NRI: INT: Robotic Shepherding for Flow Control in Uncertain Dynamic Environments

这个国家机器人计划项目将促进科学的进步，促进国家的繁荣和福利，并通过提高多机器人团队的能力来协助灾难疏散和人群控制，以保护人类，最大限度地减少伤亡和经济损失，从而确保国防安全。为了实现这一目标，该团队将开发用于网络化多机器人牧羊的分布式估计、控制和机器学习的硬件和软件。开发的算法将在各种其他应用中具有广泛的适用性，包括云机器人、机器学习和自动化履行。该项目将生产100个低成本的轮式移动的机器人，在大规模的机器人群上对算法进行实验测试。“牧羊人”和“绵羊”机器人将能够远程操作，允许人类参与异构的远程操作自主团队。该项目中开发的算法将与陆军研究实验室（ARL）合作进行测试，增加理论过渡到近期应用的可能性。本科生和研究生将获得这笔赠款的支持，他们也将有机会每年在芝加哥科学与工业博物馆向数千名学生展示他们的工作。为了实现有效的牧羊，该项目将开发新的算法来分布式估计放牧羊群的形状，分布式最优控制羊群的形状，和对鸟群动态的机器学习。这些算法将得到补充的相互作用的估计，学习和控制律的稳定性的严格分析。这项工作的宏伟目标需要在几个相关领域取得根本性进展，包括分布式优化和估计;分布式机器学习;耦合学习和控制系统的稳定性;网络可控性和可观测性。为了实现这些突破，该研究将建立在PI团队关于耦合分布式估计器和控制器的无源性理论的最新成果之上，灵活的一阶分布式优化算法。实验验证和人类互动的研究需要新一代的可远程操作的群体机器人。利用最新一代的低成本高性能硬件，该项目将创造一个由100个小型但功能强大的轮式机器人组成的群体，其价格比任何可比的商业群体机器人都要便宜一个数量级。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

This National Robotics Initiative project will promote the progress of science, advance the national prosperity and welfare, and secure the national defense by advancing multi-robot team abilities to assist in disaster evacuation and crowd control to protect humans, minimize casualty and economic damage. To achieve this goal, the team will develop hardware and software for distributed estimation, control, and machine learning for networked multi-robot shepherding. The developed algorithms will have broad applicability in a variety of other applications, including cloud robotics, machine learning, and automated fulfillment. This project will produce one hundred low-cost wheeled mobile robots to experimentally test the algorithms on a large-scale flock of robots. The ’shepherd’ and ’sheep’ robots will be capable of being teleoperated, allowing humans to participate in heterogeneous teleoperated-autonomous teams. The algorithms developed in this project will be tested in collaboration with the Army Research Lab (ARL), increasing the likelihood of transition of the theory to near-term application. Undergraduate and graduate students will be supported with this grant, who will also have opportunity to demonstrate their work to thousands of school children each year at the Chicago Museum of Science and Industry.To achieve effective shepherding, this project will develop novel algorithms for distributed estimation of the shape of the herded flock, distributed optimal control of the shape of the flock, and machine learning of the dynamics of the flock. These algorithms will be supplemented by a rigorous analysis of the stability of the interacting estimation, learning, and control laws. The ambitious objectives of this work require fundamental advances in several related areas, including distributed optimization and estimation; distributed machine learning; stability of coupled learning and control systems; network controllability and observability. To achieve these breakthroughs, the research will build on recent results by the team of PIs on passivity theory for coupled distributed estimators and controllers, flexible first order distributed optimization algorithms. The experimental validation and human-interaction studies require a new generation of teleoperable swarm robots. Taking advantage of the latest generation of low-cost high-performance hardware, this project will create a flock of one hundred small but highly capable wheeled robots, at a price point an order of magnitude less expensive than any comparable commercially available swarm robot.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.