REU: SGER: Stratified Robotic Manipulation Experimental Platform

摘要IIS-9910602 Goodwin，John W.University of Notre Dame$99,996-12 mos：分层机器人操作的实验研究本项目的主题是关于分层系统的实验研究，主要目的是通过实验确定理论分层控制方法在多大程度上有效地控制真实的机器人系统。从本质上讲，分层系统是可以具有多个物理接触的系统(如机械手或腿部机器人运动)，这导致了表征机器人配置空间的通用结构。此类属结构可用于控制目的。该项目将通过多个机器人机械手实现对对象的协调操作，从而允许部分或全部机械手间歇性地接触对象。这种方法的新奇之处在于，它以足够的数学通用性来表示，因此控制算法适用于更广泛的一类被操纵对象。PI以前已经为分层系统开发了运动规划算法；然而，它们基本上是开环的。由于机械手与物体啮合或脱离的力学模型难以精确建模，使用纯开环控制方法将很难实现非常精细和精确的运动。此外，由于分层系统的高度非线性性质，运动中的小初始误差往往会随着时间的推移而增加，这也使得精确的大规模运动变得困难。实验的目的是确定这些困难的实际程度和性质，并将感觉信息(视觉和接触传感器)纳入离散反馈控制意义上，以提高分层控制方法的有效性。此外，还将使用自适应控制方法来补充分层运动规划算法，以增强其鲁棒性。

Abstract IIS-9910602Goodwine, John W. University of Notre Dame$99,996 - 12 mos.SGER: Experimental Study of Stratified Robotic Manipulation The subject of this project is experimental research concerning stratified systems with a main goal of experimentally determining the practical extent to which theoretical stratified control methodologies are effective to control real robotic systems. Essentially,stratified systems are systems which can have multiple physical contacts (such as robotic hands or legged robotic locomotion), which results in a generic structure characterizing the configuration space for the robot. This generic structure can be exploited for control purposes. This project will implement coordinated manipulation of objects by multiple robotic manipulators allowing for intermittent engagement of the objects by some or all of the manipulators. The novelty of the approach is that it is formulated in sufficient mathematical generality so that the control algorithms are applicable to a broad class of objects to be manipulated. The PI has previously developed motion planning algorithms for stratified systems; however, they are fundamentally open loop. Because of the difficulty of accurately modeling the mechanics of a manipulator engaging or disengaging an object, very fine and precise motion will be difficult with a purely open loop control methodology. Also, because of the highly nonlinear nature of stratified systems, small initial errors in motion will tend to grow with time making accurate large scale motions difficult as well. The goals of the experiments are to determine the practical extent and nature of these difficulties and to incorporate sensory information(both vision and contact sensors) in a discrete feedback control sense to improve the effectiveness of the stratified control methodology. Additionally, adaptive control methodologies will be used to supplement the stratified motion planning algorithms to enhance their robustness.