Learning control for a closed loop system using feedback-error-learning
Hiroaki Gomi, Mitsuo Kawato · 1990
The authors propose a learning scheme using feedback-error-learning for a neural network model applied to adaptive nonlinear feedback control. After the neural network compensates perfectly or partially for the nonlinearity of the controlled object through learning, the response of the controlled object follows the desired set in the conventional feedback controller. This learning scheme does not require the knowledge of the nonlinearity of a controlled object in advance. Using the proposed approach, the actual responses after learning correspond to desired responses. When the desired response in Cartesian space is required, learning impedance control is derived. The convergence properties of the neural networks are provided by the averaged equation and Lyapunov method. Simulation results on this learning approach are presented. The proposed scheme can be used for many kinds of controlled objects, such as chemical plants, machines, and robots.>