Inverse kinematics learning by modular architecture neural networks

Inverse kinematics computation using an artificial neural network that learns the inverse kinematics of a robot arm has been employed by many researchers. However, conventional learning methodologies do not pay enough attention to the discontinuity of the inverse kinematics system of typical robot arms with joint limits. The inverse kinematics system of the robot arms, including a human arm with a wrist joint, is a multivalued and discontinuous function. Since it is difficult for a well-known multilayer neural network to approximate such a function, a correct inverse kinematics model for the end-effector´s overall position and orientation cannot be obtained by the conventional methods. In order to overcome the drawbacks of the inverse kinematics solver consisting of a single neural network, we propose a novel modular neural network architecture for the inverse kinematics model learning