Model-Based Design of Controller Programs

Abstract

The design of a purely logical controller program can be perceived as a controller synthesis problem: the discrete-event behavior of the process to be controlled and a programmable logic controller dynamically interact via sensor and actuator signals and hence form a control loop. If the behavior of the process and the desired closed-loop behavior are modeled as state automaton or formal language, then the required behavior of the controller can be computed according to the Supervisory Control Theory by R.J. Ramadge und W.M. Wonham, resulting in a controller program that is a priori correct. Yet, this method that has been known since the 1990s cannot be directly applied to praxis-relevant scenarios due to immense computational effort. This contribution therefore also explains extensions of the method towards hierarchical and modular control architectures that aim at an effective reduction of the computational effort.