UAV Maneuverability Model Based On XGBoost

The traditional four-rotor UAV control model is based on the aerodynamic model of the four-rotor drone model,and the four-rotor UAV dynamics model has the characteristics of high coupling and nonlinear relationship,so it is difficult to obtain the coefficient of the UAV dynamics model accurately.Four-rotor UAV are widely used in various fields of civil use due to their high stability and strong maneuverability,so the single and cluster of the aircraft needs to complete the task is becoming more and more complex,and needing accurate control of drone flight to complete missions.The traditional aerodynamic model is difficult to meet the needs.The study of higher precision control model is of great significance to the flight control of the four-rotor aircraft.In view of the problem that the dynamic model of the drone can’t meet the accurate control of the flight of the drone,the paper puts forward the model of the maneuverability of the drone based on machine learning.Through the data acquisition and the construction of an automatic control platform,a data set of the drone’s maneuverability was established.Using XGBoost and RNN-LSTM regression models,the maneuverability model is trained based on mobile data sets.The actual flight test was carried out through automatic control technology,which verified that the maneuverability model can be used to control the high-precision flight of the drone.The completion is as follows:1)Optimized the traditional quadrotor drone data acquisition platform,added one-key take-off and landing functions of the drone,and the platform has both the automatic flight control and data acquisition functions of the drone.The platform is divided into three parts: data display interface,data communication module,and flight strategy implementation module.Use this platform to build a UAV maneuverability data set.2)A model of UAV maneuverability based on XGBoost regression model is proposed.Compared with other machine learning regression models,this model is good at dealing with multiple input and multiple output nonlinear regression problems,and has a higher prediction accuracy rate.The XGBoost and RNN-LSTM regression models were separately trained based on the maneuverability data set,and the same test set was used to verify that the XGBoost prediction accuracy was higher.3)The actual flight test of the maneuverability model of the UAV was carried out.By analyzing the actual flight errors,it was proved that the XGBoost regression model as a maneuverability model is suitable for the UAV higher precision flight control.