| In recent years,with the rapid development of artificial intelligence,robotics technology has become the trend of modern society,and the level of robotics technology has become an important standard to measure the level of science and technology of a country.Every indoor autonomous mobile robot needs to face the problem of how to get their position and posture.Visual SLAM(Simultaneous Localization and Mapping)is a technology widely used in the field of robot indoor positioning at present,but many complex indoor environments will have a great influence on visual SLAM.So,the solution to solve these problems is of great significance to the indoor localization of robotsThis paper focuses on the robot localization technology in complex indoor environment,and proposes a multi-sensor fusion vision localization algorithm.The main research results of this paper are as follows:Firstly,explain the reasons why visual SLAM is easy to fail in complex indoor environments,i.e.indoor dynamic personnel interference,fluorescent light changes and texture loss.Secondly,In order to make up for the deficiency of traditional visual SLAM algorithm which relies too much on image information.EKF algorithm is used to fuse the position and posture information of robot wheeled odometer and IMU to assist visual odometer as the front-end part of SLAM algorithm.Thirdly,a framework of SLAM algorithm based on graph optimization is proposed,and three optimization methods,i.e.feature constrained graph optimization based on binocular vision,pose graph optimization based on fused odometer and pose graph optimization based on global plane constraint,are designed to correct the accumulated errors of the front-end localization.The theory of the optimization process is also given.Finally,the corresponding experiments are designed.The experiments of localization accuracy and localization robustness prove that the proposed algorithm has high localization accuracy and robustness. |
PDF Full Text Request