Modeling and Control of Manipulator Dynamics using Modified Self-Organizing Maps

TL;DRAbstract

A dynamic modeling of a robot manipulator by means of a neural architecture is presented. Such model is applicable to generate a decoupling and linearizing feedback in the control loop of the robot drives. A modified extended Self-Organizing Map is used to perform the needed mapping of the robot's movement state to the according joint torques. To get improved control performance by this, both correctness and smoothness of the approximated values are essential. This paper describes a suitable architecture for this, the so called Parametrized Self-Organizing Maps. A training of this neural network type by the Widrow-Hoff-learning-rule is introduced and different training methods are discussed.

Chat with Paper

AI Agents for this Paper