A reference model for distributed decision making through a multi-agent approach

During the last years the decision making processes is evolving becoming more and more distributed and asynchronous. In order to support decision-makers who are not at the same place at the same time are defined cooperation processes and a set of models able to support designers of Cooperative decision support framework. One of this, proposed in this thesis, is a Reference Model to build Multi Agent System able to represent complex and distributed system composed by several intelligent collaborative entities. Following this definition, it is possible define like collaborative network a lot of complex system such as: 1. Flexible Manufacturing Systems (FMSs) that are defined by Ciliers (Ciliers P., 1998) as complex systems in which the knowledge of the system elements composition and the interaction between the environment and the system, is not enough to understand the system functioning; 2. Industrial Clusters (ICs) that can be defined as socioeconomic entities characterized by a social community of people and a population of economic agents localized in close proximity in a specific geographic region that interact to reach similar outcomes (Marshall, 1925); 3. Supply Chains (SCs), network of suppliers, factories, warehouses, distribution centres and retailers that operate in integrated manner (Fox, 1992) with the same aim of reducing inventory and costs, adding value, extending resource and accelerating the time to market. 4. Healthcare Systems, composed by several entities interacting in a great number of critical processes. The internal dynamics of an hospital represents a complex non-linear structure hard to manage in centralized way (Harper 2002). The above mentioned list represents only a part of the numerousness examples of system that can be represented as collaborative networks and that need to be managed in distributed manner due to complexity and to great amount of information necessary to take any kind of decision. There are not many models able to represent and manage this kind of distributed systems, Petri nets represent one of this but present some limits due to: (i) the model become too large and complex even for a modest size problem (Wang J., Deng Y., 1999); (ii) the difficulty to make even a little change to a previously built model. The Multi Agent System (MAS) approach aim to overcome these limits of rigidity and computing complexity trough: (i) the possibility to solve complex problem solving a set of easier local problems; (ii) the opportunity to change some problem parameters or to substitute any system element without discard entirely the original model. The MAS are widely studied in literature like method able to represent dynamic and distributed system with several decision makers having different information domains. It is possible to observe a lack in investigate the problem solving ability of intelligent agents in a multi-agent setting and a variety of representation methods. The difficulty of apply this model approach reside into the activity of architecture design used by the system elements to speak and act. The number and the kind of relations among the system entities became the most important indicator of the model's complexity. For this reason the outcome of my researches can be presented as a reference model integrating two types of approach for the models creation in the MAS field. The main scope is to provide the guidelines able to support the designer in the system modelling like MAS, indicating also in which scenario is more convenient to adopt an approach oriented to Operation Research (OR) technique or another one. The Reference Model is evaluated and validated thanks to several application in different contests, some of whom are mentioned at the beginning of this abstract.