毕业生知识库

铁路编组站是货运铁路网络中的重要环节，负责完成货运车列的到达、编 组、出发作业。火车摘钩是编组站作业中的关键部分，目前主要由人工完成，存 在着作业环境恶劣、危险性大、人力成本高和工作效率低等问题。本文针对摘钩 作业过程中的无人化和智能化需求，开展火车摘钩机器人系统设计与关键技术 研究，针对自动摘钩过程中的摘钩提杆识别、作业末端轨迹规划等问题，提出复 杂动态环境下摘钩提杆检测与位姿匹配、摘钩机器人作业轨迹规划等方法，设计 搭建了面向铁路编组站的自动摘钩机器人系统，主要研究内容如下：

第一，对摘钩机器人研究背景、意义及国内外研究现状进行综述，从摘钩机 器人系统设计、目标检测、点云匹配及机器人伺服控制四个方面对摘钩机器人关 键技术和发展方向进行分析。

第二，通过对驼峰编组站内摘钩作业对象和需求分析，设计了自动摘钩机器 人系统，该系统在硬件结构上包括移动作业机器人与远程控制系统，其中移动作 业机器人包括基于阿克曼转向的四轮移动平台和基于协作式机械臂的自主作业 平台。针对系统复杂度高、控制难度大的问题，提出基于分层递阶结构的机器人 控制系统，实现机器人的可靠、实时控制。对摘钩机器人系统及子系统进行运动 学分析，并利用远程遥控和示教再现的方式实现了机器人摘钩作业，验证系统功 能的完备性。

第三，针对摘钩提杆表面纹理信息弱、背景特征干扰大及检测实时性要求高 等问题，设计了相机-雷达感知系统及两阶段车钩识别与位姿匹配方法。第一阶 段基于深度学习方法，实现在大范围复杂场景下的摘钩提杆快速检测；第二阶段 基于 2D 和 3D 信息融合进行摘钩提杆的实时位姿匹配，实现了大尺度复杂环境 下摘钩提杆的快速定位及位姿估计。

第四，结合自动摘钩作业任务中的轨迹规划要求，提出了摘钩机器人作业轨 迹规划方法，通过对摘钩操作点的抓取姿态计算、作业末端的位置轨迹规划与运 动空间插值，实现机器人精准抓钩与提钩的作业规划。基于前述成果进行了摘钩 机器人自动摘钩实验，通过对实验中机械臂末端运动轨迹及各关节角度、速度变 化曲线进行分析，验证了所提方法的可靠性与有效性。

Railway marshalling station is an important link in freight railway network, which is responsible for the arrival, marshalling and departure of freight trains. In the traditional marshalling station operation, the uncoupling operation still needs manual completion, there are some problems such as bad working environment, high risk, high labor cost and low working efficiency. In view of the unmanned and intelligent requirements in the process of uncoupling task, this paper carried out the research on the system design and control methods of the train uncoupling robot. Aiming at the problems of coupler rod identification and trajectory planning of manipulator in the process of automatic uncoupling task, this paper proposed coupler rod detection and pose matching, working trajectory planning of uncoupling robot and other methods in complex dynamic environment. 
The main research contents are as follows:

Firstly, the research background, significance and domestic and foreign research status were summarized, and the key technology and development direction of the uncoupling robot were analyzed from four aspects: system design, target detection, point cloud matching and robot servo control.

Secondly, through the analysis of the task requirements of uncoupling task in hump marshalling station, the automatic uncoupling robot system is designed. The hardware structure of the system includes four wheel mobile platform based on Ackerman steering and autonomous robot based on cooperative manipulator. Aiming at the problems of high system complexity and difficult control, a robot control system based on hierarchical structure is proposed. The kinematics of the robot's moving chassis and manipulator were analyzed. The uncoupling operation of the robot was realized by means of remote control and teaching reproduction, and the functional completeness of the system was verified.

Thirdly, aiming at the problems of weak surface texture information, large background feature interference and high real-time detection requirements, a camera-radar fusion sensing system and a two-stage coupler rod recognition and pose matching method were designed. The first stage is based on the deep learning method to realize the rapid detection of coupler rod in a large range of complex scenes. In the second stage, real-time position and pose matching of the coupler rod was carried out based on 2D and 3D information fusion, which realized fast positioning and pose estimation of the coupler rod in large-scale complex environment.

Fourthly, aiming at the trajectory planning requirements of automatic uncoupling task, the trajectory planning method of the hook lifting robot was proposed. By calculating the grasping attitude of the uncoupling operation point, the position trajectory planning of the end of the job and the motion space interpolation, the uncoupling robot was realized the precise job planning of the grasp and lift it. The reliability and validity of the proposed method are verified by analyzing the trajectory of the end of the manipulator and the curve of the Angle and velocity of each joint in the experiment of automatic uncoupling robot.