Research On Imitation Learning Method Of Manipulator Based On Dynamic Movement Primitives Technology

With the rapid development of robot technology,the flexibility and complexity of the manipulator will also increase.While improving the manipulator obstacle avoidance capability and range of activities,the difficulty of manipulator track planning will also gradually increased.Learning from instruction techniques can effectively reduce the difficulty of manipulating the trajectory of the manipulator and reduce the difficulty of trajectory planning.Therefore,teaching and learning gradually develops into one of the important research contents of robotics.For the robots that serve the daily life of mankind,their working environment and operating objectives are often complex and simple,simple cloning teaching action can not meet the actual needs,the manipulator with athletic learning ability,have a certain self-learning ability will be more important.So the requires gradual increase in trajectory learning methods.The traditional trajectory learning method has some shortcomings in the generalization ability and the training of motor skills,and it is limited in the field of nonlinear.On the basis of analyzing the existing learning from demonstration techniques,this paper focuses on the theory of dynamic movement primitives and uses it in teaching and learning.And the use of 5 degrees of freedom manipulator for experimental verification,achieved the writing action learning.In order to realize teaching and learning,this paper has developed a set of teaching and learning manipulator system.By analyzing the technical requirements of the manipulator to determine the manipulator teaching methods.According to the working characteristics of the teaching manipulator and the working manipulator,the corresponding driving scheme is determined respectively.In order to facilitate the expansion of teaching arm structure,this paper based on modular thinking on the teaching arm of the modular division,and the design of the corresponding modules and connectors;to obtain teaching trajectory information,this paper set up teaching arm angle acquisition system,complete the writing of control program.Based on the job design requirements,this paper completes the design of the manipulator and analyzes the deformation of the working manipulator.In order to analyze the trajectory of the manipulator,based on the D-H method,the kinematic model of the manipulator is established,and the mapping relationship between thejoint angle and the trajectory is established.At the same time,the inverse kinematics equation of the manipulator is studied,and the inverse kinematics equation of the manipulator is established based on the inverse transformation method.In this paper,the Lagrange method is used to model the working manipulator,which lays the foundation for the follow-up robot trajectory learning and motion control.Based on the basic theory of dynamic movement primitives,this paper designs the trajectory learning algorithm and completes the preparation of related programs.The validity of the algorithm is proved by the reproduction and generalization of one-dimensional teaching trajectory.The multi-degree of freedom coupling of dynamic motion primitives is realized by sharing the canonical dynamical system,and the multi-degree of freedom motion trajectory learning is realized.Through the single and multi-degree of freedom trajectory reproduction and generalization simulation experiment,this paper analyzes the parameters of dynamic movement primitives and summarizes the problems of dynamic movement primitives.This paper analyzes the influence of the basis function distribution on the tracking error of the dynamic motifs and optimizes the distribution of the basis functions,which is the problem of the position error of the learning trajectory and the teaching trajectory.In this paper,we propose a method to increase the number of iterative steps and reduce the actual target value and the setting of the learning trajectory.In this paper,we propose a method to increase the number of iterative steps,and to reduce the actual target value and setting of the learning trajectory.Value between the position error.Aiming at the problem that the dynamic trajectory is equal to the target value when the initial value of the learning trajectory is equal to the target value,the forced function is zero and the effective learning trajectory can not be obtained.In this paper,a segmented learning method is proposed.The target value is equal to the starting value And the trajectory of the multi-segment target is different from the initial value.The validity of the method is proved by reproducing the sinusoidal trajectory.In order to verify the practical effect of teaching method of dynamic motifs,this paper uses the teaching and learning robot arm system to carry out trajectory learning experiment.In order to solve the problem of jittering in the process of manipulator movement,this paper uses the cubic spline method to plan the learning trajectory to achieve the effect of reducing the speed fluctuation and tracking error.Through the teaching and learning,the manipulator mastered the writing skills of the letter "s",and proved the feasibility of the teaching methodbased on the dynamic movement primitive teaching method.