CAREER: Understanding Strategic Dynamics in the Engineering of Decentralized Systems

这项学院早期职业发展计划(Career)拨款将通过结合使用博弈论和模拟来为架构和设计决策提供信息，并在一组互动和自主的设计参与者中诊断并提供对战略动态的理解。在传统的系统工程实践中，跨基础设施、航空航天和国防领域的工程系统的设计活动更类似于集体决策过程，而不是集中化的权威。这个话题之所以令人感兴趣，是因为美国机构和公司正在积极构建具有分散决策权的系统，这些系统被描述为Internet 4.0、网络物理系统、物联网或跨领域的系统系统，包括能源、交通、制造和空间系统。这些工程系统的设计不同于其他系统的设计，因为它们规模大、寿命长、接近社会系统，引起了复杂的特征，如适应、自组织和涌现，这些特征发生在战略时间尺度上。更深入地了解战略动态如何影响理论和经验角度的设计者交互，有助于通过识别和减轻概念设计阶段的不良动态来避免代价高昂的超支和取消。系统工程的进步必须发展理论、方法和工具，以协调和促进集体设计活动。这个项目建立在一系列应用于工程设计的经济方法上，包括效用理论、决策理论、社会选择和博弈论。集体系统设计被建模为一个双层问题，其中下层决策对应于优化问题，上层策略决策对应于协调博弈。代理行为的标准模型基于经典的贝叶斯博弈论，效用函数包含风险态度等行为因素。多智能体仿真研究的是智能体之间反复交互作用下的策略演化。行为实验收集有关人类决策的经验数据，以进行验证。这项研究将为如何在概念设计阶段表征、研究和修改工程系统的战略动态提供新的知识。该教育计划开发和提供模拟活动，通过将技术建模和社会交互相结合来模拟系统问题。模拟活动就像一个高度抽象的模型系统，通过面对面的互动产生丰富的战略行为，让学生参与社会技术问题的挑战，并保持计算的易操作性，在教育环境中教授分析方法。在地球科学空间任务的背景下开发和广泛传播模拟将使广大受众了解空间系统中政府与商业相互依存的战略问题。该奖项反映了美国国家科学基金会的法定任务，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 Faculty Early Career Development Program (CAREER) grant will diagnose and provide understanding of strategic dynamics among a set of interactive and autonomous design actors through the combined use of game theory and simulations to inform architecture and design decisions. Design activities for engineering systems across infrastructure, aerospace, and defense domains more closely resemble a collective decision-making process than centralized authority in traditional systems engineering practice. This topic is of interest because U.S. agencies and firms are actively architecting systems with decentralized decision authority, described variously as Internet 4.0, cyber-physical systems, Internet-of-things, or systems-of-systems across domains, including energy, transportation, manufacturing, and space systems. The design of these engineering systems differs from that of other systems because their large scale, long lifetime, and proximity to social systems evoke complex features such as adaptation, self-organization, and emergence, which take place over strategic timescales. A deeper understanding of how strategic dynamics impact designer interactions across theoretical and empirical perspectives can help to avoid costly overruns and cancellations by identifying and mitigating undesirable dynamics in conceptual design phases. Advances in systems engineering must develop theory, methods, and tools to coordinate and facilitate collective design activities. This project builds on a line of economic methods applied to engineering design including utility theory, decision theory, social choice, and game theory. Collective systems design is modeled as a bi-level problem, where lower-level decisions correspond to an optimization problem and upper-level strategy decisions correspond to a coordination game. Normative models of agent behavior are based on classical and Bayesian game theory with utility functions incorporating behavioral factors such as risk attitudes. Multi-agent simulation studies evolution of strategies under repeated interaction among agents. Behavioral experiments collect empirical data about human decision-making for validation. The research will contribute new knowledge about how to characterize, study, and modify the strategic dynamics of engineering systems during conceptual design phases. The educational plan develops and delivers simulation activities to model systems problems by combining technical modeling and social interaction. The simulation activities behave as a highly abstracted model system to elicit rich strategic behaviors through face-to-face interaction, engage students with challenges of socio-technical problems, and retain computational tractability to teach analytical methods in educational contexts. Development and broad public dissemination of simulations in the context of Earth science space missions will expose a wide audience to strategic issues of government-commercial interdependency in space systems.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.