Abstract: A robot “hand-eye” pose autonomous coordination was regarded as a uncalibration constrained programming  problem, and a visual closed-loop control method was proposed based on second-order cone constrained programming. Firstly, the visual servoing control algorithms were constructed in the image planes and Cartesian space, respectively based on images and positions. After that, by established the path constraint and the local minimal constraint rules, and a second-order cone convex optimization model was constructed to realize the compromise optimal control of image feature trajectory and robot motion path. Moreover, the proposed second-order cone constrained programming model was embedded with an adaptive state estimator, to realize robotic Jacobian matrix online mapping learning, and to solve the unknown problems of “hand-eye” calibration parameters and visual depth information. Finally, the uncalibrated robot visual positioning experiments prove the effectiveness of the convex optimization planning model, and the real grasping tasks illustrate the feasibility of the robot pose autonomous coordination.

Key words: pose coordination, constrained programming, uncalibration visual servoing, hybrid closed-loop feedback control

摘要： 将机器人“手眼”位姿自协调视为无标定约束规划问题,提出一种基于二阶锥约束规划的视觉闭环控制方法。在图像平面和笛卡儿空间分别建立基于图像与基于位姿的视觉伺服控制算法;在路径约束和局部极小约束准则下构建二阶锥凸优化模型,实现了图像特征轨迹和机器人运动路径折中最优;将所提出的二阶锥约束规划模型嵌入自适应状态估计器，实现了机器人雅可比矩阵在线映射学习,解决了“手眼”标定参数和视觉深度信息未知问题；最后通过无标定机器人视觉定位证明了二阶锥约束规划模型的有效性,真实环境抓取操作表明了机器人位姿自协调控制的可行性。

关键词: 位姿协调, 约束规划, 无标定视觉伺服, 混合闭环反馈控制