Supporting System using Multifingered Hand-Arm Robot

我们的研究目的是开发多指手臂机器人的支撑系统。多指机械手在抓取物体时，手臂和手不得与障碍物发生碰撞。由于多指手臂机器人具有多个运动自由度，即使已知物体和障碍物的几何信息，也难以生成多指手臂机器人的抓握构型。我们提出了两种方法来生成抓握结构，并根据物体周围的环境选择一种方法。一种方法称为手形法，优先考虑手形。另一种方法称为臂形法，优先考虑臂形。我们将抓取规划分解为几个子问题。我们的抓点规划的特点是使用启发式方法，以便快速生成抓点结构。三维图形仿真结果验证了该方法的有效性。我们的支持系统的目标是多指手臂机器人可以根据手势的指示与人类在同一工作空间内工作。我们的支撑系统由多指机械手、PC机和三目立体视觉硬件组成。由于系统需要在三维工作空间中识别手的位置和姿态，因此我们采用了三目立体视觉硬件。通过距离图像获得手和物体的三维位置。手势是根据手的长度和宽度来识别的。为了快速识别手势，我们提出了将手部区域分成两块的新方法。几个实验结果证明了我们的支持系统的有效性。

The purpose of our research is to develop the supporting system using multifingered hand-arm robot. When multifingered hand-arm robot grasps an object, the arm and the hand must not collide with obstacles. It is difficult to generate a grasp configuration for a multifingered hand-arm robot even through geometric information about an object and obstacles are all known, since a multifingered hand-arm robot has many degrees of freedom of motion. We propose two methods for generating the grasp configuration and select the one method according to the environment around the object. One method is called the hand configuration method and gives priority to the hand configuration. Another method is called the arm configuration method and gives priority to the arm configuration. We separate the grasp planning into several subproblems. The feature of our grasp planning is to use a heuristics in order to generate the grasp configuration rapidly. The effectiveness of our grasp planning is clarified by simulation results with 3D graphics. The goal of our supporting system is that the multifingered hand-arm robot can work with humans in the same working space according to the instructions of hand gestures. Our supporting system is composed of the multifingered hand-arm robot, a PC and a trinocular stereo vision hardware. Since the system has to recognize the position and posture of the hand in the three dimensional working space, we use the trinocular stereo vision hardware. The three dimensional positions of the hand and object are obtained through range images. The hand gesture is recognized based on the length and width of the hand. We propose the new method that the hand area is divided into two blocks in order to recognize the hand gesture rapidly. The effectiveness of our supporting system is clarified by several experimental results.