Planning a new or extending existing automated material flow system requires a high effort for the development of the material flow control. For example, all installed material flow components must be individually integrated into the material flow control system, and all system states must be taken into account and tested. This leads to long commissioning or downtimes, especially in large plants. Due to the long service life of the systems, defective conveyor technology components can often no be replaced with identical components. In this case, as in the case of modernization or expansion, a complicated adjustment of the material flow control and the corresponding downtimes must be expected.
In order to easy to operate, maintain or modernize highly automated material flow systems, the material flow components should be automatically connected to the control system, thus reducing downtimes. The Plug and Play principle, which has been established in the electronics sector for many years, serves as a model here. According to this principle, the components of a system are ready for use immediately after connecting them to a system. This principle is to be transferred to the material flow components of a material flow system. The various components are to be described as independent hardware and software modules. These modules then have the ability to integrate themselves independently into a system after connection and to take over tasks immediately. This requires coordination with the adjacent material flow modules on the one and with the material flow control on the other hand.
In order for an independent material flow module to establish a material flow system, it must also be able to perceive its environment. The recognition of interaction possibilities with other modules is referred to here as cognition in conveyor technology. Perceptible characteristics are, for example, the identification of possible load transfer points, their conditions of use (e.g. the orientation of the material to be conveyed) or also the transferable material to be conveyed. The cognition of the individual components enables the immediate recognition of changes caused, for example, by malfunctions or planned interventions. The material flow is then immediately adapted to the current conditions.
An at least partially decentralized material flow control system should reduce the complexity of the control system, permit modular additions and ensure redundancy by means of a distributed control system. For this purpose, control modules capable of performing a plug-and-play configuration and which are able to optimally control their components in the sense of the overall system must be developed. New conveyor technology modules thus bring their own part of the control system with them in a Plug and Play-capable system, which is integrated into the existing material flow control system through communication.
The Plug and Play principle has already been adapted for material flow systems in various research projects in combination with decentralized material flow control. However, only systems with similar material flow components were considered, such as an airport baggage system. The focus of this project is on the networking and self-configuration of heterogeneous conveyor technology components, such as the combination of roller conveyors, robots, driverless transport systems and storage systems. In addition, a further project objective is the visualization of decisions made by the conveyor technology components. This should increase the acceptance of decentral controlled and self-configuring components by the users.
For the derivation of Plug and Play capable material flow modules, a module definition for the hardware and control software must be developed, which is derived from the executable functions of the components.
The material flow system is to be controlled by a multi-agent system in which the control logic is distributed decentrally among various software agents. The agents independently configure and control their own module and are able to communicate with other modules. The necessary agent architecture and communication mechanisms will be developed in this project. In addition, cost functions will be developed for different classes of conveyor technology modules, with which control decisions can be made decentrally and in collective agreement in order to enable an efficient material flow. In addition, methods of machine learning are considered, which enable the agents to optimize control decisions independently.
Basis of the self-configuration of conveyor modules is the derivation of neighborhood relationships and the identification of interaction possibilities with the environment. For this purpose, a concept is to be developed that defines interaction spaces of modules and allows an alignment of overlapping spaces. On this basis, possible interactions and their conditions will be agreed upon.
Within the framework of this project, a visualization will also be developed, which will enable the user to view the knowledge levels of the agents. Subsequently, the user can understand the decisions made by the agents.
Finally, the concept developed in this project will be implemented and validated in a demonstrator.