A Machine Learning-Based Method and System for Evaluating the Wind Resistance Performance of Small and Medium-Sized Unmanned Aerial Vehicles
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU SUPERVISION & INSPECTION INST FOR PROD QUALITY
- Filing Date
- 2025-08-15
- Publication Date
- 2026-05-26
AI Technical Summary
In existing technologies, traditional methods for characterizing the dynamic characteristics of UAVs suffer from nonlinearity issues due to Euler angle representation, leading to gimbal lock problems during large-angle maneuvers. Traditional methods also fail to adequately explore the spatiotemporal correlations of sensor data and lack effective physical constraint mechanisms, resulting in insufficient accuracy and reliability in assessments under complex wind field conditions.
Data is acquired using a nine-axis inertial measurement unit, a global positioning module, and a differential barometer. Noise is eliminated by sliding window frame processing and Kalman filtering. The least squares method is used to fit the trajectory to extract time-domain and frequency-domain features, which are then converted into Lie group spatial features. These features are evaluated using a lightweight neural network model, and the model is updated through a federated learning mechanism. The control strategy is dynamically switched to improve the evaluation accuracy and reliability.
It significantly improves the accuracy of wind resistance performance assessment and control reliability of small and medium-sized UAVs under complex wind field conditions, and solves the kinematic error problem introduced by Euler angle representation in traditional methods.
Smart Images

Figure CN121052118B_ABST