Modeling Quadrotor Flight Dynamics using Neural Networks

Long Short-Term Memory (LSTM) neural networks have been effectively used to capture complex, nonlinear, and time-varying dynamics of quadrotors. This thesis presents a comprehensive approach to modeling the flight dynamics of quadrotors using LSTM to exploit temporal dependencies in the system’s behavior. Data collected from simulated flight scenarios is used to train the LSTM architecture, which is then rigorously validated to ensure its efficacy in real-time application scenarios. The results indicate that the LSTM model outperforms traditional methods, providing a promising solution for the control and navigation of quadrotors in dynamic environments.