Study on Multi-agent systems task distribution negotiation based on loose coalition

In the distributed intelligent system composed of Multi-agent, the collaboration of multi-agent needs to carry out negotiation then makes the task distribution among tasks. The self-properties and current state of each agent, especially the properties such as friendly credit relation and cooperation history all influence the negotiation process greatly. This paper presents a task distribution negotiation method of agents based on loose coalition. The concept of loose coalition is the consociation of the agents based on mind information; the mind information can describe the friendly credit relation between each agent which is formed in previous collaboration actions. According to mind information, the multi-agent can choose a group of best co-operators by negotiating to build loose coalition. The collaboration of multi-agent based on loose coalition can reduce the blindness of negotiation, improve the efficiency and reduce the costs of communication during negotiation. When participating in the task competition, the loose coalition with high priority will become better task candidate. The negotiation based on loose coalition mechanism can solve the problem of cooperative task distribution between agents efficiently. In the occasion that the numbers of agents participating in the collaboration are large and the distribution of task is anomalous, the task distribution negotiation based on loose coalition can improve the efficiency of system obviously, and it will be more flexible and adaptive.