DESIGN OF AN INDUSTRIAL AUTOMATION ARCHITECTURE BASED ON MULTI-AGENTS SYSTEMS Cesar Bravo *, José Aguilar Castro **, Mariela Cerrada**, Francklin Rivas Echeverría ^ * PDVSA West Division. Industrial Automation. +58-264-803750 bravocn@pdvsa.com ^ Universidad de los Andes Postgrado en Ingeniería de Control y Automatización +58-274-2402846 rivas@ula.ve ** Universidad de los Andes CEMISID {aguilar,cerradam}@ula.ve Abstract: New paradigms in industrial automation aims towards intelligence distribution among components of the productive process and through the integration of several systems and applications in the companies, searching processes improvement, supporting decision-making tasks by real time information available, reliable and opportune. This work proposes industrial automation architecture based on multi-agents systems and uses the framework SCDIA to describe this architecture. Copyright © 2005 IFAC Keywords: Artificial Intelligence, Industrial Automation, Agents, Production Systems. 1. INTRODUCTION Theory of multi-agents systems seems to agree with the tendency of industrial world, especially in oil & gas industry, distributing intelligence among the components of the productive process, providing them with mechanisms that allow them being conscientious of their states, goals and actions, and therefore they can be able to do self-regulation, self- diagnosis and take actions to maintain their desired state. Multi-agents systems consist in an agents’ collection, each one with a partial vision of their environment but with social characteristics that allow them to cooperate for reaching individuals or global goals. This paper proposes a multi-agents system that represents the operations carried out in an industry, from a distributed control perspective. This system is modeled using the framework SCDIA (Intelligent Distributed Control System based on Agents, in English), which is a multi-agents system architecture, built specifically for automation and control domains (Aguilar et al., 2001, Bravo et. al, 2003). This model contemplates five agents that represent elements of a control loop, thus, on SCDIA will be agents like, controllers, observers, actuators, coordinators and specialized agents. Besides, SCDIA proposes Services Administration System (SAS), which have the responsibility to develop integration tasks between the SCDIA components and external systems. The SCDIA uses principles of Multi-Agent Systems paradigm for modeling industrial automation processes under a generic and reusable schema, useful for process control, resource administration and decisions making activities. This paper is composed by four parts. The first part proposes a multi-agents model for an industrial automation platform design. The second one presents a case of study apply to oil production industry. In the third section is presented the design of agents related to the case of study already mentioned. The last part presents the conclusions of this investigation.