Mechanical systems tracking using neural networks and state estimation simultaneously

In this paper, we analyze how to implement a nonlinear observer, which combined with a structured neural network, allows tracking of an artificial output, even when the state is not fully measurable; as an effect of this tracking, the system state converges to any selected equilibrium point. We establish and prove a theorem about the stability of the tracking error and of the state estimation error. The applicability of the approach is illustrated by simulation results