具有深度自适应估计的视觉伺服优化

Optimization for visual servoing with depth adaptation

在手眼机器人视觉伺服中,如何确定机器人末端摄像机移动的速度和对物体的深度进行有效的估计还没有较好的解决方法.本文采用一般模型法,通过求解最优化控制问题来设计摄像机的速度,同时,利用物体初始及期望位置的深度估计值,提出了一种自适应估计的算法对物体的深度进行估计,给出了深度变化趋势,实现了基于图像的定位控制.该方法能够使机器人在工作空间范围内从任一初始位置出发到达期望位置,实现了系统的全局渐近稳定且不需要物体的几何模型及深度的精确值.最后给出的仿真实例表明了本方法的有效性.

How to design the moving speed of a camera and effectively estimate the depth of the observed object have not been solved properly for visual servoing with eye-in-hand configuration. The GMC（generic model control） is introduced to confirm the output trajectory of the system, and an optimization controller is designed for controlling the moving speed of the camera. In the case of unknown object depth, an adaptive update law is proposed to estimate the depth of selected features based on the approximate depth values of these features at the initial and desired positions. Under the driving of the controller, the camera can reach any desired location from any initial position in the robot workspace while the system is globally asymptotically stable. The method does not need any knowledge of a three-dimensional model of the object. Finally, a simulation is carried out to show its validity.