In the context of the fourth industrial revolution, the trend towards individualization, and the legally driven development of new drive concepts, the production paradigm of Modular Production with decoupled stations and flexible material transport by an automated guided vehicle system has emerged in the automotive industry. To utilize the efficiency advantages of Modular Production there is a need for action in the development of effective control systems. This work addresses this need by developing a centralized, decentralized, and learning control system. In the centralized approach, a central computing unit calculates schedules for the production system in a rolling and event-driven manner using a Genetic Algorithm. In the decentralized approach, products negotiate with resources in an auction-based multi-agent system for resource allocation using the Contract Net Protocol and first-price auctions. The learning control system applies Proximal Policy Optimization from Deep Reinforcement Learning to collect interaction-based experiences, from which it learns scheduling via trial- and-error. The evaluation is conducted based on a reality-oriented production example under varying production conditions. Two recommendations for the development of control systems in Modular Production can be derived from the evaluation: First, this work recommends the development of learning control systems, especially when higher complexity levels are present in the production configuration. Second, an interlocking of planning and control is of elementary importance in the implementation of new modular production systems, because this allows the Modular Production and its control to be optimally coordinated.    «

In the context of the fourth industrial revolution, the trend towards individualization, and the legally driven development of new drive concepts, the production paradigm of Modular Production with decoupled stations and flexible material transport by an automated guided vehicle system has emerged in the automotive industry. To utilize the efficiency advantages of Modular Production there is a need for action in the development of effective control systems. This work addresses this need by devel...    »