Waveform Selection Of Cognitive Radar In Complex Environment

The cognitive radar can adjust the emission waveform according to the change of environment so that the radar emission waveform can match with environment to improve the target detection and tracking accuracy.Aiming at the tracking task of radar,this paper mainly studies the waveform selection of cognitive tracking radar,from the basic structure of cognitive radar,uses waveform parameter agility,neural network,Q learning and DQN algorithm to realize the intelligent selection of waveform,and compares and analyzes the tracking performance of various waveform selection methods in different scenes.This paper first introduces the research background of the subject and the status of research and development of cognitive radar at home and abroad,and then gives a brief overview of the various components of the cognitive radar system.Then the emission signal model of cognitive radar is established,the influence of waveform on signal-to-noise ratio and measurement is analyzed,and the relationship between the transmission signal and tracking process is constructed.The state space model,filtering algorithm,waveform library,performance Index evaluation method and simulation environment of cognitive tracking radar without clutter environment are established,and the performance based on criterion function and waveform parameter fixing method is compared,and the influence of waveform selection on tracking performance is analyzed.At the same time,the Uniform acceleration Motion model is introduced under the maneuvering target scene,and the performance based on minimum mean square error criterion and the maximum mutual information criterion and waveform parameter fixation is compared.And then the waveform parameter selection algorithm of cognitive tracking radar in clutter environment is established,and the probabilistic data association algorithm is introduced.The modified Riccati equation is introduced to approximate the covariance of the posterior estimation error,and the criterion function of the clutter scene is given,and the tracking performance of the criterion function is analyzed.This paper introduces the basic theory of neural network,establishes the waveform selection method of neural network,generates training sample data by the method of waveform parameter agility,and the neural network is trained by the generated data.The neural network learns the decision process of the criterion function.And the performance of neural network and criterion function and waveform parameter fixation is compared in the non-clutter scene and the maneuvering target scene and clutter scene and interference scene,and the tracking accuracy and computational complexity of three methods are analyzed.By introducing the reinforcement learning method to realize the waveform selection of cognitive tracking radar.This paper introduces the basic concept of reinforcement learning,and introduces the decision-making process of Markov,and deduces the optimal equation of Bellman in detail.According to the optimal equation of Bellman,the Q learning Waveform selection method is established and the related theory and process of Q learning is introduced.The performance of Q learning,criterion function,neural network and waveform parameter fixation is compared.And then the waveform selection method based on DQN algorithm is proposed,and the basic flow of DQN algorithm is introduced,and the performance of DQN algorithm,Q learning,criterion function and neural network is compared.Q Learning algorithm can not describe the entropy state of the target very well,the DQN algorithm uses the approximate function method to overcome the shortcoming and the accuracy is greatly improved compared with Q learning,and the performance of the criterion function in the interference zone is better and the calculation speed is faster.