Research On Robot Task Trajectory Learning Based On Interactive Peripheral Teaching

With the rapid development of robotics technology,the scope of robotics applications has become wider and wider,and more and more people are using them.Faced with the disadvantages of high difficulty,cumbersome process and certain requirements for professional knowledge in traditional robot teaching programming,the direct teaching method based on human-computer interaction has been proposed by people.At present,the most commonly used method of realizing direct teaching is based on the terminal six-dimensional torque sensor,but it is expensive and poor in versatility.Therefore,it is necessary to study other human-computer interaction methods to replace it,thereby reducing the teaching cost and improving the teaching method of Versatility.In response to this problem,this article uses the secondary development of the space mouse instead of the force sensor to design and develop a system that can directly teach;at the same time,considering the variability of the task environment faced by the robot,the robot is given the ability to be autonomous on the basis of force teaching.The ability to learn motion trajectories is becoming more and more important.Based on the dynamic motion primitive algorithm,this paper realizes that the robot has the ability to imitate learning and generalize the teaching trajectory.First,in order to successfully complete the realization of the teaching system and the application of the teaching learning algorithm,the kinematics analysis of the six-degree-of-freedom manipulator used in this paper is carried out.In order to meet the functional requirements of direct teaching,a space mouse is selected as a human-computer interaction device.Through its secondary development,the space mouse data acquisition was completed,and the human hand force information was acquired through the mouse induction data and human calibration experiment.In view of the problem that the robot control cabinet does not open the control authority,the control strategy of the teaching system adopts proportional control to realize the robot following force traction teaching movement;designed two teaching trajectory reproduction modes to deal with different usage scenarios and completed the design of the upper computer visual human-computer interaction interface of the teaching system.Then it is the learning of dynamic motion primitive algorithm.The algorithm is written and the simulation experiment of learning generalization of single degree of freedom and multi-degree of freedom trajectory is carried out using MATLAB.At the same time,the influence of the number of basis functions on the accuracy of learning imitation is analyzed.Based on the characteristics of the algorithm that only one trajectory sample is required to complete the learning,and in view of the fact that the direct teaching of people produces artificial jitter errors and the failure to ensure the success of one teaching,an improved method is proposed: first,use the dynamic time warping algorithm to complete the data preprocessing of the multiple teaching trajectories,then use the Gaussian mixture model to cluster the processed multiple teaching trajectories,and then use the Gaussian mixture regression to generalize the optimal trajectory to be used as that sample trajectory of the learning algorithm.Finally,it introduces and builds the basic composition of the direct teaching system,and designs related simulation work scene experiments on this basis,which verifies the feasibility of the teaching system and the effectiveness of the generalization ability of the robot's teaching learning.