A neural controller is proposed for the tracking control of flexible arms. A significant property of this proposed controller is that neither the model information nor the state of the flexible arm is required. The desired feedforward for the tracking performance is learned online based on the output of interest, and the feedback compensation of measurable output is used to ensure the Lagrange stability of the closed loop system. The analysis results of the stability show that the closed loop system is Lagrange stable, and that the tracking error of the output of interest converges to zero. A single-link flexible arm is simulated, and the simulation results are satisfied.
1993 IEEE International Conference on Neural Networks 1993, pp.749-754.