Lattice algorithms for recursive identification of general model structures

Lattice algorithms are a numerically efficient implementation of recursive identification methods. Due to an orthogonalizing basis transformation in the regressor space, they provide model estimates of several orders simultaneously, which makes them well-suited for adaptive control applications. Up to now lattice algorithms were only available for AR(X) and ARMA models. In this paper lattice algorithms are proposed for general model structures, with simplifications for ARMAX and OE models. Simulation studies show that the proposed lattice algorithms have better convergence than nonlattice implementations of recursive identification. Since also the number of computations is less, application of these algorithms in adaptive control seems very promising.