One-step fixed-lag smoothers for Markovian switching systems

Suboptimal approaches to the one-step fixed-lag smoothing problem for Markovian switching systems are examined in this paper. Two different methods for obtaining suboptimal smoothed estimates are given, where the methods differ by the sampling period upon which the state of the system is conditioned. For n models, the first method requires n² predictions and residual evaluations, while the second method requires n residual evaluations. Simulation results are presented to compare the performances of the two smoothers.