Non-holonomic constraint-assisted GNSS/SINS tight integration navigation method based on a left-invariant extended Kalman filter

Abstract Misalignment in angle errors often leads to slow convergence and poses challenges in achieving optimal precision in vehicle-integrated navigation systems utilizing integrated global navigation satellite systems (GNSS) and strap-down inertial navigation systems (SINS). Additionally, in cases where GNSS satellite signals are lost, the performance of the non-holonomic constraint (NHC) algorithm can degrade significantly. Hence, in this study, the error state model is redefined by embedding the attitude, velocity, and position within the Lie group space. A novel algorithm based on a left-invariant Kalman filter is developed to assist NHC in GNSS/SINS integration. The proposed method is validated through trials conducted in complex urban environments and open suburban settings, using both loosely and tightly coupled integration modes. The results indicate that the proposed method offers rapid convergence and high precision, even in the presence of heading misalignment errors. In challenging environments, heading accuracy converged to below 0.5° within 50 s, while in open environments, it achieved approximately 0.1°, thus providing reliable, efficient positioning and eliminating the need for expensive attitude alignment procedures.