استفاده از رویکرد چند عامل در برنامه ریزی تولید و مدل سازی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|5542||2001||8 صفحه PDF||سفارش دهید||5296 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Engineering Applications of Artificial Intelligence, Volume 14, Issue 3, June 2001, Pages 369–376
The paper describes an experimental multi-agent system developed for and aimed at a computer-supported project-oriented production planning. The system is based on a heterogeneous hierarchy of agents of three types that reflect the managerial structure of the manufacturing enterprise. To improve the system efficiency, a new formalism—called tri-base model—of the multi-agent internal communication/negotiation mechanism has been introduced. The tri-base model that has been tested in an industrial environment is treated in detail in the paper.
The research described here has been focused mainly at project-oriented production. Such type of manufacturing is characterised by the majority of labour and costs spent on project-related activities, as product quotation, project configuration and design. Whilst in manufacturing-oriented production there are thousands of items being produced once the product has been designed, in project-oriented production just a single or very limited number of pieces of one designed product get produced. As each product is unique and the factory administers a number of parallel projects, a certain type of computer assisted production planning and simulation needs to be employed in order to optimise the manufacturing processes. Tesla TV Comp., Prague, Czech Republic, has been considered as a typical instance of an enterprise the manufacturing activity of which may be classified as project-oriented production. This company manufactures TV and FM transmitters and passive transmitter elements. There is no production line and it is difficult to formalise the production as a traditional continuous process. The production is aimed at customer specific and mutually nearly independent projects of composing unique final products according to a manufacturing documentation designed for the particular business case. The production plan and factory activity schedule must be modified to reflect the large variety and complexity of the particular job order. Consequently, instead of planning and simulation of a flow of semi-products to be assembled, the project-oriented production needs a rather sophisticated information support, which could facilitate simple planning and subsequent optimisation of manufacturing of the unique final products. Currently, ill suited information flows, lack of communication among particular production units and low utilisation of available information processing solutions is generally what makes the production difficult to understand, model, plan and consequently optimise. Attempts to create global “monolithic” software solutions, no matter how well hierarchically structured, meet usually dead ends. Regardless whether re-use and integration of existing pieces of software is expected or a brand new system is being implemented, there is a strong need for a novel solution in the form of a highly distributed and heterogeneous system. This opinion is advocated by the fact that production is flexible, and frequent changes of production targets, manufacturing facilities, system knowledge and planning strategies are inevitable. Maintenance and updating of single and independent modules is highly desired. The process of production planning and simulation in project-oriented production is driven by •highly specific knowledge of a project-planning engineer who is in charge of the project design and supervision, and •potential availability of various departments, units, and machines within the enterprise. The multi-agent systems represent a new and promising solution to problems as complex as those outlined above. We suggest replacing the centralised process of production planning with the process of negotiation among the autonomous cooperating agents. The organisational structure of the respective multi-agent system shall mirror the organisational structure of the production we want to plan. We go further with suggesting an alternative process of negotiation based on agent social models maintenance that minimises inter-agent message exchange.
نتیجه گیری انگلیسی
Experiments have shown that multi-agent systems can serve as a good formalism for project-driven production planning and modelling. The classical production planning and scheduling techniques have been substituted by the process of negotiation about possible collaboration and resource sharing among different information units. This is very similar to the situation when people discuss about a joint work to be done. As the production optimisation is to be carried out through the process of negotiation among agents representing various information units within an enterprise, the types of agents must reflect the capabilities of different departments of the enterprise. Moreover, the agents must be aware as much as possible of capabilities of their peers in order to avoid blind and redundant communication. The devised tri-base acquaintance model is used within the wrappers of agents of heterogeneous nature to accomplish the desired functionality. The tri-base acquaintance model splits information on various collaborators according to the information nature within distinct bases. General problem-specific knowledge and time-dependent data are administrated separately to achieve efficiency and to ensure nearly optimal selection among the best possible plans. The proposed acquaintance model is considered as a good formalism for production planning for number of reasons: • clear separation of collaboration and problem solving knowledge, • transparent organisation of knowledge and data enabling easy maintenance, • highly dynamic behaviour with minimal communication traffic achieved through the subscription mechanism. A prototype of a multi-agent system based on the tri-base acquaintance model has been implemented. It includes agents of various nature and functionalities each representing a specific activity and functionalities of a cluster of manufacturing units within the enterprise (project planning agent, project managing agent and production agent). The particular results of the industrial application are very promising. They document high flexibility and efficiency of the approach.