This is an example from Durbin and Koopman, Time Series Analysis by State-Space Methods, 2nd ed. It's a simplified version of the yield curve analysis from Diebold, Rudebusch & Aruoba (2006) using only the latent factors and not the observable macro factors. It's the second model from the DRA paper, but takes advantage of the nature of the factors as random walks to eliminate the need for the "mu" parameters---since the random walk has no mean, it can adjust to whatever initial level fits best.