Learning-based modeling and control design for a coaxial helicopter with aerodynamic coupling

In this paper, a mathematical model of a coaxial helicopter with a new configuration is successfully constructed using the learning-based method. This nonlinear model can reflect the aerodynamic coupling characteristics between the two rotors of the coaxial helicopter. Compared with the traditional polynomial fitting method, this method significantly improves the integrity and robustness of the mathematical model. After multiple tune of the aircraft, we successfully completed the stable flight experiment using the designed attitude controller. By analyzing the real-time flight log, we verify the high accuracy of the coupled model trained on neural networks. The results of this research have brought important enlightenment to the field of aircraft design and control and proved the great potential of neural network in improving the accuracy of aircraft model and control efficiency.