Research On 2D Laser SLAM Algorithm Based On Particle Filter

As a key technology for robot autonomy,Simultaneous Localization And Mapping(SLAM)technology plays an important role in many fields such as medical care,aerospace,family services,and transportation,which greatly facilitates people's production and life.At present,the particle filter algorithms is generally used to estimate the pose of mobile robots.However,the particle filter algorithm has problems such as particle degradation,insufficient particle diversity,and large fluctuations in the effective particle number,which affects the accuracy of SLAM estimation.This paper mainly studied the 2D laser SLAM algorithm based on particle filter,and improved the Fast SLAM algorithm by using the artificial fish swarms algorithm and the Chopthin resampling algorithm,and had been verified by algorithm simulation and experiments in real environments.The result shows that the effectiveness of the algorithm has been improved greatly.The specific research content includes:(1)The SLAM problem is briefly explained,and the Bayesian filtering theory is briefly introduced.After studying the commonly used sensor models and map models in the SLAM system,it is determined that this article uses the odometer model as the motion model and the grid map as the method of map representation.(2)Aiming at the problems of poor sampling particle quality,insufficient positioning accuracy,and large number of particles required for mapping in the Fast SLAM algorithm.Based on the analysis and research of the basic principles of the particle filter algorithm and the principle of the RB particle filter,it is proposed to use the optimization feature of the artificial fish swarms algorithm to optimize the sampled particle set,and use the crowding factor in the artificial fish swarms algorithm to congest the particles.Adjust the degree to avoid the reduction of particle diversity caused by excessive optimization.Through MATLAB and ROS simulation experiments,it is shown that the improved algorithm has higher positioning accuracy than the traditional SLAM algorithm,and the number of particles required under the same mapping accuracy is significantly reduced.(3)Aiming at the problems of severe particle degradation,reduced particle diversity,and large fluctuations in effective particles in the Fast SLAM algorithm,the Chopthin resampling algorithm was applied to the Fast SLAM algorithm and compared with the adaptive resampling algorithm,through MATLAB simulation experiments and the open source data set experiment was verified.Experimental results shows that using the Chopthin resampling algorithm can effectively reduce the fluctuation of effective particles,alleviate the problem of particle degradation,and improve the positioning accuracy of the SLAM algorithm to a certain extent.(4)In order to verify the mapping performance of the improved algorithm proposed in the actual mapping,a SLAM mapping experiment based on the mobile robot platform was carried out.The two optimization strategies proposed was used to improve the Gmapping algorithm,and the calibrated robot platform was used for experiments,and the mapping results of the original Gmapping algorithm are compared.The results show that the number of particles required to build a map using the Gmapping algorithm optimized by the artificial fish swarms algorithm is reduced,while the Chopthin algorithm can effectively slow down the fluctuation of the effective particle number,and the two improved algorithms are better than the original algorithm in terms of map construction.