Radar Signal Processing in Parallel on GPU: Case Study Dual Polarization FMCW Weather Radar

Abstract

Radar signal processing is currently done digitally with DSP (digital signal processor) or FPGA (field-programmable gate array) devices. Implementation of radar signal processing at the software level can provide high scalability, but existing research still has poor performance when implemented on the CPU. Meanwhile, the development of graphics processing unit (GPU) technology provides an opportunity to perform processing of radar signals with high performance as well as scaling. This paper focus on processing radar signals using GPUs with case studies of dual polarization FMCW (frequency-modulated continuous wave) weather radar signal processing. In this research has been done the implementation of weather radar signal processing algorithm on 2 platforms, which are serialized using CPU and in parallel using GPU. The results of performance measurements on both platforms show that GPU implementations deliver performance more than 2 times faster than implementations on the CPU.