Institute of Computer-aided Product Development Systems

RPD - RAPID PRODUCT DEVELOPMENT- INTEGRATION PLATFORM FOR ACTIVE KNOWLEDGE COMMUNICATION

D. Roller

Rapid Product Development (RPD) a technique for a new Product Design and Development Process is characterised by evolutionary, iterative work of developer teams targeted to reduce the development time and costs while maximizing the quality.

The evolutionary character of RPD is reflected in the competition for the best solution with a pool of alternative solutions, their combinations, their further development and their adaptation in dynamically changing environment. The iterative character ensures the state of the development as well as the verification of the development stage during the RPD. This way it is possible to prevent faulty developments.

These RPD characteristics impose two basic requirements on an intended system. Primary the representation of knowledge related to RPD domain and secondary the support of communication and cooperation among the members of the developer team. Distributed, interdisciplinary, self organized teams are working in various processes, interacting closely, at different domains. Hence, product development is being accompanied by high degree of process parallelism and cooperation. These activities generate product data consisting of thousands of objects with lots of dependencies among them. If the semantics of the product data is not stored then information about the meaning of data is later not available anymore. Obviously, the RPD requirements are not satisfied with an only storing database system, it has to be a database system giving an explicitly possibility for semantic s representation. At the present time, there are not such database systems available on market.

Representing knowledge in a high-level form is a promising way to increase the reusability of product data. Scientifically, the difference between the terms “knowledge base” and “database” is declared to the degree on which they support representational, structural and inference capabilities.

The communication and coordination of tasks and processes as well as the requirements for knowledge integration on a common workspace lead to a creation of the Active Semantic Network (ASN). The presented approach of ASN does not only provide rich representation structures and inference capabilities to the modeller , the knowledge base is also used by internal mechanisms of the ASN to extend traditional database functions to a more intelligent behaviour. A knowledge-based component for semantic integrity enforcement allows to model and propagate dependencies inside the product model.

All these approaches integrate means of communication and cooperation to assist the responsible designers. The reason for this is that the goal of the knowledge base is not to cover the area of expert systems where often inferences take place that are uncontrolled by the users. Rather, it supports users in their design decisions by representing, controlling and synchronising parallel results of the teams. The aim of knowledge based systems for supporting rapid prototyping should be the usage of the human knowledge as a resource in order to provide means of communication and cooperation with the team members to get the information needed rather than formalising knowledge representation s of expertise [31].

These kind of services can be seen on the figure below.

Complex systems for RPD elaborate intelligent solutions using the autonomous behaviour of Multi Agent System (MAS). Using a Client-Server architecture supporting the interagent communication based on the multicast principle. After a closer examination of the requirements of RPD five different types of Agents have been developed. These are part of the MAS and are generated as necessary.

The five agent types are:

1) Monitoring Agent: monitors the ASN and notifies on changes.

2) Coordination Agent: supports coordination within the RPD as a finite state machine.

3) Transaction Agent: supports transaction protected processes and transaction protected execution of other agents within the MAS.

4) Aggregation Agent: prepares retrieved knowledge in an appropriate format.

5) Retrieval Agent: retrieves ASN knowledge.

Additionally, in our aim to search for knowledge with more intelligent tools, we developed a learning retrieval agent.