Dual-robust unscented Kalman filter for UAV state estimation under simultaneous model and measurement uncertainties

Accurate position estimation is essential for unmanned aerial vehicles (UAVs), yet conventional filters can degrade when model mismatch and non-Gaussian measurement disturbances occur simultaneously. This study proposes a dual-robust unscented Kalman filter, termed H∞-GMMC-UKF, combining an H∞-motivated prior-covariance modification with a generalized maximum mixture correntropy measurement update. The former improves tolerance to prediction errors caused by model uncertainty, whereas the latter reduces the influence of asymmetric, heavy-tailed, and impulsive measurement errors. The method was evaluated through four progressive simulation scenarios and two EuRoC real-trajectory stress tests using 100 paired Monte Carlo trials with ablation, sensitivity, and statistical diagnostics. Results show that the proposed filter remains competitive under benign conditions and provides its clearest improvement over the standard UKF when model and measurement uncertainties coexist. In the controlled dual-uncertainty simulation, mean position RMSE was reduced by 18.0%, while a 57.3% reduction was obtained in the contaminated EuRoC stress test. Ablation and sensitivity analyses further show that the two robust stages contribute through complementary mechanisms: GMMC directly attenuates contaminated measurements, while the H∞ covariance modification provides an additional prediction-side robustness margin whose effect is most evident in the tested adverse measurement-update settings. These findings demonstrate the potential of the proposed architecture for reliable UAV state estimation under challenging operating conditions.