Fixed-lag smoothing as a constrained version of the fixed-interval case

This paper studies the H/sub /spl infin// fixed-lag smoothing problem in both continuous and discrete time. The central idea is to address it as a constrained version of the fixed-interval smoothing (L/sub /spl infin// estimation) problem. This enables us to separate geometric (which are independent of the smoothing lag) and analytic constraints imposed by the problem data on the achievable performance. As a byproduct the technique provides an elegant means to find a minimal (finite) smoothing lag at which the optimal L/sub /spl infin// performance level is achieved. State-space formulae are also derived in the continuous-time case.