Research On Ground Taxiing Modeling Of High Aspect Ratio UAV

The take-off and landing stage of unmanned aerial vehicle(UAV)is usually full of kinds of accidents,as well as a key stage in the development process that requires a great deal of manpower and resources.It is difficult to get the accurate ground model that could reflect the complex dynamic characteristics of UAV during taxiing,because of many external factors such as ground force,aerodynamic force and runway conditions.Therefore,it is important to strengthen the study of ground modeling technology and improve the accuracy of model in terms of reducing the cost and quickening the pace of research on UAV.This paper takes a high-aspect-ratio UAV as the research object.Based on the detailed analysis of taxiing phase of UAV,in order to get the mathematical description of UAV’s ground motion,Newton’s laws are applied and the ground dynamics model has been established according to the dynamic characteristics of taxiing.By the analysis and comparison with actual taxiing data,an error compensation structure based on neural network is designed to figure out the nonlinear errors introduced in the modeling of supporting force,friction force,aerodynamic force and thrust.On the basis of calculating the angular acceleration by using the digital differential algorithm,this paper proposes an identification method of force and torque according to the actual running speed and flight data in the process of taxiing.Combined with the numerical solution of model and the actual taxiing data,the preparation of the neural network training sample are completed.A BP neural network compensation link is put forward based on Levenberg-Marquardt learning algorithm according to the nonlinear characteristics of force and torque compensation function model.The fitting accuracy of the error of the ground taxiing model obtained by trained neural network has been well verified by a series of fitting experiments of force and torque error which are calculated by groups of actual taxiing data.The simulation model is completed under MATLAB/Simulink environment.The simulation results show that BP neural network designed in this paper has a good error compensation effect of ground taxiing model and improve the model accuracy effectively.