The Research Of UAV Track Tracking And Planning Based On Intelligent Control

The research of UAV trails is one of the hot issues for UAV research.Since the 21 st century,UAV can be applied to military fields,civil fields,commercial fields and disaster prevention and relief fields,and have shown strong application prospects and market potential.UAV trajectory tracking and trajectory planning are the core problems for UAV application and have far-reaching significance.In this paper,theoretical research and instance simulation are combined,problems are oriented,the mathematical model of the actual problem is established,Matlab software is used to analyze the simulation system.This paper mainly studies the UAV trajectory,which is mainly divided into three levels: UAV trajectory tracking,UAV trajectory planning and UAV trajectory fitting.Aiming at solving the UAV track tracking,the mathematical model for UAV longitudinal flight is established,then for the time-varying nonlinear system with input perturbation,an integral sliding mode controller based on ADP method is used to solve the control problem of aircraft track angle.The improved integral sliding mode controller,disturbance estimator,RBF neural network observer are combined to solve the time-varying input disturbance,uncertain parts of approximation model and the system moved sliding mode surface.Optimal control theory is used to make the system stable in near the sliding mode plane,meanwhile.for HJB equation,the ADP method is used to find the optimal value of Hamilton function,and the observer based on RBF neural network with Actors-Critic structure is designed to solve the differential equation in the last step of ADP.Meanwhile the stability of the algorithm is proved by the Lyapunov method.The convergence policy is proposed to guarantee the Actors-Critic structure convergence and reduce the complexity of compute.Finally,the simulation results are presented to illustrate the effectiveness of the proposed method.Aiming at solving the path planning problem with complex constraints,a novel hybrid modified teaching learning based particle swarm optimization initialized by the normalized step cost is proposed.Specially the new method combines the canonical PSO basic policy,the teaching-learning-based optimization algorithm and the normalized step cost function in order to promote the diversity,obtain well-speed convergence and improve search ability.Meanwhile,the teaching learning optimization method are improved in four fields: Redefined mean M,A adaptive teacher factor,The learning within a team,and Individual Subject Locked(ISL)mechanism to enhance its optimal capability.Also the circular crowded sorting technique is developed to improve the diversity,reduce the computation and spread of solution when truncating the external elitism archive.The control effect of the algorithm in uav track tracking control is verified by numerical simulation.Aiming at solving the fitting problem of complex discrete path points,an improve Hyper basis function neural network regularized by soft local dimension reduction in addition to weight decay is used in this paper.Furthermore,a practical training method that is a multiple-step-size gradient algorithm based on the resilient propagation algorithm to construct Hyper BF network is presented to train this network.Hierarchal clustering is used to initialize neurons before the neural network be trained.The experiments shows that the proposed network can converge faster and smoother than the normal network by fitting the different flight trace data of different models.