2.3.2. Competence Islands

The existing large-scale production method, organised rhythmically in production halls and working in the production cycle time, will no longer be able to respond to future customer requirements. Today's "static" production and assembly lines will be replaced by a set of autonomous workplaces called competence islands (Figure 4). It can imagine as virtual production lines, formed dynamically and virtually based on real needs. The competence islands will be equipped with technologies and cooperative robots capable of working safely and reliably with people [18]. Figure 4 illustrates the vision of the competence islands.

**Figure 4.** Concept of competence islands.

New manufacturing systems should, therefore, be designed as small, highly flexible production units, which will be deployed where there is sufficient real demand. Such manufacturing systems will be designed for the production of the selected product family, which requires that their concept be built on the principles of reconfigurable manufacturing systems.

The activities of future manufacturing systems will be organised differently. Classic production and assembly lines will only be maintained where it is still economically advantageous. Future production will seem to be complete chaos to the outside observer. It will seem that materials, intermediate and elaborate production, and mobile robots are moving unplanned and chaotically. However, each of them will be guided by a strict logic of the parent level, which will enable it to be relatively autonomous. It will, therefore, be organised chaos. For production management, the principles observed from nature, which offer an evolution of proven, optimal practices, will be used.

Holding a strong position in future factories will be intelligent mobile robots and mobile robotic systems and platforms. Thousands of such robots will ensure the movement of the worked products and their processing in a seemingly chaotic world.

The production will be organised as a living organism resembling an anthill, in which the ants appear messy but are strictly organized and specialized, and each of them performs precisely defined tasks that ensure the survival of the anthill.

The product, production equipment, technology and the entire production system will be changed. Manufactured products, manufacturing equipment and mobile logistics means will become intelligent

and communicate with each other. In real-time, they will exchange and share all the necessary data and information.

Mobile robots, transporting a staged product, will move between the competence islands, while the product itself will determine the required operations and plan their order. The observer will not see the classic production line; what will be observed is the apparent physical chaos. However, there will be a hidden virtual line (its digital and virtual data model) made up of the competency islands required for the production of the customer product.

Future production will not be structured according to the production rhythm line, as is the case today, but according to the content of the work to be done. Functional relationships and not fixed cycle times will have a decisive role. This type of production environment will be suitable not only for small production companies but will be particularly advantageous for those types of products that work with high volumes, highly variants of production and which aim at high flexibility and efficiency. Such systems will be able to react more effectively to fluctuations in demand and rapid changes in the models produced by requiring different production technologies. The company Audi claims that the production islands will be much more efficient than today's linear concept.

#### 2.3.3. Virtual Manufacturing and Intelligent Agents

The simulation model, detailed, hierarchical and more leveled, containing all the significant factors of the production process, will allow a new type of management, which will be built on dynamic analysis and prediction. If we link such a model to the information sources of production and its sensory system, it will operate as a human organism and will behave adaptively while using real-time data. It will work with its own "physical map", similar to the human body. Today's experiments with in-memory computing are about such future management systems.

In the area of virtual manufacturing and intelligent agents, we meet with the solution of syntactic, semantic and pragmatic boundaries [19]. The first aspect is syntax, which is important for the machine to machine communication. The communication capabilities of agents in multi-agent systems (MAS) are characterised by data exchange mechanisms based on proprietary messages in the form of Extensible Markup Language (XML) syntax and according to MAS standard communication models, for example, defined by the Foundation for Intelligent Physical Agents (FIPA). For establishing CPS in manufacturing environments, the usage of web services is inevitable for the realisation of scalable information exchange. Thus, in addition to a language that describes the information and provides data syntax and semantics, a common underlying mechanism for transfering the information from one entity to another or to perform interactions is needed [20], so the second aspect is semantics. In semantics, we find ontology, annotations, and definitions. The semantics give a mathematical meaning to formulas that, in theory, could be used to establish the truth of a logical formula by expanding all semantic definitions [21]. To provide a proper description of an agent that is readable, understandable and interpretable by other agents in an integrative manner, the description model of each agent needs to follow common design principles, e.g., by making use of a common ontology description, fixed namespaces for agent capabilities (among others) is needed [22]. According to [23], communication between agents can be realised if all agents can find and identify each other and all agents make use of a message system with a predefined ontology, which every agent can understand. In [22], one desired goal to deal with high amounts of raw data from the shop floor would be an automatic assignment of information from the lower levels of the factory. Automated annotation of production information with context information, such as metadata, would reach both machine-readable and interpretable information for autonomous process optimisation as well as a data basis understandable by humans. The third aspect is pragmatics, which means the question of how to use axiomatics to justify the syntactic renditions of the semantical concepts of interest. That is, how best to go about conducting a proof to justify the truth of a CPS conjecture [21]. That means that we must define how to use axiomatics to justify the truth.

New sensory systems allow for the end of the transition from static monitoring systems (sampling and data collection at an interval of one day) to dynamic monitoring (sampling in microseconds, as it is done today in the process industry). The average values of output parameters (statistics) must be replaced in the new generation monitoring systems with the immediate values and trends of changes in the last, most significant periods.

The monitoring system must include a watchdog, a function that will trace (seek) potential problems, an early warning system that notifies the occurrence of potential problems and an automatic correction mechanism that resolves the potential problem before its real emergence.

If we have enough data about the production system, we can, thanks to virtual reality, create a virtual image of production (its dynamic hologram) and then in such a "reality", virtually track the effect-change factors (visualise them), observe future status and decide on the changes that will be made. In long-enough time, such a system can gradually learn, with the support of a machine-learning system and knowledge system, how to adapt to changing surroundings. If a person makes a decision instead of using computers, manual management is applied. In direct management, in automatic mode, the direct control system decides, and manual interventions are replaced by automatic steering. In the case of manufacturing control, the virtual twin of each real object will be represented by an agent. We refer to a large group of such agents and their management as multi-agent systems (MAS) [24]. Future production will be represented by two worlds: the real world's and its virtual reflection, also called the virtual world. These worlds will be mutually integrated through data. Production data will be collected and processed in real-time. Almost immediately, information about each object in the production will be available—what it is doing, in what state is it located, what is further planned, and what is lacking. The status of each product, machine, tool, device, jig, robot, or person will be immediately scanned, and the processed information will be sent to the control centre. This information will be compared with the next step in the processing of the products, the sequence of future steps will be generated and the system will make the necessary decisions for further processing of the product. The virtual world will allow, if necessary, the simulation of future status and prediction of the effects of the necessary control actions.
