Control of Robots that Move by Leaping, Bouncing, and Swinging

各种各样的机器人任务，包括移动和操纵，都可以通过在不同的接触条件之间切换来表征。例如，Biped跑步机器人的左脚可以在地面上，右脚可以在地面上，或者没有脚在地面上。机器人的运动方程控制着机器人关节处的马达的力和力矩如何导致机器人移动，随着机器人改变与地面的接触而变化。这些切换方程在动态移动腿部机器人的运动规划和控制中提出了许多悬而未决的问题，例如美国陆军正在测试的机器人骡子，以及世界各地正在开发的新一代类人机器人。控制理论是控制这些机器人的最底层软件的基础，但它跟不上机器人设计的最新发展。这项研究将缩小这一差距，为控制动态移动机器人的软件提供坚实的理论基础。这将有助于实现腿式机器人在军事、服务和家庭应用中的愿景。这项研究还将为本科生和研究生提供在芝加哥科学和工业博物馆展示他们的工作的机会，影响下一代科学家和工程师。本项目专注于具有冲击力的不确定混合机械控制系统的能量、动量和不确定性的管理。虽然机器人运动控制的经典目标是零不确定性控制机器人的状态，但这项研究明确地处理了这样一个事实：对于所有具有冲击力的不确定混杂系统，这个目标是理想化的，对于许多由于欠驱动而形式上不可能实现的混杂系统。该方法适用于一大类具有冲击力的拉格朗日控制系统，重点是运动系统。建立了具有碰撞的不确定混杂机械系统结构的状态信念传播与过滤、信念可控性、运动规划和反馈控制原则。

A wide variety of robotic tasks, including locomotion and manipulation, can be characterized by switches between different sets of contact conditions. For example, a biped running robot can have its left foot on the ground, its right foot on the ground, or no feet on the ground. The equations of motion of the robot, which govern how the forces and torques of the motors at the robot's joints cause the robot to move, change as the robot changes contacts with the ground. These switching equations raise a number of open questions in motion planning and control for dynamically locomoting legged robots, such as robotic pack mules being tested by the US Army and a new generation of humanoid robots being developed around the world. Control theory, the basis of the lowest-level software that controls these robots, has not kept up with the latest developments in robot design. This research will narrow that gap, providing a firm theoretical "footing" for the software controlling dynamically locomoting robots. This will help realize the vision of legged robots in military, service, and home applications. The research will also provide opportunities for undergraduates and graduate students to demonstrate their work at the Chicago Museum of Science and Industry, influencing the next generation of scientists and engineers.This project focuses on the management of energy, momentum, and uncertainty for uncertain hybrid mechanical control systems with impacts. While the classical goal of robot motion control is to control the state of the robot with zero uncertainty, this research deals explicitly with the fact that this goal is idealized for all uncertain hybrid systems with impact and formally impossible for many due to their underactuation. The methodology applies to a broad class of Lagrangian control systems with impacts, with a focus on locomotion systems. The work will establish establish principles of state belief propagation and filtering, belief controllability, motion planning, and feedback control for systems with the structure of uncertain hybrid mechanical systems with impacts.