Contiguration Control of the Truss-Type Parallel Manipulator by the Modular Neural Network Model

Abstract

A new control method for the truss-type parallel manipulator is proposed by a neural network approach. The proposed method solves the inverse kinematic problem involving kinematic redundancy with the multilayered neural network model. This neural network model is prepared to provide the kinematic relation of the manipulator as well as the criterion function for the configuration and/or motion control. A concept of the modular neural network is employed to cope with the immensity of the network, in which a neural network module is constructed separately for each kinematic module. This modular neural network approach enables the network system for the configuration control to be adopted for various combinations of truss modules. Some simulation studies demonstrate the effectiveness and the potential capability of the proposed method.