A Neural Compensation Technique for Control at Low Velocities

A method for adaptively identifying and compensating for certain types of nonlinearities is reported using an experimental system consisting of a DC motor with a load. The nonlinearity due to friction, which is significant at low velocities, is compensated using a neural network strategy and the technique is compared with two other adaptive ones reported in the literature. In the proposed control strategy, a feedforward neural network is used to identify on-line the unknown nonlinearities. Also, uncertainty of the parameters for the modeled dynamics is considered. An error analysis is also provided for the proposed control strategy.