A GPU-based target feature agile echo generation implementation method

Abstract

This invention relates to the field of radar signal agile echo, specifically to a GPU-based method for generating target feature agile echoes. The method includes: generating a feature agile factor sequence corresponding to the echo signal timing based on agile frequency, amplitude-phase agile mode, and agile range parameters issued by the system master control; inputting the feature agile factor sequence into the GPU memory; in the GPU memory, using a thread block corresponding to the output echo, modulating the feature agile factor sequence with the target's inherent RCS sequence according to the agile mode to generate an agile RCS sequence corresponding to each moment, i.e., the time-varying system response; convolving the time-varying system response with the radar transmitted signal to output a time-domain echo sequence; and superimposing the delay, Doppler, and power corresponding to each target in each channel onto the corresponding output time-domain echo sequence to obtain the target time-domain feature agile echo signal. This invention is applicable to radar signal agile echo processing.

Description

Technical Field

[0001] This invention relates to the field of radar signal agile echo, specifically to a method for generating target feature agile echo based on a GPU (Graphics Processing Unit). Background Technology

[0002] In modern radar systems, especially in complex applications involving electronic countermeasures, target identification, and active stealth, the radar scattering characteristics of targets no longer remain stable but exhibit highly dynamic changes, a phenomenon known as target agility. This agility includes amplitude agility and phase agility, characterized by rapid and complex changes, often strongly correlated with factors such as the radar system's measurement time and operating frequency. In this context, traditional echo simulation methods based on linear time-invariant systems (such as the frequency domain multiplication method) are no longer sufficient to accurately describe the target echo characteristics.

[0003] Existing technologies, such as the method disclosed in CN114325606A, can simulate radar echo signals of multiple systems using relatively few hardware and software resources, greatly reducing the cost and size of radar echo signal simulation systems. Furthermore, this invention utilizes agile switching via a host computer interface to simulate radar echo signals of multiple systems in the same continuous echo acquisition experiment, offering high operational flexibility. Simultaneously, this invention employs a ping-pong operation to generate radar echo signals. When the FPGA (Field-Programmable Gate Array) receives a new set of PRT echo data, it places the echo data into the RAM address of the previously output set of PRT echo data to update the existing PRT echo data in real time, and updates parameters such as the starting distance of the PRT and the number of sampling points. This ping-pong operation saves FPGA logic resources and improves time utilization.

[0004] However, real-world radar systems typically need to simulate the agility characteristics of multiple channels and multiple targets simultaneously, and each target may have an independent agility strategy. This greatly increases the computational burden. Especially under adversarial conditions, the system places higher demands on the real-time performance, accuracy, and scalability of target echo simulation, which the above-mentioned solutions cannot meet. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a GPU-based method for generating target feature agile echoes, which effectively solves the computational bottleneck of traditional methods under multi-channel and multi-target conditions and improves the real-time performance and accuracy of echo data.

[0006] The present invention achieves the above objectives by adopting the following technical solution: the present invention provides a GPU-based method for generating target feature agile echoes, comprising:

[0007] S1. Generate a characteristic agility factor sequence corresponding to the timing of the echo signal based on the agility frequency, amplitude and phase agility mode and agility range parameters issued by the system master controller, and input the characteristic agility factor sequence into the GPU memory.

[0008] S2. In the GPU memory, the feature agility factor sequence and the target's inherent RCS (Radar Cross-Section) sequence are modulated in an agility manner through the thread block corresponding to the output echo to generate the time-varying RCS sequence corresponding to each moment.

[0009] S3. Convolve the time-varying RCS sequence with the radar transmitted signal to output a time-domain echo sequence.

[0010] S4. Superimpose the delay, Doppler and power corresponding to each target in each channel onto the corresponding output time-domain echo sequence to obtain the target time-domain characteristic agile echo signal.

[0011] First, the time-varying RCS sequence of the target with non-zero gain received by the radar antenna and the signal spectrum transmitted by the radar are obtained. Then, the time-varying RCS sequence of the target with non-zero gain and the signal spectrum transmitted by the radar are vector-multiplied to obtain the spectrum of the current target echo. Then, the time-domain signal of the current target is obtained through IFFT. The target phase is multiplied to obtain the echo time-domain data of the current target. The thread grid is divided into a three-dimensional grid. The X dimension is set as the echo time-domain data of a certain target, which is divided into m / 512+1 thread blocks. Each thread block contains 512 threads, where m is the number of sampling points of the current target, the Y dimension is the number of targets, and the Z dimension is the data of different channels. The X and Y threads of the data of different channels are consistent, and the data is also consistent.

[0012] Then, the two-dimensional data of power gain for different channels and different targets are distributed into a two-dimensional thread grid, where one dimension is the data of different channels and the other dimension is the power gain of different targets.

[0013] Finally, based on the current target's location, the corresponding location among all targets is searched, and the power is multiplied and superimposed to complete the superposition of the echoes from each target.

[0014] Furthermore, in step S1, generating the characteristic agility factor sequence corresponding to the echo signal timing based on the agility frequency, amplitude-phase agility mode, and agility range parameters issued by the system master control specifically includes:

[0015] After the system master controller issues the agility frequency fh, amplitude and phase agility mode M, amplitude and phase change range, timing alignment reference, echo duration T or number of sampling points N, and sampling rate fs parameters, it first constructs a discrete time axis tn=n / fs, where tn represents the continuous time corresponding to the nth sampling point, n represents the sampling index, and fs represents the sampling rate. Then, according to the agility period Th=1 / fh, the entire timing is divided into multiple agility segments, where Th represents the agility period and fh represents the agility frequency.

[0016] Then, according to the set agility mode, a corresponding complex feature factor αk is generated in each agility cycle, wherein the amplitude and phase are selected from a specified range or distribution. Finally, the periodic αk sequence is extended into a complex sequence {α[n]} of length N aligned with the echo sampling point by zero-order hold or smooth interpolation, forming the final feature agility factor sequence. The agility mode is amplitude jump, phase jump, amplitude-phase joint jump, pseudo-random sequence or statistical distribution model.

[0017] Furthermore, step S3 specifically includes:

[0018] Zero-padding is performed on both the radar transmitted signal and the time-varying RCS sequence of the current target. Then, a fast Fourier transform is performed in parallel on the GPU to obtain their respective frequency domain representations. Next, the transmitted signal spectrum and the RCS spectrum are multiplied point-by-point in the frequency domain to achieve an equivalent convolution operation. Then, the multiplication result is converted back to the time domain through an inverse fast Fourier transform to obtain a preliminary time domain echo sequence. Finally, the effective data segment is extracted and output as the time domain echo sequence of the target.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention is based on a time-varying linear system model, which can accurately simulate the scattering characteristics and echo behavior of a target under agile conditions, thereby improving the realism of the simulation.

[0021] This invention, based on GPU parallel processing, effectively solves the computational bottleneck of traditional methods under multi-channel and multi-target conditions, and realizes real-time generation of echo data. Attached Figure Description

[0022] Figure 1 This is a flowchart of the GPU-based target feature agile echo simulation provided by the present invention;

[0023] Figure 2 This is a schematic diagram of parallel computation of the target agile RCS sequence provided by the present invention;

[0024] Figure 3 This is a schematic diagram of target echo superposition provided by the present invention;

[0025] Figure 4 This is a schematic diagram of the targetless feature-agile echo signal provided by the present invention;

[0026] Figure 5 This invention provides a target-feature agile echo signal. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0028] This invention provides a GPU-based method for generating agile echoes of target features, such as... Figure 1 As shown, it specifically includes:

[0029] S1. Generate a characteristic agility factor sequence corresponding to the timing of the echo signal based on the agility frequency, amplitude and phase agility mode and agility range parameters issued by the system master controller.

[0030] The specific generation method involves the system master controller issuing parameters such as the agility frequency fh, amplitude and phase agility mode M, amplitude and phase change range, timing alignment reference, echo duration T (or number of sampling points N), and sampling rate fs. First, a discrete time axis tn=n / fs is constructed, and the entire timing is divided into multiple agility segments according to the agility period Th=1 / fh. Then, based on the set agility mode (such as amplitude jump, phase jump, combined amplitude and phase jump, pseudo-random sequence, or statistical distribution model), a corresponding complex feature factor αk is generated in each agility period, where the amplitude and phase are selected from a specified range or distribution. Finally, through zero-order hold or smooth interpolation, the periodic αk sequence is expanded into a complex sequence {α[n]} of length N aligned with the echo sampling points, forming the final feature agility factor sequence, which is used to modulate the echo signal to simulate the dynamic scattering characteristics of the target.

[0031] like Figure 1 As shown, the time-varying factor sequence is arranged in time order as t1,…, t m ,…, t M Then it is transferred to the GPU memory via the PCIe bus.

[0032] S2. In the GPU memory, the feature agility factor sequence and the target's inherent RCS sequence are amplitude-phase modulated in an agility manner through the thread block corresponding to the output echo to generate the time-varying RCS sequence corresponding to each moment, i.e., the time-varying system response.

[0033] By using GPU parallel processing, the raw data of the target scattering points and the modulation of the agility factor are multiplied. Threads are allocated according to the start and end positions of the raw scattering point data and the agility factor sequence in the video memory. To fully utilize thread resources, 512 threads are allocated in each thread block, for a total allocation of the inserted data size / 512 thread blocks. Each thread processes one time-varying system response point and calculates each amplitude-phase modulation value. The modulation algorithm can be implemented in parallel on the GPU. To obtain multi-point data in parallel, one thread block is computed in parallel on each GPU's Streaming Multiprocessor (SM). Specifically, a GPU-based agility modulation target RCS sequence algorithm is used. Figure 2 As shown, the horizontal dimension of the thread grid represents the modulation data position that needs to be calculated for each target. The number of thread blocks in the horizontal dimension is InterpDataNums / 512, and each thread block contains 512 threads. This allocation is to better utilize resources and optimize the parallel design. The vertical dimension of the thread grid is the number of targets, N. Parallel modulation computation is used to simultaneously perform modulation operations on the scattering points of multiple targets. Based on the thread blocks of different targets and the threads within those blocks, the start and end positions of the original target data are simultaneously located. Then, modulation calculations are performed according to the data positions required by different targets, finally obtaining the agile modulated target RCS value for each target.

[0034] S3. Convolve the time-varying system response with the radar transmitted signal to output a time-domain echo sequence;

[0035] Based on the physical model of radar echoes, the target echo signal can be considered as the convolution result of the radar transmitted signal and the target's time-varying radar cross section (RCS). The time-varying RCS sequence, generated in the previous step, reflects the target's dynamic scattering characteristics under the control of frequency agility, amplitude-phase agility, and agility range. To efficiently implement this convolution operation, a frequency domain convolution method based on Fast Fourier Transform (FFT) is adopted. Specifically, the radar transmitted signal and the current target's time-varying RCS sequence are zero-padded to extend their length to a uniform length that meets the requirements of linear convolution (usually the sum of the two sequence lengths minus one, aligned to a length suitable for FFT calculation). Then, Fast Fourier Transform is executed in parallel on the GPU to obtain their respective frequency domain representations. Next, pointwise complex multiplication is performed on the transmitted signal spectrum and the RCS spectrum in the frequency domain to achieve an equivalent convolution operation. Then, the multiplication result is converted back to the time domain using Inverse Fast Fourier Transform (IFFT) to obtain a preliminary time-domain echo sequence. Finally, the effective data segment is truncated as the target's echo output. This process fully leverages the parallel computing capabilities of the GPU, employing a batch processing mode to perform FFT / IFFT operations on multiple targets simultaneously, significantly improving processing efficiency. All computations are performed in the GPU's video memory, avoiding frequent data transfers and ensuring real-time performance.

[0036] S4. Superimpose the delay, Doppler, and power corresponding to each target in each channel onto the corresponding output time-domain echo sequence to obtain the target time-domain characteristic agile echo signal.

[0037] Multiple channels transmitting beams of varying power will detect targets with different gain. For a single radar antenna, the number of targets is constant. Based on the characteristics of radar-tested targets, the detected target's attitude and delay are identical to the antenna, as the radar antenna detects the same targets. Both velocity and azimuth depend only on the target itself and are independent of the antenna's beam power. However, because the different beams emitted by the antenna have varying power, the received target beam power also varies, resulting in different gain for each target. Therefore, it is necessary to design gain settings for different channels and different targets.

[0038] Therefore, when designing a method for calculating target echoes, the first step is to obtain the time-varying RCS sequence of targets with non-zero gain received by the radar antenna and the signal spectrum transmitted by the radar. Then, the phase, delay, and power gain of all current targets, as well as the power gain of different targets in different channels, are required. Since this involves searching between the current target and all targets, data storage must be considered during the development and design process. This requires mapping the RCS position of the current target to the phase, delay, and power gain of that target among all targets. Therefore, a target RCS data array is designed, and the echoes of each target are then superimposed.

[0039] Overlay diagram as shown Figure 3As shown, the target's non-zero gain RCS is vector-multiplied with the signal spectrum emitted by the radar antenna to obtain the current target echo spectrum. Then, an IFFT is performed to obtain the current target's time-domain signal, and phase multiplication is performed to obtain the current target's echo time-domain data. Since multi-channel testing is involved, and the target's attitude and delay are consistent, the thread grid is divided into three dimensions. The X dimension represents the echo time-domain data of a specific target, divided into m / 512+1 thread blocks, each containing 512 threads, where m is the number of sampling points for the current target. The Y dimension represents the number of targets, and the Z dimension represents the data from different channels. The X and Y thread allocations in the Z dimension are consistent, and the data is also consistent. Next, the two-dimensional power gain data of different channels and different targets are allocated into a two-dimensional thread grid, with one dimension representing the data from different channels and the other representing the power gain of different targets. Finally, the corresponding position among all targets is searched according to the current target's position, and power multiplication and superposition are performed to complete the superposition of the target echoes.

[0040] The advantages of this invention will be verified through a practical data processing example below. Figure 4 It is a radar echo signal with agile target characteristics. Figure 5 It is the echo of the same transmitted signal under the target characteristic agility condition, with a signal pulse width of 25us, agility frequency of 1MHz, agility amplitude of 0~10dB, and agility phase of 0~180 degrees.

[0041] In summary, this invention extends the target modeling from a linear time-invariant system to a time-varying linear system. It generates echo signals by time-varying convolution of the target's agile response with the transmitted signal, realistically reflecting the target's dynamic characteristics during radar illumination. An agile simulator generates a sequence of target feature agile factors corresponding to the echo timing based on an agile strategy, used to dynamically adjust the target's scattering characteristics. The target feature agile simulator utilizes GPU parallel computing to perform multi-channel, multi-target feature agile echo data, achieving efficient echo generation under multi-channel, multi-target conditions.

[0042] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for generating agile echoes of target features based on GPU, characterized in that, include: S1. Generate a characteristic agility factor sequence corresponding to the timing of the echo signal based on the agility frequency, amplitude and phase agility mode and agility range parameters issued by the system master controller, and input the characteristic agility factor sequence into the GPU memory. S2. In the GPU memory, the feature agility factor sequence and the target's inherent RCS sequence are amplitude-phase modulated in an agility manner through the thread block corresponding to the output echo to generate the time-varying RCS sequence corresponding to each moment. S3. Convolve the time-varying RCS sequence with the radar transmitted signal to output a time-domain echo sequence. S4. Superimpose the delay, Doppler, and power corresponding to each target in each channel onto the corresponding output time-domain echo sequence to obtain the target time-domain characteristic agile echo signal.

2. The GPU-based target feature agile echo generation method according to claim 1, characterized in that, Step S4 specifically includes: First, the time-varying RCS sequence of the target with non-zero gain received by the radar antenna and the signal spectrum transmitted by the radar are obtained. Then, the time-varying RCS sequence of the target with non-zero gain and the signal spectrum transmitted by the radar are vector-multiplied to obtain the spectrum of the current target echo. Then, the time-domain signal of the current target is obtained through IFFT. The target phase is multiplied to obtain the echo time-domain data of the current target. The thread grid is divided into a three-dimensional grid. The X dimension is set as the echo time-domain data of a certain target, which is divided into m / 512+1 thread blocks. Each thread block contains 512 threads, where m is the number of sampling points of the current target. The Y dimension is the number of targets, and the Z dimension is the data of different channels. The X and Y threads of the data of different channels are allocated in the same way, and the data is consistent. Then, the two-dimensional data of power gain for different channels and different targets are distributed into a two-dimensional thread grid, where one dimension is the data of different channels and the other dimension is the power gain of different targets. Finally, based on the current target's location, the corresponding location among all targets is searched, and the power is multiplied and superimposed to complete the superposition of the echoes from each target.

3. The GPU-based target feature agile echo generation method according to claim 1, characterized in that, In step S1, generating a characteristic agility factor sequence corresponding to the echo signal timing based on the agility frequency, amplitude-phase agility mode, and agility range parameters issued by the system master controller specifically includes: After the system master controller issues the agility frequency fh, amplitude and phase agility mode M, amplitude and phase change range, timing alignment reference, echo duration T or number of sampling points N, and sampling rate fs parameters, it first constructs a discrete time axis tn=n / fs, where tn represents the continuous time corresponding to the nth sampling point, n represents the sampling index, and fs represents the sampling rate. Then, according to the agility period Th=1 / fh, the entire timing is divided into multiple agility segments, where Th represents the agility period and fh represents the agility frequency. Then, according to the set agility mode, a corresponding complex feature factor αk is generated in each agility cycle, wherein the amplitude and phase are selected from a specified range or distribution. Finally, the periodic αk sequence is extended into a complex sequence {α[n]} of length N aligned with the echo sampling point by zero-order hold or smooth interpolation, forming the final feature agility factor sequence. The agility mode is amplitude jump, phase jump, amplitude-phase joint jump, pseudo-random sequence or statistical distribution model.

4. The GPU-based target feature agile echo generation method according to claim 1, characterized in that, Step S3 specifically includes: Zero-padding is performed on both the radar transmitted signal and the time-varying RCS sequence of the current target. Then, a fast Fourier transform is performed in parallel on the GPU to obtain their respective frequency domain representations. Next, the transmitted signal spectrum and the RCS spectrum are multiplied point-by-point in the frequency domain to achieve an equivalent convolution operation. Then, the multiplication result is converted back to the time domain through an inverse fast Fourier transform to obtain a preliminary time domain echo sequence. Finally, the effective data segment is extracted and output as the time domain echo sequence of the target.

Images