| Project/Area Number |
03650206
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
機械力学・制御工学
|
|---|
| Research Institution | Yokohama National University |
|---|
Principal Investigator |
TODO Isao Yokohama National University, Faculty of Engineering, Professor, 工学部, 教授 (70017888)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Project Status |
Completed (Fiscal Year 1992)
|
| Budget Amount *help |
¥2,100,000 (Direct Cost: ¥2,100,000)
Fiscal Year 1992: ¥500,000 (Direct Cost: ¥500,000)
Fiscal Year 1991: ¥1,600,000 (Direct Cost: ¥1,600,000)
|
|---|
| Keywords | Automatic Control / Robotics / Mechatronics / Direct-Drive Robot / Neural Network / Digital control / Learning Control / Cooperative Control / ファジィ制御 / ニュ-ラルネットワ-ク |
|---|
| Research Abstract |
Learning control algorithms are proposed for controlling direct-drive robots. The following results have been obtained:
1. A learning control algorithm using neural networks is proposed for the grasping and the movement of an object by a pair of direct-drive (DD) robots of two degrees of freedom. The proposed algorithm has three feedback controllers and two neural networks. After the completion of learning, the outputs of the feedback controllers are nearly equal to zero, and the two neural networks play an important role in the control system. Therefore, the optimum setting of control parameters is unnecessary. In other words, the proposed algorithm does not necessitate any knowledge of the controlled system in advance. The effectiveness of the proposed algorithm is demonstrated by the experiment on the cooperative control of the parallelogram-type DD robots. Moreover, the present learning control algorithm is extended to the control problem of a pair of DD robots of three degrees of freedom in the three dimensional working space.
2. The neural networks which have the generality are complicated and require much learning time. The control algorithm is proposed for shortening the time of on-line learning. In this algorithm, the preliminary learning is carried out by a one-layer linear neural network and by off-line computations the learning results are transferred to the complicated multi-layer networks with sigmoid functions. The effectiveness is demonstrated by the experiment on the control of a DD robot.
3. A fuzzy learning control algorithm is applied to the DD robot system composed of a DD robot-arm and a DD working table.
|
