An Adaptive State-Augmented Kalman Filter for Robust UAV Altitude Control with Online Sensor Bias Correction and Dynamic Weighting in Degraded Environments

Abstract

For unmanned aerial vehicles (UAVs) to operate safely and dependably, accurate state estimation is essential. However, environmental factors that affect measurement quality and sensor biases can impair performance. This paper proposes an Adaptive State-Augmented Kalman Filter (A-SAKF) that integrates two complementary mechanisms: (i) state augmentation for online sensor bias estimation, and (ii) innovation-based adaptive adjustment of measurement covariance. Together, these features enable the filter to maintain robust state estimation performance in the presence of bias errors and uncertain measurement noise conditions. Validation through three simulation scenarios demonstrates the effectiveness of the proposed framework. In Scenario 1, the method correctly estimates and compensates for a 2.0 cm bias in the infrared sensor. In Scenario 2, the velocity estimates eliminate overshoot and reduce settling time by 18% compared to a baseline controller. In Scenario 3, under degraded foggy conditions, the adaptive weighting mechanism recovers LiDAR trust levels within 4.5 s after a 35% drop, thereby preserving altitude tracking accuracy. These results highlight the filter’s capability to address both systematic bias and dynamically varying measurement reliability.        By dynamically down-weighting the distorted LiDAR sensor data, the system demonstrates in simulation a steady and precise altitude estimate, showing improved resilience compared to fixed-covariance filters. The proposed filter demonstrates improved state estimation performance for UAVs under uncertain and biased sensor conditions, achieving lower errors than conventional EKF variants in diverse simulation scenarios. The current evidence is limited to simulation-based validation, and future work will extend testing to hardware-in-the-loop and public UAV datasets to further substantiate real-world applicability.