The place-invariants method is one of the most popular controller synthesis approaches for Petri net (PN) modeled DES. Unfortunately, the observance of the constraints is not certain in the presence of uncontrollable transitions. This thesis offers a solution to this problem for ordinary and generalized PNs. We begin by studying safe non-conservative PNs, and devising a constraint-determination technique that will always provide a set of admissible constraints for this type of model. The approach stems from the general definition of forbidden states --- that of marking vectors. In the second part of our work, we present an admissible constraint-determination technique for generalized PNs. The method is based on a special view of the system's state space. The constraints are derived from the equation of the affine hyper-plane separating the authorized- and forbidden- regions of this space. We propose an algorithm that allows the identification of the minimal maximally permissive controller.
