AI Institute: Planning: TRustworthy Autonomous Systems Engineering (TRASE)

人工智能研究所：规划：值得信赖的自主系统工程（TRASE）

• 批准号: 2020289
• 负责人: Bruno Sinopoli
• 金额: $ 50万
• 依托单位: Washington University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2020289&HistoricalAwards=false
• 关键词: AI; Institute; Planning; TRustworthy; Autonomous

In many ways, autonomy is the holy grail of artificial intelligence (AI). While tremendous advances towards this ultimate goal have been made across a number of disciplines, and for a number of applications, general-purpose autonomy remains elusive. Indeed, even when we focus on specific classes of applications, such as autonomous driving, numerous gaps remain. One fundamental challenge shared by many autonomous systems is to ensure that autonomous behavior is trustworthy: that is, safe, reliable, secure, and fair. This planning grant is a stepping stone towards an AI institute devoted to foundational and use-inspired advances in TRustworthy Autonomous Systems Engineering. The institute would be focused on four research themes: 1) trustworthy AI integration in autonomous systems, 2) trustworthy platforms for real-time AI in autonomous systems, 3) trustworthy multiagent systems, and 4) accountable autonomy. Each of these will yield important advances in understanding the design of trustworthy systems in which AI is embodied in an open-loop framework involving perception, planning, and control. Furthermore, these advances will be coupled with use-inspired research in areas such as autonomous driving and drones, creating a virtuous cycle with applications directing fundamental research, and benefiting directly from it. This project will aim to create a community of stakeholders---the TRASE Innovation Ecosystem---focusing on trustworthy autonomous systems, with implications for a broad range of applications, from autonomous driving to autonomous flight. In addition, the institute will yield transformative educational advances in the form of new materials for courses and programs that train students in the design and analysis of autonomous systems. The vision is to develop innovative educational components, such as competitions, and integrated student-industry-research teams that coalesce around common TRASE-relevant problems. Finally, educational collaborations will be established with minority-serving institutions, such as St Louis Community College (Florissant Campus) to broaden participation in computing in general, and in trustworthy autonomous systems engineering in particular.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.

在许多方面，自主性是人工智能（AI）的圣杯。虽然许多学科和许多应用程序在实现这一最终目标方面取得了巨大进展，但通用自主权仍然难以捉摸。事实上，即使我们专注于特定类别的应用，如自动驾驶，仍然存在许多差距。许多自治系统面临的一个基本挑战是确保自治行为是可信的：即安全、可靠、安全和公平。这项规划拨款是迈向人工智能研究所的垫脚石，该研究所致力于在可信赖的自主系统工程方面取得基础性和实用性的进步。 该研究所将专注于四个研究主题：1）自治系统中值得信赖的人工智能集成，2）自治系统中值得信赖的实时人工智能平台，3）值得信赖的多智能体系统，以及4）负责任的自治。其中每一个都将在理解可信系统的设计方面取得重要进展，在可信系统中，人工智能体现在涉及感知、规划和控制的开环框架中。此外，这些进展将与自动驾驶和无人机等领域的使用启发研究相结合，创造一个良性循环，应用指导基础研究，并直接从中受益。该项目旨在创建一个利益相关者社区--TRASE创新生态系统--专注于可信赖的自主系统，对广泛的应用产生影响，从自动驾驶到自动飞行。此外，该研究所将以新材料的形式为课程和项目提供变革性的教育进步，这些课程和项目将培训学生设计和分析自主系统。我们的愿景是开发创新的教育组件，如比赛，以及围绕常见的TRASE相关问题合并的综合学生-行业-研究团队。最后，将与圣刘易斯社区学院（Florissant校区）等少数民族服务机构建立教育合作，以扩大对计算的普遍参与，特别是在值得信赖的自主系统工程方面。该奖项反映了NSF的法定使命，并通过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，被认为值得支持。

项目成果

What would jiminy cricket do? towards agents that behave morally

吉米尼蟋蟀会做什么？

• 发表时间: 2022
• 期刊: Advances in neural information processing systems
• 作者: Hendrycks, Dan;Zhu, Christine;Mazeika, Mantas;Navarro, Jesus;Song, Dawn;Zou, Andy;Li, Bo;Patel, Sahil;Steinhardt, Jacob
• 通讯作者: Steinhardt, Jacob

Certified Robust Control under Adversarial Perturbations

• DOI: 10.23919/acc55779.2023.10155878
• 发表时间: 2023-02
• 期刊: 2023 American Control Conference (ACC)
• 作者: Jinghan Yang;Hunmin Kim;Wenbin Wan;N. Hovakimyan;Yevgeniy Vorobeychik

O N THE C ERTIFIED R OBUSTNESS FOR E NSEMBLE M ODELS AND B EYOND

• 发表时间: 2022
• 作者: Zhuolin Yang;Linyi Li;Xiaojun Xu;B. Kailkhura;Tao Xie;Bo Li

Neural Lyapunov Control for Discrete-Time Systems

• DOI: 10.48550/arxiv.2305.06547
• 发表时间: 2023-05
• 期刊: ArXiv
• 作者: Junlin Wu;Andrew Clark;Y. Kantaros;Yevgeniy Vorobeychik

PROVES: Establishing Image Provenance using Semantic Signatures

• DOI: 10.1109/wacv51458.2022.00307
• 发表时间: 2021-10
• 期刊: 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
• 作者: Mingyang Xie;Manav Kulshrestha;Shaojie Wang;Jinghan Yang;Ayan Chakrabarti;Ning Zhang;Yevgeniy Vorobeychik