中文摘要：

联盟形成是在多代理人系统的一个重要协作问题，并且对为代理人的合作能力的合适的描述是在处理这个问题的基本、关键的前提。在这份报纸，面向任务的合作能力的一个模型被建立，在五个面向任务的能力被提取形成合作能力向量的地方。任务需求向量也被描述。另外，有随机的机制的联盟形成的一个方法被建议减少过多的比赛。一个人工的聪明的算法被建议补偿之间的差别期望并且实际任务要求，它能为人的命令改进代理人的认知能力。模拟显示出建议模型和分布式的人工的聪明的算法的有效性。

英文摘要：

Coalition formation is an important coordination problem in multi-agent systems, and a proper description of collaborative abilities for agents is the basic and key precondition in handling this problem. In this paper, a model of task-oriented collaborative abilities is established, where five task-oriented abilities are extracted to form a collaborative ability vector. A task demand vector is also described. In addition, a method of coalition formation with stochastic mechanism is proposed to reduce excessive competitions. An artificial intelligent algorithm is proposed to compensate for the difference between the expected and actual task requirements, which could improve the cognitive capabilities of agents for human commands. Simulations show the effectiveness of the proposed model and the distributed artificial intelligent algorithm.