The design process for conventional and neural controllers -a case study

Neural nets show the ability to learn and to reproduce a given input-output behaviour without the need of a mathematical model. They can be used to mimic an operator´s behaviour when a control task can be solved manually. Therefore it seems to be easier and more cost-effective in this case to design a neural controller than a conventional controller. In a case study, both conventional and neural controllers for the lateral motion of a vision-guided vehicle were designed. This paper discusses in detail the different design processes for both types of controllers and compares the results