多Agent的层次合作模型及任务规划

在多Agent环境中，许多任务并不是一个Agent可以独立完成的，在出现能力不足或条件不足时，必须联合多个Agent成员来完成该任务．针对这种情况，文中提出了一个多Agent层次合作模型，将合作对象通过联合能力组成一个层次合作序列，以较大粒度的联合能力为规划单元，以层次合作目标为规划目标，使任务规划和任务管理变得简单和清晰．文中同时给出了确定合作对象和形成动态合作组的方法．并对合作任务的规划进行了讨论．

Layered Model and Task Planning for Multi-Agent Cooperation

In a multi-agent environment, multiple agents should be united to realize some complex tasks that are difficult to solve depending only on one agent because of being lacking in the capability or precondition of the system. Aiming at the cooperation of multiple agents, this paper proposes a layered model, in which the partners join hands to form a layered cooperation sequence via the joint capability. A large joint capability is then taken as the plan unit and the layered cooperation target is taken as the plan target. Thus, the task plan and management become simple and clear. Moreover, the methods of determining the partners and forming the dynamic teams are also presented, with the task planning being finallv discussed.