Design And Implementation Of Campus Logistics Robots Visual Navigation System Based On Hybrid Maps

With the rapid development of e-commerce in China,online shopping has become an indispensable way of life and consumption,which has led to the rapid development of the entire courier industry.Since 2015,China's courier business volume has steadily ranked first in the world for six consecutive years,and is growing rapidly at a rate of20% to 50% per year.At present,in the campus courier delivery scenario,the traditional courier delivery method has the problems of low delivery efficiency and high cost.In this paper,with intelligent campus express delivery as the research background,a hybrid map-based visual navigation system is designed for the visual navigation problem of campus logistics robots between express delivery sites.The system is implemented based on the ORB-SLAM2 framework,which can run in a small memory computer.The whole system is divided into three parts: visual localization,visual obstacle avoidance and global navigation.(1)The visual localization part mainly solves the localization problem of the campus logistics robot in the campus environment.This part uses the visual localization framework of ORB-SLAM2.For the characteristics of campus environment which is easy to change,this paper proposes a key frame matching strategy based on covidibility graph to improve the visual repositioning algorithm in the framework.Experiments show that this strategy can improve the localization stability of the original algorithm in dynamic environments without affecting the localization accuracy of the original algorithm.(2)The visual obstacle avoidance part mainly solves the local obstacle avoidance problem of the campus logistics robot in the navigation process.To address this problem,this paper uses computer stereo vision technology to detect the obstacle information of the robot's local environment and uses the VFH+ obstacle avoidance algorithm to control the robot.Experiments show that the robot can stably complete the local obstacle avoidance task.(3)The global navigation part mainly solves the visual navigation of the campus logistics robot between courier sites and the memory consumption of the global natural landmark map.For these two problems,this paper uses a hybrid map-based global navigation method to solve them.The hybrid map is divided into two layers,and the first layer consists of the campus logistics topology map,which is used to plan the global path.The second level consists of multiple local natural landmark maps constructed by ORB-SLAM2,which is used to generate global paths as well as visual localization.Experiments show that this navigation method can solve the problem of visual navigation of the robot between campus courier sites and the memory consumption of the global natural landmark map.