Quaternion-based Unscented Kalman Filter for 6-DoF Vision-based Inertial Navigation in GPS-denied Regions

TL;DR

This paper introduces a Quaternion-based Unscented Kalman Filter for accurate 6-DoF navigation using vision and inertial sensors in GPS-denied environments, validated on drone data.

Contribution

It presents a novel quaternion-based UKF tailored for 6-DoF navigation that fuses visual and inertial data, improving GPS-denied localization accuracy.

Findings

Successfully navigated a drone in 3D using the proposed filter.

Outperformed standard filtering techniques in experiments.

Validated robustness with real-world stereo image and IMU data.

Abstract

This paper investigates the orientation, position, and linear velocity estimation problem of a rigid-body moving in three-dimensional (3D) space with six degrees-of-freedom (6 DoF). The highly nonlinear navigation kinematics are formulated to ensure global representation of the navigation problem. A computationally efficient Quaternion-based Navigation Unscented Kalman Filter (QNUKF) is proposed on $\mathbb{S}^{3}\times\mathbb{R}^{3}\times\mathbb{R}^{3}$ imitating the true nonlinear navigation kinematics and utilize onboard Visual-Inertial Navigation (VIN) units to achieve successful GPS-denied navigation. The proposed QNUKF is designed in discrete form to operate based on the data fusion of photographs garnered by a vision unit (stereo or monocular camera) and information collected by a low-cost inertial measurement unit (IMU). The photographs are processed to extract feature points in 3D space, while the 6-axis IMU supplies angular velocity and accelerometer measurements expressed with respect to the body-frame. Robustness and effectiveness of the proposed QNUKF have been confirmed through experiments on a real-world dataset collected by a drone navigating in 3D and consisting of stereo images and 6-axis IMU measurements. Also, the proposed approach is validated against standard state-of-the-art filtering techniques. IEEE Keywords: Localization, Navigation, Unmanned Aerial Vehicle, Sensor-fusion, Inertial Measurement Unit, Vision Unit.