Research On Human Posture Detection And Positioning Method In Indoor Human-Robot Cooperation

Human-robot collaboration refers to the close contact between humans and robots to complete a task.The safety of personnel in the process of collaboration is directly related to the development of human-machine collaboration technology.Sending out the corresponding instruction to the robot according to the indoor position of the personnel or the minimum distance between the human body and the robot to protect humans from harm.Therefore,research needs to be carried out from two aspects:motion capture and indoor positioning.Using the inertial measurement unit(IMU)sensors to capture the human body posture.The position obtained by integrating the acceleration can achieve a higher positioning accuracy in a short time,but the error will gradually accumulate over time.Single ultra-wideband(UWB)indoor positioning technology is affected by multipath effects and non-line-of-sight(NLOS)errors which leads to large or missing ranging values,affecting positioning accuracy.Using the complementary characteristics of IMU sensor and UWB,the positioning results are fused to improve the indoor 3D positioning accuracy.The specific work of each key technology is carried out as follows:Firstly,a three-dimensional human skeleton model based on human biological characteristics and forward kinematics principle is designed.It requires that the human body cannot be deformed in the process of movement,and it can be data-driven to reproduce the human movement process.IMU is used to collect human motion data and wirelessly transmit it to the host computer software.After processing,it is transformed into the motion data of human model to drive the motion of human model and realize the real-time capture of human motion posture.Secondly,an improved pedestrian dead reckoning(PDR)method was proposed to capture the movement posture of lower limbs with pelvis as the root point.The further research is carried out from three aspects:gait detection,step length estimation and heading estimation respectively.The simulation experiments with traditional PDR method are evaluated,and the position errors under linear motion and square motion is analyzed.Since PDR algorithm does not meet the real-time requirements of human-robot interaction,a method of pelvic coordinate estimation based on IMU dynamic foot switching is proposed.The signals of foot angular velocity during walking is used to judge the situation of the foot touching the ground,and three dynamic switch mechanisms are proposed:left-foot touching the ground model,right-foot touching the ground model and both-feet touching the ground model.The knee joint coordinates and hip joint coordinates are calculated from the initial ankle coordinates,and finally the pelvic position coordinates are estimated.At the same time,the coordinates of the other ankle are estimated as the initial coordinates for the next touchdown.Experiments show that the proposed PDR method has higher positioning accuracy.Then,the label in the UWB system is placed at the pelvic position and the collected data is pre-processed to reduce the error.The IMU-based pelvic coordinates and UWB-based tag coordinates were fused.Since the motion data of the tag in the Z-axis direction of the UWB positioning system is relatively disturbed in the static state.To solve this problem,it is proposed to use the UKF algorithm to fuse the data in the XY axis direction and use the Kalman filter algorithm to smooth the Z-axis coordinate value of IMU.The host computer software is designed to receive the sensor data,complete the pose calculation and position estimation,and drive the movement of the human model in real time.Finally,the positioning accuracy of the pelvic coordinates based on IMU,the tag coordinates based on the UWB positioning system and the root point coordinates using the fusion algorithm are compared in the indoor environment.Set up the experimental environment of the base station without obstruction and with obstruction,that is the line-of sight and non-line-of-sight experimental experiment.Data processing is performed in MATLAB software,2D and 3D walking trajectory curves are drawn and the positioning errors is analyzed.Experiments show that the positioning accuracy of the fusion algorithm in non-line-of-sight and line-of-sight environments is better than that of single UWB positioning and single IMU positioning,improving UWB positioning accuracy while reducing the impact of IMU cumulative errors,verifying the stability and accuracy of the fusion IMU/UWB positioning algorithm.