This paper formalises the evolution from decision problems to disjunctive optimisation problems based on several alternative Petri Net (PN) models. This copes with the general problem of Discrete Event Systems (DES) design and operation. A new type of PN called Alternatives Aggregation Petri Nets (AAPN) is presented as a promising tool to solve this general problem in an efficient way. AAPN are constructed by aggregating the alternative PN models, which are feasible solutions to the DES design. As a result, the optimisation problem can be solved by means of classical methods like exhaustive or heuristic search in a single phase. (2010) 'The alternatives aggregation Petri nets as a formalism to design discrete event systems', Int.

Decision-making on discrete event systems with alternative structural configurations is a field with application to the efficient design and operation of many systems, ranging from manufacturing facilities to communication networks. The solution of this problem may be afforded by its transformation into an optimization problem. A variety of statements for this optimization problem can be presented by using different formalisms able to describe the model of the system. These different statements allow developing diverse optimization algorithms for solving the problem, which may be very demanding for a computer. In this paper, several approaches are presented in order to reduce the computing requirements needed by the mentioned algorithms, some of them are implemented in one processor and others are based on distributed computing. In particular, this paper presents a new distributed methodology, which associates sets of alternative structural configurations of the system to different alternative aggregation Petri net (AAPNs), regarding the number of available processors. Under certain conditions, this methodology alleviates the computational requirements for every processor and speeds up the optimization process. A case-study is presented and different techniques are applied to solve it, for illustrating diverse distributed and non-distributed methodologies, regarding the available processors, as well as for comparing their relative performance.

No abstract