CRII: CPS: RUI: Cognizant Learning for Autonomous Cyber-Physical Systems

该奖项的全部或部分资金来自《2021年美国救援计划法案》(公法117-2)。这项计算机和信息科学与工程(CEISE)研究启动计划(CRII)的目标是为网络物理系统(CPS)开发一个认知学习框架，其中包括对风险敏感和非理性的决策。两点意见说明了建立这样一个框架的必要性。首先，自动驾驶汽车等CPS将与其他CPS和人类用户共享一个环境。人类驾驶员对速度的变化表现出高度的敏感性，并且可以很容易地适应环境和路况的变化，这使得CPS具有识别非理性行为的能力是必不可少的。其次，CPS在运行过程中生成的大量数据以及对其环境模型的有限访问可能会使CPS依赖机器学习算法进行决策，以满足可达性和安全性等性能要求。我们的研究将基于改善自动驾驶汽车在现实交通情况下的行为。这一努力的结果将有助于发展统一控制、学习和行为经济学的研究范式。以本科为主的学院的学生将受益于直接参与研究过程的所有方面。研究任务将包括以垂直整合的方式由本科生组成的团队，经验更丰富的学生将指导新团队成员。拟议的努力包括两个方面。推力1将构建实用程序，以编码与风险敏感和非理性决策的实用模型一致的CPS绩效目标。策略将通过制定和解决强化学习问题来学习，以最大限度地提高这一效用。将制定方法，使已掌握的战略能够充分考虑评价和执行因《战略行动方案》的实际组成部分而产生的行动之间的延误。当多个CP在同一环境中运行时，推力2将设计算法来学习分散的认知策略。为了在不确定的环境中或在反馈稀疏的情况下提高可靠性，将确定确定每个CP对共享公用事业的贡献的技术。解决方案方法将通过广泛的实验进行经验评估，并通过确定概率性能保证进行理论评估。PI将在西华盛顿大学(WWU)开发CPS和机器学习方面的研究议程和新的本科课程。该项目的研究和教育目标将通过CALA自动驾驶汽车研究模拟器和F1/10自动驾驶汽车平台整合在一起。该项目的多学科范围将在通过学生外展服务和WWU的STEM俱乐部的外展努力中得到强调，以鼓励和扩大传统上代表性较低的学生群体的参与。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2). The objective of this Computer and Information Science and Engineering (CISE) Research Initiation Initiative (CRII) proposal is to develop a cognizant learning framework for cyber-physical systems (CPS) that incorporates risk-sensitive and irrational decision making. The necessity for such a framework is exemplified by two observations. First, CPS such as self-driving cars will share an environment with other CPS and human users. Human drivers demonstrate a heightened sensitivity to changes in speed and can easily adapt to changes in the environment and road conditions, which makes it essential for a CPS to have an ability to recognize non-rational behaviors. Second, large amounts of data generated during their operation and limited access to models of their environments can make a CPS reliant on machine learning algorithms for decision making to meet performance requirements such as reachability and safety. Our research will be grounded on improving behaviors of autonomous vehicles in realistic traffic situations. Outcomes from this effort will contribute to the development of a research paradigm unifying control, learning, and behavioral economics. Students at a Primarily Undergraduate Institution will benefit by being directly involved in all aspects of the research process. Research tasks will involve a team of undergraduate students in a vertically integrated manner where more experienced students will mentor newer team members. The proposed effort comprises two thrusts. Thrust 1 will construct utilities to encode CPS performance objectives consistent with practical models of risk-sensitive and irrational decision making. Strategies will be learned by formulating and solving a reinforcement learning problem to maximize this utility. Methods to enable learned strategies to adequately consider delays between evaluation and execution of actions arising from the physical components of the CPS will be developed. Thrust 2 will design algorithms to learn decentralized cognizant strategies when multiple CPS operate in the same environment. To improve reliability in uncertain environments, or when feedback is sparse, techniques to identify contributions of each CPS to a shared utility will be identified. Solution methodologies will be evaluated empirically through extensive experiments and theoretically by determining probabilistic performance guarantees. The PI will develop a research agenda and new undergraduate curriculum in CPS and machine learning at Western Washington University (WWU). Research and educational goals of the project will be integrated through the CARLA simulator for autonomous vehicle research and the F1/10 Autonomous Vehicle platform. The multidisciplinary scope of the project will be emphasized in outreach efforts through Student Outreach Services and STEM Clubs at WWU to encourage and broaden participation from traditionally underrepresented student groups.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.