Based On Improved Pose Estimation And Loop Detection Algorithm Mobile Robot SLAM Research

SLAM(Simultaneous Localization and Mapping)can not only realize autonomous localization in unknown environments of robots,but also obtain real-time environmental information using ranging sensors,which has become a current research hot spot.Pose estimation is the top priority of SLAM technology,and the state of the pose can be estimated by the optimizations.Aiming at the problem that the cartographer algorithm uses low-cost sensors to create images in large scene environments,a new 2D lidar positioning method is proposed to solve the problem of real-time and robust SLAM under low-cost hardware conditions.The algorithm in this paper is developed based on the mobile robot software platform ROS,and the speed model,odometry model and laser sensor model of the mobile robot are modeled at the same time.First of all,the multi-sensor fusion scheme is established in this paper.Perform error analysis on the original odometry data,and use the EKF algorithm to fuse the original odometry and the original IMU data to obtain new odometer data.In order to optimize the initial predicted pose,the pose difference method is used to compensate the current calculation time,so a more accurate predicted pose can be obtained.Secondly,in this paper a new laser matching method is proposed to optimize and update the predicted pose obtained in the previous chapter.The map grid is updated by the laser scanning point,and the five nearest grids near the current laser point cloud data are determined in the same thread.If the linear condition is met,a straight line fitting is performed,and then the current laser point cloud and its nearby point cloud Iterative optimization is performed on the distance of the current laser frame to minimize the sum of the distances from all the point clouds in the current laser frame to the matching straight line nearby.The pose obtained at this time is the optimal observation pose.Finally,the back-end optimization of the pose error that still exists in the front-end is performed to minimize the global error.Here,the nearest neighbor search algorithm is used to find the looping point,and the pose detection during the stationary process of the robot and the pose detection of short-distance motion are considered to ensure that the robot can find the matching point normally after the loop.At the same time,the branch and bound method is adopted in this paper to accelerate the matching of the current calculation.In order to verify the feasibility of the algorithm,the test was performed on a physical mobile robot.The EAI-G4 lidar was used in this test platform to test in three scenarios.This article tested three scenarios to compare the effects with cartographer at the same level.From the analysis of the results,the obtained pose is more accurate and robust,which verifies the rationality of the scheme used in this article.