考虑障碍物尺寸信息的机械臂避障路径学习方法  

Motion Planning Algorithm for Autonomous Driving Based on Dynamic Motion Primitives

作　　者：毛飞鸿 冀晓春 黄开启 苏建华[2] Mao Feihong;Ji Xiaochun;Huang Kaiqi;Su Jianhua(School of Electrical Engineering and Automation,Jiangxi University of Science and Technology,Ganzhou 341000,China;Key Lab of Complex System and Intelligence Science,Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]江西理工大学电气工程与自动化学院,江西赣州341000 [2]中国科学院自动化研究所多模态人工智能系统全国重点实验室,北京100190

出　　处：《机电工程技术》2024年第10期136-142,共7页Mechanical & Electrical Engineering Technology

摘　　要：针对机械臂在存在障碍物的复杂工作场景中运动轨迹规划难的问题,提出了一种考虑障碍物尺寸信息的多关节机械臂避障路径学习方法。首先采用动态运动基元根据人工示教数据生成初始的避障运动轨迹,借鉴人类经验生成全局最优的运动路径;然后,定义了包含障碍物尺寸、机器人尺寸及二者之间的距离的避障修正量,克服了点障碍物丢失物体几何信息、超二次曲面对感知信息要求较高等缺点;提出了根据示教轨迹运动趋势转换修正量方向的方法,解决了机械臂避障运动不平滑且加速度大等问题。在与常见修正量避障方式的对比实验中,本方法能较好地保留示教轨迹的特征,同时生成的避障轨迹更加平稳安全。最后,通过ROS平台对UR5机器人执行地面清扫任务的场景进行仿真实验,以此验证了所提出方法的实用性和可靠性。To address the challenges of planning motion trajectories for robotic arms in complex working environments with obstacles,a learning method for obstacle avoidance paths for multi-joint robotic arms that incorporates obstacle size information is proposed.Initially,dynamical movement primitives are utilized to generate initial obstacle avoidance motion trajectories based on manually demonstrated data,drawing on human experience to generate globally optimal motion paths.Then,an obstacle avoidance correction quantity is defined,which includes the sizes of the obstacles,the size of the robot,and the distance between them,overcoming the shortcomings of point obstacles losing geometric information and the high requirements for perception information of superquadric surfaces.A method for changing the direction of the correction quantity based on the motion trend of the demonstrated trajectory is proposed,addressing issues of unsmooth obstacle avoidance movements and high acceleration of the robotic arm.In comparative experiments with common correction quantity obstacle avoidance methods,this method can better preserve the characteristics of the demonstrated trajectory,while the generated obstacle avoidance trajectory is smoother and safer.Finally,through simulation experiments on the ROS platform with a ground cleaning task scenario for the UR5 robot,the practicality and reliability of the proposed method are verified.

关 键 词：动态运动基元 避障规划 修正量 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]