Neural Network Model Predictive Control Applied to Parafoil under Turbulence Wind Field

In the field of parafoil flight control, a critical challenge is managing the complex control requirements and mitigating the effects of turbulenceinduced disturbances. This paper explores the intricacies of parafoil control and the influence of the turbulent wind field on system performance. To address the challenges, we propose a Neural Network Model Predictive Control (NNMPC) strategy that utilizes extensive historical operational data, including flight attitude records from multiple flights and control responses under various wind conditions, to develop a neural network model as an alternative to traditional 3-degree-of-freedom (3DOF) models. The proposed approach integrates a NARX (Nonlinear AutoRegressive with Exogenous inputs) neural network into the Model Predictive Control(MPC) framework, leveraging its temporal modeling capabilities to improve prediction accuracy and enhance disturbance rejection. Experimental validation was performed in a simulated turbulence environment using the Dryden turbulence model, with carefully designed control groups in which both traditional MPC and NNMPC were tested under identical initial conditions and target trajectories. The results demonstrate that NNMPC achieves superior trajectory tracking performance, indicating its potential for robust parafoil control in complex atmospheric conditions.OPEN ACCESS Received: 06/03/2025 Accepted: 06/05/2025 Published: 22/09/2025