Communication solution for industrial control applications with multi-agents using OPC servers Eugen Diaconescu*, Cristian Spirleanu † *University of Pitesti, Romania, eugen.diaconescu@upit.ro † IMSAR Bucuresti, Romania, cristian.spirleanu@generalserv.ro Abstract - This paper presents a concrete way of linking the JADE multi-agent system with the equipment (eg PLC, DCS, SCADA, HMI) comprised into a distributed industrial control system based on agents, using OPC servers. The differences between traditional control and agent based control approach also briefly showed. Industrial applications of multi-agent technology are limited, among others, especially due to the difficulties of communication between agents development environments and heterogeneous set of control devices, sensors and actuators which can be found in an industrial process. The solution involved the use of OPC standard. By using an OPC client written in Java, the connection between JADE multi-agent system development also written in Java, and OPC servers, was made; that allow access from JADE agent to process variables. The concrete steps for developing a JADE agent with the ability to connect it to the OPC server were presented. The content of this paper refers to a part of a dedicated application for monitoring, collection and archiving data of a manufacturing process in the automotive industry. The data are used by the maintenance planning system for carrying out checks and repairs on monitored equipment and machinery according to real functioning duration. I. INTRODUCTION The deterministic mechanisms using a centralized control system can not optimally handle the dynamics of a complex system. Moreover, an increased complexity leads to the impossibility of using, in some cases, of the deterministic algorithms. This is also visible in case of media production/ manufacturing areas which are very dynamic because of changing situations: malfunction/failure of providing utilities, some old orders are canceled, new orders arriving, delays in the supply of materials, disruption of supply, etc. [1][2][3]. The using of traditional models for hierarchical control or centralized production, has proved to be effective in some situations where the main objective is to achieve good yields, mainly due to the intrinsic caracteristics to determine a good organization. However, the dynamic and adaptive response to change is now the key to competitiveness, and the traditional approaches of manufacturing control software determines the construction of monolithic and centralized systems, requiring a huge effort and higher costs for implementation, maintenance and reconfiguration of control applications. These approaches are not appropriate because do not effective ensure the support for the current requirements imposed on the manufacturing systems, particularly in terms of flexibility, cost, agility and reconfigurability [2][4]. Therefore, we need a new class of intelligent and distributed control systems for the production in order to fill the gaps created by the centralized approaches. Several aspects are characterizing this new class and the requirements that are determined [5]: - use a distributed approach; a complex problem can be divided into several smaller problems to be solved using block control units; - each control unit is autonomous with its own objectives, knowledge and skills and incorporates smart features, however, none of these units has a global vision system; - overall control decisions (eg planning, monitoring and diagnosis) are determined by several control units (more than one), which means that the control units must work together, interacting in a collaborative way to reach a decision; - some control units are connected to physical automation devices such as CNC machines and robots. - control units are characterized by robustness, ability to reconfigure, reuse, connectivity and real time learning. A production control system that meets the above requirements, will operate in a completely different way when it is compared to traditional centralized control systems. The change from traditional centralized approach to addressing newer intelligent and distributed approach, is illustrated in Figure 1, adapted from [6]. MASTER SLAVE SLAVE SLAVE SLAVE SLAVE SLAVE SLAVE SLAVE AGENT AGENT AGENT AGENT AGENT AGENT AGENT Classic control (centralisation) Agent based control (cooperation) Fig. 1The difference between approaches for the traditional control and the control based on agents The multi-agent control and the holonic control of production are two adequate examples that address this new class of distributed and intelligent control of the manufacturing. These paradigms, introducing in practice the artificial intelligence techniques, have the ability to promptly and accurately respond to changes and are different compared to conventional approaches due to their inherent capacity to adapt to emerging without external intervention. A multi-agent system can be considered a computer system that is concurrent, asynchronous, stochastic and distributed. The above characteristics of a multi-agent