USING AGENTS IN THE EXCHANGE OF PRODUCT DATA Udo Kannengiesser and John S. Gero Key Centre of Design Computing and Cognition, University of Sydney Abstract: This paper describes using agents in the exchange of industrial product data when predefined translators are not available. A major problem with standard translators is that a seamless data transfer instantly fails when not every translator implements a mapping into or from the standard format. This is frequently the case for large design projects that involve the use of a multitude of heterogenous tools, possibly in evolving configurations over time. This approach to using agents aims to flexibly provide product models in a form adapted to the need of the particular tools when a common data format is not readily available. Experiments show the feasibility of this approach as well as its efficacy and efficiency. Key words: product data exchange, product modelling, interoperability 1. INTRODUCTION There has been an increased use of computational tools to support various tasks in product development. Examples include computer-aided drafting (CAD) and manufacturing (CAM) systems and a number of specialised tools for analyses such as finite element analysis (FEA) and spreadsheet analysis. Most computational systems have been developed independently from one another to address the specific needs of each task and use different product data representations. However, industrial product development is a process that involves a complex network of interrelated activities, each of which needs information produced or manipulated by the other. Interoperability – the ability to move data from one representation of a product to another to allow other computational processes to operate on it – has become an area of growing concern as the cost of such interchanges increases (NIST 1999).