Research On GPU-based Time Division MIMO Radar Signal Processing Technology

Multiple-input multiple-output(MIMO)radar is a new type of radar that transmits orthogonal signals through multiple transmitting antennas,and multiple receiving antennas receive echo signals.MIMO radar is mainly divided into time-division working mode and frequency-division working mode.Slot of time-division MIMO radar is transmitted by an array element each time,so the signals cannot be superimposed in space.The volume of data received by MIMO radar is huge.The traditional CPU serial calculation is difficult to meet the real-time processing requirements of the radar system.The outstanding computing power of the graphics processing unit(GPU)brings new vitality to typical computation-intense problems.This thesis mainly analyzes and designs parallel time-division MIMO radar signal processing algorithms,and builds a time-division MIMO radar parallel simulation system based on CPU/GPU architecture.The main work of this thesis is as follows:1.The overall framework of the time-division MIMO radar system is designed,which mainly includes the echo system and the signal processing system,and the key technologies for system realization are introduced.After pointing out the necessity of building a timedivision MIMO radar system,the overall processing flow of the time-division MIMO radar system is designed.Taking into account the functions and processing speed of each part of the system,the software framework of the time-division MIMO radar system is designed based on the overall processing framework.The key technology introduces the CPU parallel Open MP technology and GPU parallel CUDA technology briefly,mainly analyzes the Open MP programming model and memory model,conducts the CUDA programming model,software system and storage structure,and also introduces the function library related to CUDA programming.Such as CUBLAS library,CUFFT library,CURAND library and Thrust library.2.The time-division MIMO radar echo system is an important part of the overall system.This thesis designs and implements the specific processing flow of the MIMO radar echo system,introduces the basic principles of MIMO radar,and focuses on the analysis of the plane array antenna pattern and its beam direction,virtual array layout situation.The echo system is divided into transmitter module,antenna module,echo generation module,and receiver module.The principle of each module is analyzed,and MATLAB simulation modeling of the time-division MIMO radar echo system is carried out.On this basis,the echo system is transplanted to the C++ platform.In view of the large amount of echo data,this thesis also performs de-chirp and down-sampling processing on the echo.It is convenient to use CUDA to do GPU parallel acceleration processing of echo data.3.Based on the design and implementation of the time-division MIMO radar echo system,in view of the time-division characteristics of the MIMO radar,this thesis designs the overall flow of the MIMO radar signal processing system and the GPU parallel acceleration scheme,and improves the algorithm in order to increase the parallel processing speed.The three parts of windowing,MTI,and MTD are combined into one module for processing.For MIMO beam forming,the array element number index is proposed to rearrange the antenna elements.Therefore,the time-division MIMO radar signal processing is divided into four modules,including MTI/MTD module,digital beam forming module,constant false alarm detection module,and sidelobe hiding module.Subsequently,the algorithm principle of each module is studied,the feasibility of parallel processing of each module algorithm is analyzed.The main parallel function modules of the GPU-based time-division MIMO radar signal processing system are introduced,and the GPU running results are compared with the MATLAB platform running results to verify the correctness of the system.The acceleration ratio results show that the GPU is compatible with the time-division MIMO radar signal.The processing has a good acceleration effect.4.In order to further explore the correctness and system performance of the GPU-based time-division MIMO radar signal processing system to detect targets,this thesis designs a GPU-based data merging module and a speed and distance information extraction module,and analyzes the parallel results of the two modules Analyze the correctness from the four aspects of target distance,speed,wave position angle,and spot trace results.Then,by analyzing the performance of the data copy module and using stream parallelism,the CPU+GPU hybrid parallel scheme is designed and realized,and the acceleration effect is compared and analyzed with the GPU acceleration scheme.The results show that the CPU+GPU hybrid parallel processing has better acceleration effect.Finally,the GPU-based time-division MIMO radar signal processing system is verified by measured data.