Integration of Vision and Force Feedback for the Synthesis and Execution of Error-Tolerant Robot Motion Plans

该项目的目标是整合视觉和力反馈，以综合和执行容错机器人的运动计划。具体地说，探索了视觉约束曲面的纯几何表示。视觉约束曲面是由对象特征(例如边缘)及其在监控摄像机的图像平面上的投影定义的直纹面。计划控制方案以促进：符合视觉约束表面的运动；在接触视觉约束表面时终止的运动；以及同时受视觉和物理约束表面约束的运动。对现有前像式运动规划器的扩展将使综合利用视觉和力传感器的执行时间反馈的运动计划成为可能。虽然在力反馈的运动规划和控制以及视觉反馈的运动控制方面已经做了很多工作，但到目前为止还没有报道的系统能够基于视觉反馈创建运动规划或者能够将力和视觉传感相结合来进行运动控制。这项研究解决了其中的每一个问题。

The objective of this project is the integration of vision and force feedback for the synthesis and execution of error-tolerant robot motion plans. Specifically, a purely geometric representation of visual constraints in terms of visual constraint surfaces is explored. A visual constraint surface is a ruled surface defined by an object feature (such as an edge) and its projection onto the image plans of a supervisory camera. Control schemes are planned to facilitate: motion that complies to a visual constraint surface; motion that terminated on contact with a visual constraint surface; and, motion that is simultaneously constrained by visual and physical constraint surfaces. Extensions to existing preimage style motion planners will make possible the synthesis of motion plans that exploit execution-time feedback from vision and force sensors. While much work has been done in the areas of motion planning and control with force feedback and motion control using visual feedback, to date there has been no reported system capable of creating motion plans based on visual feedback or capable of integration of both force and vision sensing for motion control. This research addresses each of these issues.