Research And Implementation Of Tracking And Obstacle Avoidance Algorithm For Quadruped Robot In Complex Environments

It is well known that the earth’s land surface is about 149 million square kilometers and that more than half of the land surface is rugged and unstructured,while quadruped mammals have excellent mobility and agility in similar environments.Therefore,a series of quadruped robots with quadruped motion characteristics have emerged.The quadruped robot has the short board advantage of other mobile robots in a complex environment,which has high flexibility,high mobility,high environmental adaptability,and high dynamics.It is one of the research hotspots of special robots in recent years.Nowadays,with the research on the quadruped bionic robot,its structure and stable motion control technology have become more and more mature,and improving the intelligence of the quadruped robot has become one of the current hotspots.To be able to integrate with people and achieve natural interaction,quadruped robots are indispensable to have the ability to perceive the environment.Research in this regard has become one of the core contents of the research field of quadruped robots.Therefore,this thesis carried out research on the environmental perception of quadruped robots,and the related research can be divided into autonomous tracking,personnel tracking,and autonomous obstacle avoidance.Mainly includes the following:(1)Based on the physical platform of a small electric-driven quadruped robot,an autonomous tracking algorithm with good tracking performance is designed on the basis of integrating color cameras and related processing units.Combined with the directional motion control strategy of the quadruped robot,the autonomous tracking of the robot is realized,which provides a certain algorithm application basis for the subsequent improvement of the robot’s environmental perception ability.(2)An autonomous tracking and obstacle avoidance algorithm for the quadruped robot based on a depth camera is designed and implemented.The quadruped robot is equipped with a Real Sense D435 depth camera sensor and a high-performance supercomputer Jetson TX2 processor.The color image information and depth image information are collected by the depth camera,and then the YOLOv3 convolutional neural network is used to identify the target personnel and obtain their position coordinates.The pixel coordinates of the obstacle image are obtained by image processing.Based on this information,the following steering algorithm and obstacle avoidance algorithm are designed to realize the single target tracking and obstacle avoidance function of the quadruped robot,and the feasibility and effectiveness of the algorithm are verified on the physical platform of the quadruped robot.(3)In order to further improve the ability of human tracking and autonomous obstacle avoidance of the quadruped robot and make up for the lack of specific target tracking in multi-target based on vision,a method of fusing 3D laser radar and UWB(Ultra Wide Band)positioning system is designed.This method first constructs three UWB base station site models and obtains the actual distance information between the base station and the tag through the UWB ranging method based on bilateral flight time.Based on the original trilateration method,this thesis proposes an improved weighted trilateration algorithm to determine the positions of the tracking personnel holding UWB tags and the quadruped robot carrying UWB tags.Then in order to detect obstacles of different sizes and heights as much as possible,3D laser radar is used to collect obstacle point cloud information.The original point cloud is preprocessed by a statistical filtering algorithm,and the ground point cloud is eliminated by the RANSAC(Random Sample Consensus)algorithm.Then the UWB tag position and obstacle point cloud are mapped by a two-dimensional grid map.Finally,the A* path planning algorithm based on the sliding window is introduced to plan the optimal path in real-time,and the personnel tracking and obstacle avoidance of the quadruped robot are completed.The feasibility and effectiveness of the algorithm are verified on the quadruped robot physical platform.