Self-organizing neural networks for learning inverse dynamics of robot manipulator

Fast and accurate trajectory tracking of a robot arm primarily depends on the knowledge of its explicit inverse dynamics model. Online learning of inverse dynamics using a supervised learning algorithm is difficult in the absence of a priori knowledge of command error. On the other hand, a self-organizing neural network employing an unsupervised learning scheme does not depend on the command error. These networks are suitable for both off-line and online schemes of learning the inverse dynamics. The present paper proposes two schemes based on unsupervised learning algorithms, namely, Kohonen´s self-organizing topology conserving feature map and “neural-gas” algorithm. Simulation results on a single link manipulator confirms the efficacy of the proposed schemes