针对合同网协议在解决大规模多智能体系统的任务分配问题时所表现出的协商效率低、信息量大等不足,提出了一种基于投标阈值和可用度的合同网协议改进方案。该方案对参与者的投标数量设置了阈值,以限制其不必要的投标,并通过最大化参与者接受适当数目的任务授权的概率来确定阈值;给出了当参与者能力为任何值时均适用的可用度定义,并将其引入到发起者评价过程中,与任务完成质量加权综合,形成新的标书评价函数;并基于系统中总任务数和总能力值的分析,得到可用度权重的计算公式。模拟实验结果表明,该改进方案能有效地提高大规模多智能体系统中合同网协议的性能。
To overcome the shortcomings such as low efficiency and large amounts of information in the process of using contract net protocol to solve the task allocation problem of large-scale multi-agent system, an improved scheme of contract net protocol based on bidding threshold and availability was proposed. Threshold was set to eliminate unnecessary biddings. And to determine the threshold, an approach was designed by maximizing the probability of participants receiving appropriate number of task authorization. Besides, a definition of availability applicable to all cases of participants' ability was also presented, which was introduced to the process of bids evaluation and comprehensive weighted with the quality of service, as a result, a new evaluation function for biddings was generated. And then, a formula to calculate the weight of availability was proposed based on the analyses of the total number of tasks and ability in the system. Finally, simulation experiments were conducted. The results showed that the performance of contract net protocol in large-scale multi-agent system could be enhanced remarkably by the improvement.