بر روشهای سیستماتیک برای حذف مانیتور مزاحم از زنده بودن- اجرای سرپرستان خالص

Petri nets based deadlock prevention for flexible manufacturing systems has received much attention over the past decade, primarily due to the seminal work of Ezpeleta et al. in 1995. A Petri net based deadlock prevention mechanism is usually implemented by adding monitors or control places to a plant Petri net model such that liveness can be enforced. The significance of this methodology lies in that both a plant model and its supervisor are in a same formalism-Petri nets. Due to the inherent complexity of Petri nets, in theory, the number of additional monitors that have to been added to achieve liveness-enforcement purpose for an uncontrolled plant model is exponential with respect to the size of the model. This paper first proposes a systematic method to minimize the number of additional monitors in a liveness-enforcing Petri net supervisor such that the resultant net system has the same permissive behavior while liveness can still be preserved. Furthermore, for the liveness-enforcing Petri net supervisors of flexible manufacturing systems, which have some particular property, an algorithm is developed such that more permissive liveness-enforcing Petri net supervisors can be obtained after liveness-restrictive monitor removal. Compared with the existing techniques of eliminating redundant monitors in the literature, the complete state enumeration of a supervisor is avoided, which implies the high computational efficiency of the methods in this paper. Flexible manufacturing examples are used to demonstrate the proposed approaches.

In a flexible manufacturing system (FMS), different types of raw parts enter the system at discrete points of time and are processed concurrently, sharing a limited number of resources such as machine tools, AGVs, robots, buffers, and fixtures. In such a system, every raw part follows a preestablished production sequence through the set of system resources. These production sequences are executed concurrently and therefore they have to compete for the set of shared resources. This competition for shared resources can cause deadlocks that are a highly undesirable situation, where each of a set of two or more jobs keeps waiting indefinitely for the other jobs in the set to release resources (Viswanadham, Narahari, & Johnson, 1990).

Behavior permissiveness, computational complexity, and structural complexity are the three major problems when we are concerned with the design of liveness-enforcing Petri net supervisors for discrete event systems. In the framework of supervisory control theory of discrete event systems originally proposed by Ramadge and Wonham (1989), behavior permissiveness is usually characterized by the number of reachable states of the controlled system, i.e., the supervisor of a plant model. Given a particular system, a more permissive supervisor means that it has a larger number of reachable states. In theory, the measure of permissiveness is independent from the model that we establish for the system to be investigated. The behavior permissiveness underlying a modeling formalism is measured by the number of reachable states of the controlled system of a plant model whatever the modeling formalism is adopted.