ERI: Design of Reliable Autonomous Engineering Systems with Active Failure Prevention

该工程研究启动（ERI）项目的目标是创建一个设计框架，使用计算机模拟来设计可靠的自主工程系统（AES），该系统可以在运行过程中主动防止灾难性故障。自主工程系统，如自动驾驶汽车和工业机器人，是新兴的工程系统，正在社会的许多方面取代人类。未来AES不仅要具备自主完成任务的能力，而且要具备主动预防运行故障的能力（即主动故障预防）。本研究通过不确定性下基于仿真的设计，将设计后主动故障预防纳入AES的设计前期，从而降低设计的AES投入运行时的故障概率。该研究将为实现以下目标奠定理论基础：(1)在存在不确定性的情况下，高效准确地分析具有主动故障预防功能的AES的可靠性；(2)共同设计AES物理和网络系统，以满足特定可靠性要求的主动故障预防。本文的研究为AES的设计提供了可靠的依据。它将通过减少设计过程中的可靠性认证工作来加速AES的发展。该项目的其他成果包括改进可靠性和与设计相关的课程，为本科生和研究生提供研究经验，以及为学生和少数族裔提供外展活动。该项目的总体目标是创建一种新的设计范式，从传统的基于可靠性的工程系统设计（通过被动可靠性保证）转变为基于模型的可靠AES设计（同时考虑群体层面的被动可靠性保证和个体层面的主动故障预防）。个性化的主动故障预防和自主性使AES有别于传统的工程系统，并使当前基于不确定性方法的设计不适用于AES。该研究利用基于模型的设计、统计学、机器学习和优化来填补当前方法的空白。它将通过第一通道方法和联合模拟来解决计算挑战，从而在可靠性分析方面进行创新，同时考虑主动故障预防，以及通过解耦和分散基于可靠性的协同设计问题来实现设计优化方面的创新。该研究具有广泛的社会影响，通过考虑可靠性的主动故障预防设计AES，加速了下一代可靠自主工程系统的开发，以取代在安全关键或危险任务/工作环境中的人类。综合教育计划包括课程改进，让本科生参与研究，传播研究成果，以及向K-12学生和教师推广。该奖项反映了美国国家科学基金会的法定使命，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。

The objective of this Engineering Research Initiation (ERI) project is to create a design framework using computer simulations to design reliable autonomous engineering systems (AES) that can proactively prevent catastrophic failures during operation. Autonomous engineering systems, such as autonomous vehicles and industrial robots, are emerging engineering systems that are replacing humans in many aspects of society. It is envisioned that AES in the future should have the capability of not only autonomously accomplishing a mission, but also proactively preventing failures in operation (i.e., active failure prevention). Through simulation-based design under uncertainty, this research will incorporate post-design active failure prevention into the early design stage of AES, thus reducing the probability of failure when the designed AES is put into operation. The research will establish a theoretical foundation that enables: (1) efficient and accurate analysis of the reliability of AES equipped with active failure prevention in the presence of uncertainty; and (2) co-design of AES physical and cyber systems for active failure prevention subject to specific reliability requirements. This research will enable the design of AES with assured reliability. It will accelerate the development of AES by reducing the reliability certification effort in the design process. Additional deliverables of this project include improvement of reliability and design-related curricula, research experiences for undergraduate and graduate students, and outreach initiatives for students and under-represented minorities.The overarching goal of this project is to create a new design paradigm shift from the conventional reliability-based design of engineered systems through passive reliability assurance to the model-based design of reliable AES to account for both passive reliability assurance at the population level and active failure prevention at the individual level. Personalized active failure prevention and autonomy set AES apart from conventional engineered systems and make current design under uncertainty methodologies inapplicable to AES. The research draws upon model-based design, statistics, machine learning, and optimization to fill the gap in the current methods. It will create innovations in reliability analysis by tackling the computational challenges through a first-passage method and co-simulations, while accounting for active failure prevention, along with innovations in design optimization by decoupling and decentralizing reliability-based co-design problems. The research has broad societal impacts by enabling the design of AES with active failure prevention with the consideration of reliability, which accelerates the development of next-generation reliable autonomous engineering systems to replace humans in safety-critical or dangerous mission/working environments. The integrated education plan involves curriculum improvement, engaging undergraduate students in research, dissemination of research results, and outreach to K-12 students and teachers.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.