Due to unpredictable environmental factors, the measurement noise in INS/GNSS integration can be affected by outliers or exhibit statistical uncertainty. A single adaptive or robust filter may not be suitable for all noise scenarios. To address this issue, a new approach called the measurement noise covariance matrix (MNCM) ’R’ modified centered error entropy cubature Kalman filter (RMCEECKF) is proposed in this study. In this method, the MNCM is adjusted based on the innovation sequence, and outliers are detected using the Mahalanobis distance. If outliers are identified, the CEE criterion with strong robustness is applied for the posterior update. Simulation results on INS/GNSS integration demonstrate that the RMCEECKF offers higher estimation accuracy compared to existing methods in scenarios involving outliers, uncertain noise covariance, and outliers under uncertain noise covariance. The inclusion of outlier detection also enhances the computational efficiency of RMCEECKF when compared to the CEECKF.