权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

CPS: Small: Collaborative Research: Information Design and Price Mechanisms in Platforms for Cyber-Physical Systems with Learning Agents

CPS：小型：协作研究：具有学习代理的网络物理系统平台中的信息设计和价格机制

基本信息

批准号：
1931718
负责人：
Lillian Ratliff
金额：
$ 29.88万
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：
2019-10-01 至 2023-09-30
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1931718&HistoricalAwards=false
关键词：
CPS Small Collaborative Research Information

项目摘要

The goal of this project is new theory for architectures for crowdsourced markets for cyber-physical systems (CPS), such as power systems and shared vehicle systems. A technology-enabled platform digitally mediates transactions between buyers (e.g., energy consumers/riders) and sellers (e.g., generators/drivers) helping to determine prices and pair buyers and sellers. Large-scale CPS impose fundamental physical constraints than traditional 'cyber' platforms, including environmental uncertainty, spatial difference, delay, and limited resources. The broader impacts of the work include mentoring with high-school students and teachers.The project aims to develop guidance for optimal platform design. A central challenge is the complex behavior of agents, which possess limited information and receive limited feedback about how their actions translate to payoffs, and learn continuously over time to discover optimal strategies. The focus of the project is on the design of two significant levers available to the platform: information feedback and price-setting mechanisms. The research agenda is divided into two thrusts: (1) characterization of the dynamical behavior of agents under limited feedback; and (2) platform design through information sharing and pricing. In the first thrust, each agent is assumed to use a reasonable, `natural' learning algorithm and the goal is to understand how mechanisms available to the platform impact the learning process and eventual collective behavior. Outcomes from the first thrust will feed directly into the second, where the primary aim is to design information mechanisms to supplement pricing in order to shape the aggregate behavior of agents.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.

该项目的目标是为网络物理系统（CPS）的众包市场提供新的架构理论，如电力系统和共享车辆系统。技术支持的平台以数字方式调解买家之间的交易（例如，能量消费者/乘客）和销售者（例如，发电机/驱动程序），帮助确定价格和配对买家和卖家。大规模CPS比传统的“网络”平台施加基本的物理约束，包括环境的不确定性，空间差异，延迟和有限的资源。这项工作的更广泛影响包括对高中学生和教师的指导。该项目旨在为最佳平台设计提供指导。一个核心的挑战是代理的复杂行为，这些代理拥有有限的信息，并收到有限的反馈，关于他们的行动如何转化为回报，并随着时间的推移不断学习，以发现最佳策略。该项目的重点是设计平台可用的两个重要杠杆：信息反馈和定价机制。研究议程分为两个重点：（1）有限反馈下代理的动态行为的表征;（2）通过信息共享和定价的平台设计。在第一个推力，每个代理被假定为使用一个合理的，“自然”的学习算法和目标是了解如何提供给平台的机制影响学习过程和最终的集体行为。第一阶段的成果将直接反馈到第二阶段，第二阶段的主要目标是设计信息机制，以补充定价，从而塑造代理商的总体行为。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

期刊论文数量（9）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Multiplayer Performative Prediction: Learning in Decision-Dependent Games

多人表演预测：决策相关游戏中的学习

DOI：
发表时间：
2023
期刊：
Journal of machine learning research
影响因子：
6
作者：
Narang, Adhyyan;Faulkner, Evan;Drusvyatskiy, Dmitriy;Fazel, Maryam;Ratliff, Lillian J.
通讯作者：
Ratliff, Lillian J.

Global Convergence to Local Minmax Equilibrium in Classes of Nonconvex Zero-Sum Games