Improved robust and adaptive filter based on non-holonomic constraints for RTK/INS integrated navigation

Abstract

The signal quality of a global navigation satellite system can easily be affected in a complex urban environment. Signals from inertial navigation systems (INSs) may also be affected by hardware and structures. To solve the problems of excessive robustness and mishandling of INS errors in the conventional robust and adaptive Kalman filter (CRAKF), an improved non-holonomic robust and adaptive Kalman filter (NRAKF) is proposed in this paper. The NRAKF algorithm first judges whether the INS contains gross errors according to non-holonomic constraint, then selects the order of calculation of the covariance of robust measurement noise and adaptation to state noise covariance, and finally uses an extended Kalman filter (EKF) to obtain the position, velocity, and attitude of the real time kinematic (RTK)/INS integrated navigation system. To demonstrate the performance of the NRAKF, two land vehicle tests were conducted; the results indicate that the tightly integrated model is more robust than the loosely integrated model. Compared with the EKF, both the CRAKF and the NRAKF have stronger robustness. Compared to the CRAKF, the NRAKF improves the position and velocity accuracy in the east and north directions by more than 36.85%, 33.13%, 36.46%, and 22.65% under the loosely coupled RTK/INS integration, and by 61.5%, 57.42%, 73.94%, and 33.53% under the tightly coupled RTK/INS integration. Therefore, the NRAKF performs better than the CRAKF for both loosely and tightly integrated models, and can effectively avoid the phenomena of excessive robustness and mishandling of INS errors.