Abstract:Accurate prediction of aerodynamic forces in real-time is crucial for autonomous navigation of unmanned aerial vehicles (UAVs). This paper presents a data-driven aerodynamic force prediction model based on a small number of pressure sensors located on the surface of UAV. The model is built on a linear term that can make a reasonably accurate prediction and a nonlinear correction for accuracy improvement. The linear term is based on a reduced basis reconstruction of the surface pressure distribution, where the basis is extracted from numerical simulation data and the basis coefficients are determined by solving linear pressure reconstruction equations at a set of sensor locations. Sensor placement is optimized using the discrete empirical interpolation method (DEIM). Aerodynamic forces are computed by integrating the reconstructed surface pressure distribution. The nonlinear term is an artificial neural network (NN) that is trained to bridge the gap between the ground truth and the DEIM prediction, especially in the scenario where the DEIM model is constructed from simulation data with limited fidelity. A large network is not necessary for accurate correction as the linear model already captures the main dynamics of the surface pressure field, thus yielding an efficient DEIM+NN aerodynamic force prediction model. The model is tested on numerical and experimental dynamic stall data of a 2D NACA0015 airfoil, and numerical simulation data of dynamic stall of a 3D drone. Numerical results demonstrate that the machine learning enhanced model can make fast and accurate predictions of aerodynamic forces using only a few pressure sensors, even for the NACA0015 case in which the simulations do not agree well with the wind tunnel experiments. Furthermore, the model is robust to noise.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use a small number of pressure sensors to achieve real - time and accurate aerodynamic force prediction in unmanned aerial vehicle (UAV) autonomous navigation. Specifically, the paper proposes a machine - learning - enhanced real - time aerodynamic force prediction model based on sparse pressure - sensor inputs. This model aims to overcome the problem of degraded navigation performance in existing methods when GNSS signals are missing, while reducing the need for additional sensors, thereby reducing system cost and complexity. ### Main contributions of the paper: 1. **Data - driven aerodynamic force prediction model**: - The model combines linear terms and nonlinear correction terms. The linear term is achieved by reconstructing the surface pressure distribution with a reduced basis, while the nonlinear term is corrected using an artificial neural network (NN) to improve prediction accuracy. 2. **Sensor position optimization**: - Use the discrete empirical interpolation method (DEIM) to optimize the position of sensors, ensuring accurate prediction results even when the number of sensors is limited. 3. **Model robustness**: - The model not only performs well under noiseless conditions but also maintains high prediction accuracy when there is noise in the sensor inputs. ### Technical details: - **Linear term**: - Based on reconstructing the surface pressure distribution with a reduced basis, where the basis functions are extracted from numerical simulation data, and the basis coefficients are determined by solving the linear pressure reconstruction equation. - The surface pressure distribution is reconstructed by integration, thereby calculating the aerodynamic force. - **Nonlinear term**: - Use an artificial neural network (NN) for nonlinear correction. In particular, when the DEIM model is constructed based on low - fidelity numerical simulation data, the NN can bridge the gap with the true value. - The input of the NN is the pressure - sensor values at selected positions, and the output is the aerodynamic force prediction error. - **Sensor position selection**: - Optimize the sensor positions through the DEIM algorithm to ensure the best surface pressure reconstruction effect with a limited number of sensors. ### Application verification: - **2D NACA0015 wing dynamic stall**: - Use experimental data and URANS numerical simulation data to verify the effectiveness of the model. The results show that the DEIM + NN model can quickly and accurately predict the aerodynamic force and can also perform well even in the dynamic stall region. - **3D UAV dynamic stall**: - Further verify the applicability and robustness of the model in three - dimensional cases. ### Conclusion: The DEIM + NN model proposed in the paper can achieve real - time and accurate aerodynamic force prediction with a small number of pressure sensors, and is suitable for aerodynamic force estimation in UAV autonomous navigation. The model is highly efficient and robust and can cope with the influence of sensor input noise.

Machine learning enhanced real-time aerodynamic forces prediction based on sparse pressure sensor inputs

Machine-Learning-Enhanced Real-Time Aerodynamic Forces Prediction Based on Sparse Pressure Sensor Inputs

Incorporating Physical Models for Dynamic Stall Prediction Based on Machine Learning

Machine learning-assisted sparse observation assimilation for real-time aerodynamic field perception

Physics-Infused Machine Learning Based Prediction of VTOL Aerodynamics with Sparse Datasets

Rapid prediction of unsteady aerodynamic characteristics of flapping wing based on GRU

Deep Learning Predictions of Unsteady Aerodynamic Loads on an Airfoil Model Pitched Over the Entire Operating Range

Deep Learning for Multi-Fidelity Aerodynamic Distribution Modeling from Experimental and Simulation Data

Fast Pressure Distribution Prediction of Airfoils Using Deep Learning

Deep learning reduces sensor requirements for gust rejection on a small uncrewed aerial vehicle morphing wing

Unsteady aerodynamic prediction using limited samples based on transfer learning

A Novel Model with an Improved Loss Function to Predict the Velocity Field from the Pressure on the Surface of the Hydrofoil

Fast Prediction of Airfoil Aerodynamic Characteristics Based on a Combined Autoencoder

A new nonlinear lifting line method for aerodynamic analysis and deep learning modeling of small unmanned aerial vehicles

Transferable Machine Learning Model for the Aerodynamic Prediction of Swept Wings

Sparse sensor reconstruction of vortex-impinged airfoil wake with machine learning

Artificial Hair Sensor Placement Optimization on Airfoils for Angle of Attack Prediction

Comparative Study of Machine Learning Modeling for Unsteady Aerodynamics

NeuroBEM: Hybrid Aerodynamic Quadrotor Model

Deep learning based real-time energy extraction system modeling for flapping foil

Modelling the unsteady lift of a pitching NACA 0018 aerofoil using state-space neural networks