Research On Positioning And Mapping Technology Of An Indoor Security Robot

Mobile robot is a hot research topic in the field of artificial intelligence.How to realize simultaneous localization and mapping(SLAM)in unknown environment is the core content of current research.It is the prerequisite for robot navigation and path planning.This paper mainly studies the localization and mapping of indoor robots in the structured environment,using a RGB-D camera and a 2D lidar as sensors,focusing on visual localization and combined localization.The first problem to be solved in this paper is the synchronization of multi-sensor data in time and space.Firstly,we need to know the start time and the end time of the current laser data in the odometry queue,and obtain the odometry pose at corresponding time by interpolation.Then,the corrected lidar data is obtained by using multiple interpolation methods while the robot is assumed to perform uniform motion in a short time.In RGB-D localization algorithm,pose estimation based on line feature matching has some problems such as easy loss of features and mismatching.In view of this difficulty,we consider the discontinuity of line feature caused by intersection and occlusion of line segments from the point of view of how to extract line features accurately,and merge line segments according to the difference of direction and distance of segments.At the same time,considering that the dynamic line features brought by pedestrians in the images will cause mismatching,the static weight method is used to solve this problem.Firstly,the weight of the feature line segment is calculated,and then the static line segment is judged according to the weight size.If not,the static line segment is eliminated.So that it can reduce the impact of dynamic objects on pose estimation.In order to solve the problem of combining RGB-D with laser odometer,a multi-filter fusion algorithm based on federated filtering is used to fuse the local optimal estimation of visual location with the local optimal estimation of laser location,and the information distribution coefficient is determined according to their own error covariance.The failure of visual localization due to the lack of features is also considered.Local estimates are filtered by the difference of two positions to ensure the accuracy of the global optimal estimates.The comparative experiments show that the combined localization results have great advantages.