Identifying Autonomous Agents for Capitalizing Knowledge in R&D environments
|
Abstract:
Whilst we have a number of techniques for building multiagent systems, such techniques assume that the number and role of each agent in the system is known a priori. Unfortunately it is rarely the case in practice, in particular when trying to build multiagent knowledge management systems in R&D environments exhibiting a high rate of turnover. Thus, we need specific analysis and design guidelines. The paper presents a new method, SAAS, for identifying and selecting a collectively agreed upon set of autonomous agents necessary to implement a system for capitalizing collective knowledge. The method borrows concepts from another method proposed by Grundstein for identifying strategic knowledge in production processes. SAAS was used to determine the number of agents necessary for capitalizing the knowledge of a research group in the MEMOLAB project at UTC. In 1994 we were developing the OSACA 1 environment as a set of tools for facilitating the construction of open systems of cognitive agents [Scalabrin et al 96]. The OSACA approach was to be tested in the design of a system for capitalizing the knowledge of a research laboratory, in the
