سال انتشار: ۱۳۸۶

محل انتشار: هفتمین همایش انجمن هوافضای ایران

تعداد صفحات: ۷

نویسنده(ها):

A. Mahmoodi – Department of Aerospace Engineering Department, Amirkabir University of technology, Tehran, Iran
M.B. Menhaj – PhD Student, Electrical Engineering Department, Amirkabir University of technology, Tehran, Iran
M. Sabzehparvar – professor, Department of Aerospace Engineering Department, Amirkabir University of technology, Tehran, Iran

چکیده:

The authors conducted a research regarding position control of Stewart platform by replacing the nonlinear inverse and forward kinematics functions with neural networks (NN). Two 2-layered perceptron type neural networks are implemented with BP training algorithm to function as inverses of each other. By comparing the direct analytical solutions and that of the NN, one can easily see the effectiveness of the NN. While its results are close enough to analytical solution, it makes considerable decrease in calculations and avoids singularities faced in analytical solution in forward kinematics it.