CAREER: Co-Design of Information and Incentives in Societal-Scale Cyber-Physical Systems

职业：社会规模网络物理系统中信息和激励的协同设计

基本信息

批准号：
1844729
负责人：
Lillian Ratliff
金额：
$ 50万
依托单位：
University of Washington
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
2019
资助国家：
美国
起止时间：
2019-06-15 至 2025-05-31
项目状态：
未结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1844729&HistoricalAwards=false
关键词：
CAREER Co Design Information Incentives

项目摘要

The objective of this CAREER project is to develop a theoretical and computational framework for the co-design of information and incentive mechanisms targeted at humans in Societal-Scale Cyber-Physical Systems (SCPS) in order to encourage efficient shared resource consumption while mitigating unintended consequences. The application focus is on intelligent transportation systems, a prototypical SCPS with humans in the loop, rapid technology adoption, and emerging mobility markets. CPS and internet of things (IoT) infrastructure is pervasive in the mobility landscape, allowing operators and intelligently augmented humans to make decisions rapidly as they interact with one another and engage with the system. Market mechanisms that support interactions occur on multiple timescales, are constrained by CPS dynamics, and are exposed to exogenous uncertainties, information asymmetries, and behavioral aspects of human decision-making. Through the design of algorithms with guarantees for learning in and shaping of SCPS, this project will address two fundamental components missing in the state-of-the-art: (i) models that capture the interactions and learning processes of different SCPS stakeholders, and (ii) certifiable algorithms with high-probability guarantees for the co-design of adaptive information and incentive mechanisms that achieve measurable improvement in system-level performance while ensuring individual-level quality of service and avoiding discriminatory policies. The validation approach uses a data-informed experimental platform with simulation and living lab components. The research agenda will aid in revising the design of operational mechanisms for both private and public CPS-enabled mobility platforms to include efficiency and measurable fairness as valued criteria. The proposed agenda includes an integrated research and education plan: (i) course development leveraging the experimental platform; (ii) undergrad research in which students aid in building out the experimental platform, and engage with municipal/industry partners; (iii) development of Girls In Research Labs (GIRLs), a week-long summer program in which middle school girls explore research labs across campus through hands-on projects.Contributions to the science of CPS will be made through the study of intelligent infrastructure, with a special focus on behavior unique to human-in-the-loop CPS, and applications to multi-modal transportation systems. The technical plan is based on fundamental methods in decision sciences (control theory, game theory, behavioral economics, and mechanism design), statistics, and online learning. The research agenda is organized along two key thrusts: (i) algorithms for learning in SCPS and (ii) algorithms for shaping SCPS via incentives and information. The proposed tool set will enable analysis of multi-timescale decision-making of autonomous agents, including humans, coupled with CPS infrastructure in resource constrained environments, and will allow for the certifiable design of algorithms for learning and control (e.g., co-design of slow policy changes and real-time control). The modeling, synthesis, and validation approach will provide a principled, scientific basis for SCPS engineering design and operations, and supports CPS education by providing a platform for future engineers to discover realities associated with real-world implementation (e.g., socio-technical constraints). The unique perspective of co-designing information and incentives will also lead to new tools for modeling risk and uncertainties and thus, expose potentially new approaches to resilience in the engineering of CPS.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该职业项目的目标是开发一个理论和计算框架，用于在社会规模的网络物理系统（SCPS）中共同设计针对人类的信息和激励机制，以鼓励有效的共享资源消耗，同时减轻意外后果。应用重点是智能交通系统、人类参与循环的典型 SCPS、快速技术采用和新兴移动市场。 CPS 和物联网 (IoT) 基础设施在移动领域中无处不在，使操作员和智能增强人类能够在彼此交互和参与系统时快速做出决策。支持交互的市场机制发生在多个时间尺度上，受到 CPS 动态的约束，并受到外生不确定性、信息不对称和人类决策行为方面的影响。通过设计保证 SCPS 学习和塑造的算法，该项目将解决最先进技术中缺失的两个基本组成部分：（i）捕获不同 SCPS 利益相关者的交互和学习过程的模型，以及(ii) 可认证的算法，具有高概率保证自适应信息和激励机制的共同设计，实现系统级性能的可衡量改进，同时确保个人级服务质量并避免歧视性政策。验证方法使用具有模拟和生活实验室组件的数据知情实验平台。研究议程将有助于修改私人和公共 CPS 支持的移动平台的运营机制设计，将效率和可衡量的公平性作为有价值的标准。拟议的议程包括综合研究和教育计划：（i）利用实验平台进行课程开发； (ii) 本科生研究，学生协助建立实验平台，并与市政/行业合作伙伴合作； (iii) 开发女孩研究实验室 (GIRL)，这是一个为期一周的夏季项目，中学生通过实践项目探索校园内的研究实验室。将通过智能基础设施的研究为 CPS 科学做出贡献，特别关注人在环 CPS 的独特行为以及多式联运系统的应用。该技术计划基于决策科学（控制论、博弈论、行为经济学和机制设计）、统计学和在线学习的基本方法。研究议程按照两个关键主旨进行组织：(i) SCPS 学习算法和 (ii) 通过激励和信息塑造 SCPS 的算法。所提出的工具集将能够分析包括人类在内的自主代理的多时间尺度决策，再加上资源有限环境中的 CPS 基础设施，并将允许用于学习和控制的算法的可认证设计（例如，共同设计缓慢的政策变化和实时控制）。建模、综合和验证方法将为 SCPS 工程设计和运营提供原则性的科学基础，并通过为未来工程师提供一个平台来支持 CPS 教育，以发现与现实世界实施相关的现实（例如，社会技术限制）。共同设计信息和激励措施的独特视角还将带来用于风险和不确定性建模的新工具，从而揭示 CPS 工程中潜在的弹性新方法。该奖项反映了 NSF 的法定使命，并被认为值得通过以下方式获得支持：使用基金会的智力价值和更广泛的影响审查标准进行评估。

项目成果

期刊论文数量（21）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

A SUPER* Algorithm to Optimize Paper Bidding in Peer Review

DOI：
发表时间：
2020-06
期刊：
Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data
影响因子：
0
作者：
Tanner Fiez;Nihar B. Shah;L. Ratliff
通讯作者：
Tanner Fiez;Nihar B. Shah;L. Ratliff

Online Learning in Periodic Zero-Sum Games

周期性零和博弈中的在线学习

DOI：
发表时间：
2021
期刊：
Advances in neural information processing systems
影响因子：
0
作者：
Fiez, Tanner;Sim, Ryann;Skoulaskis, Stratis;Piliouras, Georgios;Ratliff, Lillian J.
通讯作者：
Ratliff, Lillian J.

On Gradient-Based Learning in Continuous Games