On Multi-agent Cooperative Control of Conflict Workflow Tasks

摘要

With the increase of capabilities of the World-Wide-Web, the necessity to build tools for cooperative work has been more and more obvious. Also, we think that workflow systems must evolve to parallel workflow systems (PWS). After a discussion on the nature of parallel workflow activity processes, we address the theoretical issues related to the use of a formal knowledge representation model, showing how the notions of dependence, validity, coherence can help in understanding, organizing and formalizing fundamental distinctions between traditional workflow processes and parallel workflow ones. We then present a software agent architecture which addresses the needs of these systems by encapsulating several primitive agents in the same structure. The three key aspects of our agent model are: parallelization of agent tasks, reuse of agent components and partial mobility of agent code. As a practical example, we report our ongoing experience in building a technical specification writing system in the telecommunication domain.