Distributed Self-Optimization of Modular Production Units: A State-Based Potential Game Approach.

Area of Science:

• Engineering
• Computer Science
• Control Theory

Background:

• Industrial production units require efficient optimization strategies.
• Distributed systems offer flexibility but pose coordination challenges.
• Integrating artificial intelligence with control theory is crucial for modern manufacturing.

Purpose of the Study:

• To develop a novel distributed optimization approach for modular production units.
• To leverage potential game (PG) theory and machine learning for autonomous system development.
• To enhance the adaptability and efficiency of production processes.

Main Methods:

• Modeling the production environment as a state-based potential game (PG).
• Assigning actuators as agents aiming to maximize utility through learned optimal behavior.
• Developing a novel learning algorithm based on a global interpolation method.
• Applying the approach to a laboratory-scale modular bulk good system.

Main Results:

• Demonstrated the effectiveness of state information in dynamic game environments.
• Gained insights into the learning dynamics and process behavior.
• Achieved encouraging results in optimizing a modular bulk good system.
• Validated the potential for IEC 61131 conforming PLC implementation.

Conclusions:

• The proposed distributed optimization approach offers plug-and-play functionality and online capability.
• The system demonstrates fast adaptation to changing production requirements.
• Potential game theory combined with machine learning provides an intelligent solution for autonomous production systems.