基于运动目标状态预测的机械臂跟踪技术研究

发布时间：2018-06-12 09:49

本文选题：目标状态预测 + 粒子滤波　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：传统的工业机器人在进行生产制造的过程中主要依靠离线编程手段完成任务,存在可执行任务单一、自主适应环境能力弱等问题,而利用视觉引导机器人工作可以有效弥补上述不足。在实际的应用中,利用视觉信息控制机器人抓取运动的目标,具有重要的现实意义和广阔的应用前景。针对运动目标的抓取,涉及许多环节和难点,如控制策略的设计复杂、视觉反馈信息易受噪声影响等,这对系统动态性能有很大影响。本课题针对上述问题进行研究,通过仿真得出结论,并通过实验对结论进行了验证。基于现实中的复杂场景,实现了一套视觉引导抓取系统。获取目标参数的滤波算法选用了粒子滤波,并通过MATLAB仿真分析验证其相对于广泛采用的扩展卡尔曼滤波在强机动目标抓取中的优越性。同时,进一步针对机械臂抓取操作中需要的目标位置、速度或加速度等参数采用了边缘粒子滤波算法,进一步提高参数精度;对于视觉反馈信息的获取,选择了可获得深度信息、操作空间更大的双目视觉系统;执行机构部分选用六自由度机械手臂,并对其建立了运动学模型,完成其运动学解的计算,通过仿真进行了验证;控制策略部分,为了应对设备性能不足、噪声、控制器延迟等因素的影响并缓解粒子滤波的实时性缺点,采用了预测控制的策略,通过状态观测器估计目标运动参数,以此来指导机器人执行抓取操作,最后通过实验验证仿真结论。
[Abstract]:The traditional industrial robot mainly relies on off-line programming to complete the task in the process of production and manufacture, which has the problems of single executable task, weak ability to adapt to the environment and so on. Using vision to guide robot work can effectively compensate for the above deficiencies. In the practical application, it has important practical significance and broad application prospect to use visual information to control the robot grasping moving target. There are many links and difficulties involved in moving target capture, such as the complexity of control strategy design, the visual feedback information is easily affected by noise, which has a great impact on the dynamic performance of the system. This paper studies the above problems, draws a conclusion by simulation, and verifies the conclusion by experiment. Based on the complex scene in reality, a visual guidance grab system is implemented. Particle filter is used to obtain target parameters, and the superiority of extended Kalman filter (EKF) in strong maneuvering target capture is verified by MATLAB simulation. At the same time, the edge particle filter algorithm is used to improve the precision of the parameters, such as the target position, velocity or acceleration, which is needed in the manipulator grasping operation. The binocular vision system which can obtain depth information and have more operation space is selected, and the manipulator with six degrees of freedom is selected as the executive mechanism, and the kinematics model is established to complete the calculation of its kinematics solution, which is verified by simulation. In the part of control strategy, in order to deal with the influence of equipment performance, noise, controller delay and other factors, and to alleviate the real-time shortcomings of particle filter, the predictive control strategy is used to estimate the moving parameters of the target by state observer. Finally, the simulation results are verified by experiments.
【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】：TP391.41;TP241

【参考文献】