Скачать или смотреть Neural Network-Based Controller for Trajectory Tracking of a Rotary-Wing UAV in AirSim

The design and implementation of an advanced trajectory tracking controller in a rotary-wing unmanned autonomous vehicle. The main objective is to enhance the accuracy and stability during trajectory tracking, overcoming
the limitations of the AR.Drone’s controller. Two algorithms were developed: one for generating random trajectories and another focused on efficiently tracking these trajectories. The evaluation is performed using the AirSim simulator in Python. A neural network is also designed and trained to complement and optimize the controller’s performance. Experimental results
demonstrated that the integration of the neural network not only improved the accuracy and stability of trajectory tracking, but also outperformed the capabilities of the original internal controller. This approach highlights the potential of neural networks as a key tool to enhance the performance of autonomous vehicles in critical applications.