A Model ofthe Distríbuted Constraint Satísfactíon Problem and an Algorithm for Configuratíon Design

Abstract

Abstract. The objective of this research is to develop an approach to representing and satisfying constraints during the cooperative configuration design of complex objects with hierarchical structure. Designers often collaborate in product development, so our constraint satisfaction mechanism is based upon multi-agent technology, permitting communication among participating applications coupled with localized solution methods. A system to be configured is divided into fragments; each of them based on local knowledge about template component compatibility. In this case each fragment development task is treated as an agent with embedded constraintsatisfaction facilities. An agent is considered as a computational process with expertise about a Iimited portion of a design problem, capable of achieving specific goals and communicating with other agents. Configuration design process is formulated as a distributed dynamic constraint-satisfaction problem, accomplished by a number of agents in parallel. The algorithm of distributed search on the dynamic constraint network is discussed. The developed algorithm is well suited to collaborative design because it operates incrementally and without global information about the constraint network. The architecture of Concurrent Configuration Design Advisor -a distributed agent-based expert system for configuration design of complex objects with hierarchical structure -is presented. The discussion is iIlustrated with the examples from FMS configuration design application domain. Current and future work on the expert system implementation is considered.