NRI: FND: COLLAB: Hierarchical Safe, and Distributed Feedback Control of Multiagent Legged Robots for Cooperative Locomotion and Manipulation

NRI：FND：COLLAB：用于协作运动和操纵的多智能腿机器人的分层安全分布式反馈控制

• 批准号: 1924526
• 负责人: Aaron Ames
• 金额: $ 37.5万
• 依托单位: California Institute of Technology
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-15 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1924526&HistoricalAwards=false
• 关键词: NRI; FND; COLLAB; Hierarchical; Safe

The project aims to realize legged co-robots that cooperatively work with each other or people to achieve a variety of tasks in complex environments. One of the most challenging problems in deploying the next generation of ubiquitous co-robots is mobility in complex environments. More than half of the Earth's landmass is inaccessible to wheeled vehicles this motivates utilizing legged co-robots to access these environments and thus bring robots into the real world. Legged robots that are augmented with manipulators can form co-robot teams that assist humans in different aspects of their life. Although important theoretical and technological advances have enabled the development of distributed controllers for complex robot systems, including multiagent systems composed of collaborative robotic arms, multifingered robot hands, aerial vehicles, and ground vehicles, understanding how to control cooperative legged agents is an open problem. The challenges in achieving coordination in this domain stems from the fact that legged robots are inherently unstable, and the evolution of legged co-robot teams is represented by high-dimensional and complex hybrid dynamical systems which complicate the design of distributed control algorithms for control and coordination. There is a fundamental gap in knowledge of distributed control algorithms for safety-critical control of these inherently unstable, underactuated, and complex hybrid dynamical systems. The overarching goal of this proposal is to create a formal foundation, based on hybrid systems theory, scalable optimization, and robust and safety-critical control, to develop distributed and hierarchical feedback control algorithms for cooperative legged co-robots with manipulators to achieve a variety of tasks in complex environments. The proposed research will have broad societal impact through the formally principled and safety-critical deployment of ubiquitous collaborative legged robots in scenarios where robots can assist humans, e.g., disaster response. The integrated educational plan will have broad impact by designing a new course based upon the results, utilizing robots for STEM-based outreach for K-12 students, teachers, and under-represented minorities.The project aims to develop resilient and versatile algorithms that address cooperative locomotion and manipulation of high-dimensional hybrid models of legged co-robot teams in a safe, stable, and reliable manner. These algorithms will further enable legged co-robot teams to adapt to new tasks and environments with minimal modification to software. It will advance knowledge in the largely unexplored field of distributed control of large-scale hybrid system models of legged co-robots through specific objectives and key innovations in Scalability and Customizability. Intelligent and optimization-based motion planning algorithms will be created for hybrid models of legged co-robots to adapt to a wide variety of complex environments and new situations. Distributed and hierarchical control algorithms, based on nonlinear, robust and predictive controllers, together with scalable convex optimization, will be developed for coordination of multiagent legged robotic systems to enable agile locomotion patterns while manipulating objects in a dexterous manner. Finally, safety-critical control methods, based on set invariance and convex optimization, will be integrated with the hierarchical and distributed controllers for obstacle avoidance. To bridge the gap between theory and implementation, the proposed research will transfer the theoretical innovations into practice through experiments with a co-robot team consisting of multiple quadruped robots and one humanoid robot working collaboratively.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.

该项目旨在实现腿式协作机器人，这些机器人相互合作或与人合作，以在复杂环境中完成各种任务。部署下一代无处不在的协作机器人最具挑战性的问题之一是复杂环境中的移动性。地球上超过一半的陆地是轮式车辆无法进入的，这促使利用腿式合作机器人进入这些环境，从而将机器人带入真实的世界。带有操纵器的腿式机器人可以组成合作机器人团队，在生活的不同方面帮助人类。虽然重要的理论和技术进步，使分布式控制器的复杂机器人系统，包括多智能体系统组成的协作机器人手臂，多指机器人手，飞行器和地面车辆的发展，了解如何控制合作腿代理是一个开放的问题。 在这一领域实现协调的挑战源于这样一个事实，腿机器人是固有的不稳定，腿合作机器人团队的演变是由高维和复杂的混合动力系统，复杂的分布式控制算法的设计控制和协调。这些固有的不稳定，欠驱动，复杂的混合动力系统的安全关键控制的分布式控制算法的知识是一个根本的差距。 该提案的总体目标是建立一个正式的基础，基于混合系统理论，可扩展的优化，鲁棒性和安全关键控制，开发分布式和分层反馈控制算法的合作腿合作机器人与机械手，以实现各种复杂环境中的任务。拟议的研究将通过在机器人可以帮助人类的场景中部署无处不在的协作腿式机器人，产生广泛的社会影响，例如，灾难响应。该综合教育计划将根据结果设计新课程，利用机器人为K-12学生、教师和代表性不足的少数族裔进行基于STEM的外展，从而产生广泛的影响。该项目旨在开发有弹性且通用的算法，以安全、稳定和可靠的方式解决腿协作机器人团队的高维混合模型的协作运动和操作。这些算法将进一步使腿协作机器人团队能够以最小的软件修改来适应新的任务和环境。它将通过可扩展性和可定制性方面的具体目标和关键创新，在腿式协作机器人的大规模混合系统模型的分布式控制这一基本上未探索的领域中推进知识。智能和基于优化的运动规划算法将被创建用于腿式协作机器人的混合模型，以适应各种复杂环境和新情况。分布式和分层控制算法，非线性，鲁棒性和预测控制器的基础上，连同可扩展的凸优化，将开发多智能体腿机器人系统的协调，使敏捷的运动模式，同时以灵巧的方式操纵对象。最后，安全关键的控制方法，基于集合不变性和凸优化，将集成与分层和分布式控制器避障。为了弥合理论和实践之间的差距，该研究将通过由多个四足机器人和一个人形机器人组成的协作机器人团队进行实验，将理论创新转化为实践。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。

项目成果

Coupled Control Lyapunov Functions for Interconnected Systems, With Application to Quadrupedal Locomotion

互连系统的耦合控制李亚普诺夫函数及其在四足运动中的应用

• DOI: 10.1109/lra.2021.3065174
• 发表时间: 2021
• 期刊: IEEE Robotics and Automation Letters
• 影响因子: 5.2
• 作者: Ma, Wen-Loong;Csomay-Shanklin, Noel;Kolathaya, Shishir;Hamed, Kaveh Akbari;Ames, Aaron D.
• 通讯作者: Ames, Aaron D.

Data-driven Characterization of Human Interaction for Model-based Control of Powered Prostheses

人类交互的数据驱动表征，用于基于模型的动力假肢控制

• DOI: 10.1109/iros45743.2020.9341388
• 发表时间: 2020
• 期刊: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS
• 作者: Gehlhar, Rachel;Chen, Yuxiao;Ames, Aaron D.
• 通讯作者: Ames, Aaron D.

Online Learning of Unknown Dynamics for Model-Based Controllers in Legged Locomotion

• DOI: 10.1109/lra.2021.3108510
• 发表时间: 2021-10
• 期刊: IEEE Robotics and Automation Letters
• 影响因子: 5.2
• 作者: Yu Sun;Wyatt Ubellacker;Wen-Loong Ma;Xiang Zhang;Changhao Wang;Noel Csomay-Shanklin;M. Tomizuka;K. Sreenath;A. Ames

Learning to Control an Unstable System with One Minute of Data: Leveraging Gaussian Process Differentiation in Predictive Control

• DOI: 10.1109/iros51168.2021.9636786
• 发表时间: 2021-03
• 期刊: 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
• 作者: I. D. Rodriguez;Ugo Rosolia;A. Ames;Yisong Yue

Nonholonomic Hybrid Zero Dynamics for the Stabilization of Periodic Orbits: Application to Underactuated Robotic Walking

• DOI: 10.1109/tcst.2019.2947874
• 发表时间: 2020-11-01
• 期刊: IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY
• 影响因子: 4.8
• 作者: Akbari Hamed, Kaveh;Ames, Aaron D.
• 通讯作者: Ames, Aaron D.