Asymmetric Gait Generation for Legged Locomotion in Complex Environments via Off-Line Model Reduction and Real-Time Optimal Control

这个机器人基础研究（FRR）项目将创造新的控制工具，以提高有腿机器人在现实环境中的运动能力。目标是为具有高度关节的腿式机器人提供一个可推广和广泛适用的控制框架。这些新的控制工具将解决机器人的复杂性，由间歇性地面接触引起的混合动力学，以及小型电机提供的有限控制权限。虽然目前的人形机器人在直线匀速移动方面非常熟练，但在需要意想不到的速度和方向变化的情况下，它们可能会受到挑战。克服这一挑战将促进有腿机器人在家庭、办公室、工厂和仓库的主流应用中得到更广泛的采用。该项目还将通过本科生研究、顶点项目以及向科学、技术、工程和数学领域的少数族裔推广机器人技术来培养下一代工程师和机器人专家。研究最广泛的有腿机器人控制方法是使用对称的力/扭矩分布，这导致匀速直线运动的对称步态。然而，为了以可变速度移动和转向，机器人需要使用不对称的扭矩/力剖面，这导致了步态的不对称。腿部系统的高自由度使得不对称步态生成的计算要求很高，而欠驱动（驱动器少于自由度）使得步态稳定具有挑战性。拨款的中心思想是使用数据驱动的方法来近似步态中关键实例之间的动态。这些近似是用低阶多项式完成的，它们捕捉到了步态的不对称性。当将这些降阶模型整合到优化框架中时，可以实时求解。最后，计划在现实场景中对高维双足机器人进行广泛的实验验证和验证。该项目由跨部门机器人基础研究项目支持，由工程（ENG）和计算机与信息科学与工程（CISE）联合管理和资助。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识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 Foundational Research in Robotics (FRR) project will create new control tools to improve the locomotion capabilities of legged robots in real-world environments. The goal is a generalizable and broadly applicable control framework for legged robots with highly articulated limbs. These new control tools will address the complexity of the robots, the hybrid dynamics caused by intermittent ground contact, and the limited control authority provided by small motors. Although current humanoid robots are highly proficient at moving at a constant speed in a straight line, they can be challenged in situations requiring unanticipated changes in speed and direction. Overcoming this challenge will facilitate the broader adoption of legged robots for mainstream applications in homes, offices, factories, and warehouses. This project will also train the next generation of engineers and roboticists through undergraduate research, capstone projects, and robotics outreach to minorities in Science, Technology, Engineering, and Mathematics.The most extensively investigated control approach for legged robots is to use symmetric force/torque profiles, which leads to symmetric gaits characterized by a constant speed, straight-line movement. However, to move at variable speed and to steer, robots need to use asymmetric torque/force profiles, which leads to asymmetric gaits. The high degrees of freedom of the legged system makes asymmetric gait generation to be computationally demanding and the underactuation (fewer actuators than degrees of freedom) makes gait stabilization to be challenging. The central idea of the grant is to use data-driven methods to approximate the dynamics between key instances in the gait. These approximations are done with low-order polynomials and they capture the asymmetry of the gaits. These reduced-order models, when incorporated within an optimization framework, can be solved in real-time. Finally, extensive experimental verification and validation is planned on a high-dimensional bipedal robot in real-world scenarios. This project is supported by the cross-directorate Foundational Research in Robotics program, jointly managed and funded by the Directorates for Engineering (ENG) and Computer and Information Science and Engineering (CISE).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.