Technology marketing for industrial agents

Industrial agent technology to control production system has now a history of ten years in intensive research. However the technology is far away from being extensively exploited in industry. The present contribution shows the main barriers for agent systems to be implemented as an industrial solution and gives perspectives how this can be overcome in future research activities. The results presented are based on fundamental strategies of technology marketing, related research projects and a series of industrial experiments carried out to test virtual, industrial implementation. Technology marketing is a relatively unknown discipline of technology management. It deals with the development and application of methods to identify, characterize and evaluate future markets for innovations. Innovations in this sense are not only technical and product but also business strategies and administrative procedures. Technology marketing is a necessity in industrial planning where long term investments have to be supported by definitive arguments and also in strategic research planning In particular where public money is involved. The overall purpose of a continuous technology marketing process is to control research and development to avoid financial losses as a consequence of unfeasible technical characteristics of products, lacking user acceptance, social and political resistance and usability deficiencies. Typical technology marketing processes contain three main elements and derivates of them: Scenario techniques to get a model of the technology, Roadmapping to plan the research and development and identify the key elements and forecasting to research the markets. All of these elements normally involve a number of external competences to ensure the integrity of results. Since the beginning of the evolution of industrial agents, IPA undertook a series of technology marketing activities for agent technologies which escort the institute's technical developments in this field. The findings presented here are based on this technology marketing. Around 1990 a car manufacturer set up a project with equipment suppliers to realize an ultra-flexible manufacturing line for power-train components. The aim was to use a number of CNC-machines instead of the transfer lines in series manufacturing to produce single parts. This single part production was necessary to provide spares, prototypes and in production ramp-up. With several hundred different active parts and the fact that not all machines can perform all operations it turned out, that the logistics in the line and the sequence control could not be realized in conventional sequence control technology. Finally the solution could be achieved by implementing an agent system based on communication between virtualized products, the transport system and CNC machines. This success was the beginning of a series of cooperation projects in international dimension to advance agent technology for industrial manufacturing control. In important projects like Holonic Manufacturing Systems HMS and PABADIS developed generic models and industrial test applications. From 1996 on, the foundation for intelligent physical agents cares about standards and implementation rules. From 2000 on there is very less known about the technology development. This is why the IEEE technical committee on industrial agents has started now a review to collect actual information on the status of the various activities. Agent controls have, according to the projects results, many advantages. Robustness and flexibility seem to be extraordinary. Production ramp-up and reconfiguration are highly simplified. The question is whether this resolves a significant problem in industrial production and represents a strong motivation to change the whole control concept. At least some of the principle problems have been analysed by the various projects and were summarized as basic argumentation for agent controls. The main aspect is to regard any production as a process chain interconnected by a material flow. While the process controls are usually encapsulated in the respective devices the line operation is performed by decentralized sequence controls. Traditionally this sequence control is performed by decentralized PLCs. Programming the logic of the sequence is for large systems very difficult; in particular because all exemptions have to be foreseen and appropriate strategies have to be implemented. Around 65% of the program code in the PLCs is for exemption handling but still it is not sure that all possible exemptions are covered. As with all software around 3% of the code is faulty; undiscovered until the exception happens after years. Therefore robustness is an issue in large lines and one of the key motivations for agent controls. The number of changes in production lines with impact to the sequence control is significant. They range from about 300 in large process industry, e-g. breweries to several thousands in large automotive lines. These changes, often initialized by a product adaption, are extremely expensive because a large part of the sequence control code is affected and changes there often have unpredictable effects. The traditional sequence control is very flexible from an individual point of view, but if networked to large systems it becomes very rigid and reformation resistant. Improved flexibility for line sequence control is obviously an unresolved problem in many production areas. Never touch a running system is the strategy here. From a number of generic researches from technology marketing it is known that in most production areas the processes change by at least 3 % per year. Key drivers are new materials, quality control and eco trend. Installing or replacing processes in a line is difficult because many changes in the sequence control are necessary. Obviously, traditional lines control limit not only the flexibility but also the reconfigurability. As a side effect, the introduction of new processes is very often lagged until a new line is installed. These finding clearly show that a radical innovation in industrial production control methods are necessary and agent technology obviously can solve the key problems.