Neural Robust Control for Perturbed Crane Systems

In this paper, we present a new control methodology for perturbed crane systems. Nonlinear crane systems are transformed to linear models by feedback linearization. An inverse dynamic equation is applied to compute the system PD control force. The PD control parameters are selected based on a nominal model and are therefore suboptimal for a perturbed system. To achieve the desired performance despite model perturbations, we construct a neural network auxiliary controller to compensate for modeling errors and disturbances. The overall control input is the sum of the nominal PD control and the neural auxiliary control. The neural network is iteratively trained with a perturbed system until acceptable performance is attained. We apply the proposed control scheme to 2- and 3-degree-of-freedom (D.O.F.) crane systems, with known bounds on the payload mass. The effectiveness of the control approach is numerically demonstrated through computer simulation experiments.