Special Issue "Fuzzy/Neural Network Applications to Dynamics and Control of Mechanical Systems". Learning Control of an Inverted Pendulum Using Neural Networks.
Toru Kumagai, Satoshi AKITA, Mitsuo WADA · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series C · 1992
We present a new learning control method to control a multi-input-output system. In conventional learning controllers using a neural network, it is difficult to treat a multi-input-output system because of the difficulty in designing reference models. Hence, we propose to divide a complex plant into sub-plants and to use a learning controller for each one. We use our method to the regulation problem of the inverted pendulum that is a one-input, two-output system. In the simulation and the experimental system, we regulate the inverted pendulum and show the effectiveness of this method. We also show that our learning control system can regulate a system that has a time lag between the input and the output signals. Moreover, we show that a reference model of order lower than the plant order is available.