基于网络直角坐标机器人视觉伺服系统研究

为了实现定位抓取任务,提出基于网络的直角坐标机器人视觉控制系统。针对机器人运动控制的非线性与强耦合特性,采用神经网络控制器,构建了图像偏差与运动控制量之间的对应关系。通过对图像增强、边缘提取、特征提取等图像处理方法的综合分析,提出了一套优化组合图像处理法。在计算机网络环境下,采用自定义协议实现图像处理器与运动控制器协调控制,并将远程监控应用到机器人控制中。实验结果表明,该系统能够在视野范围内自动实现定位抓取动作。

Network based visual servo system of the orthogonal coordinate robot was proposed,in order to implement grasping job.The neuron network controller is designed to create the relationship between the image errors and the movement-control quantities,because that the robot control has the nonlinear and strong coupling character.By the conclusion obtained by comparing the various methods,such as image enhancement,edge check and image detect,the optimal image processing method is found.The harmonic control between the image processor and the movement controller is realized by the user-defined protocols.The remote-supervisory control is applied in the robotic control.These results for position control experiments indicate that the robotic system can wonderfully grasp the object in field of vision.

中国科学院沈阳自动化所机器人学研究室基金资助项目(RL200202)