Research On Mobile Robot UFastSLAM Based On Local Sampling

The application level and studies degree of mobile robotics represents the level of development for a country's industrial automation,and it means an important strategic significance for national defense,social and scientific technology.To achieve autonomous navigation in a variety of unknown and complex environments is the basic for mobile robot to complete the task,Simultaneous Localization and Mapping(SLAM)is the basic and key factor for mobile robotics to achieve fully autonomous navigation,and it has been an active research topic in mobile robotics.In order to solve the SLAM problem of robot in unknown environment,this paper focuses on the probabilistic SLAM method:(1)According to the analysis of SLAM theory,the mathematical models of mobile robot motion control and environment description are established.Including:motion model of mobile robot,sensor observation model,noise model and so on,which build a unified platform for the research of subsequent related SLAM methods.(2)Two kinds of probability based SLAM methods are deeply studied:1)the SLAM method based on Kalman filter,which only applies to the nonlinear Gaussian model.Firstly,the theory of Kalman filter is introduced,and then the principle and process of Extended Kalman Filter SLAM(EKF-SLAM)method and Unscented Kalman Filter SLAM(UKF-SLAM)method are described in detail.And the validity of the two methods is verified on the MATLAB simulation platform which show that the UKF-SLAM is more accurate than the EKF-SLAM in estimating the location and map.2)the SLAM method based on Particle Filter,which applies to any nonlinear nonGaussian model.Firstly,the theory of Particle Filter is introduced,and then the principle and process of FastSLAM method and Unscented FastSLAM(UFastSLAM)method are described in detail.The simulation results show that the UFastSLAM method has better estimation effect and more reliable,but it takes longer uptime.(3)Considering the problem that UFastSLAM method has long uptime and poor real-time,an UFastSLAM method based on partial sampling is proposed.Firstly,the consistency problem is analyzed,and it is proved that the local sampling and the global sampling estimation precision of the UKF are the same when the scale parameter of the Sigma point is selected properly.Then,on the basis of local sampling,the covariance matrix between the variables in the filtering process is expressed by the linear product of the equivalent Jacobian matrix,which is used to reconstruct the related term in the UKF formula,so that the UKF method has a linear structure similar to EKF,computational complexity is reduced.The idea is applied to calculate the UKF process of the proposed UFastSLAM,and the improved UFastSLAM method is obtained,so that the proposed method and UFastSLAM method have uniformly optimality in the estimation results,the computational efficiency is greatly reduced and the running efficiency is improved,which is beneficial to improve the real-time performance.Finally,the simulation and analysis results show that the UFastSLAM based on local sampling can improve the efficiency of operation.