Engineering with Computers(1994) 10:112-123 9 1994Springer-Verlag LondonLimited Engineering Co~nputers A Tandem Architecture for Cooperating Knowledge-Based Concurrent Engineering Systems Guo Q. Huang and Derek Sheldon Engineering DesignCentre,Coventry University,UK Abstract. This paper proposes a tandem architecture for cooperating heterogeneous expert systems. Two levels of recta and working expert systems are involved. The working-level expert systems (W-ES), which may be implemented in their own computational environments and in private proof languages, are mainly for application computations. The recta-level expert systems (M-ES), using a common argument language, are mainly responsible for cooperation. The prototype AGENTS system is described for constructing M-ES. Interaction among W-ES has been transformed into two forms: communication between M-ES through ordinary AGENTS messages and communication between M-ES and the corresponding W-ES using the Deductive Inference Language (DIL). DIL predicates are provided for defining DIL: messages, actuators and converters for interpreting DIL queries and instantiating variables. By this approach, stand-alone capability of individual systems is retained at the working-level and cooperation is achieved effectively with minimum embellishment at the meta-level. Keywords. Agent-based framework; Architecture; Concurrent engineering; Cooperating expert systems; Tandem integration 1. Introduction Efforts are being continuously made to develop computational frameworks to support concurrent engineering. New problems are being incrementally ascertained. Appropriate prescriptions for tackling these problems are being sought. One of the most notable approaches is due to the work by Talukdar and Fenves [1] based on the concept of TAO graphs. Many fundamental difficulties arising from hetero- geneity (differences in languages, styles, residing machines, data formats, user interfaces, etc.) have been dealt with in their prototype implementations. One of Correspondence and offprint requests to: Guo Q. Huang/Derek Sheldon, EngineeringDesign Centre, CoventryUniversity,Priory Street, Coventry CV1 5FB, UK. the inspirations from their work can be summarised as a principle of third-party development. By this it is meant that: (1) individual expert systems can be developed rela- tively independently of each other by different groups using different environments; (2) each of them can be used as a stand-alone system without the existence of others; (3) they can participate in a cooperative session with minor customizations at the interface level but not internally. This paper is arranged from a work package's delivery report of an on-going research project. The objective of the project has been to prototype a computational framework for integrating proprietary design tools, already existing or yet to appear. That is, it aims to exploit accumulative expertise with little expensive re-writing. Early research at Cardiff [2, 3] led to a prototype form of the AGENTS system. It was soon recognised that AGENTS could not satisfy the principle of third-party development to a sufficient extent, and extensions should be sought. Our evalua- tion of previous approaches to system integration, such as blackboard and message passing, has led us to experiment with the idea of a tandem integration architecture (TIA) at Coventry. Two levels of meta and working expert systems are involved in a tandem integration architecture. The working-level expert systems, which may be imple- mented in their own computational environments and in private proof languages, are mainly for applica- tion computation. The meta-level expert systems, using a common argument language, are mainly responsible for communication (a) among themselves and (b) between the two levels. By this approach, stand-alone capability of individual systems is retained at the working-level and cooperation can be obtained effectively with minimum embellishment at the meta- level. As this paper unfolds, it becomes clear that TIA