CAREER: Game Theoretic Methods for Multiagent Coordination

一个多智能体系统的特征在于一个相互作用的子系统的集合，每个子系统根据本地信息做出本地决策。有许多多智能体系统的例子跨越工程和社会科学，例如，风电场、监控系统和交通网络。不管具体的域，在这样的系统中的主要目标是设计可接受的机制，用于协调各个子系统的行为，以确保紧急的集体行为相对于系统级目标是可取的。 这个项目的首要目标是开发多智能体协调的博弈论方法，广泛适用于社会和工程系统。 在社会多智能体系统的背景下，该项目将专注于在社会对这种机制的敏感性未知的情况下协调集体行为的最佳机制的推导，例如，交通网络的税收机制。 在工程多智能体系统的背景下，本计画将著重于（i）分散式控制演算法的推导，以确保多主体系统呈现最佳的相关行为，以及（ii）关于信息模式如何影响多智能体系统中可实现的性能保证的基本关系。更广泛的影响这个项目的一个中心目标是促进对多智能体系统的普遍理解，工程社区和超越。 毫无疑问，工程师们将负责设计各种各样的多智能体系统，其例子涵盖社会的各个方面。成功控制了这样的？复杂？系统要求工程师是多学科的，精通工程和社会科学。该项目将寻求通过与K-12学生的直接互动，公开讲座，在社会科学和工程科学的交叉点传播课程材料，以及与行业的直接互动来实现这一目标，以说明所开发方法的适用性。

A multiagent system can be characterized by a collection of interacting subsystems each making local decisions in response to local information. There are numerous examples of multiagent systems spanning both the engineering and social sciences, e.g., wind farms, surveillance systems, and transportation networks. Regardless of the specific domain, the primary goal in such systems is to design admissible mechanisms for coordinating the behavior of the individual subsystems to ensure that the emergent collective behavior is desirable with respect to a system-level objective. Intellectual Merit The overarching goal of this project is to develop game-theoretic methodologies for multiagent coordination with broad applicability to both social and engineering systems. In the context of social multiagent systems, this project will focus on the derivation of optimal mechanisms for coordinating the collective behavior in situations where the society's sensitivity to such mechanisms is unknown, e.g., taxation mechanisms on traffic networks. In the context of engineering multiagent systems, this project will focus on the derivation of (i) distributed control algorithms that guarantee that the multiagent systems exhibits the optimal correlated behavior and (ii) fundamental relationships pertaining to how informational patterns impact achievable performance guarantees in multiagent systems.Broader ImpactsA central goal on this project involves promoting the general understanding of multiagent systems to the engineering community and beyond. Undoubtedly, engineers will be responsible for the design of a wide array of multiagent systems with examples spanning all aspects of society. Successfully controlling such ?complex? systems requires that engineers be multi-disciplinary with a strong proficiency in both the engineering and social sciences. This project will seek to achieve this goal through direct interactions with K-12 students, public lectures, dissemination of course material at the intersection of social and engineering sciences, and direct interactions with industry to illustrate the applicability of the developed methodologies.