EMERGENCE OF SELF ORGANIZATION AND SEARCH FOR OPTIMAL ENTERPRISE STRUCTURE: AI EVOLUTIONARY METHODS APPLIED TO AGENT BASED PROCESS SIMULATION

摘要

Enterprise simulation allows what-if analysis and helps in business process re-engineering. There are mainly two approaches to simulation: process based, which is strictly deterministic and generally used to model well known parts of enterprises or mechanical/electronic systems and agent based, which allows to study the emergence of aggregate behaviour, through the creation of models, known as artificial societies. In order to simulate enterprises where the environment and human factor are relevant, a hybrid formalism is proposed: Agent Based Process Simulation. Colonies of intelligent agents, modelled using evolutionary methods derived from the AI field, are put side by side with the representation of processes, modelled as symbolic and deterministic agents built using paradigms derived from Propositional and Modal Logic. The main goal of this work is to study how this approach can allow the emergence of aggregate behaviour and, by using some performance parameters, can help to find the optimal organization for an enterprise. In particular, agents built using Genetic Algorithms and Classifier Systems can evolve to find local maximum of functions representing situations whose rules are not entirely known a priori, such as the ones behind the optimal organization of an enterprise.