Constructing an adaptive fuzzy neural network for solving the inverse kinematics problem of the of multilink mechanisms

Ganin P. E., Kobrin A. I.

The paper considers the construction of an adaptive fuzzy neural network for a motion control system of a multi-link manipulation mechanism with a serial connection of links. It is proposed to introduce an optimality criterion based on an estimate of the reaction time and accuracy. The results of studies of the influence of network parameters on the quality of the resulting solution are presented. We consider such parameters as: the size of the training sample, the number of training epochs, the number and type of membership functions.

Keywords:inverse kinematics problem, adaptive fuzzy neural network, simulation, real-time control system.