General purpose graphic processing unit implementation of adaptive pulse compression algorithms

Yan Zhang,Jingxiao Cai

doi:10.1117/1.jrs.11.035009

Abstract

This study introduces a practical approach to implement real-time signal processing algorithms for general surveillance radar based on NVIDIA graphical processing units (GPUs). The pulse compression algorithms are implemented using compute unified device architecture (CUDA) libraries such as CUDA basic linear algebra subroutines and CUDA fast Fourier transform library, which are adopted from open source libraries and optimized for the NVIDIA GPUs. For more advanced, adaptive processing algorithms such as adaptive pulse compression, customized kernel optimization is needed and investigated. A statistical optimization approach is developed for this purpose without needing much knowledge of the physical configurations of the kernels. It was found that the kernel optimization approach can significantly improve the performance. Benchmark performance is compared with the CPU performance in terms of processing accelerations. The proposed implementation framework can be used in various radar systems including ground-based phased array radar, airborne sense and avoid radar, and aerospace surveillance radar.

Highlights

There are many existing applications of graphic processing unit (GPU)-based radar processing implementations, such as synthetic aperture radar processing[1,2] and constant false alarm rate processing.[3,4] Many of the applications show the potential of acceleration using graphical processing units (GPUs) collaborating with CPUs,[5,6] such as one thousand times acceleration of image formation over using CPUs alone.[7]
The results indicate that the length of the waveform has a larger impact on processing time compared with the length of ground truth, which is implied in the description of its computation complexity listed in Table 2, and the GPU-based platform performs better when the length of either parameter mentioned above is larger, as it can be seen that 10 times acceleration is expected when the data size is sufficiently large
We explored the feasibility of GPGPU-based implementation of advanced pulse compression algorithms, which is a key element and usually a bottleneck of an end-to-end radar data processing chain

Summary

Introduction

There are many existing applications of graphic processing unit (GPU)-based radar processing implementations, such as synthetic aperture radar processing[1,2] and constant false alarm rate processing.[3,4] Many of the applications show the potential of acceleration using GPUs collaborating with CPUs,[5,6] such as one thousand times acceleration of image formation over using CPUs alone.[7]. APC algorithms are based on existing pulsed compression (PC) algorithms in solid-state radar and offer reduced sidelobes and enhanced resolution. As such, they are important for downwardlooking high-altitude airborne and space radars.[11] There are multiple versions of APC algorithms, including reiterative minimum-mean-square error (RMMSE)[8] and RMMSE based on matched filter (MF-RMMSE) output.[10] Implementation of APC algorithms in real time will allow for fast remote sensing image formation for weather observation.[12] The challenge of this implementation with parallel computing has been the reiterative nature of APC algorithms, as well as the latencies and memory constraints for matrix inversions.

Objectives

Results

Conclusion