MIMO Radar Signal Processing And Parallel Implementation Based On GPU

Compared to the phased array radar,the collected MIMO radar has features of more degrees of freedom at the transmit end and flexible operation model,it's suitable for targets detection task under multiple targets and complex environment,has developed to an important area of radar technology.At the same time,the MIMO radar signal processing is with high algorithm complexity and computational burden,which raises a big challenge for the computational ability of the radar signal processing system.In recent years,the graphics processor(GPU)is advancing in terms of computation performance and makes breakthroughs in important indicators such as core number,core frequency,memory bandwidth and memory capacity.The CUDA is efficient to development and the software maintenance is convenient.Therefore,studying the MIMO radar signal processing program and parallel implementation based on GPU is of great significance.The specific items about GPU based parallel implementation of MIMO radar signal processing are inducted as follows: 1.The processing flow of MIMO radar echo signals is studied.The two signal processing structures of MIMO radar are briefly described.The computational burden of the two kinds of signal processing is analyzed.The related algorithms involved in the two signal processing structures are analyzed.The parallel implementation of algorithms,including matched filtering,digital beamforming,clutter suppression and constant false alarm detection,are studied deeply.At the same time,the measured data are used and the simulation experiments are carried out using two processing structures.2.The GPU based parallel technology is briefly introduced and the significance of the development of GPU parallel technology for radar signal processing is analyzed.From the hardware logic structure of the GPU,its parallel computing principle is described and the CUDA programming model is further described.And then,from the interdependent API library,the hardware interface is described.Finally,several libraries that are most closely related to radar signal processing,such as the CUFFT library,the CUBLAS library,and the CURAND library is introduced.3.The echo signal processing implementation of GPU-based MIMO radar is deeply studied.Using the modular design idea,the MIMO radar signal processing flow is divided into full virtual array element pulse compression module,full wave position digital beam forming module,clutter suppression module and the constant false alarm detection module.The parallel implementation of each module respectively introduces its implementation flow chart and related kernel functions.The comparison of the computational performance between the GPU and the MATLAB platform is shown.4.The research of different radar target tracking algorithm are presented,including probabilistic data association filtering algorithm,dynamic programming based TBD algorithm and Hough transform based TBD algorithm.Based on the simulation results,the necessity of parallel acceleration is analyzed.The implementation scheme of the GPU based parallel acceleration method is proposed and verified.