Research On The Human-like Trajectories Planning Of Six-DOF Robots

At present,six degrees of freedom robots have been widely used in practical production,and gradually liberate human beings from heavy and repetitive tasks.However,in some workplaces where robots are needed to cooperate with humans,robots need to be used as tools to expand and extend human capabilities.Working together with human beings requires a high degree of safety for robot trajectory.In order to work together with humans,high requirements for the safety of robot motion trajectories are proposed.The purpose of this paper is to predict unknown trajectories based on the characteristics of a set of wrist motion trajectory data.A method for solving the motion trajectory of humanization is established by studying the characteristics of the known motion trajectory data of the wrist.The coordinates of the whole trajectory are planned when the terminal coordinates of unknown trajectory are entered.Firstly this paper introduces the theory of robot kinematics and the arm model of human that corresponding to the robot structure.Then,the motion trajectory of the wrist is captured by Kinect.On this basis,a trajectory planning algorithm based on K nearest neighbor is proposed.When the data of a given target point taken as input information,the algorithm can directly output the humanization trajectory after planning.After the inverse solution algorithm of the robot,we can obtain the robot's control method according to the planning motion trajectory and realize the humanization of the robot's end effector motion trajectory.Experiments show that the algorithm can learn online and it is applicable to processing small scale data sets.The result of the algorithm is consistent with the trajectories of the wrists.When the number of trajectories of the training data sets is large,the running time will increase significantly.Aiming at the deficiency of K nearest neighbor algorithm in dealing with large scale data,a trajectory planning algorithm based on extreme learning machine is proposed in this paper.This algorithm learns the motion trajectory of the wrist and trains a model that matches the motion trajectory of the wrist.The training data sets do not need to be called during the trajectory planning process,and it can be solved directly according to the model.It has faster response speed when dealing with a large number of training data sets.In addition,because the grasping path to the same target point of the robot executing mechanism is not unique,this paper uses the conditional preference network(CP-nets)to solve the optimal trajectory.According to the wrist motion preference model solved by CP-nets,the characteristics of human motion preference are analyzed.Even if the optimal trajectory is not collected,it can be solved by preference model.Experiments show that the motion trajectories obtained by CP-nets have higher humanization and comfort.In summary,the trajectory planning algorithm based on K nearest neighbor and the trajectory planning algorithm based on the extreme learning machine have their own advantages in dealing with data sets of different sizes.The optimal trajectory planning method based on CP-nets realizes the optimization of humanization trajectory.The above method can realize humanization of the trajectory of robot end effector and improve the human-machine interaction ability of robot.