Today, end consumers increasingly want to have a direct influence on the products they order: from a self-designed logo and a suitable product name to individual recipes and, ultimately, direct delivery to the desired address. These wishes can open up new markets and sales channels for manufacturers. In practice, however, operators of highly specialized and optimized production facilities in particular are finding it difficult to make the leap into the new age.

The interdisciplinary research project 'RoboFill 4.0' has therefore specifically chosen the beverage and bottling industry, a sector whose highly automated systems have to achieve the highest possible output. The 'RoboFill' concept envisages that the end consumer can put together their own personal mixed beer drink via an online customer portal and provide it with a freely designed label. The individual customer configuration is passed on to production as an order. The individual production modules are designed in such a way that they can work independently. A linear transport system and industrial robots with intelligent gripper systems are used for handling at the processing stations and transport within the system, while at the same time breaking up the linearity. Labeling and printing on the bottles is carried out by a special inkjet printer. The result is customized production for the beverage industry that can be adapted to any discrete production process.

AI in the multi-agent system

Simplified structure of a multi-agent system including the hierarchical structure of the agents. The distribution across several containers creates a decentralized MAS.

© Infoteam Software

A multi-agent system (MAS) is the fundamental building block that enables decentralized production driven by the product itself. The starting point is physical objects - such as industrial controllers, sensors and actuators - whose digital shadows are mapped as software agents and thus become cyber-physical objects. These can communicate with each other and exchange information via a communication infrastructure. Production control focuses on the products to be manufactured, which are also mapped as cyber-physical objects. The digital shadows ensure that simple objects, such as drinks bottles, are given artificial intelligence (AI) for the production process and are able to communicate. To this end, each object is assigned to a digital shadow (software agent) via a unique identifier, such as a QR code.

The AI of the agents is reflected in the combination of status, goals, communication and problem-solving methods. The agent achieves its goals independently using configurable strategies. To do so, it uses all the information available to it and communicates both with its environment and with other agents. A decisive advantage of decentralized, virtual agent intelligence is the dynamic compensation in the event of failure of individual components: The agents then find new solutions independently. This means that production can continue even if a subsystem fails.

The iAgent framework

Comparison of RAMI 4.0, the Industry 4.0 component and iAgent, where the administration shell is realized as a software agent.

© Infoteam Software

The RoboFill project is implemented using 'iAgent' from Infoteam Software, a framework for the development of decentralized MAS in an industrial environment. It consists of the InfoteamRuntime as a runtime environment for containers that contain different agents. A decentralized MAS is created by distributing the containers across several execution environments (hosts). The Message Transport Service (MTS) is available in each container for communication between the agents. In addition to ready-made, ready-to-use FrameworkAgents, each container also has so-called ApplicationAgents. These serve as a template for users to implement their own agents and adapt the MAS to user and system-specific requirements. The basic types MachineAgent and TransportAgent are used to implement cyber-physical objects. They represent the digital shadow of system and production components. The core element within product-driven production is the ProductAgent, which represents the digital shadow of each product to be manufactured. It has all the information on the necessary production steps and independently organizes the production of the assigned product.

The InterfaceAgents category describes agents for implementing interfaces to and from the MAS. As the connection between the real object and the software agent, the LinkAgents implement the communication interface to external components. Their tasks include converting to specific protocols, controlling access rights and forwarding the respective communication requests between agents and machines. Which interfaces and protocols are required for communication with an external component depends on the system-specific conditions.

OPC UA as an access interface

OPC UA can serve as an access interface for data exchange or method calls between agents and physical objects. If the machine is able to offer its own OPC UA server, the described coupling agent can be implemented to connect this machine. If the machine does not have its own OPC UA server, it is possible to process the necessary information via an additional hardware component (IoT gateway).

This means that the fieldbus protocols already in use can still be used at machine level. In addition to communication between machines and MAS, the coupling agent takes on other tasks, such as implementing security-relevant aspects such as encryption and authentication. It can also integrate other external systems such as databases, web stores or interfaces for visualization and configuration.

Integration into RAMI 4.0

The framework developed by Infoteam Software specifically for use in the industrial environment can be integrated into RAMI 4.0 (Reference Architecture Model Industry 4.0) and also maps the Industry 4.0 component - a model that describes the fundamentals of Industry 4.0 in more detail. RAMI 4.0 is based on a three-dimensional coordinate system with the following axes:

• Hierarchy Levels: The hierarchy levels from IEC 62264 are shown on this axis. These describe the functionalities within production.
• Life Cycle & Value Stream: This axis represents the life cycle of systems and products. It is based on IEC 62890 on life cycle management.
• Layers: The 'Layers' axis describes the digital shadow and therefore the cyber-physical object of a machine or product.

The iAgent framework can be used to create MAS for industrial production by assigning a digital shadow (software agent) to each real component. All elements of the three-dimensional coordinate system of RAMI 4.0 are taken into account.

The Industry 4.0 component

The MAS Designer is used to configure and parameterize individual agent systems.

© Infoteam Software

The Industry 4.0 component is the first model based on RAMI 4.0 and describes the fundamentals of Industry 4.0 and the properties of cyber-physical systems. In this model, the asset administration shell forms the digital part (i.e. the digital shadow), while the associated component, for example a machine, represents the real part. If the components in production fulfill the characteristics of the Industry 4.0 component, they become Industry 4.0-capable. Throughout the entire life cycle, the production data can be stored in the electronic container - in 'iAgent' in the software agent - and thus made available to every instance involved in the value creation process. The asset administration shell, including the electronic container, forms the digital shadow of real objects.

The central properties of the asset administration shell of an Industrie 4.0 component include

• Communication: For Industry 4.0-compliant communication between real objects and their digital shadows, all components must be capable of communication and communicate using a uniformly defined language. The connection between the asset administration shells is established via Industry 4.0 communication. Communication in 'iAgent' is based on a message schema in which agents and containers exchange information according to a defined ontology concept.
• Data: All data relevant during the entire life cycle must be stored within an electronic, secure container. This container is part of the asset administration shell in the Industry 4.0 component model. In summary, all relevant data that is generated during production forms the digital shadow of a component. All this data can be stored in a software agent.
• Functions:Various functions, for example for configuration, operation and maintenance, must be providedinthe asset administration shell. These and other functions, such as for project-specific business logic, are provided by 'iAgent'.
• Services: The asset administration shell data can be made available in the company network or in the cloud, for example, and made transparently accessible to every user and for every use case via IT services. In 'iAgent', all relevant data is stored in the software agents and can be passed on to other services such as database or cloud systems via standardized interfaces.
• Integration: The combination of Industry 4.0-compliant communication protocols and the data available in the asset administration shell enables horizontal and vertical integration in production - one of the key aspects of Industry 4.0. This is exactly what the MAS solution makes possible.
• Seamless knowledge: Information is seamlessly available for engineering, operation and maintenance via the asset administration shell - in 'iAgent' via the software agent.
• Modularity: Information on machine parts and components is also stored in the asset administration shell. This also applies to production components that do not have their own data interface. This concept means that every component involved in production, including the product, for example, becomes a smart part of networked production. Software agents can be defined so granularly that, in addition to machines and stations, individual components of a machine can also be mapped as a digital shadow. In this way, even components without a data interface and without intelligence become smart components thanks to the intelligence within the agent.

Focus on the production planner

With iAgent, multi-agent systems leave the academic field of application and move into modern production - up to batch size 1. Production planners are therefore able to create an intelligent production system driven by the product based on the configuration of a MAS and the associated components. The combination of standard agents and their specific configuration, parameterization and description creates a MAS that is specially designed to meet the needs of the respective production. iAgent provides defined solution algorithms that can be extended with your own logic modules to overcome the challenges within production.

Graphical editors and user interfaces support the production planner during configuration. At the beginning, he defines the scope of the system by selecting generic standard agents. The production planner must then configure and parameterize these with regard to the special conditions and requirements of the respective production. Typical configurations and parameters include the data within the agents, interfaces to other systems or agents and the definition of services that agents offer or require. This definition enables the agents to communicate and solve complex issues in cooperation.

The digital shadows are connected to the machines via the OPC UA interface as standard. Communication parameters and data points can be configured via user interfaces. The production planner can also configure the distribution of agents to decentralized execution instances, including the necessary network and security settings.

After the initial configuration of the MAS, tasks such as maintenance, servicing and expansion of the system can be carried out in a few simple steps.

Authors:
Alexander Urban is a System Engineer at Infoteam Software;
Angela Ringlein is Marketing Communications Manager at Infoteam Software.