Combination control of robot based on the neural network

The movement of the robot arm must be smooth in most industry manufacture. If vibration occurred in any case, the consequences would be disastrous. The movement characteristic of the robot arm is determined by the controller of the robot and the neural networks control is a new kind control scheme. Whenever the input of the control system is mixed with the interference signal, the output of the control system would be vibrational. This phenomenon would be very serious especially in the neural network control system based on the self-learning scheme of PD. The vibration phenomenon can be controlled using some special learning scheme. A reasonable thought mi-ht be as follow: The input signal is not very pure, in other words, some disturbance sinal must be contained in the input signal. But. in most cases. the disturb sinal has random characteristic and the average value would nearly be zero. In the other hand, the real input signal has some continue characteristic and the average value is not zero. According to the thought above, a control scheme can be abstained. When the input signal is large enough, the control scheme will be: self-learning control based on PD. When the input is small to a certain range, the self-learning scheme of the controller is changed to the integral filter scheme. In this cast, the output of the controller will act according to the real input signal and the disturb signal would be filtered away, because the average values of the disturb sinal is near to zero. A robotics arm controller based on the above thought is established. The behavior of the system is simulated on a microcomputer and the results are very satisfactory. Furthermore, a robotics arm driven by hydraulic system is controlled by this controller. The results of the control procedure are very satisfactory either.