Negotiation and Cooperation in Multi-Agent Environments

多Agent环境中的协商与合作

• 批准号: 9820657
• 负责人: Jonathan Wilkenfeld
• 金额: $ 33.42万
• 依托单位: University of Maryland, College Park
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-08-01 至 2003-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9820657&HistoricalAwards=false
• 关键词: Negotiation; Cooperation; Multi; Agent; Environments

The research objective is to identify strategies and learning techniques for automated agents that will be able to participate successfully in auctions and negotiations. Auctions and negotiations can be used to resolve conflicts in a wide variety of multi-agent domains. Our main application domain will be electronic commerce. We will also continue to develop our simulation of document allocation in distributed information servers. These agents have insufficient information upon which to base strategic decisions. Thus, we proposeto identify learning methods for agents participating in auctions. We will use well known learning techniques and will focus on the questions: how accurate are the agent's expectations based on the learned information? and how does the learning process influence the negotiation and the auction outcome? Finally, human-computer experiments will be conducted focusing on the extent to which the information collected and evaluated by automated agents will be accessed and utilized by human participants in auctions. The significance of this research lies in its contribution to the development of autonomous agents capable of reaching mutually beneficial agreements efficiently in complex environments characterized by uncertainty and incomplete information. Our results will be particularly applicable to the emerging areas of electronic commerce and digital libraries.

研究目标是为自动代理确定能够成功参与拍卖和谈判的策略和学习技术。拍卖和谈判可用于解决各种多代理域中的冲突。我们的主要应用领域将是电子商务。我们还将继续开发我们在分布式信息服务器中的文档分配模拟。这些代理人没有足够的信息来作为战略决策的基础。因此，我们建议为参与拍卖的代理确定学习方法。我们将使用众所周知的学习技术，并将重点放在以下问题上：基于学习的信息，代理的预期有多准确？学习过程对谈判和拍卖结果有何影响？最后，将进行人机实验，重点是自动代理收集和评估的信息将在多大程度上被拍卖中的人类参与者访问和利用。这一研究的意义在于它有助于自主智能体的发展，使其能够在以不确定性和不完全信息为特征的复杂环境中有效地达成互利协议。我们的结果将特别适用于电子商务和数字图书馆等新兴领域。