Research Of Human Motion Imitation System Of Robot Based On Posture Perception

As an important indicator to measure the modern technological level,robotics technology gets more and more attention at home and abroad.At present,the robot research has made great progress in motion ability,human-robot interaction,artificial intelligence,etc.,but still faces many technical problems.The imitation of human motion can improve the robot's ability of the human-robot interaction,learning and adapt,and allow the robot to serve human better.Existing studies on imitation of human motion have limitations on motion acquisition,mobility and effects.Therefore,the motion information of human obtained by body perception is used to control humanoid robots and industrial robots.Firstly,the right leg model of the humanoid robot NAO is established according to pose description,DH coordinate system and kinematics equation.The analytical method and BP neural network based on LM algorithm are adopted to solve the inverse kinematics respectively.A motion imitation system with humanoid robot is build,which consisting of NAO,Computer and a Microsoft somatosensory sensor Kinect.According to the respective characteristics of human skeleton and NAO joints,an analytical method is designed and implemented to transform human pose to NAO.Then,the positional relationship between COM projection and support polygon is adopted to judge the robot pose stability.The method is proposed to maintain robot's stability,which adjusting the target position of feet and ankles of unstable posture,and using the LM algorithm to solve inverse kinematics.Through posture sequence's filtering,speed limit and other steps,a complete control process is designed and implemented.The joint angles and link directions are respectively used to evaluate the pose similarity between human and robot quantitatively.Experimental results show that,the proposed pose transforming algorithm and controlling process could make robot imitate human's pose and motion similarly and stably.And the similarity evaluation demonstrates the effectiveness of motion imitation algorithm.In addition,a gesture-based SCARA robot writing system,using Kinect to obtain body posture information,is designed and implemented.It uses the palm's relative position of human hand to control the drawing trajectory of robot,and uses palm's status to control the pen up or down.In order to improve the controlling efficiency,a method fixing palm's trajectory based on error limitation is proposed.Experimental results show that,this system is easy to operate,and could make robot imitate human's wtring motion quickly and accurately.