Adaptive robust UKF based on IMM for single-axis rotation-INS/GNSS integration system

Robust navigation and localization of unmanned surface vehicles (USVs) in dynamic marine environments remains critically challenging. This paper proposes a novel interacting multiple model (IMM)-based adaptive robust unscented Kalman filter (UKF) to enhance single-axis rotation inertial navigation system/global navigation satellite system (rotation-INS/GNSS) integration. To correct kinematic model errors and mitigate measurement outliers, adaptive multiple fading factors and robust multiple enhancing factors are introduced. In contrast to conventional single-factor scalar schemes, this multi-factor method realizes high-dimensional error decoupling—independent channel-wise adjustment preserves error-free data while selectively isolating corrupted signals. To resolve the conflict between asymmetrical multi-factor scaling and the positive-definiteness constraint of the predicted covariance matrix, a dynamic secondary regulator is introduced to maintain its positive definiteness within strict bounds and avoid Cholesky decomposition failures. Finally, the IMM framework exploits sub-filter structural homogeneity to achieve transient-free probability fusion, with concurrent adaptability and robustness. Extensive simulations, along with vehicle- and ship-borne experiments, demonstrate that the proposed method achieves high-precision navigation and positioning under complex operational conditions.