Precise and fine sensing of manipulation task state by a robotic hand with Multi-jointed fingers

本研究的目的是培养多指机械手对操作任务状态的精确感知能力。研究成果如下：提出了一种基于位置/力混合控制的微型力/力矩传感器主动触觉感知未知表面形状的方法。提出了一种利用曲面上点的三维数据和主动触觉获得的表面法线信息对未知曲面物体进行曲面重建的方法，并针对滚动接触多指精密操作中物体表面指尖滑移的检测和测量提出了一种新的方法。该方法基于视觉、手指关节编码器和指尖力/力矩传感器的多传感器融合。将指尖接触的切向滑动位移与指尖滚动运动引起的位移分离，提出了几种多指操作抓取的规划方法。为了从可能的候选人中挑选出更好的把握，引入了质量措施。规划者成功地产生了抓取，使抓取对象的运动范围最大化。已经建立了钉洞任务的任务模型。该模型由一系列任务状态转换、运动基元和确定任务状态的过程组成。

Objective of this research is to develop a precise and fine sensing ability of manipulation task state by a multi-fingered robotic hand. Followings are the results of the research.A method has been developed for active tactile sensing of shape of unknown surfaces using a miniature force/torque sensor mounted on by sliding motion of the fingertip based on position/force hybrid control. A method of surface reconstruction has been developed for an object with unknown curved surface using 3-D data of points on the surface and the surface normal information obtained by active tactile sensing.A new method has been developed for detection and measurement of fingertip slip on a surface of a manipulated object in a multi-fingered precision manipulation with rolling contact. The method is based on the multi-sensor fusion of vision, joint-encoders of the fingers, and fingertip force/torque sensors. Tangential slip displacement of the fingertip contact is separated from the displacement by rolling motion of the fingertip.Several planning methods of grasp for multi-fingered manipulation have been developed. Quality measures are introduced to select better grasp among possible candidates. The planner succeeded to generate a grasp which enables maximizing the range of motion of the grasped object.Task model for the Peg-in-Hole task has been developed. The model is composed of a sequence of task state transition, motion primitives, and procedures for determining task state.