🤖 AI Summary
To address low attitude estimation accuracy, high computational cost, and dynamically varying sensor reliability in visual-inertial odometry (VIO) for unmanned aerial vehicles operating in complex environments, this paper proposes a loosely coupled hybrid filtering framework. The method integrates an error-state Kalman filter (ESKF), a scaled unscented Kalman filter (UKF), IMU preintegration, and robust visual feature tracking. Its core innovations include: (1) a quaternion-focused hybrid error-state EKF/UKF architecture; and (2) a dynamic confidence mechanism leveraging multi-dimensional features—including image entropy and motion blur—to adaptively tune observation noise covariance. Evaluated on the EuRoC dataset, the approach achieves 49% and 57% improvements in position and attitude accuracy over standard ESKF, respectively, attaining accuracy comparable to a full UKF implementation while reducing computational overhead by approximately 48%. This yields significant gains in robustness, accuracy, and real-time performance under challenging conditions.
📝 Abstract
This study presents an innovative hybrid Visual-Inertial Odometry (VIO) method for Unmanned Aerial Vehicles (UAVs) that is resilient to environmental challenges and capable of dynamically assessing sensor reliability. Built upon a loosely coupled sensor fusion architecture, the system utilizes a novel hybrid Quaternion-focused Error-State EKF/UKF (Qf-ES-EKF/UKF) architecture to process inertial measurement unit (IMU) data. This architecture first propagates the entire state using an Error-State Extended Kalman Filter (ESKF) and then applies a targeted Scaled Unscented Kalman Filter (SUKF) step to refine only the orientation. This sequential process blends the accuracy of SUKF in quaternion estimation with the overall computational efficiency of ESKF. The reliability of visual measurements is assessed via a dynamic sensor confidence score based on metrics, such as image entropy, intensity variation, motion blur, and inference quality, adapting the measurement noise covariance to ensure stable pose estimation even under challenging conditions. Comprehensive experimental analyses on the EuRoC MAV dataset demonstrate key advantages: an average improvement of 49% in position accuracy in challenging scenarios, an average of 57% in rotation accuracy over ESKF-based methods, and SUKF-comparable accuracy achieved with approximately 48% lower computational cost than a full SUKF implementation. These findings demonstrate that the presented approach strikes an effective balance between computational efficiency and estimation accuracy, and significantly enhances UAV pose estimation performance in complex environments with varying sensor reliability.