EAGER: Ensemble Design of Resource-Aware Control Strategies for Multi-Agent Robotic Systems

本项目将物理和化学的平均场方法推广到多智能体机器人系统的可扩展、网络资源感知、分布式控制策略的集成设计中，旨在开发保留底层多智能体机器人系统显著特征的宏观模型，并将这些模型用于分布式控制策略的设计。对于复杂的网络物理系统，这有望提供一种新的设计方法，这种方法可能适用于大型系统，因此，将为整个社区提供基础知识。这个高风险，高回报的项目整合了物理，化学，控制理论和机器人技术的思想，为多智能体机器人系统的协调策略的设计，验证和改进开发新的理论基础。这个项目吗?S的智力优势在于对网络物理系统的设计、验证和改进的集成方法。平均域方法提供了底层分布式系统的系统级抽象，同时保留了各种代理级交互的显著特征。将这些模型推广到相互作用的工程系统的集成中，为设计分布式控制器提供了新的方法，这些控制器对变化的网络资源敏感，并且可以预测和调整其性能以达到所需的短期和长期性能规格。更广泛的影响：这个项目的更广泛的影响是双重的。首先，平均场方法考虑了网络资源的使用和管理，为设计可扩展的分散控制和协调策略提供了一个集成策略。其次，与受生物学启发的方法不同，平均场方法能够设计分布式协调策略，其性能可以系统地预测和调整，以满足详细的性能规范。这有可能统一各种现有的多代理协调方法。研究成果将通过技术会议和期刊的出版物传播，并纳入PI？美国现有的本科和研究生课程以及旨在提高女性参与STEM领域的K-12外展工作。

This project, generalizing mean-field approaches from physics and chemistry for integrated design of scalable, network resource aware, distributed control strategies for multi-agent robotic systems, aims to develop macroscopic models that retain salient features of the underlying multi-agent robotic system and use these models in the design of distributed control strategies. For complex cyber physical systems, this promises to provide a novel design methodology that is potentially applicable to a large class of systems and, therefore, will result in foundational knowledge of use to the community at large. This high-risk, high-reward project integrates ideas from physics, chemistry, control theory, and robotics to develop new theoretical foundations for the design, validation, and improvement of coordination strategies for multi-agent robotic systems.The project?s intellectual merit lies in the ensemble approach towards the design, validation, and improvement of cyber physical systems. Mean-field methods provide a system-level abstraction of the underlying distributed system while retaining the salient features of the various agent-level interactions. The generalization of these models to ensembles of interacting engineered systems provides new methods for designing distributed controllers that are sensitive to changing network resources and whose performance can be predicted and adjusted to achieve both the desired short-term and long-term performance specifications.Broader Impacts: The broader impacts of this project are twofold. First, the mean-field approach takes into account network resource usage and management, providing an integrated strategy for designing scalable decentralized control and coordination strategies. Second, different from biologically-inspired approaches, the mean-field approach enables the design of distributed coordination strategies whose performance can be systematically predicted and tuned to meet detailed performance specifications. This has the potential to unify various existing multi-agent coordination approaches. The research outcomes will be disseminated through publications in technical conferences and journals and incorporated into the PI?s existing undergraduate and graduate curriculum and K-12 outreach efforts targeted at increasing female participation in STEM fields.