Combining Product- and Resource-Related Reasoning for Agent-Based Production Automation

Beschreibung
The use of distributed control systems is widely regarded as one possible approach for designing cyber-physical production systems (CPPS) in the sense of Industry 4.0. The essential goals of such systems are to establish an autonomous flexible behavior and to reduce the effort for engineering and commissioning. One approach to realize a distributed control system is to apply the paradigm of multi-agent systems for creating intelligent modules. The basic principle of agents is to solve control tasks by forming distributed software entities that collaboratively realize problem-solving during runtime.

This thesis presents the concept, implementation, and evaluation of a hierarchical agent architecture, which is based on cybernetic principles. Furthermore, it investigates how intelligence can be distributed cohesively between the product and the resource level of an automation plant. Therefore, two agent prototypes are proposed, which possess the ability to anticipate the system’s behavior a priori based on discrete and continuous knowledge bases. This foundation enables the agents to realize reasoning and decision-making during runtime for operating the plant with the ability to compensate breakdowns as well as to achieve optimality regarding multiple criteria.
The approach was implemented as a real-time capable control architecture based on MATLAB/Simulink and by incorporating EtherCAT as a field bus system. Evaluation studies with a real and a simulated production plant revealed that by combining agents on the resource and the product level, an autonomous behavior could be achieved, leading at the same time to optimal and robust operation.