Comparison of Delay Compensation Strategies in ESKF Low-Latency Navigation for More-Autonomous Aircraft

• Andrei Freiberger ITA http://orcid.org/0009-0009-5167-5204
• Stiven Schwanz Dias Group Embraer S.A. / TU Eindhoven http://orcid.org/0000-0002-6285-4103
• Marcos R. O. A. Máximo ITA http://orcid.org/0000-0003-2944-4476

Resumo

Accurate state estimation is critical for the safe and reliable operation of autonomous aerial vehicles. However, the inherent delay in sensor measurements, such as those from vision-based systems, poses a significant challenge. This study presents a comparative analysis of two delay compensation techniques, Filter Recalculation (FR) and Measurement Extrapolation (ME), for an Error-State Kalman Filter (ESKF) designed for inertial navigation, particularly in GNSS-denied environments. We developed a simulation environment in C++ to model a navigation system fusing IMU and camera data, subjecting it to different measurement delays. The results indicate that the FR method achieves superior accuracy and robustness, nearly eliminating delay-induced errors, but at a significant computational cost. The ME method offers a more computationally efficient alternative but exhibits numerical instability with larger delays. This analysis provides crucial insights for selecting an appropriate delay compensation strategy based on system-specific accuracy and latency requirements.

Palavras-chave: Extrapolation, Accuracy, Measurement uncertainty, Aircraft navigation, Robustness, Delays, Computational efficiency, Kalman filters, State estimation, Robots, State Estimator, Navigation, Error-State Kalman Filter, ESKF, Delay Compensation