Experiments in On-Line Learning Neuro-Control

Abstract

We apply our on-line learning neural network approach to control of non-linear systems; first to joint level trajectory control of an industrial welding robot, and then to one front leg of a hexapod walking robot. The first application demonstrates the ability of neuro-control to model complex non-linear components of a plant, thereby yielding impressive improvements in accuracy compared to the original robot’s controller. The second application illustrates the strength of an on-line learning approach in coping with disturbances to a plant’s characteristics. In order to set this work in context we briefly review our on-line learning neuro-control method, used for both sets of experiments. It has a strict theoretical basis including guarantees of the whole system’s stability.