Research On Signal Processing Algorithms For Millimeter-wave Automotive Anti-collision Radar

With the popularity of cars,road traffic safety has become the focus of people’s attention in recent years.The development of science and technology drives the innovation of automobile industry.Millimeter-wave automotive anti-collision radar,which provides driving assistance,has become the mainstream solution of automotive sensors because of its small size,low power consumption,high accuracy,strong anti-interference ability and other advantages.In this paper,millimeter-wave automotive anti-collision radar will be taken as the research object to study the key technologies in signal processing of the millimeter wave radar.The specific and effective improvement algorithms will be put forward to solve the existing problems.Finally,a complete and feasible signal processing flow of millimeter-wave automotive anti-collision radar will be formed.Specific work of this paper is as follows:(1)The key algorithms of millimeter-wave automotive anti-collision radar in target detection and parameter estimation are studied.For the selection of window function before fast fourier transform(FFT),the frequency response of several window functions is analyzed and compared,and the mechanism of selecting window function parameters according to the characteristics of automotive radar target is proposed.For the problem that constant false alarm rate(CFAR)detection has a serious influence on target masking under the condition of low signal-to-noise ratio(SNR),an improved adaptive CFAR detection algorithm based on interference elimination is proposed,which effectively improves the detection probability and reduces the leaking-detecting probability.For the high computational complexity in tradition direction of arrival(DOA)estimation,a fast DOA estimation algorithm based on spatial FFT is proposed,which associates the peak sequence number of spatial spectrum signal with azimuth angle without matrix operation.(2)The DOA estimation algorithms for near-field targets of millimeter-wave automotive anti-collision radar are studied.For the fact that the incident wave of the near-field source cannot be assumed as parallel wave in the far-field condition,the near-field model is established and the steering vector of the near-field source signal is derived.Then,three algorithms are proposed to solve the problem of DOA estimation in near field.One of them is the peak sequence look-up table algorithm,which needs a large amount of data storage.The other two algorithms are date-driven DOA estimation algorithms based on support vector machine(SVM).The difference between the two SVM-based algorithms is that the process of generating decision functions.Simulation shows that the complexity of DOA estimation algorithm based on decision-tree-hierarchy SVM in training and classification stage is lower than that based on One-versus-rest(OVR)-SVM.In addition,decision-tree-hierarchy SVM algorithm achieves more robust detection results under the condition of insufficient training sample data.(3)The design and improvement of millimeter-wave automotive anti-collision radar based on TIAWR1642 chip in hardware system level are realized.Firstly,the hardware structure and the parameters and characteristics of TI-AWR1642 are introduced.Then,for the velocity ambiguity caused by the parameter limitation of millimeter-wave automotive radar,two algorithms of velocity ambiguity resolution are proposed,which have their own advantages.The peak value comparison algorithm based on discrete fourier transform(DFT)factor does not need additional target matching operation,but it will cause a large performance degradation under the condition of low SNR.However,the algorithm based on sliding window has a better adaptability to SNR,but it needs extra multi-target matching operation.Combining the advantages of the two algorithms,the DFT factor algorithm is adopted in near-field mode and the sliding window algorithm in middle-field mode with the multi-mode and non-interference characteristics of TI-AWR1642 chip.This scheme improves the performance of system in precision and complexity without affecting the data refresh rate of system.Finally,according to the proposed signal processing flow of millimeter-wave automotive anti-collision radar,a frame structure with clear hierarchy and reasonable structure is proposed.The frame structure fully considers the change of waveforms and antennas,which is suitable for the millimeter-wave automotive anti-collision radar system with wide detection range and high detection accuracy.