基于ROS的堆垛机器人运动规划研究

【作者】 王鑫；

【导师】 袁庆霓；

【作者基本信息】 贵州大学 ， 机械工程， 2021， 硕士

【摘要】 随着电子商务和货运物流行业的飞速发展,仓储作为物流行业中的重要组成部分,对货物的堆垛和分拣的要求越来越高,而在传统的仓储行业中,人工堆垛和分拣货柜中的物品已经成为仓储物流的瓶颈,因此需要实现仓储自动化促进智能仓储的产业升级。本课题针对仓储货柜堆垛问题,对基于ROS的堆垛机器人运动规划进行研究,包括堆垛过程中的机器人的运动学分析、手眼标定方法和路径规划等内容,具体如下:(1)针对本课题选取的仓储机械臂,进行运动学分析。首先,建立UR5型机械臂的D-H坐标系和D-H参数表,计算相邻杆件间的齐次变换矩阵,建立正向运动学方程,求解机械臂末端相对基座坐标系的位姿;然后,建立逆向运动学方程,运用代数法进行反向求解机器人各关节角度变化信息;最后,建立基于UR5的仓储机械臂运动学模型。(2)针对于机器人的定位和抓取问题,建立了眼在手上(Eye-in-Hand)的手眼标定系统。首先,研究标定相机的成像原理;其次,通过相机标定获取相机内、外参数和抓取目标相对于标定相机的位姿信息;然后,建立眼在手上的手眼标定系统,推导出标定相机和机器人基座的相对位置关系;最后,使用Kinect V2相机进行标定实验,对相机标定和手眼标定方法进行实验验证,作为机械臂运动规划的基础。(3)针对快速拓展随机树算法(RRT)在机械臂路径规划中具有随机性高、搜索效率低、路径冗余等问题,不能在货柜堆垛场景中取得相对最优的光滑路径,提出了一种改进RRT-人工势场法混合算法对货柜堆垛运动进行路径规划。首先,引入目标搜索,对传统RRT算法进行改进,并使用改进后的RRT算法进行全局路径规划;其次,分别以斥力势场范围为阈值和机械臂末端执行器至末位置点影响修正引力函数和斥力函数,改进人工势场法,使用改进的人工势场法对局部路径进行优化;然后,使用三次非均匀B样条曲线对路径平滑处理并得到最终路径,最后,通过Python对算法进行了比较分析和数字模拟场景下模拟验证。(4)针对UR5机械臂在仓储堆垛运动的规划问题,搭建了ROS环境下货柜堆垛仿真平台。通过Python接口将改进后的RRT-人工势场法相结合的混合算法公有继承到OMPL类中,并在Gazebo中进行货柜堆垛运动仿真,验证了机器人能够对移动的物体进行识别定位,并抓取物体运动至货柜的指定位置,为机器人规划出光滑无冲击的无碰撞路径。更多还原

【Abstract】 With the rapid development of e-commerce and freight logistics industry,warehousing,as an important part of the logistics industry,has higher and higher requirements for the stacking and sorting of goods.In the traditional warehousing industry,manual stacking and sorting are required.The items in the picking container have become the bottleneck of warehousing logistics,so it is necessary to realize warehousing automation to promote the industrial upgrading of intelligent warehousing.This topic focuses on the problem of warehouse container stacking,and studies the motion planning of ROS-based stacking robots,including the kinematic analysis of the robot during the stacking process,hand-eye calibration methods and path planning,etc.The details are as follows:(1)Perform kinematics analysis on the storage robot arm selected for this topic.First,establish the D-H coordinate system and D-H parameter table of the UR5 manipulator arm,and the homogeneous transformation matrix between adjacent rods.Establish forward kinematics equations to solve the position and attitude information of the end of the robot arm relative to the base coordinate system;then,establish the inverse kinematics equations on the basis of forward kinematics,and use algebraic methods to solve the angle change information of each joint of the robot;Finally,establish the UR5 kinematics model.(2)Aiming at the problem of robot grasping and positioning,an Eye-in-Hand calibration system was established.First,study the imaging principle of the calibration camera;secondly,obtain the internal and external parameters of the camera and the posture information of the captured target relative to the calibration camera through camera calibration;then,establish the eye-in-hand hand-eye calibration The system derives the relative positional relationship between the calibration camera and the robot base;finally,the Kinect2 camera is used for calibration experiments to verify the camera calibration and hand-eye calibration methods as the basis of the robotic arm motion planning.(3)Aiming at the problems of high randomness,low search efficiency,and path redundancy in the path planning of the robotic arm,the rapid expansion random tree algorithm(RRT)cannot obtain a relatively optimal smooth path in the container stacking scene.An improvement is proposed.RRT-artificial potential field method hybrid algorithm for path planning of container stacking movement.Firstly,the target search is introduced,the traditional RRT algorithm is improved,and the improved RRT algorithm is used for global path planning;secondly,the range of the repulsive force potential field is used as the threshold and the end effector of the robot arm is used to influence the correction gravitational function and The repulsion function,the improved artificial potential field method,and the improved artificial potential field method are used to optimize the local path;then,the path is smoothed using the cubic non-uniform B-spline curve and the final path is obtained.Finally,the algorithms are compared through Python Analysis and simulation verification under digital simulation scenarios.(4)Aiming at the planning problem of the UR5 collaborative robot arm in the warehouse stacking movement,a simulation platform for container stacking under the ROS environment was built.The hybrid algorithm combining the improved RRT-artificial potential field method is publicly inherited into the OMPL class through the Python interface,and the container stacking motion simulation is performed in Gazebo,which verifies that the robot can identify and locate moving objects and grab them The object moves to the designated position of the container,and a smooth,impact-free,collision-free path is planned for the robot.更多还原