O. Bochmann. Probabilistic Approaches in Multi-Agent Systems for Manufacturing Coordination and Control. PhD thesis, Katholieke Universiteit Leuven, Department of Mechanical Engineering, Leuven, Belgium, 2005.
This dissertation addresses the goals of robust, adaptive, scalable, dynamic and distributed agent systems and their application in manufacturing control. A synthesis of methods from dynamical systems theory, computation theory and inductive inference has been developed that suggests a constructive approach to address the problem of emergent computation. Most applications of multi-agent systems in manufacturing control use the term emergence in a rather vague fashion, describing a process where macro-level behavior arises from the interaction of some micro-level components. The actual cause for this emergent phenomenon - the intrinsic computation of the process - has not yet been taken into account.
In this context, the multi-agent system is understood in the sense of an agent based model where each thing is represented as an agent in a population of distinctive agents. Agents observe their environment and reconstruct a model based on patterns that reflect the structure in which the process stores and transforms information. This unique model - it is minimal in complexity and maximal in predictive power - is used in forecasting the manufacturing process. This pattern that allows for a large reduction in complexity a small reduction in accuracy is the emergent phenomenon in forecasting.
- Current State of Play in Manufacturing
- Multi-Agent Manufacturing Coordination and Control
- Probabilistic Methods: Modelling Uncertainty in Manufacturing Systems
- Complexity and Complexity Measures
- Emergence and Self-Organization
- Forecasting in Manufacturing Processes
- Distributed Decision Making in Manufacturing Control
- Results and Discussion