Trajectory Tracking Of A Quadrotor Based On Gaussian Process Model Predictive Control

Quadrotor aircraft has the characteristics of simple mechanical structure and strong carrying capacity,which can realize vertical take-off and landing.Therefore,quadrotor aircraft have been widely used in civil,commercial and military fields.However,it is a nonlinear,under-driven and strongly coupled system.At the same time,quadrotor is susceptible to many interference factors in actual flight,and it is difficult to obtain accurate quadrotor model,which makes flight control of quadrotor more difficult.This paper focuses on the problems of accurate establishment of dynamic model and trajectory tracking control of quadrotor,the main research contents are as follows:(1)The modeling method of the quadrotor under wind field disturbance is studied.The simplified position and attitude dynamics model of the quadrotor is established by Newton Euler formula.Then the linear part of the two parts is defined as the nominal dynamic model,and the Gaussian Process(GP)model is used to learn the nonlinear part and the unmodeled dynamic system,so that the dynamic model is established based on the GP learning.This data-driven modeling method not only makes the modeling more accurate,but also describes the uncertainty of the model in the form of covariance.Simulation results show the effectiveness of the proposed modeling method.(2)The trajectory tracking control method of the quadrotor under wind field disturbance is proposed.A hierarchical control scheme is adopt,the inner loop is the attitude subsystem,and the outer loop is the position subsystem.A model predictive control(MPC)based on the GP is designed according to the nominal model of the position and attitude subsystem and the learning GP model respectively.Simulation results show the effectiveness of the proposed method.At the same time,the proposed method is compared with the GPMPC without considering the nominal model.The simulation results show that the proposed GPMPC based on the nominal model has improved the trajectory tracking control accuracy.