by Francisco P. Maturana and Douglas H. Norrie
Manufacturing industries are facing increasing competitive challenges in both maintaining their existing markets and improving their capability to respond efficiently to marketplace needs. New architectures are required for next generation manufacturing systems which must be developed to meet these challenges. Innovative research in Distributed Artificial Intelligence and Intelligent Manufacturing Systems is now redefining the essential characteristics of manufacturing organizations. These attempts aim for system integration across the heterogeneity barriers created by the systems' components. This integration must satisfy new requirements for integrability, configurability, adaptability, extendibility, agility, and reliability. Research centers, consortia, and manufacturing companies are exploring multi-agent systems for suitable solutions. Such multi-agent systems will require flexible control mechanisms for both creating and managing such agent communities. This paper introduces a novel approach for creating and managing such agent communities. These virtual communities are supported by mediator agents which coordinate actions to satisfy local and global objectives. The manufacturing system is thus populated by heterogeneous agents and structures of control which operate autonomously during the planning and execution of the manufacturing tasks.
by Francisco P. Maturana and Douglas H. Norrie
The next-generation intelligent manufacturing system (IMS) will be a multi-agent system containing planning, control, and application entities that dynamically collaborate to satisfy both local and global objectives. This multi-agent system will be populated by numerous heterogeneous agents which should desirable be coordinated through "virtual clusters" (dynamic agent coalitions grouping). These virtual clusters should be dynamically created, modified, and destroyed as needed for collaborative planning and action on tasks and sub-tasks. They will each contain a number of dependent agents and can be created by top level mediator agents (static and dynamic members). This structure of control facilitates both intra-agent and inter-agent communication, thus allowing for integrating heterogeneous multi-agent platforms concurrently. Such virtual clusters will provide a heterarchical distribution of resources within themselves (heterarchical communities of agents), and will be arranged heterarchically within the multi-agent community. An architecture and implementation mechanisms for virtual clustering is described. Keywords: Agents, Clusters, Coordination, Mediator, KQML.
Please send any comments/suggestions/criticism to: maturana@enme.ucalgary.ca