| Summary: | 碩士 === 國立中央大學 === 電機工程學系 === 85 === Abstract In the paper, we discuss the weights learning method
of neural network that were usually applied to control system.
In general, if a priori knowledge of system dynamics is known,
we usually use supervised learning method for the controller
design; on the contrary, if none or a little knowledge of system
dynamics is known, we usually use the reinforcement learning
method. Here, we investigate the effects of the combination of
these two methods for the control design to acquire more
advantages. To verify the results, we use the learning method to
the cart-pole balance system simulation, and compare their
results. We show that the combination learning method that has
many better performances from the comparison between these
simulation results. Finally we use it to control the seesaw
system for demonstrating the practicability.
|
|---|
