| Summary: | 碩士 === 逢甲大學 === 自動控制工程研究所 === 84 === This thesis studies two nonlinear identification methods,
the radial basisfunction method and the Kolmogorov-Gabor
polynomial method, applied on thecontrol of atwo-link robot
manipulator. It is well known that two-link robotmanipulators
are highly nonlinear systems. A feedback linearization method
isapplied to linearize the system. After that, a PID
controller is employed tocontrol the linearized system. In the
linearization loop, nonlinear functionmust be given.
Usually, these nonlinear functions are obtained by
thederivation of the system model. In contrast, this thesis
uses two nonlinearidentification methods to get these nonlinear
functions without knowing thesystem parameters. Simulation
results show that both nonlinear identificationmethods work
well for the set-point control of robot manipulators. However,
asimple decoupling test is performed and shows that the
Kolmogorov-Gaborpolynomial method is better than the radial
basis function method. This mayhint that, in many nonlinear
identification methods, whether one can select amethod whose
bases' functions are close to the nonlinear system is
veryimportant.
|
|---|
