Influence Of Network Structure And Agent Property On System Performance

System structure can affect or decide the system function. Many pioneers have analyzed the impact of system’s macro-statistical characteristics, such as degree distribution and giant component, on system performance. But only few research works were conducted on the relation of mesoscopic structure and agent property with system task performance. In this paper, we designed a scenario that, in a multiagent system, agents will try their best to form a qualified team to fulfill more system tasks under the requirements from agent property, structure and task. The theoretical and simulation results show that the agent link network, agent properties and task requirement will co-affect the dynamic team formation and at last have serious effects on a system’s task completion ratio and performance. Some factors such as network density and task introduction period have positive influence. Task execution time and team size have negative influence. Some factors show a counter-intuitive influence. The clustering coefficient has not much influence as people expected and the task publicity time isn’t bigger the better. Notably, system performance is affected by the coupling effect, instead of the independent effects of all factors. The effect of system structure on system function conditionally relies on the support from agent ability and task requirement.