Research On Parameter Estimation Algorithm For Automotive Millimeter-wave Radar And Its System Implementation

With the new round of technological and industrial revolution,advanced assisted driving system(ADAS),as an extremely important active safety technology,has greatly improved the safety of people’s travel,guaranteed the safety of life and property,and has become a research hotspot of the automotive industry in recent years.Millimeter-wave(mm-wave)radar is regarded as one of the core sensors of perception layer in ADAS due to its small size,low cost and resistance to harsh environment.The signal processing scheme of automotive mm-wave radar is constrained to the limited hardware scale and cost considerations,and it faces two major problems.On the one hand,the automotive mm-wave radar faces the situation that the echo signal-to-noise ratio(SNR)level deteriorates seriously in the medium and long distance range,resulting in the failure of the parameter estimation algorithm.On the other hand,the realization of high-precision detection and parameter estimation in the short distance range usually brings a huge amount of computation and memory consumption,which puts forward a high demand on the system’s computational power,and it is difficult to meet the real-time requirements of the scene.In order to achieve a more robust and accurate target detection and parameter estimation performance,some key technologies are investigated in this paper as follows:Firstly,the range measurement scheme in the automotive mm-wave radar system is studied.Aiming at the problem of inaccurate estimation of peak frequency in the existing range measurement scheme,a method of multiple modulation zoom spectrum analysis is proposed to partially refine the spectrum near the peak frequency of the beat signal,and the Jacobsen algorithm is utilized to reduce the straddle loss caused by fence effect and improve the estimation accuracy of peak frequency.Aiming at the problems of ignoring range-velocity coupling and fast-slow time coupling in the existing range measurement scheme based on peak frequency estimation,a simple frequency-domain interpolation correction scheme is proposed to remove the fast-slow time coupling,and joint range-doppler processing scheme is used to get ride of the range-velocity coupling and obtain a more accurate range estimation.Then,the direction of arrival(DOA)estimation scheme design in the automotive mm-wave radar systems is studied.For the defects of the existing multiple signal classification(MUSIC)DOA estimation scheme,which has a large memory consumption and high computational complexity in the case of large antenna arrays,the framework of low-complexity beamsapce MUSIC is proposed and the beamformer is optimized specifically.To solve the most challenging problem of the low SNR level and single snapshot,a novel beamspace MUSIC estimator is proposed via exploiting the relationship between the subspace model and a signal-plus-interference model of sample covariance.Additionally,to compensate the sample covariance estimation error resulted from single snapshot,especially for two closely located targets,we propose a joint processing of sub-matrix averaging and Toeplitz structure recovery.Simulations are presented to validate the effectiveness of our proposed scheme and its performance superiority over existing MUSIC DOA schemes.Finally,the implementation of 77 GHz multi-chip cascade automotive mm-wave radar systems based on TI-AWR1243P chips is studied.For the extra antenna array resources offered by multi-chip cascade model,an antenna array arrangement method using virtual array technology is designed,which well balances the power burden and system performance.In view of the complex structure,unclear hierarchy and no specific corresponding performance indicators in the current automotive mm-wave radar systems’ frame design,a multi-level,dual-mode and low-complexity frame structure is planned,which is concise and easy to implement in the actual system.Aiming at the problem of antenna phase error in the actual situation,an antenna self-correction scheme is proposed.Furthermore,to overcome the problem that the original velocity ambiguity resolution scheme fails due to the decrease of the SNR level of the far-range echo signal,an improved velocity ambiguity resolution scheme with simultaneous transmission of multiple antennas is proposed.Practical tests and data analysis of the multi-chip cascade automotive mm-wave radar systems based on TI-AWR1243P chips verify the effectiveness and feasibility of the system.