A Modified Adaptive Square-root Cubature Kalman Filter for GNSS/INS Integration

Abstract

In the GNSS/INS integrated navigation system, the filtering accuracy of Square-root Cubature Kalman Filter (SCKF) will reduce when the measurement noise statistics is not precisely known. To solve this problem, a Modified Adaptive SCKF (MASCKF) method is proposed. At first, the maximum likelihood estimation of the innovation covariance with a moving window is calculated to adaptively adjust the measurement noise statistics. Then, a new adaptive filter framework is designed, which utilizes a plurality of moving window estimators with different widths. Finally, the corresponding weights are set according to the different innovation covariance, which can optimize the innovation covariance and reduce the filtering errors due to the improperly selected width of moving window. To evaluate the performance of this algorithm in the GNSS/INS integrated navigation system, a vehicle test is conducted. The test results show that the proposed MASCKF can improve the estimation accuracy and robustness compared to SCKF and ASCKF. The proposed method can effectively improve the adaptive ability and performance of the GNSS/INS integrated navigation system.