动态种群下诚实多Agent序贯决策的可实现性
Implementability of Honest Multi-Agent Sequential Decision-Making with Dynamic Population
摘要:
研究了动态环境下多智能体系统资源分配决策机制的设计问题。个体对资源的偏好会随着时间的推移而变化,并且由于采用了停止规则,种群是动态的。该模型通过指定一个分配规则和三个支付规则来设计参与动态机制的Agent的相遇规则,以引出Agent的诚实偏好报告和多个时段最优停留的耦合决策。该机制提供了一个特殊的标价支付规则,只依赖于每个代理的实现停止时间直接影响人口动态。这封信的重点是理论上的可实施性的规则在完美的贝叶斯纳什均衡和特征的必要条件和充分条件,以保证代理人的诚实的均衡行为在一定时期内。我们提供的设计原则来构建支付的分配规则,并确定设计者的能力,影响人口动态的限制。所建立的条件使设计者的问题,找到多个规则,以确定一个最佳的分配规则。
英文摘要:
We study the design of decision-making mechanism for resource allocations over a multi-agent system in a dynamic environment. Agents' privately observed preference over resources evolves over time and the population is dynamic due to the adoption of stopping rules. The proposed model designs the rules of encounter for agents participating in the dynamic mechanism by specifying an allocation rule and three payment rules to elicit agents' coupled decision makings of honest preference reporting and optimal stopping over multiple periods. The mechanism provides a special posted-price payment rule that depends only on each agent's realized stopping time to directly influence the population dynamics. This letter focuses on the theoretical implementability of the rules in perfect Bayesian Nash equilibrium and characterizes necessary and sufficient conditions to guarantee agents' honest equilibrium behaviors over periods. We provide the design principles to construct the payments in terms of the allocation rules and identify the restrictions of the designer's ability to influence the population dynamics. The established conditions make the designer's problem of finding multiple rules to determine an optimal allocation rule.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。