Research And Implementation Of Mobile Robot Navigation System Based On Stereo Vision

With the continuous development of robotics and artificial intelligence technology,robots have gradually entered the lives of people from the original industrial scene.For robots,self-localization and navigation has always been one of its core issues.Early robots mainly relied on GPS,IMU,lidar for localization.With the development of computer vision,stereo cameras are increasingly being used in the field of robot localization and mapping because of their low power consumption,low cost and abundant data.In this paper,the stereo vision sensor is used to supplement the traditional mobile robot system relying on GPS localization and navigation.So that the mobile robot can simultaneously create accurate environment maps during the movement process,which improves the navigation accuracy.Specifically,this paper analyzes the visual odometer,back-end optimization and loop closing algorithms involved in the mainstream algorithm framework ORB-SLAM2,and improve the problem of low feature extraction efficiency.Due to the inability to perform dense 3D reconstruction in the original framework,this paper proposes a method for dense 3D reconstruction using octree structure,which improves the practicality of the framework in mobile robot platforms.In the aspect of mobile robot navigation,this paper proposes an improved smooth A* path planning algorithm,which improves the local obstacle avoidance effect of mobile robots.Then we propose different navigation schemes for indoor rich texture scenes and outdoor less texture scenes,so that the robots can maintain high motion efficiency when they are in different environments.Especially in the outdoor environment,this paper proposes to use the open street map processing method to reduce the probability of obstacle avoidance of mobile robots.Finally,we use the wheeled robot experimental platform to perform modular performance tests on the system and compare the performance of the improved algorithms.Then we performed the actual navigation performance test of the entire mobile robot system in indoor and outdoor environments.Experiments show that the system can complete the autonomous navigation task of mobile robots in indoor and outdoor environments,and has good practicability and robustness.