Learning sequences with Neural Gas for robot motion planning
Ignazio Aleo, Paolo Arena, Luca Patanè
The aim of this paper is to investigate novel solutions for motion sequence learning based on an extension of the Neural Gas with local Principal Component Analysis (NGPCA) algorithm. As an abstract Recurrent Neural Network (RNN) this model is able to complete a partially given pattern. Under this point of view it is possible to generalize the model as a dynamical system in which for a given actual configuration and a particular task the desired state variables are retrieved as outputs converging to a particular state iteratively. The developed architecture has been tested in the control of a redundant manipulator in simple forward and inverse kinematic problem solving and in motion sequence reproduction.