DEVELOPMENT OF AN INTEGRATED DOMAIN REPRESENTATION FOR CONFIGURATION, ACTION PLANNING, AND LAYOUT PLANNING OF ASSEMBLY SYSTEMS

During the past fifteen years, robot work cells have become a standard component of many industrial production systems. Up to now, the design of a robot work cell or a automated work station for a specific manufacturing process is a demanding task, although tools have been already developed for different design steps. The problem is that each tool is suitable only for one design step such as assembly planning, cell configuration, layout planning, and robot programming. The mutual effect among these design steps cannot be considered, although it is necessary for a high quality design. Thus, the common way is the manual design by human designers who have to work cooperatively for this task. This paper presents a new theoretical approach for an integrated design process consisting of process planning, component selection and work station planning. This article presents the structure of a domain representation for an integrated planning system for assembly stations. This domain representation allows to integrate the main steps of the design: configuration, action planning and layout planning. Each selection of a manufacturing resource constrains the selection of further manufacturing resources of the same assembly station or influences the course of action which has to be taken. For example: A specific joining process requires specific manufacturing resources and a defined course of action. In addition, it must be possible to arrange all the manufacturing resources in the working space of the handling system. The selection of manufacturing resources and their spatial arrangement has strong influence on cycle times and the investment costs of assembly cells. Therefore, the planning of assembly systems is a combined configuration, action planning and layout planning problem. In order to achieve optimal results, it is required to deal with these dependencies during the planning process. Action planning is a classical application area of artificial intelligence. Therefore, a number of methods and systems have been developed. STRIPS [1] is the classic action planning system. Afterwards, nonlinear planning, hierarchical planning [2], resource availability [3], and formal checks for correctness and completeness [4, 5] were introduced. Ongoing research work focuses on proves for the correctness of plans and the development of powerful representation of real problem domains. The area of configuration focuses more on knowledge acquisition aspects. This is often caused by the complex structure of the application domains of configuration systems. In a lot of cases, rules are used to represent expert knowledge: R1/XCON [6], SYLLOGIST [7] and PLAKON [8]. A further method for configuration problems is the 'propose and revise' method applied in VT [9] and SALT [10] and 'resource oriented configuration' [11].