移动机器人开锁作业的视觉定位与末端执行器设计

发布时间：2018-05-30 20:02

本文选题：移动机器人 + 目标识别　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：随着机器视觉和机器人控制技术的快速发展,越来越多的机器人应用到航天、深海以及核电站等极限作业环境中。当核电站受到外界因素的影响导致核泄漏时,将对人类的生存环境带来灾难性的影响,而进行救灾作业也会对救援人员的健康造成伤害,因此,具备移动作业能力的机器人成为核电站救灾作业的急需装备。利用机器人打开防火门的门锁是核电站救灾作业中十分重要,也是难度较大的一项任务。本文通过对开锁作业过程以及视觉定位误差的分析,研制了一款具备较高柔顺性能的开锁末端执行器,并对其中的一些关键技术进行了深入研究。通过对开锁作业的任务需求进行深入分析,明确了末端执行器需实现的功能,基于钥匙插入锁孔过程的相关约束分析,建立了趋向钥匙与锁孔配合目标的行为运动学模型,进而给出了参数化的定位抓持捕获区域的计算方法,通过对不同的钥匙插入工况进行分析,得到了相应的容差范围,确定抓持过程中末端执行器所需要的柔顺参数,为末端执行器的设计提供了理论依据。为实现核电站条件下视觉引导的自动开锁任务,开展了机器视觉有限分辨率条件下对目标图像的识别方法研究,应用HOG-SVM方法实现了锁孔以及门把手等目标的识别,并在此基础上提出一种融合目标物理特征的识别方法,提高了对目标锁孔识别的准确率。在锁孔的识别返回的目标图像基础上进行动态跟踪研究,针对传统的Meanshift跟踪方法的不足,提出了一种融合最大稳定极值区域特征的目标跟踪方法,降低了跟踪均值误差,对动态跟踪获得的最终目标锁孔图像进行了定位及方向的判别,并与激光跟踪仪测量得到的位置数据进行比较,对视觉定位误差值进行了评估。基于开锁过程柔顺参数和视觉定位误差,提出了具备较高柔顺性能的欠驱动自适应开锁末端执行器总体设计方案,并通过仿真软件对重要部件进行了拓扑优化。在对开锁作业仿真分析后,搭建了开锁试验平台,在力学测试评估基础上,提出了自动开锁策略,并对开锁末端执行器进行了综合测试,结果表明,开锁末端执行器能够在允许容差范围内,完成开锁任务。
[Abstract]:With the rapid development of machine vision and robot control technology, more and more robots are applied to space, deep sea and nuclear power plant. When a nuclear power plant is affected by external factors, it will cause a catastrophic impact on the human living environment, and disaster relief operations will also cause harm to the health of rescue workers. The robot with mobile operation ability becomes the urgent equipment of the nuclear power station disaster relief operation. It is very important and difficult to use robot to open the door lock of fire proof door in the disaster relief operation of nuclear power station. Based on the analysis of the process of unlocking and the error of visual positioning, a kind of terminal actuator with high compliance performance is developed in this paper, and some key technologies are deeply studied. Through the thorough analysis of the task requirement of the unlocking operation, the function of the terminal actuator is clarified. Based on the analysis of the relevant constraints in the keyhole insertion process, the behavioral kinematics model of the orientation key and the keyhole matching target is established. Then a parameterized calculation method of the capture area is given. By analyzing the different key insertion conditions, the corresponding tolerance range is obtained, and the compliance parameters of the terminal actuator in the grasping process are determined. It provides a theoretical basis for the design of the end actuator. In order to realize the task of automatically unlocking under the condition of nuclear power plant vision, the recognition method of target image under the condition of finite resolution of machine vision is studied, and the recognition of keyhole and doorknob is realized by using HOG-SVM method. On the basis of this, a method of fusion target physical feature recognition is proposed, which improves the accuracy of target keyhole recognition. Based on the target image from the keyhole recognition and return, the dynamic tracking method is studied. Aiming at the shortcomings of the traditional Meanshift tracking method, a target tracking method combining the maximum stable region feature is proposed, which reduces the mean error of tracking. Finally, the location and orientation of the final target keyhole image obtained by dynamic tracking are analyzed and compared with the position data measured by laser tracker, and the visual positioning error is evaluated. Based on the compliance parameters of the unlocking process and the visual positioning error, the overall design scheme of the underactuated adaptive unlocking end actuator with high compliance performance is proposed, and the topology optimization of the important components is carried out through the simulation software. After the simulation analysis of the unlocking operation, the open lock test platform is built. Based on the mechanical test and evaluation, the automatic unlock strategy is put forward, and the comprehensive test of the open locking end actuator is carried out. The results show that, The unlocking end actuator can complete the unlocking task within the allowable tolerance.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【参考文献】