学术论文

      人机协作下的机械臂轨迹生成与修正方法

      Trajectory Generation and Adjustment Method for Robot Manipulators in Human-Robot Collaboration

      摘要:
      针对日常生活中多障碍物环境下的服务机器人机械臂轨迹规划问题,提出了一种人机协作下的轨迹生成与修正方法。首先,基于动态动作基元(DMP)模型,设计了一个能生成与示教轨迹形状相似的路径的方法。该方法针对多自由度耦合产生的轨迹形状畸变问题,通过将3维目标点投影于示教轨迹平面,再利用罗德里格旋转公式生成3维路径,保证了生成轨迹在各个方向都具有较稳定的形状特征。其次,针对存在多种形状障碍物的复杂操作环境,提出了通过插入交互点对轨迹形状进行修正的方法。并采用双抛物线插值算法使修正后的轨迹平滑。最后,基于人机协作的思想,在ROS(robot operating system)环境下搭建一个交互界面。从而操作者能够直观地辅助机械臂完成无障碍物和有障碍物的情况下的末端轨迹生成和修正。实验结果验证了该方法的直观性和灵活性,该方法适用于具有多种障碍物的复杂日常环境。
      Abstract:
      For trajectory planning of service robot manipulators in daily living environment with obstacles, a trajectory generation and adjustment method in human-robot collaboration is proposed. At first, an approach is designed to produce a trajectory with similar shape to the demonstrated one based on dynamic movement primitive (DMP) model. Here, the problem of shape distortion caused by multi-degree-of-freedom coupling is solved by projecting the 3D target point on a plane where the demonstration trajectory falls, then generating the 3D trajectory by using Rodrigues’ rotation formula. Thus the shape character of the produced trajectory can be ensured in all directions. Secondly, the trajectory can be adjusted by inserting interactive points to meet the operation requirements in cases of complex environments with obstacles in different shapes, and then it is smoothened by dual parabolic interpolation algorithm. Lastly, an interactive interface is built in ROS (robot operating system) under the idea of human-robot collaboration. Operators can intuitively help a manipulator to generate and adjust the 3D trajectory of the end-effector in the environment with or without obstacles. Obstacle-avoidance experiments validate the intuitiveness and flexibility of the proposed approach, which can adapt to complex daily-living environment with multiple kinds of obstacles.
      Author: LIU Weihui CHEN Diansheng ZHANG Lizhi
      作者单位: 北京航空航天大学机械工程及自动化学院,北京,100191
      刊 名: 机器人 ISTICEIPKU
      年,卷(期): 2016, 38(4)
      分类号: TP242.6
      在线出版日期: 2016年9月26日
      基金项目: 北京市科技计划重大项目(D141100003614002).