Design of High-Performance Millimeter-Wave Radar Digital Processing Accelerator

This research topic aims at designing a high-performance millimeter-wave radar digital processing accelerator. According to the current millimeter-wave radar point cloud generation algorithm, the internal computing components of the accelerator mainly include Fast Fourier Transform(FFT), Constant False Alarm Rate(CFAR), windowing, and other submodules. To minimize the interference of the central processor with this accelerator as much as possible, this research designs an operational register that is highly versatile, reusable, and circular. After parameter programming by the user, the accelerator can independently and continuously complete the operation of the point cloud generation algorithm. The experiments show that the accelerator can independently calculate complex sample data of 256*32 size without external interference. After the operations of 2DFFT and ABS, the relative error remains below 10% compared to MATLAB. The computational results of 2D-CFAR are entirely consistent with the simulated results from MATLAB.