摘要
针对动态运动基元(Dynamic movement primitives,DMP)轨迹学习方法在机器人示教轨迹学习过程中轨迹位置收敛精度低的问题,提出了一种改进的动态运动基元机器人轨迹学习方法。首先采用操作空间动态运动基元对示教轨迹进行泛化,然后利用高斯函数在少数示教轨迹型值点处建立位置误差吸引力势场函数,并耦合在标准动态运动基元转换系统函数中。将提出的方法与标准动态运动基元方法追踪同一条轨迹进行对比仿真。仿真结果表明,所提方法能够有效提高机器人在轨迹学习过程中的轨迹位置收敛精度。
Aiming at the problem of low convergence accuracy of dynamic motion primitives(DMP)trajectory learning method in robot teaching trajectory learning process,an improved dynamic motion primitives robot trajectory learning method is proposed.Firstly,the dynamic motion primitive of operation space is used to generalize the teaching trajectory.Then,the Gaussian function is used to establish the potential field function of position error attraction at a few value points of teaching trajectory,which is coupled into the standard dynamic motion primitive transformation system function.The proposed method is compared with the standard dynamic motion primitive method to track the same trajectory.Simulation results show that the proposed method can effectively improve the trajectory position convergence accuracy of the robot in the trajectory learning process.
作者
张焕鑫
万俊
孙薇
吴洪涛
ZHANG Huan-xin;WAN Jun;SUN Wei;WU Hong-tao(Jiangsu University of Technology,Changzhou Jiangsu 213001,China;Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 210016,China)
出处
《计算机仿真》
2024年第8期476-480,共5页
Computer Simulation
基金
国家自然科学基金面上项目(51975277)。
关键词
动态运动基元
机器人
轨迹学习
吸引力势场函数
Dynamic motion primitives
Robot
Trajectory learning
Attractive potential field Function