The Research On Algorithms Of Mobile Robot Simultaneous Localization And Mapping

With the improvement of social information level, the intelligent mobile robot has been playing an important role in social development. Simultaneous localization and mapping(SLAM) is the core content for a mobile robot to realize autonomous navigation, and it also serves as a key link of a primary robot evolving to real autonomous robot. The main starting point of this paper is to improve the accuracy and robustness of SLAM algorithms. The content can be summarized as follows:First, various experimental models and standard data sets for the mobile robot simultaneous localization and mapping are introduced.Second, to deal with the problems that the Kalman filter based SLAM algorithms tend to show low robustness to disturbance, a cubature extended H?filter based SLAM algorithm is proposed. By using statistical linear error propagation method, the cubature transform technique can be embedded into the EH?F framework, and the resulting estimator named CEH?F has the advantages of both CKF and EH?F. CEH?F is used to calculate SLAM posterior probability density, which improves SLAM accuracy and robustness.Third, for feature map based SLAM problems, a Jacobian-free CH?FastSLAM algorithm is proposed. The characteristic is that the cubature extended H?particle filter is used to estimate robot path, which avoids the accumulation of linearization errors and at the same time improves robustness; landmarks are estimated by CKF, which is easy to implement and has high estimation accuracy.In addition, based on Rao-Blackwellized particle filter framework, a high-accuracy spherical simplex-radial cubature FastSLAM algorithm(SSRCFastSLAM) is proposed.In this algorithm, 3rd-degree spherical simplex-radial rule is utilized to calculate the nonlinear Gaussian weighted integral; spherical simplex-radial cubature kalman filter is used to design the proposal distribution and estimate feature positions, thus not only the Jacobian operation and linearization errors are avoided, but also the accuracy of robot localization and map building is improved.All the proposed algorithms in this paper are tested by different simulation environment and by standard dataset in SLAM region, and the results verify the effectiveness and superiority of these algorithms.