Research On Visual Navigation System Of Indoor Mobile Robo

Map construction and path planning are key technologies in indoor mobile robot navigation,and how to achieve real-time map construction and accurate path planning in complex indoor scenarios is the key to the indoor mobile robot navigation problem.When a robot is in an indoor location where both people and obstacles are present,the robot’s ability to analyse and understand such complex scenarios is still deficient,resulting in untimely responses to temporary obstacles during path planning.Therefore,the study of real-time semantic map construction and path planning for indoor mobile robots has important theoretical significance and practical application value.This paper will take visual navigation of mobile robots in indoor scenes as the application context,and propose a robot perception framework based on vision and Robot Operating System(ROS).This framework can build semantic maps by improving the mobile robots’ analysis and understanding of complex scenes,while improving the accuracy of path planning,and finally,this framework will be applied to the robot hardware platform for effect validation,the specific research is divided into the following parts.(1)Based on the functional requirements of visual navigation,this paper designs an indoor mobile robot visual navigation system by building a hardware platform with an indoor mobile robot cart and a software platform with a Linux operating system and a robot operating system(ROS).This system first establishes the semantic map in the mobile robot navigation framework,followed by path planning.(2)Based on the laser SLAM algorithm,this paper proposes a semantic map construction scheme.This scheme combines LIDAR and depth camera,uses Cartographer algorithm based on graph optimisation to sense the robot’s position and the surrounding spatial information,uses YOLOv5 algorithm for target detection,and performs point cloud segmentation by RANSAC algorithm to obtain the semantic information of the object,and then determines the position of the object in the map.At the same time,this paper proposes an improved scheme for model envelopment,which reduces the complexity of point cloud marking and improves the efficiency of map construction than the original scheme.The effectiveness of the proposed scheme is also verified through simulation experiments.(3)Based on the path planning algorithm,this paper adopts the combination of global path A-star algorithm and local path TEB algorithm for the navigation of the mobile robot,and proposes a path planning scheme to complete the autonomous navigation process of the indoor mobile robot together with the semantic map constructed by the system.(4)Finally,the above functions are transplanted to a small indoor mobile robot ROS cart for experimental testing in a real environment simulation,verifying that the visual navigation system designed in this paper has good real-time performance and accuracy.