Neural Dynamic Multi-rotor Aircraft Controller: Design And Simulation Verification

Multi rotor UAV is widely used in search,rescue,surveillance and 3D modeling areas.Tracking control is the basis of UAV application in various fields.In order to achieve good tracking performance,this thesis studies the use of neural dynamic method to design unmanned aerial vehicle(UAV)controllers,and improves the performance of UAV controllers based on neural dynamic method from three perspectives: anti-disturbance performance,flexible performance and precise control.Specifically,this thesis first introduces the neural dynamics method.To study the convergence and robustness of the neural dynamics method,first an exponential varying-gain neural dynamic network is analyzed for its robustness.Secondly,because the neural dynamic UAV controller design process is based on the controlled object model,and the model will inevitably be affected by the disturbance in the environment,which will degrade the performance of the UAV controller.In order to overcome the influence of disturbance,this thesis combines neural network adaptive algorithm with neural dynamic method,so as to accurately estimate the unknown disturbance.The UAV can still track time-varying tasks with high precision under the disturbance.Then,in order to solve the problem of insufficient flexibility of the UAV caused by only using the pitch and roll angles to realize the horizontal movement of the UAV in the traditional UAV controller framework,this thesis proposes a quadratic programming based neural dynamic UAV controller.This method uses quadratic programming to make three attitude angles(i.e,pitch angle,roll angle and yaw angle)cooperate to realize the horizontal movement of the UAV,so that the UAV can meet additional performance indexes or perform subtasks.Finally,different form traditional UAV controllers based on neural dynamics methods that use force and torques as control variables,this thesis uses the input of UAV motors as control variables,and applies neural dynamic method to the position layer,speed layer,torque layer and motor layer of a UAV.A more practical multi-layer neural dynamic UAV controller is obtained.Both theoretical analysis and numerical simulation results verify the effectiveness,convergence,stability and accuracy of the proposed neural dynamic UAV controllers in tracking time-varying trajectories.