Study On Multi-dimensional Domain Processing And Development For Near-Field Search Based On Networked Radar

In today’s era,radar is an essential means of perceiving the battlefield situation on the modern battlefield.With the emergence of new cluster combat targets such as Unmanned Aerial Vehicles(UAVs),the requirements for airspace resolution are increasing day by day.However,the aperture resolution of traditional radar cannot meet the needs of the high resolution of the target colony.In this context,networked radar is proposed as a novel form of radar.The networked radar can be either a centralized radar with self-transmitting and self-receiving,a dual/multi-base radar with multiple transmissions and multiple receivers,or a antenna array radar with a sparse large aperture arrays.It has many advantages such as simultaneous multi-tasking,on-demand reconfigurability,flexible working mechanism,and strong survival ability.Usually,the antenna array radar with a sparse large-aperture arrays has a significantly larger aperture than the traditional radar.Thus,it has a good potential for improving the airspace resolution performance of the radar for the target colony.Networked radar comprises multiple widely distributed radar nodes(each node consists of several antenna elements).Because of its wide-area distribution characteristics,the possibility of the target located in the near-field region of the large-aperture array radar is greatly increased,and the traditional far-field array receiving signal model is no longer applicable.Until now,research on the near-field detection of networked radar comprises signal processing technology,performance evaluation,and system designs is hardly given in the state-of-theart literature.On the other hand,many concepts and critical techniques in networked radar still need to be further studied.Additionally,to achieve high-resolution detection for nearfield target colony,the sparse large-aperture array is utilized to improve the resolution of the target colony.Then,the Frequency Division Single Tone(FD-ST)signal is used to improve the problem of high sidelobes,which occurs the combined airspace transceiver beamforming.So further research on the near-field target detection method in networked radar is performed.In addition,based on the characteristics of multi-task configuration according to requirements and distributed computing efficiently for networked radar,the parallel development of near-field search processing of networked radar is considered via the hardware platform of the Graphic Processing Unit(GPU).The research focuses on the key problems in the multi-target high resolution of networked radar and the bottleneck problem that is difficult to meet the requirements of efficient realtime.The main work can be summarized as follows:1.Based on the near-field spherical wave propagation model,the near-field array receiving signal model is constructed.The networked radar near-field echo signal equation under FDST waveform is derived.Then,the near-field search processing for FD-ST signal coupled with distance is qualitatively analyzed.Finally,the phase reference accumulation conditions of the near-field echo signal in the space-time-frequency domain are given.2.Based on the coupling property of actual target distance and spatial multi-parameter in near-field detection,the formula for matching processing of networked radar in the nearfield multi-dimensional domain is deduced.A networked radar with an FD-ST signal is then proposed for multi-dimensional domain search and processing in the case of near-field situations.Several simulation results have been performed to validate the effectiveness of the processing method,and the grid mismatch in the matching processing is further considered.More specifically,the performance loss for multi-dimensional domain processing is quantitatively analyzed in grid mismatches in range,azimuth,and elevation,respectively.3.Aiming at the problem of large data volume and computation in the multi-dimensional domain processing of networked radar,a modular and hierarchical software framework for multidimensional domain processing of networked radar near-field search was constructed based on GPU parallel computing hardware platform combined with C++ language features.Subsequently,modular designs such as emission signal separation,multi-dimensional domain matching,accumulation synthesis,Moving Target Detection(MTD),and Constant False-Alarm Rate(CFAR),etc.are developed;The arrangement of input and output data for each module and its the techniques used in compiling the Kernel function are also discussed in detail.Moreover,the correctness of each processing module in the near-field multidimensional domain is verified by comparing the running results of GPU and MATLAB.Then,considering the platform characteristics of single-server multi-GPU or multi-server multi-GPU,a variety of distributed deployment schemes for each GPU including receiving channel division and spatial parallel division are proposed.The optimization results of the distributed deployment scheme for each GPU are compared under various hardware conditions.Finally,a test environment is established using simulated sources to verify the overall performance of multi-dimensional domain processing in the near field.