EAGER: Advancing the Engineering of Complex Systems through Automated Mechanism Design for Complex Network Formation

传统的工程方法已被证明对设计复杂的系统非常有效，如汽车和计算机，在这些系统中，部分的总和等于整体。然而，这些方法对于设计复杂的系统（如全球供应链、物联网、社交网络和智能电网）并不是很有用。这些系统在美国社会、工业和经济基础设施中发挥着迅速增长和关键的作用，需要以一种内在高效、稳健和可持续的方式设计复杂系统的新原则。探索性研究的早期概念资助（EAGER）奖支持基础研究，为设计此类复杂系统提供工程原理。来自计算机科学、博弈论、优化和机器学习等多个学科的知识是本研究不可或缺的一部分。此外，对复杂系统的强调将有助于拓宽工程研究的兴趣，并对工程教育产生积极影响。许多物理的、虚拟的和网络物理的复杂系统的高度相互依赖的性质，以及我们对它们的日益依赖，需要一个健全的基础来设计它们。本研究将为复杂系统的设计提供原则，将该问题视为战略贝叶斯网络形成博弈背景下的自动机制设计之一，以自动设计激励，从而实现多个全局设计目标。许多复杂系统固有的局部性和随机性激发了基于行为的网络形成规则视角，允许考虑更现实的场景，并可能对网络形成本身提供更深入的了解。研究团队将在不同的系统设计目标下进行模拟，设计模型并进行分析推导，以建立隐藏的关系，改变复杂系统的大小以更好地理解适应和进化，考虑固定的系统交互以揭示遗留或骨干基础设施对未来系统和机制的作用，并集成在线机器学习作为提供反馈的手段，从而允许自适应机制。

Traditional engineering methodologies have proven highly effective for designing complicated systems, such as automobiles and computers, where the sum of the parts equals the whole. However, these methodologies are not very useful for designing complex systems that are greater than the sum of their parts, such as global supply chains, the Internet of Things, social networks and smart power grids. The rapidly increasing and critical role such systems play in U.S. social, industrial and economic infrastructures necessitates new principles for designing complex systems in an inherently efficient, robust and sustainable fashion. This EArly-concept Grant for Exploratory Research (EAGER) award supports fundamental research to provide engineering principles for designing such complex systems. Knowledge from several disciplines including computer science, game theory, optimization and machine learning is integral to this research. Moreover, the emphasis on complex systems will help broaden interest in engineering research and positively impact engineering education. The highly interdependent nature of many physical, virtual and cyber-physical complex systems and our increasing reliance upon them, demand a sound basis upon which to base their design. This research will provide principles for designing complex systems by considering the problem as one of automated mechanism design within the context of strategic Bayesian network formation games to automatically devise incentives such that multiple global design objectives are achieved. The local and stochastic nature inherent to many complex systems motivates a behavior-based perspective on network formation rules, permitting the consideration of more realistic scenarios and may provide deeper insight into network formation itself. The research team will perform simulations, devise models and perform analytical derivations under different system design objectives in order to establish hidden relationships, vary complex system size to better understand adaptation and evolution, consider fixed system interactions to reveal the role of legacy or backbone infrastructure on future systems and mechanisms, and integrate online machine learning as a means for providing feedback thus allowing for adaptive mechanisms.