Research On Rapid Perception Of Indoor Corridor Environment Based On Multi-sensor

With the rapid development of the intelligent industry,SLAM as an emerging technology has received extensive attention.As the market field involved in intelligent robots becoming wider and wider and the application scenarios becoming more complex,the requirements for the speed and accuracy of map construction and path planning are higher.This article focuses on the situation of positioning offset and low feature matching when mobile robots use lidar or depth camera alone to build maps in a corridor environment with high scene similarity,which causes problems such as map drift,low accuracy,and global path planning algorithms.The planned path has many problems such as the big number of inflection points,the path is not smooth,and the local optimum is easy to fall into.Based on the Robot Operating System(ROS)designing and build a multi-sensor environment perception platform,the multi-sensor fusion SLAM under the corridor environment is studied Algorithms and path planning algorithms.The main contents of this paper are as follows:First,in order to solve the problems of light interference and insufficient field of view when the laser sensor or vision sensor is used alone in the corridor environment,the current stage of the corridor environment multi-sensor fusion algorithm is analyzed.The EKF algorithm based on weighted observation fusion realizes the information fusion of the depth camera and the lidar,constructs the state equation of the filtered depth camera point cloud information and the lidar laser information,and calls the weighted least square method to weight the state equations to solve the weighting matrix and state equation.Multi-sensor information fusion can reduce the positioning error of the robot and increase the integrity of the environment and obstacle information.Secondly,the principle of Cartographer algorithm is analyzed and researched.When the Cartographer algorithm constructs a map in a corridor environment with similar environmental geometric features,long distances and equal widths,there will be problems such as loop detection errors leading to map offset and divergence.The fusion information of lidar and depth camera as the front-end input of cartographer algorithm and optimization of its back-end.Adding a verification mechanism to the closed-loop detection of the Cartographer algorithm to calculate the collection of node poses when the total error of each fusion node poses and IMU node pose closed-loop constraint is the smallest,then eliminating error feature points,this can reduce cumulative errors and improve mapping accuracy and coverage.Finally,in order to avoid collisions with obstacles such as glass partition walls that cannot be recognized by single-line lidar and depth cameras,as well as obstacles within the blind zone of the two sensors,an ultrasonic sensor is added to detect obstacles in the environment before planning the path.When an obstacle is not recognized by the camera,the obstacle avoidance strategy is invoked to avoid the obstacle.At the same time,to solve the problems of unsmooth path,low planning efficiency and easy to fall into local minimum when planning the path in the corridor environment,the current global path planning algorithm suitable for the corridor environment was analyzed and researched,and selecting the A*algorithm of static network with few inflection points and short time-consuming.When obstacles or dynamic obstacles that do not exist in the map appear in the environment,the DWA dynamic window algorithm is applied to realize autonomous obstacle avoidance.