NRI: Planning, Collaborative Guidance and Navigation in Uncertain Dynamic Environments

动态、不确定环境中的导航是搜救、协同移动、分布式监控等应用中普遍存在的难题。协作解决方案非常有前景，因为它们有潜力利用人类和自动化的优势。然而，协作导航中的主要挑战不仅包括底层自动化有效地处理设计阶段可能没有考虑到的新场景和不断变化的环境的能力，还包括人-自动化交互需求。能够应对这些挑战的方法和工具的设计可以从根本上实现协作人类-机器人系统的新功能。这项研究的创新之处在于将控制理论、运动规划和人类引导相结合，为高动态和不确定环境中的导航提供高效的解决方案。拟议的项目还将创造机会，让代表不足的少数群体参与K-12外联活动以及本科生和研究生研究，促进跨学科协作，并开发一个新的跨学科研究生课程。该建议旨在开发一个通用的框架，用于复杂环境中的协作导航，该框架可以容纳数百个移动的障碍物(可能具有随机动力学)、非平凡的静态障碍物和循环中的人类。我们建议a)评估高动态环境中用户和自主系统短期和长期信息的权衡，b)将我们现有的算法技术扩展到更高复杂性的环境，例如多机器人和非平面环境，c)设计和测试几个用户界面，它们满足用户可观察性和用户可预测性的预定条件，以提高安全导航的有效性，以及d)通过Android应用程序实验验证我们现有的通过Android应用程序在动态、不确定环境中进行协作导航的设置。我们将开发紧密耦合的规划和控制工具，将人工指导和决策与自动化工具相结合，完成对高度动态和不确定环境中的安全的严格分析。开发的方法将在多个环境、人类受试者和微型机器人试验台上进行验证。

Navigation in dynamic, uncertain environments is a difficult yet ubiquitous problem in diverse applications such as search and rescue, coordinated movement, distributed monitoring and surveillance. Collaborative solutions are highly promising because of their potential to exploit strengths of both the human and the automation. However, major challenges in collaborative navigation include not only the ability of the underlying automation to effectively handle novel scenarios and changing environments that may not have been considered at the design stage, but also human-automation interaction requirements. The design of methods and tools that can address these challenges could enable fundamentally new functionality in collaborative human-robot systems. The novelty of the proposed research is in the integration of control theory, motion planning, and human guidance to provide highly effective solutions for navigation in highly dynamic and uncertain environments. The proposed project will also create opportunities to involve under-represented minorities in K-12 outreach and in undergraduate and graduate research, to facilitate interdisciplinary collaboration, and to develop a new interdisciplinary graduate course. This proposal aims to develop a generic framework for collaborative navigation in complex environments that can accommodate hundreds of moving obstacles (with possibly stochastic dynamics), non-trivial static obstacles, and humans in the loop. We propose to a) evaluate the tradeoff between short-term and long-term information for both users and autonomous systems in highly dynamic environments, b) extend our existing algorithmic techniques to environments of higher complexity, e.g., multi-robots and non-planar environments, c) design and test several user-interfaces, which satisfy pre-determined conditions for user-observability and user-predictability, for their effectiveness in improving safe navigation, and d) experimentally validate our existing setup for collaborative navigation in dynamic, uncertain environments via an Android app. We will develop tightly coupled planning and control tools, integrate human guidance and decision making with automated tools, and complete a rigorous analysis of safety in highly dynamic environments with uncertainty. The developed methods will be validated in multiple environments, with human subjects, and on a micro robot testbed.