TOWARDS SYMBOLIC PROCESS-CONTROL

Much recent work in Intelligent Control has introduced and analyzed va rious approaches to the implementation of expert controllers. But very little attention has been given to a fundamental question: ''How can a controller be designed by abstracting rules from observations of a r eal process?'' This paper gives an overview of the concept of symbolic control and briefly compares it with expert control. A method for des igning symbolic controllers is proposed, based on qualitative measures of state coordinates and their derivatives, which allows the symbolic to numeric interface to be expressed using a system of recurrence equ ations. The technique is illustrated with two examples, symbolic proce ss control of a first order process and symbolic control of a simulati on representing a car on a defined track.