Comparison of massively parallel algorithms on graphics processing unit for MIMO radar

Abstract

This paper proposes a method for accelerating an enhanced resolution 3D Multiple Input Multiple Output (MIMO) radar on a Graphics Processing Unit (GPU). Due to the size of the data required for range, bearing, and doppler processing, computations for the MIMO radar are extensive and seldom permit real-time operation without performance compromises. Current methods for achieving reasonable frame rates include reducing the scope of the radar (i.e., limiting the number of dimensions, the field of view, or the ranges of interest), choosing efficient but coarse algorithms (i.e., the FFT for range, velocity, and bearing estimation), or offloading the computations on task specific hardware, DSP, or FPGA. The proposed framework enables real-time operation of the MIMO radar by performing the signal processing on a GPU without compromising the radar coverage, while replacing the widely used 3D FFT with an enhanced resolution alternative. This paper compares the execution times of the various algorithms when performed on a Central Processing Unit (CPU), and when performed on the GPU.

Keywords

• MIMO radar
• Parallel programming
• Graphics processing unit
• Chirp Z transform