Research On Robot Navigation And Obstacle Avoidance Based On Binocular Vision In Indoor Envionment

Robot industry is an important symbol of the development of science and technology in the 21 st century.With the rapid development of the robot industry,the application of indoor-serving robots has become more and more common,which poses increasing challenges for mobile robot technology.In robotics,navigation and obstacle avoidance technology is one of the key technologies.Given a goal in the map of the environment,the technology can give a plan of a global static viable path,capture the depth of the environment and color information with the binocular sensors and use local path planning technology in the actual walking to avoid moving obstacles.Based on the study of global static path planning and local obstacle avoidance algorithm,this paper proposes some improvements to the existing technologies as follows:In this paper,the existing SLAM technology is researched firstly,and the binocular sensor is used to collect the surrounding environment information.Through experiments,the map of the indoor environment is successfully established and the robot is positioned correctly in the map.Based on the grid map,some common global path planning algorithms are introduced,and a path planning algorithm called jumping point search algorithm(JPS)is emphatically studied.The theory and experimental results show that the proposed algorithm has better performance in Efficiency compared to A-star algorithm.Then,the concept of path redundancy is proposed and the path redundancy detection algorithm is proposed according to the path angle problem planned by the JPS algorithm.Based on this,a jumping point search and post processing(JPS-PO)algorithm is proposed to reduce the length of the actual path at the cost of some more time.In addition,this paper also studies the local obstacle avoidance algorithm based on the dynamic window approach.Considering the dynamic pedestrian obstacle problem which often appears in the scenario of serving robots application,we find that the traditional dynamic window approach can not deal with the pedestrian with motion trails in advance.Based on this,this paper uses the HOG + SVM method for pedestrian detection,using Kalman filter to consider the pedestrian movement rules before and after the camera frame,and predicts the following motion trajectory of the pedestrian obstacle.And then proposes a concept called collision factor to evaluate the possible collision between the robot trail and the pedestrian prediction trajectory.This factor is added to the objective function in traditional dynamic window approach.The experiment proves that this method can avoid the moving pedestrian obstacle in the environment in advance.Finally,the use of global path planning or local path planning alone can not meet the needs of the actual application situation,so this paper combines the two algorithms together.Based on the ROS platform in the actual scene,this paper completes the task of robotic navigation and obstacle avoidance,which proves the proposed method's feasibility.