采用Kinect的多臂协调操作系统

Kinect-based dobby coordinate operating system

为完成基于Kinect的多臂协调精细操作的任务,搭建了一套双机械臂操作系统,利用Kinect作为视觉传感器对场景进行实时检测,并利用基于工作空间的RRT算法对其中一台七自由度机械臂末端进行路径规划完成目标的自主抓取。根据手眼协调控制技术,利用另外一台六自由度机械臂末端的摄像机采集的图像误差控制机械臂的运动,并利用粒子滤波算法对目标进行实时跟踪。通过设计一套双臂协作完成物体交接的实验系统,完成了多臂协同操作的任务,并验证了实验方法的可靠性。

Using Kinect and mechanical arm,we build a dual-arm accurate operation system in which Kinect is used for real-time scene dection and a work-space based RRT algorithm is proposed to plan the path for the 7-DOF arm＇s target grabbing. A camera at the end of the other 6-DOF arm is used to control the arm tracking the target.The method used for optical target tracking is particle filter-based algorithm and the difference between the target position in the image and the image center is used for controling the arm. In order to test the system,a dual-arm cooperation object transfer experiment was proceed and the successful experient result proved the reliability and efficience of the system.