Attitude and Altitude Control of Quadcopter Maneuvers using Neural Network–Based Direct Inverse Control

Abstract

In this study, a neural network–based direct inverse control (DIC) approach was developed and simulated to control various maneuvers of an unmanned aerial vehicle (UAV) quadcopter. The aim was to propose an inner loop control algorithm for UAV quadcopter maneuvers using a neural network–based DIC system. The appropriate connection weights of neurons in the controller were determined through a backpropagation learning algorithm using real quadcopter maneuver flight data. The neural network–based DIC was trained and then tested using a trajectory dataset different from the training dataset. The experimental results showed that the neural network–based DIC could follow the maneuvers of the testing trajectory dataset with excellent performance, as indicated by an overall mean squared error (MSE) of 1.461 and attitude MSEs of 3.104 for roll, 0.889 for pitch, 1.834 for yaw and 0.018 for altitude. These results indicate that the proposed artificial neural network–based DIC can be used to control the attitude and altitude of the quadcopter during maneuvers.

Keywords

• backpropagation
• direct inverse control
• maneuver
• neural network
• quadcopter