Abstract
Aiming at the problem that traditional attitude measurement cannot meet the requirements of small and unmanned aerial vehicles (UAVs) in terms of weight, volume, and power consumption, this paper proposes a UAV attitude measurement system based on MEMS accelerometers, gyroscopes, and magnetometers. The system utilizes a hybrid approach combining Euler angles and quaternions for attitude solution. The Three-Axis Attitude Determination (TRIAD) algorithm constructs the measurement model, while data fusion is performed via the Unscented Kalman Filter (UKF) to ensure accurate static and dynamic measurements. Experimental results indicate static errors below 1° and dynamic errors below 1.3°, demonstrating the system's suitability for practical applications.
Introduction
The rapid development of small unmanned aerial vehicles (UAVs) in industrial and civil sectors has created a demand for high-precision navigation systems that are lightweight, compact, and energy-efficient. Traditional high-precision systems, such as liquid-floating or electrostatic gyroscopes, are prohibitively bulky and expensive for these platforms. Micro-Electro-Mechanical Systems (MEMS) offer a viable alternative due to their low cost and small footprint, but they suffer from significant noise and drift. This research addresses these challenges by implementing an advanced data fusion algorithm to enhance the reliability of MEMS-based attitude measurement.
System Methodology
The proposed system integrates data from a three-axis MEMS gyroscope, accelerometer, and magnetometer. To overcome the limitations of individual sensors and mathematical singularities like "gimbal lock," a multi-stage processing architecture is employed.
Data Fusion and Filtering
The core of the system is the Unscented Kalman Filter (UKF). Unlike the standard Extended Kalman Filter (EKF), which linearizes nonlinear systems and may lose accuracy, the UKF uses the Unscented Transform. By propagating a set of sigma points through the actual nonlinear equations, the UKF captures higher-order accuracy in the state distribution, which is crucial for the highly nonlinear dynamics of UAVs.
The Three-Axis Attitude Determination (TRIAD) algorithm is used to construct the measurement model by processing reference vectors from the accelerometer (gravity) and magnetometer (geomagnetic field). For dynamic state estimation, a fourth-order Runge-Kutta algorithm solves quaternion differential equations, which are then converted to Euler angles for final output. This hybrid approach avoids mathematical singularities while maintaining computational efficiency.
Experimental Results and Analysis
The system was evaluated using a three-axis turntable to simulate both static and dynamic flight conditions. The tests measured the deviation of the system's output from the reference orientation over a 35-second period.
Performance Metrics
- Static Performance: In steady-state tests, the roll and pitch errors remained under 0.5°, with heading (yaw) error staying below 1°.
- Dynamic Performance: During active rotation, the system maintained high accuracy, with pitch and roll errors under 1° and heading errors under 1.3°.
These results confirm that the UKF-based fusion of MEMS sensor data effectively mitigates noise and drift, providing a stable attitude reference for small UAVs.
Conclusion
This research successfully demonstrates a high-precision attitude measurement system tailored for the constraints of small UAVs. By leveraging the UKF and a hybrid mathematical model, the system achieves sub-degree accuracy in static conditions and maintains robust performance during dynamic maneuvers. Future work will focus on improving resistance to external magnetic interference and evaluating long-term sensor drift in diverse environmental conditions.
Summary updated