A Symbolic Regression Method For Dynamic Modeling And Control Of Quadrotor UAVs

The quadrotor UAV has broad prospects in civil and military applications due to its advantages such as simple structure,rapid maneuvering,hovering and vertical taking off and landing(VTOL).Therefore,the modeling and control of quadrotor UAV has become a topic of considerable interest.However,there are some problems in the existing modeling methods of quadrotor UAV,such as over simplification of model,inaccurate modeling and model not analyzed,which lead to low accuracy modeling and difficult application of model-based control methods.Moreover,the characteristics of dynamic system,such as nonlinearity,underactuation and strong coupling,bring considerable challenges to the controller design.Thus,this paper proposes a symbolic regression method for dynamic modeling and control of quadrotor UAVs.The main research contents and relevant contributions of this paper are as follows:(1)For the modeling of quadrotor UAV with little of prior knowledge,parameter measured difficultly and the corresponding dynamic model not analyzed,this paper utilizes symbolic regression to construct the quadrotor dynamic model.By introducing symbolic regression to analyze the potential relationship between control input data and state output data of the model,an analytical model describing the physical characteristics of the system is constructed,which solves the problems of lack of prior knowledge and difficulty in parameter measurement.The subsequent control scheme design revolves around the analytical model.In order to achieve position trajectory tracking capability of the underactuated and coupling system,a double closed-loop cascade control scheme is proposed consisting of PI controller and backstepping controller.The PI controller for position subsystem is designed to obtain the desired roll and pitch angles,and the backstepping controller for attitude subsystem is developed to ensure that the altitude and Euler angles track a desired time-varying trajectory with fast convergence.For the attitude control of quadrotor,the backstepping method is proposed to achieve real-time tracking.Finally,the effectiveness of the position and attitude control scheme is verified by comparing with the corresponding PID control scheme.(2)Considering the influence of external disturbances and model uncertainties,we adapt symbolic regression to re-model the quadrotor dynamic system,and obtain the analytical form of the disturbances and partial model uncertainties.And a backstepping method is proposed to compensate for the above mentioned disturbances and uncertainties.In order to demonstrate the superiority of the above control scheme in robust control,the sliding mode control scheme is introduced as a comparison.Further experimental results show the high accuracy of the proposed attitude control strategy.