Research On Human-Following Method Of Mobile Robot Based On LiDAR

Human-Robot Interactions(HRI)is the trend of mobile robots.Compared with traditional manual labor in the logistics industry,Human-Following Robot(HFR)could be flexible to complete tasks,efficient to transport cargoes,and harmonious to work in a Human-Robot-Cooperated Intelligent System(HRCIS).Concerning issues of poor efficiency and high-frequency repetition in transportation,the person following is an essential task to serve in logistics.At the same time,the purchase cost of Li DAR is decreasing in recent years,thus high-precision and high-frequency Li DAR is suitable for mobile robots in environment perception and target detection.Therefore,this paper studies the human-following method of mobile robots based on Li DAR.The main topics of this paper are as follows:Basic platform and control model in the human-following system.In order to verify the innovative method proposed in this paper,laser scanners,servo motors,and control computers of human-following robots are tested and debugged under the ROS system.All of these software and hardware platforms of human-following system are built with the idea of modularization.Meanwhile,in order to describe the control model of the human-following robot,the kinematic model and dynamic model of the human-following system are established in this paper.To unify the coordinate system,the robot coordinate system and Li DAR coordinate system are integrated into the world coordinate system.In this paper,the control model of the human-following robot is derived based on an integrated system.In the field of target identification,target detection methods of the human-following system are discussed.In order to realize the target recognition of the robot in the multi-person scene,a target detection method based on Support Vector Data Description(SVDD)is proposed.This method can map the 2-dimensional laser point data to the hypersphere to achieve high-precision target detection in a multi-person scene.The experimental results show that the detection accuracy of detecting targets is up to 80.7%.To predict the position and posture of a target at the next moment,a target prediction method based on Kalman Filter is discussed.In this method,the motion intention of the following target is estimated.In the field of motion control,target-following methods of the human-following system are discussed.In order to control the angle and speed of the robot system,this paper studies the target-following method based on the PD controller.Although the high accuracy of this method,a human-following robot often shakes and loses target.In order to improve the adaptive ability of human-following robots,this paper proposes the following method based on an improved spring model.In this method,there is a virtual spring between the robot and the target,so that robot could follow the target as smoothly as the virtual spring.The experimental results show the variance of the target trajectory is 25.37 m~2,the variance of the robot trajectory is 0.84 m~2,and the Root Means Square Error(RMSE)between the target and the robot is 1.69 m~2.