Summary: | 碩士 === 中原大學 === 機械工程研究所 === 82 === The prosperity of robotic manipulators in industrial automation
is due to their flexibility. However, their structural
complexity becomes a burden during generating control command.
For currently Von Neuman computer structure still executes
computation codes line-by-line sequentially. A robot system
with high degrees of freedom will demand a considerable
computation time to evaluate its dynamic equation and hence its
control force. Thus, it produces control delay and speed down
the robot. For this reason parallel process is intuitively an
effective approach to reduce the control delay. There are two
methods of parallel process are trying to optimize program
code, or using a parallel structure to process more computation
simultaneously. For this purpose, we employ neural network to
the controller which have such character. Neural network is a
new theorem which has capacity of emulating nonlinear function.
Inside network also is a simple structure to success parallel
process. Take these advantages of network to design control
algorithem will cut down control delay and speed up the robot.
In order to verify our design, several computer simulations
focus on three link SCARA manipultor are executed. From that,
it shows that neural network controller has same behavor with
traditional numerical controller. That''s what this paper
address.
|