Research On Mobile Robot Localization Algorithm Based On Multi-sensor Fusion And Scanning Matching

With the rapid development of artificial intelligence technology,mobile robots,as one of the important development directions of artificial intelligence,are developing and innovating at an unprecedented speed.In the research of the key technology of mobile robots,due to the continuous improvement of the localization accuracy requirements of the robots under various tasks,the robots are required to achieve high-precision localization to meet the requirements of the tasks.This paper studies and analyzes the mobile robot localization algorithm based on multi-sensor fusion and scan matching.Based on the Robot Operate System(ROS)as the platform,the localization algorithm is targeted at the limitations of single-sensor localization and poor localization accuracy in complex environments.Improved,through the design of simulation experiments,the optimization of the localization algorithm has a good effect,and the localization accuracy has been greatly improved.The main research contents of this paper are as follows:(1)Research on multi-sensor fusion technology based on wheel odometer,inertial measurement unit(IMU)and lidar.Firstly,a fusion algorithm of wheeled odometer and IMU based on Extended Kalman Filter(EKF)is proposed,and the motion model of odometer is established to predict the pose of the mobile robot,and the precise angle information of IMU is used to establish observations.The equation modifies the predicted pose.The EKF fused pose is used as the sampling source of the Adaptive Monte Carlo Localization(AMCL)algorithm motion model to obtain the predicted particle set,and the laser likelihood domain ranging model is used as the observation information to update the particle set to obtain AMCL integrates the pose data of odometer,IMU,and laser.(2)Because the laser model is limited by the complex environment,the pose accuracy provided by the AMCL localization algorithm is limited.This article adds scan matching and Discrete Fourier Transform(DFT)processes to optimize the localization algorithm.The weighted average output of AMCL is used as the initial value of scan matching.By constructing the matching function model of the lidar observation point and the prior map,using Gauss Newton's method to optimize the solution,the small jitter at the localization is finally filtered through DFT filtering,which improves Robustness of the system.(3)Design simulation experiments to verify the performance of the localization algorithm under the ROS platform.Simulate the real indoor environment,build a robot model,and judge the accuracy of the algorithm based on the average value of repeated localization errors.The experimental results show that the average position error and the average angle error of the AMCL algorithm are 0.043 m and 0.023 rad,respectively;the average error of the optimized algorithm is reduced to 0.022 m,0.010 rad.Experiments show that the optimized localization algorithm in this paper improves the accuracy of localization.