A Genetic Algorithm-based Approach to Holon Virtual Clustering Yuefei Xu, Robert W. Brennan, Xiaokun Zhang and Douglas H. Norrie Dept of Mechanical and Manufacturing Engineering, University of Calgary, 2500 University Drive NW, Calgary, Alberta, T2N-1N4, Canada Tel. 1-(403)-220-2991; Fax. 1-(403)-282-8406 E-mail: [yfxu | brennan | xkzhang | norrie] @enme.ucalgary.ca Abstract This paper focuses on the dynamic re-configuration and task optimization of holonic manufacturing systems (HMS). The concept of dynamic virtual clustering is extended to the control process of a holarchy or holonic organization. A task-oriented clustering mechanism and a corresponding optimization algorithm are presented as an efficient approach to the holonic control in the HMS domain. The mediator-based dynamic virtual clustering mechanism is presented firstly. Then a negotiation strategy based on the Contract Net protocol is proposed for cooperative action among holons. Finally, a genetic algorithm-based optimization algorithm for task allocation is described to support the optimum organization of a holarchy. This approach is illustrated through an example. Keywords: Holonic Manufacturing Control, Dynamic Clustering, Genetic Algorithm, Optimum Control. 1. Introduction Recently, new concepts from agile manufacturing system, intelligent manufacturing systems, and holonic manufacturing systems have led to new manufacturing architectures and control approaches. Autonomous-ability, cooperative-ability, reactive-ability, integrated-ability and architecture recursive- ability are now being considered to be the key requirements for new manufacturing systems (Gou, et. al, 1998). One such system, the Holonic Manufacturing System (HMS) uses emerging system technologies to design, implement and maintain distributed organizations and control systems within the manufacturing domain. A HMS is composed of different kinds of holons, which are identifiable entities within a manufacturing system. Autonomy, cooperation and recursiveness are considered as the basic characteristics of holons. The architecture and related properties of holons, including autonomy, cooperation and recursion have been considered by Bussmann (1998) and Van Brussel et al. (1998). The concepts of the Holarchy and the Cooperation Domains were described by Christensen (1994). Each holon participates in one on more Cooperation Domains. Multiple holons and non-holons are aggregated recursively to form a Holarchy, which is viewed as a recursive structure of holons, organized into logical and physical hierarchies whose constituent entities collaborate primarily through Cooperation Domains. How holons are dynamically organized into collaborative groups during run-time is one of the key research issues for the HMS. However, the task-oriented approach to this problem in which a holarchy is dynamically configured through the process of task allocation among cooperative holons, appears to be a promising solution. This paper focuses on the dynamic re-configuration of holarchies and optimum task allocation. The concept of dynamic virtual clustering (Maturana, 1997) is here extended to the organizational control of a holarchy. A task-oriented clustering mechanism and a corresponding optimization algorithm are presented as an effective approach to holonic control in the HMS domain. The paper is organized in three parts. The mediator-based dynamic virtual clustering mechanism is discussed in section 2. Next, in section 3, the negotiation strategy based on the Contract Net is applied to cooperative activity among holons in a community. In section 4, on the basis of holon/task coupling, a genetic algorithm (GA) based optimization algorithm is designed to facilitate optimal organization of a holarchy. The reproduction, crossover, reverse and mutation algorithms are given in this section. An illustrative example is presented to demonstrate this approach. 2.Dynamic Virtual Clustering Dynamic Virtual Clustering is a dynamic mechanism for organizational reconfiguration of a manufacturing system during run-time. This mechanism was shown to be an effective approach in the manufacturing domain by Maturana (1997). Dynamic organization can use virtual clustering as a grouping technology to continually reconfigure entities corresponding to changing requirements. In a HMS, holons may form holarchies whose members collaborate through Cooperation Domains. Using the mechanism of virtual clustering, holons can be dynamically involved in different clusters (holarchies) and cooperate through a Cooperation Domain. The cluster exists for the duration of its cooperation tasks and disappears when the tasks are completed. A Cooperation Domain can be implemented and maintained through the creation of a mediator holon. In order to initialize and manage a cluster, a mediator holon can be created dynamically by a primary holon. This mediator holon has the capability of searching for potential cooperation holons and sending an initial invitation to those holons that are potentially suitable for the cooperation tasks. The process cycle for virtual clustering can be described as follows: 1) The primary holon gathers some or all of its contracts into a new local task. After re-planning and analysis, the cooperation requirements are listed as cooperation tasks.