is that Viterbi works on a discrete state space (DNA base pairs, bits in a convolutional code, etc), while the Kalman filter works in a continuous space (e.g. position). https://en.wikipedia.org/wiki/Recursive_Bayesian_estimation https://en.wikipedia.org/wiki/Recursive_Bayesian_estimation