【作者】 王丽丽；

【导师】 范晋伟；

【作者基本信息】 北京工业大学 ， 机械工程， 2017， 博士

【摘要】 果实采摘作业是农业生产中最耗时、最费力的一个环节,属于劳动力密集型的工作,急需实现果实自动化采摘以提高劳动效率。因此,本文主要对番茄采摘机器人的底盘行走控制方法、采摘机械臂、番茄自动识别与定位、系统集成等关键技术进行了系统研究,研发了具有自主导航移动、果实目标自动识别与定位、机械手自动采摘功能的番茄采摘机器人。主要完成了以下研究:(1)总体设计了番茄采摘机器人系统。根据成熟番茄的采摘作业要求,分析了番茄采摘机器人的工作原理和基本结构,运用机电工程设计方法,融合传感与检测、自动控制、多源信息融合等技术,总体设计了番茄采摘机器人,系统由四轮结构型式的行走底盘、四自由度关节型机械手结构、双目视觉识别与定位系统、激光自动导航系统组成。(2)采用模块化设计方法和嵌入式控制系统理论,研制了四轮独立转向行走平台。对移动平台进行了运动学分析,建立了番茄采摘机器人速度、角速度和偏转角度之间的运动学模型。在考虑动力合理分配和控制精度的前提下,采用模块化机器人系统设计方法设计了四轮独立转向驱动控制系统,运用基于低速阿克曼转向原理实现了番茄采摘机器人的四轮独立转向。(3)研制了基于激光传感技术的激光自动导航控制系统。分析了激光传感器的工作原理、测量误差、数据融合算法和导航控制方法,提出了基于改进的UKF算法的数据融合方法和基于自适应PID算法的激光自动导航控制方法。(4)研制了串联式灵巧型四自由度机械臂。在分析番茄采摘作业空间的基础上,设计与优化了番茄采摘机械臂各关节参数,基于蒙特卡洛法求解了番茄采摘机械臂工作空间范围。在建立D-H坐标系的基础上,对番茄采摘机械臂进行了运动学分析与仿真及逆运动学计算、基于关节空间的机械臂运动轨迹规划,设计了机械臂电气结构和软件控制界面。(5)研究了基于双目视觉的番茄自动识别与定位技术。通过番茄相关生物特征的测试和建库,确定了利用颜色特征识别番茄的技术途径。在分析番茄果实特性和图像分割方法的基础上,提出了基于自适应调整修正系数的Niblack算法的番茄目标提取方法,研究了番茄果实的椭圆模板匹配方法、三维重建方法、采摘路径规划方法。(6)开展了番茄采摘机器人各部分控制系统性能试验和整体试验。进行了移动平台转角控制、机械臂定位精度测定、机械臂末端承载能力测定、成熟番茄识别成功率和成熟番茄定位误差等性能试验,试验结果表明:移动平台在0°~360°旋转过程中,移动轮转角控制的最大平均绝对误差为0.10°,对应标准差最大不超过0.03°,控制较稳定,转角控制具有较高精度。机械臂定位精度最大误差2.5mm,平均误差1.1mm,带载能力3.5kg。双目立体视觉系统识别成功率99%,当识别距离小于600mm时,定位误差小于10mm。进行了激光导航控制系统和番茄采摘成功率等整体试验,试验结果表明:激光导航控制系统响应速度快,偏航误差小于8cm。番茄采摘机器人采摘单个番茄平均耗时15s,采摘成功率大于86.7%。更多还原

【Abstract】 Fruit picking is the most time-consuming and laborious work during the agricultural production.It belongs to labor-intensive work,and urgently needs to realizethe automatic operation to improve its labor efficiency.Thus,some key technologies for tomato picking robot were systematically investigatedin this paper,such as chassis control method,picking robot arm,tomato automatic identification and positioning.A novel picking robot with the functions of automatic navigation,automatic identification and positioning to fruit target was finallydeveloped.The main contents and results were as follows:(1)The tomato picking robot system was overall designed.According to the picking requirements for the mature tomato,the basic structure and working principle of the tomato harvesting robot were analyzed;Using mechanical and electrical engineering design methods,sensor and detection technologies,automatic control and multi information fusion technology,the overall structure of the tomato harvesting robot system was designed,which consisted of four wheel type walking system,four-degree-of-freedom manipulator,binocular visual recognition and positioning system,laser automatic navigation system.(2)Using the modular design method and the embedded control system theory,the four-wheel independent steering platform was developed,and the kinematics analysis of the mobile platform was carried out.The kinematic model of the speed,angular velocity and deflection angle of tomato picking robot wasestablished.Especially,under the consideration of reasonable power allocation and control accuracy,a four-wheel independent steering control system was designed by using the modular robot system design method.Based on the low speed Ackerman steering principle,the robothad a function of four wheel independent steering.(3)The laser navigation control system was developed.The working principle and measurement error of laser sensor,sampled data fusion algorithm and navigation control method were analyzed.A method of data fusion based on improved UKF algorithm waspresented,and an automatic laser navigation control method based on adaptive PID algorithm was proposed.(4)Asmart tandem typeof four-degreeoffreedom manipulator was developed.Based on the analysis of tomato picking work space,the parameters of each joint were calculated and the specific sizes of each joints of robot arm weredetermined.Based on Monte Carlo method,the working space of tomato harvesting manipulator wasdetermined.On the established D-H coordinate system,the kinetic analysis,simulation analysis and inverse dynamics calculationof the tomato picking manipulator werecarried out.Based on the joint motion planning of the joint space,the electrical structure and the software control interface of the manipulator were designed.(5)The tomato automatic identification and positioning technology were studied.With the aid of testing and construction of tomato related biological characteristics,the technical way of recognizing tomato with color characteristics was determined.Based on the analysis of tomato fruit characteristics and image segmentation method,a Niblack algorithm based on adaptive adjustment correction coefficient was proposed to extract tomato target.The fruit elliptic template matchingmethod,the three-dimensional image reconstruction method and the method for picking path planning of tomato fruit were studied.(6)The individual and overall performance test for each function modules and the whole tomato picking robot system respectively.The individual performance tests included mobile platform corner control test,robot arm positioning accuracy test,bearing capacity test ofmechanical arm,error test ofmature tomato identification.The experimental results showed that the maximum absolute error of the moving wheel angle control was 0.10° during the rotation of 0~360°,and the corresponding standard deviation was not more than 0.03°.The control system was verystable and the corner control hada high precision.The maximum error of robot positioning accuracy was 2.5mm;the average error was 1.1mm,theload capacity of machine could reached to 3.5kg,and the recognition rate of binocular stereoscopic vision system was99%.When the recognition distance was less than 600 mm,the positioning error was less than 10 mm.The overall test focused on the laser navigation control system test and the tomato picking success rate test,whose resultindicated that the response speed of laser navigation control system was so fast that the yaw error was less than 8cm.The average time to pick a single tomato was within15 s,the picking success rate was more than 86.7%.更多还原