To meet the application requirements of anti-jamming and enhanced robustness for autonomous navigation systems in complex environments， this paper proposes a Robust Cubature Kalman Filter method based on the Interacting Multiple Model （IMM-RCKF） framework. The method integrates measurement uncertainty quantification and nonlinear filter estimation， leveraging the filter innovation before measurement interruption to dynamically adjust the posterior probability matrix weights through prior knowledge. By optimizing the parametric error model of the robust cubature Kalman filter and adopting a hybrid static-dynamic filtering strategy for interactive fusion， a high-order nonlinear multi-model filter is developed. This architecture significantly enhances the environmental adaptability and interference robustness of the filter. The proposed method is implemented in the autonomous navigation system of a hypersonic aerospace vehicle. Experimental results demonstrate that the IMM-based multi-model filtering approach effectively improves the convergence speed and estimation accuracy of the navigation system. Compared with conventional robust Kalman filtering methods， the position accuracy is increased by approximately 14.3% and the velocity accuracy by 11.2%. Furthermore， the system exhibits superior anti-interference capability， successfully addressing the long-standing challenge of traditional Kalman filters in handling model errors under dynamic uncertainties.