J. Jarmulak, E.J.H. Kerckhoffs, L.J.M. Rothkrantz, “A workbench for neural control in a simulation environment”, in: EUROSIM’95, Simulation Congress, Proc. of the 1995 Eurosim Conference (Vienna, September 11-15, 1995), F. Breitenecker and I. Husinsky (Eds.), Elsevier Science, Amsterdam, 1995, ISBN 0-444-82241-0; pp. 1155-1160.
- pdf – the layout is bad in places due to using new WordPerfect version with a very old WP file
1. INTRODUCTION: NEURAL CONTROL
Neural control promises to offer a solution to the problem of controlling complex systems, which are difficult or impossible to model and control using traditional (non-AI) techniques. Neural networks can be used to model and control such systems because of their ability to approximate broad classes of nonlinear functions. In a control system two mappings are evident candidates to be represented by neural networks. The first one is the process itself, which may be difficult or impossible to model using other techniques. If one is able to obtain a proper neural network model (the so-called identifier) of the process, it can be later used to find the correct process inputs. Another mapping that can be implemented by means of a neural network is the controller.