Neural networks to solve nonlinear inverse kinematic problems

In making neural networks learn nonlinear relations effectively, it is desired to have appropriate trainingsets. In the proposed method, after a certain number of iterations, input-output pairs having worseerrors are extracted from the original training set and form a new temporary set. From the followingiteration, the temporary set is applied to the neural networks instead of the original set. In this case,only pairs with worse errors are used for updating the weights until the mean value of errors decreasesto a desired level. Once the learning is conducted using the temporary set, the original set is appliedagain instead of the temporary set. The effectiveness of the proposed approach is demonstrated throughsimulations using kinematic models of a leg module with a serial link structure and an industrial robot.