Research On Key Technologies Of Multi-station Passive Location Based On TDOA

Multi-station passive positioning based on Time Difference of Arrival(TDOA)refers to receiving and processing non-cooperative signals from the same target radiation source through multiple receiving stations,thereby obtaining non-cooperative signals of the target radiation source to reach different receiving.The time difference of the station is based on the time difference information to establish a correlation positioning equation including the position of the radiation source to realize the solution of the position of the target radiation source.In the two-dimensional plane,the arrival time difference measurement information corresponds to a hyperbola,and the intersection of the hyperbola obtained between the plurality of receiving stations is the position of the target radiation source.In this paper,the key technologies in multi-station passive location based on TDOA are studied.The three key problems of non-cooperative signal identification,high-precision estimation of delay and multi-station passive location solving method are studied.The main research completed is as follows:(1)The basic principles and basic processes of multi-station passive positioning based on TDOA are summarized.The platform architecture and signal receiving and processing module of multi-station passive positioning are introduced.The basic signal identification and delay estimation are given.And a mathematical description model of the positioning solution.(2)For the target radiation source signal identification problem,this paper focuses on the signal recognition method based on deep learning,and studies the feature extraction of the data set by LSTM deep neural network to realize the identification of the communication radiation source signal.In the simulation environment,the recognition accuracy of the LSTM network is the highest,the overall recognition rate is87%,the accuracy of the CNN method is second,and the recognition rate of the Inception and Resnet networks is about 80%.In addition,for the LSTM network,when the number of convolution layers is 3,the recognition rate is the highest.(3)For the high-precision time delay estimation problem,this paper introduces the basic principles of two traditional adaptive time delay estimation methods,and studies the constrained adaptive time delay estimation method.By limiting the delay estimation,the gain control is added.It can achieve good delay estimation in a low signal to noise ratio environment.Taking the BPSK signal as an example,the simulation results show that the time delay estimation result obtained by the constrained method is more accurate.When the SNR is 20 d B,the estimated deviation obtained by the constrained method is 0.001 s,while the traditional method is estimated to be 0.01 s.And the constrained method has better performance under low SNR conditions.(4)For the position solving algorithm,this paper studies the method of constrained weighted least squares(CWLS)and two-step weighted least squares(TSWLS)based on analytic method.The simulation results show that TSWLS has better effect.The positioning accuracy is higher.When the SNR is 25 d B,the RMSE value of TSWLS is0.59 m,and the RMSE value of CWLS is 1.92 m.In addition,this paper studies the solution method based on extreme value search,studies the solution method based on genetic algorithm(Gene),and proposes the solution method of joint weighted least squares and firefly optimization algorithm.The simulation results show that the proposed joint algorithm can A higher precision is obtained.When the SNR is 25 d B,the RMSE value is 0.07 m,which is lower than the RMSE value of 2.38 m of the Gene method.