Research On Real-time Motion Simulation Of Humanoid Robot

With the development of science and technology,robots play an increasingly important role in various fields such as human-computer interaction,artificial intelligence,and medical care.At present,the research on robots still faces many problems,such as low robot intelligence and weak anthropomorphism.Through the research of robot human motion imitation,the robot interaction ability can be improved,the robot can be anthropomorphized,and the robot can better serve human beings.At home and abroad,we have achieved relatively rich results in this field,but there are still limitations in the accuracy of human motion acquisition,imitation and real-time imitation.Based on this,this thesis uses the Kinect device and Vicon device to realize the real-time imitation of human body action by the anthropomorphic robot.The specific work is as follows:(1)A system based on the Kinect-based Nao robot to mimic the movement of the upper limbs of the human body.Firstly,the position data of the human skeleton points are obtained by Kinect,and the vector mapping is performed according to the mechanical structure of the robot.Secondly,the forward kinematics model of the robot is constructed by using the D-H matrix.Since most of the robots are driven by joint angles,it is necessary to solve the joint angles by inverse kinematics.In order to solve this inverse kinematics problem,this paper presents two solutions for analytic method and numerical method.In the analytic method,an inverse kinematics solution for solving multiple joint chains is proposed.In the numerical method,a Levenberg-Marquard(LM)algorithm based on geometric method is proposed.Solve the solution scheme of inverse kinematics.After obtaining the joint angles,in order to ensure the stability of the imitation,joint angle filtering and self-collision avoidance are added.In order to prove the performance and advantages of these two inverse kinematic solutions,this paper compares the similarity of the joint angle configuration and the similarity of the end trajectories with the geometric method and the traditional iterative method,and analyzes the realtime imitation.The experimental results prove that the two schemes designed in this paper can achieve real-time human upper limb imitation with high performance.(2)A system based on the Vicon-based Nao robot to mimic human motion.Because Kinect has low precision at the wrist joints,and must face the limitations of Kinect during the imitation process,it is difficult for Kinect to achieve wrist imitation and walking imitation.In view of this,use Vicon to solve the above problem.Specifically,based on the simulation of Kinect's upper limbs,the realization of the wrist joint simulation based on the geometric method was added,and the sensitivity and performance of the imitation were analyzed from the aspects of slight movement and overall movement.In addition,in order to enhance the freedom of imitation,the robot can reach any position and direction through the form of body sense.This paper also proposes a scheme to realize walking imitation through the somatosensory control robot.In walking imitation,this paper gives the switching mode of walking imitation and human upper limb imitation,and fully considers the influence of robot steering on walking in walking imitation,and gives corresponding solutions.The experimental results show that the scheme designed in this paper can achieve walking imitation well.