CAREER: Bridging Self-Organized and Algorithmic Approaches to Multi-Robot Systems

这个项目包括研究解决机器人群体编程的传统方法的局限性，以便更容易地编程它们来解决团队中的问题。这项工作有助于实现这样一个未来：机器人可以处理重要的应用，比如那些与生命、生态和国家战略要素有关的应用（例如，制造业、机器人农业、行星探索）。为了做到这一点，这项研究正在为思考非常大的数据集而开发的方法、物理学家为小规模现象而发明的数学模型和今天的机器人群之间建立新的联系。正在探索的一项特别任务是以最少的人为干预管理碳封存的可行性；如果成功地扩大规模，这将对生态产生巨大的积极影响，提高全球的生活质量。在过去的几十年里，两种不同的观点已经主导了对多机器人系统的思考，每种观点或范式都有自己的哲学、工具、模型，甚至出版场所。本研究探索的观点是，范式的现有分离是退化的，源于早期人工智能关于表征的问题，为了取得进展，方法和工具必须适应混合两种范式特征的系统。这项工作是通过沿着两个方向桥接和巩固范式来提高多机器人系统的可扩展性、性能、鲁棒性和模型可预测性。第一部分使用一类新的分布式算法在范式之间的空间中引入并建立了一个新的位置，这些算法将亚线性时间逼近技术扩展到具有消息传递的通信设置。第二个推力通过应用重整化群变换方法来描述多尺度系统行为来发展和应用跨越既定边界的理论。

This project consists of research to address the limitations of traditional ways of programming groups of robots in order to make it easier to program them to solve problems in teams. The work is helping realize a future where robots address important applications such as those with life-saving, ecological, and national strategic elements (e.g., manufacturing, roboticized agriculture, planetary exploration). To do this, the research is establishing new connections between methods developed for thinking about very large data sets, mathematical models invented by physicists for small-scale phenomena, and today's robot swarms. One particular task being explored is the feasibility of managing carbon sequestration with minimal human intervention; if successfully scaled up this could have a huge positive impact ecologically, improving quality-of-life globally.Over the last couple of decades, two disparate perspectives have come to dominate thinking about multi-robot systems, each perspective or paradigm having its own philosophy, tools, models, and even publication venues. The idea being explored by this research is that the existing separation of the paradigms is vestigial, arising out of early AI questions about representation, and that for progress to be made it is essential that methods and tools accommodate systems that mix the characteristics of both paradigms. The work is improving scalability, performance, robustness, and model predictability for multi-robot systems by bridging and consolidating the paradigms along two thrusts. The first introduces and establishes a new position in the space between the paradigms with a new class of distributed algorithms that extend techniques for sublinear time approximation to communication settings with message-passing. The second thrust develops and applies theory that spans the established boundaries by applying renormalization group transformation methods to characterize multi-scale system behavior.