2005 Fiscal Year Final Research Report Summary
Study on application of fuzzy / neural network based control to tumbling motion reduction by means of cushion damper
| Project/Area Number |
14550855
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Aerospace engineering
|
|---|
| Research Institution | Tokyo Metropolitan University (2005)
Tokyo Metropolitan Institute of Technology (2002-2004) |
|---|
Principal Investigator |
KOJIMA Hirohisa Tokyo Metropolitan University, Dept of sysem design, Associate Professor, システムデザイン学部, 助教授 (50322350)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
FUJII Hironori Tokyo Metropolitan University, Dept of system design, Professor, システムデザイン学部, 教授 (30070650)
|
|---|
| Project Period (FY) |
2002 – 2005
|
| Keywords | adaptive control / visual servo / wavelet / neural network / fuzzy control |
|---|
| Research Abstract |
We constructed an in-space light simulator and investigated a method of extracting the edges of a satellite using wavelet transform and self-organizing map techniques. In addition, we compared this method with other conventional edge extraction methods, and found that the wavelet transform is better than the other three edge extraction methods.
A ground based nutation simulator was also developed to study the impulsive thrust method proposed by Kawamoto. A three-axis gyro is installed on the top of the developed experimental to measure the angular velocity changes when the impulsive force is added by throwing a soft ball to the experimental device. The angular velocity change is also estimated from the recorded images. By comparing the angular velocity change estimated by the image processing and measured by the gyro, a good agreement was confirmed. Therefore, it can be said that even if the target satellite is an incorporative satellite or the angular velocity cannot be obtained directly from the communication with the target satellite, the momentum reduction can be estimated from the velocity change of the particle thrown to the tumbling target satellite, the thrown particle mass, and the collision position.
We discuss a fuzzy control applied to reduction of tumbling motion of the non-cooperative object by the repetitive impulses produced by projecting small particles to the non-cooperative object. The parameters of the fuzzy controller which determine the timing and velocity of the projection are tuned by means of learning algorithm. The proposed method has advantage in robustness to variation of target motion and initial values of the parameters.
|
