多Agent系统集成方案的选择能极大程度地影响多Agent系统的性能。目前大多数的多Agent系统集成方案只关注于系统级的行为与性能。本文在关注多Agent系统性能的同时，也关注各个Agent的收益。本文提出一种基于双向选择的多Agent系统集成方案，在此方案中，Agent根据自身的意愿选择合适的角色进行申请，同时角色根据对各个Agent的信任值选择合适的Agent来承担其任务。实验表明，随着若干次系统学习，多Agent系统协作任务完成时间能较快地下降，Agent在单位时间内的收益逐步提高。

The integration process can have a significant effect on the performance of MutiAgent systems (MAS). At present, most of the integration approaches of MAS focus on the behaviour and performance at the system level. In this paper, we are concerned about the performance of multiagent systems as well as the revenues of individual agents. This paper presents an approach to the integration of multiagent systems based on twoway selections. In the approach, agents apply for the roles accroding to their capabilities and how well the  roles meet their desires, while the  roles choose agents based on how well the agents fulfill tasks. In the process of twoway selections, the roles will evaluate the degree of trustworthiness of the agents dynamically and adjust the trust degrees in time. Experiments show the performance of multiagent systems can improve rapidly and the revenues of agents can gradually increase in a given interval.