Broadband And Narrowband Radar Signal Processing Based On GPGPU

With the customizable demand of reconfigurable hardware and software,the research on radar technology gradually shift from the customized monopolization to the generic openness.The application of the heterogeneous parallel platform on radar real-time signal processing is analyzed,and the software architecture is designed simultaneously.Firstly,the heterogeneous computing architecture and parallel multi-task programming model is summarized.The computing platform based on GPU assisted with the hybrid programming model of MPI(Message Passing Interface),Open MP(Open Multi-Processing)and CUDA(Compute Unified Device Architecture)is established.This model can maximize its advantages of the multi-level storage and multi-granularity parallel on the heterogeneous computing platform to obtain the highest computing performance.Secondly,a parallel algorithm based on this heterogeneous platform is designed for typical narrowband and broadband radar.SIMT(Single Instruction Multiple Thread)is used to implement the parallel PC(Pulse Compression)and CFAR(Constant False Alarm Rate)algorithm for narrowband PD(Pulse Doppler)radar,which obtains 700+ times speedup.Additionally,a logarithmic-step algorithm for range alignment in broadband ISAR(Inverse Synthetic Aperture Radar)radar is proposed and obtains 4+ times speedup.The performance and the accuracy of these parallel algorithm are verified in cross-validation with the embedded platform based on TI multi-core DSP.At last,a software architecture framework is designed for this heterogeneous parallel computing platform base on GPGPU(General-Purpose GPU),and a simple discussion on two key techniques is given.By breaking down the framework into five layers,including Hardware,Hardware Abstraction Layer,Middleware,Core Framework and Componentbased Application,the development process is decoupled,and more professional and general products are developed.These lay the foundation for open architecture of radar industry.