On the Design of Observer-based Dynamic Visual Servo Systems via Nonlinear H_∞ Model Predictive Control

在本研究中，我们考虑采用非线性H_∞模型预测控制设计基于观测器的动态视觉伺服系统。首先，将三维视觉反馈控制问题建立为机器人控制问题和摄像机与目标物体之间相对刚体运动估计问题；针对这一问题，我们提出了视觉伺服系统的非线性H_∞模型预测~7e控制律。具体来说，所提出的控制律的合适能量函数在控制设计中起着重要的作用。其次，采用高性能DSP和实时系统，由机械手、摄像机和图像处理板构成动态视觉伺服系统。在提供实时软件将所提出的控制律实现到实际系统的同时，考虑了最优问题的算法，以便由实际系统计算出控制器。最后，对所构建的动态视觉伺服系统的控制性能进行了实验验证。通过高性能工作站对实验结果进行了精确分析。为了用实验结果系统地验证所提出的控制律，不仅考虑了平均状态，而且考虑了最坏情况下的状态。通过仿真和实验，对优化问题的实时软件和算法进行了重新思考。

In this research, we consider the design of observer-based dynamic visual servo systems via nonlinear H_∞ model predictive control. Firstly, the 3-D visual feedback control problem is established as the robot control problem and the estimation problem of the relative rigid body motion between the camera and the target object. For this problem, we propose a nonlinear H_∞ model l)redicti~7e control law for the visual servo systems. Specifically, the appropriate energy function to the proposed control law plays an important role in the control design.Secondly, the dynamic visual servo system is constructed from a robot manipulator, a camera and an image processing board with a. high-performance DSP and a. real-time system. While we provide real-time software for implementing the proposed control law to the real system, the algorithm for optimal problems is considered in order to calculate the controller by the real system.Finally, we have experiments for evaluating the control performances using the constructed dynamic visual servo system. Experimental results are analyzed exactly by a high-performance workstation. Not only the states of the average but also the states of the worst case are considered in order to verify the proposed control law with experimental results systematically. Moreover, the real-time software and the algorithm for optimal problems are reconsidered via, simulations and experiments.