Comparative study of two on-line trained neural approximators of a robot inverse kinematics

In this article are proposed two on-line trained neural approximators of a robot inverse kinematics. The first one consists of an inverse neural model of the robot forward kinematics, and the second one consists of a forward neural model of the robot forward kinematics and an optimization procedure. Simulation research is conducted of the neural approximators working separately and simulation research is conducted of the control system working with these neural approximators. For that purpose is used a two-link planar robot.