CISE PostDoc: A Compositional Approach to Vision-Based Manipulation

9805817 Hager，Gregory耶鲁大学 CISE PostDoc：一种基于视觉操作的合成方法 在机器人手眼协调中，目前控制级视觉伺服和高级任务规范之间存在差距。 拟议的研究计划的目标是弥合这一差距的原则，但实际的方式。 要研究的主要问题是如何以及何时可以从简单的可证明正确的控制级组件组成一个给定的复杂的操作任务，以保证任务性能的正确性。 一个相关的问题是如何大规模的运动可以分为更小的部分，使同伦方法可以用来序列这些短的视觉伺服任务。 所提出的研究的另一个目标是开发用于指定手眼协调任务的方法，这些任务是使用对象附加坐标的机械手和环境不变的。这种视觉任务规范的方法是通过实施一个人机界面，操作员交互式地指定以对象为中心的操作任务进行实验测试。 该项目的影响是实现了PI在大型复杂手眼系统背景下的先前理论发展。 反过来，这些发展可以预期会激发更多的理论和实践问题，涉及到高层次的规范手眼协调任务。

9805817 Hager, Gregory Yale University CISE PostDoc: A Compositional Approach to Vision-Based Manipulation In robotic hand-eye coordination there is currently a gap between control-level visual servoing and high-level task specification. The objective of the proposed research program is to bridge this gap in a principled yet practical manner. The primary question to be studied is how and when a given complex manipulation task can be composed from simple provably correct control-level components in such a way as to guarantee correctness of task performance. A related question is how large-scale motions can be divided up into smaller segments so that a homotopy method can be used to sequence these short visual servoing tasks. Another objective of the proposed research is to develop methods for specifying hand-eye coordination tasks which are manipulator and environment invariant using object-attached coordinates. This approach to visual task specification is to be experimentally tested through implementation of a human-machine interface where the operator interactively specifies object-centered manipulation tasks. The impact of this project is the realization of previous theoretical developments by the PI within the context of a large, complex hand-eye system. These developments can in turn be expected to motivate additional theoretical and practical questions related to high-level specification of hand-eye coordination tasks.