Evaluation of robust Kalman-based estimators for quadcopter systems under measurement disturbances

Accurate state estimation in quadcopters is a key component, especially when contaminated with noise. Researchers have widely utilized the Extended Kalman Filter (EKF), but they have not considered the causes of estimation deviation, namely gross errors such as outliers, bias, and drift. Robust methods such as the Iterated Extended Kalman Filter (IEKF), Mahalanobis-EKF, maximum correntropy criterion EKF (MCC-EKF), Clipping, and Median-EKF have been developed, but most have limitations in computational efficiency and the types of disturbances they handle. This study discusses MC-EKF because it is effective in correcting measurement residuals. Seven estimator methods will be compared in a simultaneous disturbance scenario. Simulation results show that although MCC-EKF yields the smallest attitude errors, the proposed MC-EKF provides the best balance between estimation robustness and real-time efficiency. By achieving fast computation while maintaining competitive accuracy under mixed disturbances, MC-EKF offers the most favorable trade-off for embedded quadcopter applications. To obtain accurate state estimation under non-ideal measurement, the MC-EKF is very suitable for real-time application on quadcopters. This work contributes a clear operational characterization of MC-EKF: (1) a computationally lightweight residual correction that explicitly separates and compensates outliers, biases, and drifts before the EKF update; (2) a set of practical tuning rules that enable immediate deployment in embedded systems; and (3) an empirical complexity analysis showing MC-EKF’s O(n) per-step cost compared with iterative Iterated Extended Kalman Filter (IEKF) and kernel-based maximum correntropy criterion (MCC) solutions. We further specify the operating envelope under which MC-EKF is preferable, real-time platforms with moderate measurement corruption but limited CPU resources.