Automated design of a Petri net feedback controller for a robotic assembly cell

Abstract

This paper has presented an extended example demonstrating a particularly simple and computationally efficient method for constructing feedback controllers for untimed Petri nets, even in the face of uncontrollable and unobservable plant transitions. The method is based on the idea that specifications representing desired plant behaviors can be enforced by making them invariants of the controlled net, and that simple row operations on a matrix containing the uncontrollable/unobservable columns of the plant incidence matrix can be used to eliminate controller use of illegal transitions. The significance of this particular approach to Petri net controller design is that the control net can be computed efficiently and automatically based on the plant constraints. The method shows promise for controlling large, complex systems, or for recomputing control laws online due to some plant failure, such as the loss of a required resource, the break down of an actuator, or the corruption of a sensor.