SLAM Based on Double Layer Cubature Kalman Filter
Feng Yang,Bo Jin,Mengting Yan,Yujuan Luo
DOI: https://doi.org/10.23919/fusion43075.2019.9011369
2019-01-01
Abstract:In the simultaneous localization and mapping (SLAM), the challenge is the large computational cost, low accuracy and instability. In SLAM system, cubature Kalman filter (CKF) has shown good performance. However, in terms of algorithm accuracy and stability, double layer cubature Kalman filter (DLCKF) is better than CKF. It calculates the predicted state at next moment through the inner layer CKF, which is more accurate than the predicted value obtained directly through motion model. The outer layer CKF then updates the predicted state with the measurement to obtain a more accurate estimate. Combined with the advantages of DLCKF, an innovative filter-based SLAM algorithm based on double layer cubature Kalman filter was established in this paper to solve the above problems. Simulation results are presented that the positioning error of the proposed algorithm is significantly reduced and the accuracy of mapping is greatly improved compared with traditional EKF-SLAM, UKF-SLAM and CKF-SLAM.