An approach to constructing evolutionary agent structure for workflow management system based on simulation

Multi-agent coordinate mechanism research has attracted increasing attention in recent years. Researches on the problem mainly focus on how to organize and coordinate relations between agents. The composition of different agents is an issue that must be faced by developers. This paper introduces an automatic agent combination method oriented to workflow which considers both taskpsilas dynamic workload and agentpsilas evolving cognitive ability. It composes agent structure through three steps: 1) Clusters the tasks according to their resources requirements by using decision tree, which helps to define the corresponding agent set. 2) Calculates the ability and cost of agent executing workflow based on information about task workload and duration with uncertainty model. 3) Search for the optimal agentspsila composition with the objective to maximize the speed of workflow execution while balancing the workload among agents under the constraint of agent ability, workload threshold and execution cost based on performance analysis of simulation result. Experimental results show that this method has a good performance by identifying the optimal agent configuration to execute workflow scenario.