Passive Separation Approach to Adaptive Visual Tracking for Robotic Systems
Passive Separation Approach to Adaptive Visual Tracking for Robotic Systems
复制标题
DOI：
10.1109/tcst.2017.2748061
复制
发表时间：
2018-04
影响因子：
4.8
通讯作者：
H. Wang;C. Cheah;W. Ren;Yongchun Xie
中科院分区：
文献类型：
--
作者：
H. Wang;C. Cheah;W. Ren;Yongchun Xie
运动学和动力学回路的分离对于工业/商业机器人的应用(即设计运动控制方案)和简化控制器结构是重要的，但由于缺乏这种分离，文献中的大多数视觉伺服算法很难被应用于大多数具有隐藏内部控制回路的工业/商业机器人系统。在本文中，我们研究了在没有图像空间速度测量的情况下，如何利用无源性和非线性反馈来实现对具有参数不确定性和时变深度的机器人系统的视觉跟踪的分离目标。我们提出了两种基于联合参考速度而不是联合速度的被动图像空间观测器，并基于这些观测器设计了两个自适应控制器，它们不需要图像空间速度测量，更重要的是实现了运动学和动力学回路的分离。通过在控制器和观测器中采用自适应逆雅可比控制和非线性反馈来实现回路分离，得到了两种适用于具有封闭结构的机器人的自适应运动学方案，但允许设计关节速度(或位置)命令。此外，所提出的第二自适应控制器不需要以使用潜在的高增益反馈为代价来求取估计深度的逆。数值仿真结果表明了该控制器的性能，并讨论了该控制器在工业/商业机器人中的应用问题。
Separation of the kinematic and dynamic loops is important for industrial/commercial robotic applications (i.e., for designing kinematic control schemes) and also for simplifying the controller structure, but most visual servoing algorithms in the literature, due to the lack of such separation, are hard to be justified as applied to most industrial/commercial robotic systems with a hidden inner control loop. In this brief, we investigate how passivity and nonlinear feedback are used to realize the objective of separation for visual tracking of robotic systems with parametric uncertainty and with time-varying depth, in the case of no image-space velocity measurement. We propose two new passive image-space observers that rely on the joint reference velocity rather than joint velocity, and based on these observers, we develop two adaptive controllers that do not require the image-space velocity measurement and more importantly achieve the separation of the kinematic and dynamic loops. The loop separation is achieved by resorting to the adaptive inverse-Jacobian-like control and nonlinear feedback in the controller and observer, yielding two adaptive kinematic schemes applicable to robots with a closed architecture yet admitting the design of the joint velocity (or position) command. In addition, the proposed second adaptive controller does not require inversion of the estimated depth at the expense of using a potentially-high-gain feedback. The performance of the proposed controllers is shown by numerical simulations and the implementation issues concerning the application to industrial/commercial robots are also discussed.