Research On Interference Classification And Mitigation Methods For Automotive Millimeter Wave Radar

With the development of automatic driving technology,intelligent sensors have become a key part of automatic driving,such as industrial cameras,lidars,millimeter wave radars,etc.Compared with other sensors,millimeter wave radar has all-time and all-weather working characteristics,and has stronger environmental adaptability.Therefore,millimeter wave radar has been widely promoted in the automotive field.However,if multiple radars share one frequency band,there will be interference between radars,which will lead to reduced radar detection performance,missed target detection and radar false alarm.As more and more vehicles are equipped with radars,and each vehicle is equipped with multiple radars,the problem of inter radar interference becomes more and more serious.Therefore,inter radar interference mitigation has become an urgent problem to be solved.At the same time,each radar on the car transmits different waveforms,so different types of interference will be generated between radars.In order to better improve the effect of interference mitigation.First of all,it is necessary to classify the types of interference,and then,for different types of interference,different interference suppression measures are taken to improve the radar detection performance.The specific research contents of this paper are as follows:(1)The echo model of LFMCW(linear frequency modulated continuous waveform)without interference is analyzed.The principle of mutual interference between radars and the power of interference signal is much higher than that of target signal are explained through classical scenes.The mathematical models of five different interference signal types(long single frequency signal,slow frequency modulation signal,fast frequency modulation signal,short single frequency signal,step frequency signal)are derived.The integrated jamming suppression module of millimeter wave radar AWR2944 of TI is introduced,and the effect of phase coding method on interference mitigation is analyzed.(2)In order to solve the problem of radar interference signal classification,Res Net residual neural network is proposed to classify radar interference signal.According to the mathematical models of different interference signals,the data were simulated.Fourier transform was performed in the fast and slow dimensions respectively to obtain the range-doppler graph data sets.Res Net residual neural network was used to classify different types of interference signals.The experimental results show that compared with SVM and CNN,the classification accuracy of Res Net residual network is not only greatly improved,but also can converge well.(3)Aiming at the problem of interference between LFMCW radars,an interference mitigation method using EMD(empirical mode decomposition)and(KF)Kalman filtering in time-frequency domain is adopted.First,the time-domain interference signal is converted to the frequency-domain through short-time Fourier transform,and then the time-domain signal is subjected to empirical mode decomposition.The interference part of the interference dominated eigenmode function is zeroed,and the eigenmode after the interference is zeroed is added to the eigenmode dominated by the target signal.Finally,the signal is reconstructed through Kalman filtering.The experimental results show that this method can not only reduce the noise in the frequency domain,but also improve the signal-to-noise ratio of the target and increase the probability of target detection.