On Deadlock Prevention Policies For A Class Of FMS

For a class of Petri nets, RCN-merged nets, a deadlock prevention policy is developed by using elementary siphons. First the problematic siphons are distinguished in an RCN-merged net by elementary and dependent ones. Therefore, by the deadlock prevention policy, a liveness-enforcing Petri net supervisor can be produced. Furthermore, a comparison study is conducted with the existing methods, which deal with the deadlock problems in RCN-merged nets. The results indicate that this method is much better in terms of permissive behavior. Targetting a class of systems where the deadlocks in their Petri net models result from empty siphons, this thesis also develops a formal design methodology for the deadlock prevention problems based on elementary siphons of Petri nets and the theory of regions, and programming for the analysis of the reachability graph. The approach proposed consists of two main stages: siphon control and the theory of regions. The first stage work significantly lowers the computational cost at the second stage compared with the approach where the theory of regions is alone used.