Research On Teleoperation Of Humanoid Robot Based On The Detection Of Human Motion Posture

With the increasing innovation of 5G technology,the new generation of communication technology undoubtedly brings more possibilities to various fields,especially in the field of artificial intelligence.Teleoperation is a kind of remote control technology realized by high technology.People hope to combine this technology with our robot,so that human can control the robot to solve various problems remotely.According to the research and investigation,robots developed by teleoperation technology are often far from human's shape structure and unable to really have the same activity ability as human,whose application scenarios are relatively narrow.This thesis presents a humanoid robot that can be operated remotely and the robot has the same structure as the human body.The joints meet the standard of adults.The lower part of the body is equipped with omnidirectional crane and can adapt to various venues.After obtaining the depth information,the joint data optimization algorithm can accurately locate the three-dimensional position coordinates of each joint,and through the dynamic interpolation algorithm and the inter frame data to greatly optimize the jitter problem in the motion process.Finally,the angle information is mapped to the joint actuator to realize the remote operation control.Firstly,based on the research purpose,this thesis designs the mechanical mechanism of the humanoid robot according to the proportion of the human upper body,which carries twelve MX series steering gears to control the motion of each joint.It is equipped with head,neck,shoulder,big arm,small arm,wrist,waist and other limb joints,and its mobility and range of motion are completely consistent with that of human beings.The pull rod design adopted by the forearm reduces its weight for stability.Most of the whole body adopts carbon fiber technology.Finally,we designed the head shell and the display device used to display the eyes,making the whole machine more close to human beings.Due to the complexity of the surrounding environment and the difference of the detected human body,the data obtained by the depth camera is not accurate,which may lead to severe jitter.Therefore,this thesis proposes a filtering algorithm which can not only predict the wrong data,but also smooth the joint data.This method can greatly improve the accuracy of joint data and solve the influence of the wrong data on teleoperation.Due to the limitation of frame rate,the captured trajectory is not completely continuous,which can not meet the requirements of high-precision steering gear.In order to solve this problem,a dynamic interpolation algorithm is proposed.Some angle information is added to improve the smoothness of joint motion according to the movement trend.