A hierarchy of self-organized multiresolution artificial neural networks for robotic control

Abstract

Summary form only given, as follows. A robotic control system based upon the CMAC, and an enhancement to this architecture using a hierarchy of CMAC neural networks, are discussed. The overlapping input domain cells of each of the layers in the hierarchy are organized using a simple Kohonen network. Using this novel approach, the manipulator input domain has been discretized into cells that have varying placement and size as well as retaining coarse coding generalization. This scheme was evaluated using a computer simulation of a robotic system and has shown significant improvement in the network's overall performance.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>