Cooperative Learning Model of Agents in Multi-cluster Grid

The idle computational resources of CSCW environment that is composed of computer clusters are mined to construct the multi-cluster grid in order to support the computation-intensive tasks. For fitting the state changes of idle computing resources during the computing process and the migration process, the dynamic rule mechanism of agents are proposed. By using of the grid techniques, the independent agent, the cooperative agent, the state space, the action space, the dynamic rule space, the reinforcement learning, the migratory task, the data parallel computation and the fuzzy relation theory, a cooperative learning model of agent was designed and implemented. The experimental results show that this model can increase the use rate of idle resources in CSCW. It can be fit for the computation in CSCW based Intranet or Internet.