Convergence Accelerator Phase I (RAISE): Safe Skill-Aligned On-The-Job Training with Autonomous Systems

融合加速器第一阶段 (RAISE)：利用自主系统进行安全的技能协调在职培训

• 批准号: 1936997
• 负责人: Siddharth Srivastava
• 金额: $ 99.86万
• 依托单位: Arizona State University
• 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: 2019-09-01 至 2021-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1936997&HistoricalAwards=false
• 关键词: Convergence; Accelerator; Phase; RAISE; Safe

The NSF Convergence Accelerator supports team-based, multidisciplinary efforts that address challenges of national importance and show potential for deliverables in the near future. The broader impact/potential benefit of this Convergence Accelerator Phase I project stems from empowering the global competitiveness of our future workforce. Although staying competitive requires the productivity of US manufacturing to increase with the utilization of artificial intelligence (AI) and autonomous robotics, advanced robotic systems currently require a workforce with, at the least, 4-year degrees in engineering, computer science or AI. The problem is amplified by the fact that current robotics technologies do not inherently support adaptability, safety and explainability. This project will address these issues by creating autonomous on-the-job training platforms that use safe, self-explaining, adaptive robots. In order to achieve these objectives, our team employs a convergent approach drawing upon ideas and tools from research on intelligent tutoring systems (ITS), AI, robotics, manufacturing processes, human systems engineering, and cognitive science. The project's team also includes experts on resolving legal and socio-ethical challenges of bringing advanced technology to existing social infrastructure. This project synthesizes these diverse approaches in coordination with multiple industrial partners to enable an autonomous on-the-job training platform that would empower our national workforce for working with autonomous systems. This Convergence Accelerator Phase I project aims to initiate the invention, development and evaluation of intelligent training systems and self-explaining autonomous systems for providing safe on-the-job training for work with autonomous systems. Although autonomous systems have immense potential for empowering a highly productive workforce, this potential cannot be realized with the current state of the art. Training for work with today's autonomous systems presents unique challenges not addressed by current training paradigms, which are designed for operational systems characterized by fixed functionality and behavior. In contrast, AI systems will, by definition, change from day to day in their functionality and behavior. This interdisciplinary project will utilize safe and taskable self-explaining autonomous systems to develop a new class of intelligent tutoring systems that provide on-the-job training for work with autonomous systems. In the process, it will also advance methods for creating self-explaining autonomous systems, for the automated synthesis of task-specific robot behavior that is safe and compliant with workplace regulations, and for the evaluation of collaborative human-autonomy teamwork. In addition, it will advance research in these areas through the development of reproducible testbeds in consultation with industry experts in the domain of collaborative human-robot advanced manufacturing.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.

NSF融合加速器支持以团队为基础的多学科努力，以应对国家重要性的挑战，并在不久的将来展示可交付成果的潜力。 这个融合加速器第一阶段项目的更广泛的影响/潜在利益源于增强我们未来员工的全球竞争力。虽然保持竞争力需要美国制造业的生产力随着人工智能（AI）和自主机器人技术的利用而提高，但先进的机器人系统目前需要至少拥有4年工程，计算机科学或AI学位的劳动力。当前的机器人技术本质上并不支持适应性、安全性和可解释性，这一事实加剧了这个问题。该项目将通过创建使用安全、自我解释、自适应机器人的自主在职培训平台来解决这些问题。 为了实现这些目标，我们的团队采用了一种融合的方法，借鉴了智能教学系统（ITS）、人工智能、机器人、制造工艺、人类系统工程和认知科学研究的思想和工具。该项目的团队还包括解决将先进技术引入现有社会基础设施的法律的和社会伦理挑战的专家。 该项目综合了这些不同的方法，与多个工业合作伙伴协调，以实现一个自主的在职培训平台，使我们的国家劳动力能够使用自主系统。该融合加速器第一阶段项目旨在启动智能培训系统和自我解释自主系统的发明、开发和评估，为使用自主系统的工作提供安全的在职培训。虽然自主系统具有巨大的潜力，使一个高生产力的劳动力，这种潜力无法实现与当前的艺术状态。培训与今天的自主系统的工作提出了独特的挑战，没有解决目前的培训模式，这是专为操作系统的特点是固定的功能和行为。相比之下，根据定义，人工智能系统的功能和行为每天都会发生变化。这个跨学科项目将利用安全和可分配任务的自我解释自主系统来开发一类新的智能辅导系统，为使用自主系统提供在职培训。在此过程中，它还将推进创建自我解释的自主系统的方法，用于安全且符合工作场所法规的特定任务机器人行为的自动合成，以及用于评估协作的人类自主团队合作。此外，该基金会还将通过与人机协作先进制造领域的行业专家协商，开发可重复的试验台，推进这些领域的研究。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

RADAR: automated task planning for proactive decision support

• DOI: 10.1080/07370024.2020.1726751
• 发表时间: 2020-03-19
• 期刊: HUMAN-COMPUTER INTERACTION
• 影响因子: 5.3
• 作者: Grover, Sachin;Sengupta, Sailik;Kambhampati, Subbarao
• 通讯作者: Kambhampati, Subbarao

TLdR: Policy Summarization for Factored SSP Problems Using Temporal Abstractions

TLdR：使用时间抽象对因子式 SSP 问题进行策略总结

• 发表时间: 2020
• 期刊: Proceedings of the International Conference on Automated Planning and Scheduling
• 作者: Sarath Sreedharan, Siddharth Srivastava

Unifying Principles and Metrics for Safe and Assistive AI

• DOI: 10.1609/aaai.v35i17.17769
• 发表时间: 2021-05
• 作者: Siddharth Srivastava

Risk-Bounded Control Using Stochastic Barrier Functions

使用随机屏障函数的风险有界控制

• DOI: 10.1109/lcsys.2020.3043287
• 发表时间: 2021
• 期刊: IEEE Control Systems Letters
• 影响因子: 3
• 作者: Yaghoubi, Shakiba;Majd, Keyvan;Fainekos, Georgios;Yamaguchi, Tomoya;Prokhorov, Danil;Hoxha, Bardh
• 通讯作者: Hoxha, Bardh

Human, AI, Robot Teaming and the Future of Work: Barriers and Opportunities for Advancement

人类、人工智能、机器人团队和工作的未来：进步的障碍和机遇

• DOI: 10.1177/1071181320641018
• 发表时间: 2020
• 期刊: Proceedings of the Human Factors and Ergonomics Society Annual Meeting
• 作者: Chiou, Erin K.;Holder, Eric;Dolgov, Igor;McDowell, Kaleb;Menthe, Lance;Roscoe, Rod D.;Zaveri, Shivam
• 通讯作者: Zaveri, Shivam