Complex Decision-Making in Economy and Finance. Pierre Massotte
limits.
1.4.4. Message flows in complex information systems
1.4.4.1. Distributed information processing
This case study was taken as a typical example of a situation that many people are familiar with: it is a Global Information System Network of which the Internet, or Intranet, is one of the elements. This heterogeneous network includes a large number of server and client centers. Each center or agent has its own strategies and can perform different routing or control tasks, as shown in Figure 1.3.
Knowledge and information are distributed throughout the network [MAI 94]. The evolution and growth of such a system with tens of thousands of nodes cannot be ensured, controlled or planned from a central computer. In such a network there is an “apparent” anarchism; each node (agent) is an autonomous computer system: it has the possibility to direct traffic according to predefined rules and the saturation state of the network. It can also manage information flows according to their nature and the state of nearby cells. Indeed, and as we have seen, even though some cells, or a group of cells, have chaotic behavior, there is often a smoothing of chaos at the global level.
Figure 1.3. A dynamic industrial system with nonlinear interactions
In this figure, we see that it is cellular automata (CA) with independent but interacting agents who do not have knowledge of the overall consequences of their actions. The probabilistic data and incomplete or inaccurate information they manipulate, combined with processing delays, result in the emergence of various attractive states such as fixed points, oscillations or even deterministic chaos and auto-catalytic mechanisms that converge them into particular collective states and behaviors. There is the emergence of a collective “intelligence” that cannot be predicted and controlled in advance and that highlights the fact that reductionist approaches cannot be referred to. For these industrial, dynamic and nonlinear interaction systems, the development of models based on evolution equations makes it possible to characterize and study them.
1.4.4.2. Emergence of collaborative work
As already mentioned, chaos and fractals are part of the same field of mathematics and underlie the principles of autonomy and self-organization. These properties are exploited in cellular automata, involving stochastic functions; solutions can therefore emerge from systems composed of communicating entities and functions that rapidly evolve into simple – periodic or quasi-periodic – and strange attractors. Their properties can therefore directly influence control systems, management methods and organization. The impact on new skills requirements, people’s education, structure and social aspect in the firm have been particularly studied in industry in Germany [WAR 93].
1.5. Applications of new concepts in industrial systems
1.5.1. New features and functionalities to consider
In the field of complexity, the presence of deterministic chaos in electronic circuits and signal processing is known to many automation engineers [EEA 95]. Without going into detail, we have simplified the representation of the new paradigms to be implemented and positioned them in a graph. The aim here is to switch from the “standard” system to “OKP” (One-of-a-Kind Production). In this context, the concepts developed above will have an impact on the methods and tools used to control and manage production systems. As indicated in Figure 1.4, there is an opposition of characteristics, resulting in a conflict between capabilities and goals.
Figure 1.4. Positioning of paradigms in production management
In view of the new constraints observed in industry and the changing needs of consumers, it will be necessary to increase both the possibility of producing specific devices, also known as “attributes” (and no longer “finished products”), personalized, in small quantities and on demand, with maximum efficiency.
In short, and this is a change, clients are becoming inflexible, while production systems and products must be more flexible and adaptable. In case of difficulties, we will even say that it is a supply crisis and not a demand crisis. The initial approach consists of developing and using information technologies as a factor of innovation and resolution. However, these only concern process automation and are based on concepts and information theories that have certainly evolved and led us to JIT (Just-In-Time), CIM (Computer-Integrated Manufacturing), FMS (Flexible Manufacturing Systems) and so on. Thanks to robotics too, we have been able to improve the flexibility of production systems, scheduling techniques, etc.
Processes have therefore benefited from these developments and have seen their performance increase. However, as systems evolve towards greater complexity and these evolutions have their limits, it follows a simple observation: the difficulties we encounter are not technological in nature, but involve, when they are resolved, a change in logistical and organizational approaches. In fact, it will be necessary to introduce a paradigm shift such as fractal chaos or self-organization.
1.5.2. Design of complex industrial systems management tools
When we consider the new structure of production workshops, we must instead speak of production networks because production systems are, in fact, made up of cells and resources that communicate and interact with each other in such a way as to constitute self-configuring systems. Such systems have been extensively studied at the VTT Laboratory in Helsinki, Finland [RAN 93].
Autonomy and self-organization are essential characteristics of future production systems. Combined with these concepts, and to produce customized products, with a high reactivity to meet demand, it is important to implement new principles of planning, scheduling, piloting and control. However, limiting the ability of systems to be flexible and adaptable through the improvement of their operation or production control functions leads us to play on other potentialities such as interactions between functions, autonomy and system dynamics. This is mainly due to the increasing difficulties encountered in solving problems of scheduling, synchronization and development of the right ranges: as everyone knows, it is indeed very difficult to determine good scheduling in multi-product processes and multi-processes under conditions of nonlinearity and uninterpretable discontinuities.
Of course, there has long been an attempt to simplify processes and improve their flexibility by developing increasingly complicated strategies and algorithms (in a study conducted in 1993 in the industry, more than 430 scheduling algorithms corresponding to specific problems had been identified, around the production control community, by our team in IBM. Some of them are now considered as algorithms for AI; this demonstrates the many various approaches in machine learning and solution elaboration in planning and scheduling). The question then is: should we continue to try to solve each new problem in a traditional way? How far will we go in the level of complexity to be understood? Can we not work on new and more original approaches? Can we not take better advantage of the intrinsic properties of the systems concerned? How can we exploit new architectures or properties? Here again, we are tackling a new paradigm, which will be part of the familiar domains to come and which are called: fractal factory, virtual factory, agile manufacturing, etc.
Future production systems will not only have to be adaptable, but also have to be able to change the configuration of their own structure as well as the structure and functional principles of the production system itself.