Multiple steps prediction with nonlinear ARX models, Report no. LiTH-ISY-R-2793

NLARX (NonLinear AutoRegresive with eXogenous inputs) models are frequently used in black-box nonlinear system identi cation. Though it is easy to make one step ahead prediction with such models, multiple steps prediction ids far from trivial. The main di culty is that in general there is no easy way to compute the mathemaical expectation of an output conditioned by past measurements. An optimal solution would require intensive numerical computations related to nonlinear ltering. The purpose of this paper is to investigate simple non-optimal prediction merhods. It is shown that caution must be paid when using such methods, since the prediction behaviors may be radically di erent, depending on some detailed choices.