Collaborative Research: Resilient Decentralized Estimation and Control for Cooperative Rigid Body Multivehicle Systems

该合作研究项目将通过整合现实车辆和通信模型，以及弹性分散估计和控制策略，显著提高协作多车系统的自主导航和控制能力。结果将适用于例如微型飞行器群和航天器编队。强大的分析技术可以控制单个车辆，表示网络系统的相互作用，适应通信延迟，并解释不确定性。该项目将解决阻碍将这些不同结果整合到一个统一框架中的重大技术障碍。利用大量互联车辆执行实际操作的能力在勘探、搜索和救援、农业、天气监测、监视、卫星大地测量和环境监测等应用中越来越有用，甚至是必要的。综合教育工作包括向高中学生演示机器人车辆和多智能体网络控制模拟。当前的多车辆估计和控制设计方法调用了强烈的简化假设，例如通过点质量或单或双积分器对单个车辆动力学进行建模；假设所有车辆具有相同的动力学和不确定性特性；忽视沟通延迟，或者假设它们是已知的或恒定的。所得到的理想控制器在实际条件下具有有限的功能。本项目力求通过使用鲁棒、非线性和混合控制理论、图论和几何力学等多种技术来克服这些限制。多车系统由一个独特的、全局的、无奇点的表示来描述，作为一个异构刚体网络，在特殊欧几里得群SE(3)的独立副本上演化，即在三维空间的刚性平移和旋转空间上。车辆间通信拓扑由具有异构、连接相关时延的时变图指定。全局、鲁棒和弹性的估计和控制方案将基于Lyapunov-Morse-Krasovskii函数方法，并使用分散算法实现，也就是说，每辆车只需要局部信息来计算其控制动作。结果将通过实验室规模的实验进行验证，并将推进分散估计和控制的系统理论基础，并实现更有能力、更健壮和更自主的多车系统。

This collaborative research program will significantly advance autonomous navigation and control capabilities for cooperative multivehicle systems, by incorporating realistic vehicle and communication models, and resilient decentralized estimation and control strategies. The results will be applicable, for example, to swarms of micro air vehicles and formations of spacecraft. Powerful analytic techniques exist to control individual vehicles, to represent the interactions of networked systems, to accommodate communication delays, and to account for uncertainty. This project will address the substantial technical obstacles that stand in the way of integrating these various results in a unified framework. The ability to exploit large numbers of interconnected vehicles to perform practical operations is increasingly useful -- even necessary -- in applications including exploration, search and rescue, agriculture, weather monitoring, surveillance, satellite geodesy, and environmental monitoring. Integrated educational efforts include demonstrations to high school students of robotic vehicles and multi-agent network control simulations. Current approaches to multi-vehicle estimation and control designs invoke strong simplifying assumptions, such as modeling individual vehicle dynamics by a point mass or by a single- or double-integrator; assuming that all vehicles have identical dynamics and uncertainty characterization; and neglecting communication delays, or assuming that they are known or constant. The resulting idealized controllers have limited functionality under real-world conditions. This project strives to overcome these limitations by using diverse techniques from robust, nonlinear, and hybrid control theory, graph theory, and geometric mechanics. The multivehicle system is described by a unique, global, and singularity-free representation, as a network of heterogeneous rigid bodies evolving on separate copies of the special Euclidean group SE(3) -- that is, on the space of rigid translations and rotations in three dimensions. The inter-vehicle communication topology is specified by a time-varying graph with heterogeneous, connection-dependent time delay. Global, robust and resilient estimation and control schemes will be based on a Lyapunov-Morse-Krasovskii functional approach, and implemented using decentralized algorithms, that is, each vehicle will require only local information to compute its control action. The results will be validated with laboratory-scale experiments, and will advance the system-theoretical foundations of decentralized estimation and control, and enable more capable, robust, and autonomous multi-vehicle systems.