Abstract: In image-based visual servoing control, image change is directly interpreted to camera motion, there isn’t direct control over the Cartesian velocities of the robot end-effector. As a result, the robot executes trajectories can be indirect or seemingly roundabout in Cartesian space. This paper proposes a new control scheme that overcomes the problem of the conventional visual servoing. The proposed approach separates the rotation from general motion, and control robot to rotate firstly, till current image feature is coincident in direction with goal image feature, then control robot to execute the rest motion by common method. As there is not interferences of translation and rotation of image, the performance of this approach is more advantage in the setting times and the generating optimal motion trajectories of Cartesian space than classical IBVS. Camera retreat is interpreted with the control law of classical IBVS and the perspective imaging model of camera. We illustrate new control scheme with a variety of simulation results.

Key words: robot, visual servoing, camera retreat, image feature.

摘要： 基于图像的视觉伺服方法,图像的变化直接解释为摄像机的运动,而不是直接对机械手末端实现笛卡尔速度控制,导致机械手的运动轨迹迂回,产生摄像机回退现象.针对这一问题,提出了将旋转和平移分离并先实现旋转的视觉伺服方案.该方案计算量小,系统响应时间短,解决了图像旋转和平移间的干扰,克服了传统基于图像视觉伺服产生的摄像机回退现象,实现了时间和路径的最优控制.并用传统IBVS的控制律和摄像机成像模型解释了回退现象的产生原因.二维运动仿真说明了提出方案的有效性.

关键词: 机器人, 视觉伺服, 摄像机回退, 图像特征.