Research On Multi-Agent Based Shop Floor Control Architecture And Technique

Globalization and information development is changing the manufacturing environment, and in turn profoundly influencing a wide range of areas including traditional management, control mode, organization structure and decision-making rules. As an important part of the manufacturing system, shop floor affects the agility and intelligence of the whole enterprise. Its control architecture essentially determines the enterprise's production efficiency and competition ability in the market. Computer network and distributed artificial intelligence provide fundamental technologies for research on shop floor control mode. In this thesis, we studied multi-agent based control architecture and scheduling for networked shop floor.In the thesis, the background of agent is introduced and the characteristics of Multi-Agent System (MAS) based on academic studies of single agent are summarized. The essential characteristics of agent, such as apperception, autonomy, cooperativity are clarified and emphasized. Communication and cooperation among agents are analyzed, and the modeling rules of MAS applied in manufacturing system are presented.We proposed an order-oriented shop floor control model, which is based on MAS and dynamic_logical cell to overcome the weakness of the existing distributed and hierarchical form. The model is divided into three layers, namely the shop floor layer, the dynamic cell layer and the equipment layer. The tasks of shop floor layer are to receive, decompose and administrate orders. The dynamic_logical cell is the core of the dynamic cell layer, which organizes manufacturing resources in the level of process planning of parts. The equipment layer is composed of resource agents which are categorized by their characteristics and scheduled by the schedule agent of the corresponding type.In the existing floor shop, manufacturing resources are usually disposed with either of two modes, One is the GTjype cell. In this type manufacturing resources are organized by group technology; Another type of cell is organized by machine tool cluster. According to the principle and methodology proposed in this thesis, regardless of its cell, any resource can be integrated in to our dynamic_logical cells owning to theapplication of Network and MAS technology, thus displaying great advantage in improving dynamic reconfiguration and resource sharing at shop floor.At the same time, software architecture and dynamic scheduling strategies related to this control model are also presented in this thesis. Contract Network is used in the negotiation among agents. Synthesized scheduling criteria are proposed, which includes processing time, machining cost and efficiency as well as transfer routes. Optimization of dynamic scheduling is achieved by adjusting the weights of these parameters. Partheno-genetic algorithm is employed in simultaneous invite-bid problem to avoid illegal chromosome in inheritance.Shop Floor Markup Language based on XML is presented as a describing language in communication among agents with extension of new performatives of shop floor based on KQML.On the basis of a program of advanced manufacturing R&D initiative funded by the State Development and Planning Commission, the thesis designed a prototype of shop floor control architecture, constructed software framework with UML and provided a scheduling example.