空间机器人自主视觉伺服控制策略

Strategy of autonomous visual serving control for space robots

针对传统的空间机器人遥操作控制方法受时延影响严重的问题,基于最新研制的4自由度空间机器人,提出一种基于多目相机的自主复合控制策略.该策略包含了眼在手和眼到手2种相机配置结构,每个相机均具有独立的位姿计算能力,前者保证了伺服控制的精度以及操作的灵活性,后者能够在观察到全局视景的情况下做出伺服控制.为满足机器人视觉实时控制的需求,对视觉处理算法进行了特别考虑.首先,设计了基于多边形形状拟合的特征识别方法,提出曲线矢量数据贪婪算法来处理图像遍历拟合过程中计算密集的问题;其次,结合特征识别结果和模型目标的空间信息,提出基于弱透视模型的单目位姿估算及优化算法;最后,依据所提策略在实验室环境中完成了自主导航及捕获任务,验证了在较低层次进行这种复合控制的可行性和有效性.

As time delay will affect the control of teleopration for traditional space robots,a multi-camera hybrid visual serving control method was present based on the newest developed 4DOF space robot.The hybrid scheme consists of an eye-in-hand camera and eye-to-hand camera configuration and each camera has the capability of determining pose of target independently.The first one guarantees good accuracy and the ability to explore the workspace;the second one ensures a panoramic sight of the workspace.Suitable vision process algorithm was studied to realize the real-time visual serving control.Firstly,based on polygonal fitting method,the object feature recognition was designed.Furthermore,a greedy algorithm based on curve vector was proposed to reduce the calculation time of Traversal fitting.Secondly,by integrating feature recognition and modular object,monocular pose estimation and optimization algorithm were studied.Finally,an experiment of autonomous navigation and grasp verified the feasibility and effectiveness of proposed hybrid control method from a low level.