Virtual Constraints: A New Paradigm for the Control of Motion

本提案涉及研究非线性动力系统运动控制的新范例。运动控制问题涉及设计算法，使一个或多个动力系统表现出规定的行为。运动控制问题在所有工程领域都存在，包括多足机器人的运动、自主飞行器的飞行控制以及地面、空中和太空飞行器的编队控制。传统的运动控制方法对环境中的干扰和不确定性具有有限的鲁棒性，因此不足以解决多腿机器人运动和航天器编队协调等新兴应用。在过去十年中，出现了一种新的运动控制范式，它有可能克服传统方法的局限性，并可能实现新一代运动控制算法。这种范例依赖于虚拟约束的概念，这是一种对控制系统状态的约束，它在物理上并不存在，但可以通过反馈控制来实施。虚拟约束的概念已经成功地用于诱导双足机器人的稳定行走步态，但要使虚拟约束范式超越双足机器人的运动，并变得足够普遍，以适用于大量的工程问题，还有很多研究要做。拟议的研究计划将为一类机械系统和自动驾驶车辆的虚拟约束系统理论奠定基础，并将研究其在卫星编队控制和机器人远程操作方面的应用。本研究计划的最终目标是一套通用工具来自动生成一组运动原语。每个运动原语将与虚拟约束相关联，并将对应于期望的行为。然后，一个不熟练的用户将能够选择一系列运动原语来诱导动力系统中的复杂行为。这项研究项目的成果可能在机器人、航空航天工程和康复工程等领域具有相当大的实际意义。

This proposal concerns the investigation of a new paradigm for motion control in nonlinear dynamical systems. The motion control problem involves designing algorithms to make one or more dynamical systems exhibit a prescribed behaviour. Motion control problems abound in all areas of engineering, and include locomotion in multi-legged robots, flight control of autonomous aircrafts, and formation control of terrestrial, aerial, and space vehicles. The traditional motion control methodology has limited robustness against disturbances and uncertainties in the environment, and it is therefore inadequate to address emerging applications such as locomotion in multi-legged robots and coordination of spacecraft formations. In the last decade, a new paradigm for motion control has emerged which has the potential to overcome the limitations of traditional methods, and may enable a new generation of motion control algorithms. This paradigm relies on the concept of virtual constraint, a constraint on the states of a control system that does not physically exist, but can be enforced via feedback control. The notion of virtual constraint has been used with great success to induce stable walking gaits in biped robots, but much research remains to be done to make the virtual constraint paradigm go beyond biped locomotion, and become general enough to be applicable to a vast array of engineering problems. The proposed research program will lay the foundations for a systematic theory of virtual constraints for a class of mechanical systems and autonomous vehicles, and will investigate its applications to the control of satellite formations and robotic teleoperation. The ultimate goal of this research program is a set of general tools to automatically generate a collection of motion primitives. Each motion primitive will be associated with a virtual constraint, and will correspond to a desired behaviour. An inexpert user will then be able to select a sequence of motion primitives to induce complex behaviours in the dynamical system. The outcomes of this research program may have considerable practical implications in areas as diverse as robotics, aerospace engineering, and rehabilitation engineering.