Holonic and Multi-Agent Systems for Manufacturing

Abstract

This PhD thesis addresses the shop-oor control level in manufacturing and industrial production at some important points for the future. The main contributions are principles and software relating to the control of production systems and hereunder production devices. Specifically challenging at the production device level is the real-time, applicationoriented and sensor-integrated motion control of industrial robots. At the higher level, the production control system for a shop-oor, considered as an ensemble of independently controllable production devices, is addressed. The challenge at this higher level regards the representation of all production devices of a shop-oor by suitable control interfaces for a distributed deployment of the control system. The control of industrial robots and their integration into a production system is proposed to be troubled by each robot manufacturer developing their own native application platform. Each native platform oers some pre-chosen paradigm for programming and its own restricted set of technologies available for application development. For migrating the real-time motion control of industrial robots onto application platforms oering higher exibility, or simply the right technologies, than that found in the native controllers, several software frameworks and platforms exist. The work underlying this PhD thesis has produced such a software framework, dubbed PyMoCo. It is entirely implemented in the high-level, interpreted programming language Python, and allows fast and highly exible real-time motion control and application integration of industrial robots. At the shop-floor control level, it is proposed by a segment of the research community dealing with production control, that holonic and multi-agent control architectures are suitable paradigms for bringing highly intelligent, computationally powerful resources into the direct control loops in production systems. The use of autonomous holonic or agent-based systems for distributed real-time shop-oor control presents some inherent difficulties regarding formal validation and verification. Hence, in the research community it has been suggested to use real-time, realistic emulators of the production systems as platforms for experimenting, developing, and validating holonic and agent-based production control systems. This PhD thesis presents principles for using the Blender Game Engine for implementing a physically realistic real-time emulator for the collection of production devices of a production shop-floor. An emulator has been designed and implemented for an extended version of a prototype shop-floor system set up in a laboratory, and an experiment control system has been implemented to validate the emulator. The presented emulator is developed to a state of being suitable for use as a realistic platform for development of a distributed, autonomous multi-agent system for control of the extended prototype shop-floor system.