SHF: Medium: Collaborative Research: Formal Analysis and Synthesis of Multiagent Systems with Incentives

SHF：媒介：协作研究：带激励的多智能体系统的形式分析与综合

• 批准号: 1704883
• 负责人: Moshe Vardi
• 金额: $ 80万
• 依托单位: William Marsh Rice University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-07-01 至 2022-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1704883&HistoricalAwards=false
• 关键词: SHF; Medium; Collaborative; Research; Formal

The project develops automated methods for the tool-aided design and analysis of multi-agent systems with incentives. These systems are natural models for real-world situations in which collections of actors interact with one another in an autonomous and self-interested manner. For example, such a system can model a set of software agents that participate in an internet-based protocol, such as an advertisement auction or crypto currency; a collection of robots that share physical or digital infrastructure; or a set of cells that participate in an evolutionary process. The project combines game-theoretic and logical methods to develop techniques for formal modeling, analysis, verification, and synthesis of such systems. The end objectives of the project include reliable engineering of protocols that govern interactions among autonomous agents, and computer-aided understanding of naturally occurring game-theoretic interactions.The technical approach of the project has three dimensions. The first is the development of new formal models, correctness requirements, and abstraction and reasoning principles for multi-agent systems with incentives. Research directions include richer notions of equilibria in multi-agent systems, and effects of interaction and randomization on equilibria. Second, the project studies algorithmic tools for analysis and verification of multi-agent systems with incentives with respect to desired requirements regarding system behaviors and equilibria. The third direction is to automatically synthesize mechanisms so as to guarantee desired properties. Applications from a range of areas, including financial protocols, robotics, and biology, are used to guide and evaluate the research.

该项目开发了具有激励措施的多代理系统的工具辅助设计和分析的自动化方法。这些系统是现实情况下的自然模型，在这种情况下，演员的集合以自主和自私的方式相互作用。例如，这样的系统可以建模一组参与基于Internet的协议的软件代理，例如广告拍卖或加密货币；共享物理或数字基础架构的机器人集合；或一组参与进化过程的单元。该项目结合了游戏理论和逻辑方法，以开发用于形式建模，分析，验证和综合的技术。该项目的最终目标包括对自主代理之间交互的协议的可靠工程，以及对天然发生的游戏理论相互作用的计算机辅助理解。项目的技术方法具有三个维度。首先是开发新的正式模型，正确性要求以及具有激励措施的多机构系统的抽象和推理原则。研究方向包括多代理系统中更丰富的平衡概念，以及相互作用和随机化对平衡的影响。其次，该项目研究算法工具，用于分析和验证有关系统行为和平衡的所需要求的多代理系统的验证。第三个方向是自动合成机制，以确保所需的特性。包括财务方案，机器人技术和生物学在内的各个领域的应用都用于指导和评估研究。

项目成果

Synthesis from Satisficing and Temporal Goals

从满意目标和暂时目标综合

• DOI: 10.1609/aaai.v36i9.21202
• 发表时间: 2022
• 期刊: Proceedings of the AAAI Conference on Artificial Intelligence
• 作者: Bansal, Suguman;Kavraki, Lydia;Vardi, Moshe Y.;Wells, Andrew
• 通讯作者: Wells, Andrew

Comparator Automata in Quantitative Verification

• DOI: 10.1007/978-3-319-89366-2_23
• 发表时间: 2018-01-01
• 期刊: FOUNDATIONS OF SOFTWARE SCIENCE AND COMPUTATION STRUCTURES, FOSSACS 2018
• 作者: Bansal, Suguman;Chaudhuri, Swarat;Vardi, Moshe Y.
• 通讯作者: Vardi, Moshe Y.

Public and Private Affairs in Strategic Reasoning

战略推理中的公共和私人事务

• DOI: 10.24963/kr.2022/14
• 发表时间: 2022
• 期刊: Proceedings of the 19th International Conference on Principles of Knowledge Representation and Reasoning
• 作者: Fijalkow, Nathanaël;Maubert, Bastien;Murano, Aniello;Rubin, Sasha;Vardi, Moshe
• 通讯作者: Vardi, Moshe

Control Regularization for Reduced Variance Reinforcement Learning

• 发表时间: 2019-05
• 作者: Richard Cheng;Abhinav Verma;G. Orosz;Swarat Chaudhuri;Yisong Yue;J. Burdick

OneVision: Centralized to Distributed Controller Synthesis with Delay Compensation

OneVision：具有延迟补偿的集中式分布式控制器综合

• 发表时间: 2022
• 期刊: 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS
• 作者: J. Wei, T. Li