Constructing petri net models using genetic search

The problem considered is that of constructing a Petri Net model of a particular device, process, or system described by observations of its interactions with its environment. The algorithm proposed for achieving a solution employs the principles of genetic search. Also presented is a detailed investigation of its operation, thereby affording a theoretical justification of its design. pressive and representational power of Petri Nets renders them ideally suited to the modelling of many complex event systems. As a modelling language, they directly support the intrinsically difficult concepts of concurrent and parallel activities, synchronisation of events and the distribution of various resources. However, the development of a model of a significant system is laborious and circumstances of limited knowledge of the systemʹs internal structure compound the difficulty. posing the perceived stimuli and responses of the system under study as a set of behavioural requirements, the Petri Net construction problem can be examined. The inherently difficult nature of this problem renders most other approaches inapplicable or inadequate; the quest for a satisfactory solution leads instead to the development of an algorithm employing genetic search technology. around a genetics-based machine learning architecture, the first algorithm developed shows a deficiency in its dependence on the particular order in which various options are explored. Alleviating this difficulty through a simple modification may also hold a lesson for other classifier systems.