Dynamic Stable Grasp and Object-Manipulation by Multi-Fingered Hands with Soft Tips Based on Sensory Feedback

这项研究项目一开始就注意到，人类可以稳定地抓住一个物体并灵活地操纵它，而让多指机器人的手做同样的事情是相当困难的，尽管它们的机制类似于人类的手，并且每个手指关节都配备了一个优越的执行器。我们还注意到，有大量文献涉及使用多个无摩擦手指的2D或3D对象的形状闭合或不可动性问题。众所周知，三个无摩擦的手指足以固定任何三角形的2D对象，但固定平行六面体需要四个手指。在开始这项研究之前，我们发现，如果手指末端具有合适半径的半球形状，并且手指末端与物体表面之间产生滚动，那么在动态意义上，只需两个手指就足以实现对平行平坦表面的刚性物体的安全抓取。主要原因在于单个…在本研究项目中，首先从理论上验证了具有半球形指端的多自由度机器人手指能够稳定地抓取具有非平行或平行平面的刚性物体，并基于仅基于手指运动学和手指关节测量以及物体旋转角度的感知反馈信号来构建感觉-运动协调来控制其方位角。结果还表明，整个手指-物体系统是欠驱动的，总的动态范围。这个系统是多余的。尽管动力学存在欠驱动和冗余性，但研究表明，“流形上的稳定性”这一新概念在实现动态力/力矩闭合(稳定抓取)以及对象定向的并发控制方面起着至关重要的作用。其次，基于CSM(约束稳定化方法)的仿真结果验证了所提出的感觉-运动协调方法的有效性，并给出了在不同反馈项下确定反馈增益的准则。实验结果表明，在不知道物体大小和质心位置的情况下，可以实现对刚体物体的同时抓取和姿态控制。该理论现在正在扩展到通过使用两个或三个具有球形末端的多个手指来抓取和操纵3D多边形物体的情况。较少

This research project started from noting that human can grasp an object stably and manipulate it dexterously while it is quite difficult to let multi-fingered robot hands to do the same, even though their mechanisms resemble human hands and each of their finger joints is equipped with a superior actuator. We also noted that there is a vast literature concerning the problem of form closure or immobilizability of 2D or 3D objects by using a number of frictionless fingers. It is known that three frictionless fingers suffice to immobilize any 2D object with triangular shape but four fingers are necessary for immobilizing a parallelepiped. Just before starting this research we found that only two fingers are enough to realize secure grasp of a rigid object with parallel flat surfaces in a dynamic sense if finger ends have a hemispherical shape with appropriate radius and thereby rollings are induced between finger ends and object surfaces. The principal reason lies on the fact that a singl … More e rolling contact generates two contact constraint forces, one in the direction normal to the object surface at the contact point and another in the direction tangent to the object surface.In this research project, we firstly verified theoretically that a pair of multi-degrees of freedom robot fingers with hemispherical finger ends can grasp stably a rigid object with non-parallel or parallel flat surfaces and control its orientation angle by means of sensory-motor coordination based on sensory feedback signals constructed on the basis of information of only finger kinematics and measurements of finger joints and the rotational angle of the object. It is also shown that the overall fingers-object system is under-actuated and the total d.o.f. of the system is redundant. Notwithstanding such underactuation and redundancy of the dynamics, it is shown that a new concept of "stability on a manifold" plays a crucial sole in realizing a dynamic force/torque closure (stable grasp) together with concurrent control of the object orientation. Secondly, simulation results on the basis of the CSM (Constraint Stabilization Method) verified the effectiveness of the proposed method of sensory-motor coordination and present guidelines for determining feedback gains in various feedback terms. Finally, experimental results show that concurrent grasp and orientation control of a rigid object can be realized without knowing the size of the object and location of the object mass center. The theory is now under extending to the case of grasp and manipulation of 3D polygonal objects by using two or three multi-dof fingers with spherical ends. Less