Estimation Of Direction Of Arrival Of Large Digital Array In Interference Environment

As a key technical problem in array signal processing,the estimation of direction of arrival has been paid more and more attention in the past decades.Due to the requirement of ideal environment without interference and high computational complexity,the traditional method is increasingly difficult to be applied in practice under the trend of large array and complex electromagnetic environment.Because the main sidelobe interference corresponds to different anti-jamming algorithms,and some anti-jamming algorithms will destroy the main lobe structure,which will have a bad influence on the subsequent angle measurement.The combination of anti-jamming algorithm and angle measurement algorithm can not solve the problem of estimation of the direction of arrival in the presence of interference,and the high computational complexity makes the traditional method unable to deal with large array data.In order to solve this problem,this paper proposes to use machine learning to directly realize the mapping from the echo signal to the target angle,and solve the problem of the direction of arrival estimation in the interference environment end-to-end.The specific research content is as follows:(1)Construct the array echo signal model,and carry out theoretical analysis and simulation experiments based on this,and select the combination of the best traditional anti-jamming and direction of arrival estimation methods.Firstly,the array echo models of line array and planar array are constructed.Then,theoretical derivation and experimental simulation are carried out for the classical anti-sidelobe interference and anti-main lobe interference algorithms respectively.Then,the principle model of the estimation of the direction of arrival is derived,and two classical estimation algorithms are simulated and analyzed.Finally,through comprehensive analysis,the optimal combination of traditional anti-jamming and angle measuring methods is selected to provide a comparison object for the subsequent improved methods.(2)The optimization model of the estimation of the direction of arrival in the interference environment is proposed,and the machine learning optimization solution based on fully connected neural network is proposed.For the traditional two-step problem,this paper constructs it as a complete constrained optimization problem by introducing the parameter model according to the optimization idea.Then according to machine learning theory,data-driven optimization learning is carried out on the fully connected neural network as a parameter model by constructing data sets.The experimental results show that the method based on fully connected neural network not only outperforms the traditional method in angle measurement accuracy,but also improves the computing speed by thousands of times,making it possible to calculate the direction of arrival in real time under interference environment.(3)Further optimize the computing speed in terms of both network model and computing device.Based on the theoretical advantages of convolutional neural networks in two-dimension signal processing,a convolutional neural network model is proposed to estimate the direction of arrival of large arrays under interference conditions.Then the advantages of GPU in neu-ral network optimization are demonstrated by comparing with CPU theory and simulation.Simulation results under different environments and model parameter settings show that the proposed convolutional neural network method can achieve a breakthrough millisecond level accurate estimation of the direction of arrival under the harsh conditions of tens of thousands of elements,single quick beat and interference.