A GPU-accelerated sidelobe cancellation computation method

Abstract

The application provides a GPU acceleration-based sidelobe cancellation calculation method, and steps are as follows: data transmission from a CPU end to a GPU end, interference sample point acquisition, weight value calculation, interference projection and elimination, and data transmission from the GPU end to the CPU end. The acceleration calculation method can significantly improve the sidelobe cancellation calculation speed, solves the problems of excessive radar echo sample points, large weight value calculation amount and heavy burden of the CPU, and has practical application value in the field of radar signal processing technology.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing, particularly to the field of radar anti-jamming, specifically to a GPU-accelerated sidelobe cancellation calculation method. Background Technology

[0002] Electronic warfare plays an increasingly important role in modern warfare; radars without anti-jamming technology are completely rendered incapable of detecting enemy targets. With the rapid development of electronic technology, sidelobe cancellation technology has been widely applied to new radar systems. It possesses strong anti-sidelobe jamming capabilities, minimizing output interference power by adjusting the weights of auxiliary antennas, thus suppressing interference. However, the increasing types and number of jamming points currently limit CPU processing speed, making it impossible to eliminate sidelobe interference signals in real time.

[0003] With the rapid development of radar and jamming systems, the architecture of modern radar signal processing platforms has undergone a transformation. Traditional CPU-based radar anti-jamming implementations suffer from low efficiency and struggle to meet the diverse needs of modern technology, necessitating a new method to improve the efficiency of radar anti-jamming processing. The rapid advancements in semiconductor technology, integrated circuit manufacturing processes, and computer technology have led to continuous upgrades and improvements in real-time radar anti-jamming processing. Radar anti-jamming processing has shifted from the most widely used CPU implementation to implementation on high-speed programmable logic chip development platforms. While this solves the speed problem, developers need extensive prior knowledge and familiarity with the chip's built-in hardware resources to allocate them appropriately according to algorithmic logic requirements in engineering practice, resulting in a lengthy development cycle. Furthermore, system development and hardware platform are highly specific and coupled. For radars with different applications, this method presents significant development challenges. Summary of the Invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention aims to propose a GPU-accelerated sidelobe cancellation method to reduce the computational complexity of weighting and solve the real-time problem in radar signal processing.

[0005] To achieve the above objectives, the implementation scheme of the present invention includes the following:

[0006] A GPU-accelerated sidelobe cancellation calculation method, characterized by the following:

[0007] 1) Generate simulated radar echo signals, the echo signals including target signals and jamming signals, wherein the target signal generation includes linear frequency modulation signals and pulse Doppler signals, and the jamming signal generation includes deception jamming and suppression jamming;

[0008] 2) Obtaining interference samples:

[0009] The simulated echo signal generated in step 1) is transferred from CPU memory to GPU memory. Then, a system clock is added, and each thread executes at irregular time intervals based on parallelism. This allows the number of currently executing threads to be calculated, and a decision threshold is used to filter out those containing interference. Data from 10 sample points;

[0010] 3) Weight calculation:

[0011] For the first part containing interference, filtered out by step 2) Weights are calculated for each sample point.

[0012] 4) Interference projection and elimination:

[0013] By multiplying the weight matrix and the interference matrix, the interference can be projected onto the main lobe beam direction. Subtracting the interference projection value from the received echo signal in the main lobe direction allows for sidelobe cancellation calculation.

[0014] Based on the above technical solutions, the present invention may also employ the following further technical solutions, or combine these further technical solutions:

[0015] The parameters for the linear frequency modulated signal and pulse Doppler signal in step 1) are set as follows: The sampling rate for both radar signals is set to 10. The sampling distance is 2 The radar carrier frequency is 5.5. The number of sampling points is 200,000, and the number of pulses for pulse Doppler is... .

[0016] Target signal generation in step 1)

[0017]

[0018] In the formula, Indicates the distance to the target. Indicates the radial velocity of the target. Indicates the carrier frequency. Indicates baseband signal, Represents the speed of light. Indicates time, This indicates the pulse repetition period.

[0019] In step 2), the acquisition of interference samples is specifically implemented in parallel processing within the GPU kernel function. Here, a system clock needs to be added as a delay to allow a time difference in the calculation results of the parallel computing threads, so as to ensure that the first N points containing interference can be calculated, and then the weights are calculated using these N points.

[0020] The formula for calculating the weights in step 3) is as follows:

[0021]

[0022] Using the least squares method Calculate weights

[0023] In the formula, Indicates gain. Indicates the target signal. Indicates interference signal. Indicates the number of interfering signals. Indicates the auxiliary path interference gain. Indicates the echo analog signal. This indicates the interference signal received by the auxiliary circuit. Indicates the weight.

[0024] By employing the technical solution of this invention, the present invention has the following specific advantages:

[0025] (1) This invention transforms the computation method executed serially on the CPU into parallel processing on the GPU.

[0026] (2) By using the system clock to introduce the thread execution time difference during parallel processing, the first N interference sample points containing interference can be calculated, reducing the amount of computation and significantly improving the time for calculating the cancellation weights. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the implementation of the present invention.

[0028] Figure 2 This is a schematic diagram of the thread's working mode after adding a system clock according to the present invention.

[0029] Figure 3 It is the generated target linear frequency modulated signal.

[0030] Figure 4 It is a simulated interference signal.

[0031] Figure 5 It is a linear frequency modulated signal superimposed with an interference signal.

[0032] Figure 6 It is the target signal after the sidelobes of the radar echo signal are canceled.

[0033] Figure 7 It is the generated target pulse Doppler signal.

[0034] Figure 8 It is a simulated interference signal.

[0035] Figure 9It is a pulse Doppler signal superimposed with interference signal.

[0036] Figure 10 It is the target signal after the sidelobes of the radar echo signal are canceled. Detailed Implementation

[0037] Reference Figure 1 This invention discloses a GPU-accelerated sidelobe cancellation calculation method. It utilizes the generated radar echo signal to acquire interference samples on the GPU, then performs weight calculations, and finally projects and cancels the interference to achieve sidelobe cancellation. The specific implementation steps are as follows:

[0038] Step 1: Generate a radar simulation signal.

[0039] The target signal is generated as follows:

[0040]

[0041] In the formula, Indicates the distance to the target. Indicates the radial velocity of the target. Indicates the carrier frequency. Indicates baseband signal, Represents the speed of light. Indicates time, This indicates the pulse repetition period.

[0042] For suppression interference, a baseband random signal and parameters (amplitude modulation coefficient, phase modulation coefficient, frequency modulation coefficient) are generated to produce a corresponding interference signal.

[0043] For repeater-type interference, a baseband signal is generated, and then the baseband signal is periodically extended according to the frame reception period to form a repeater-type signal, which is then multiplied with Gaussian white noise to generate an interference signal.

[0044] Step 2: Obtaining interference samples.

[0045] The simulated echo signal generated in step 1 is transmitted from the CPU memory to the GPU memory via PCIe. Here, a system clock delay needs to be added (this delay needs to be greater than or equal to the clock speed of each thread execution divided by the system clock) to create a time difference in the calculation results of the parallel computing threads, so as to ensure that the first N sample points containing interference can be calculated (here, N is obtained by accumulating the threads after parallelization through the delay). In this experiment, N is set to 16. When N is less than 16, the error of weight W is large. When N is too large, it will affect the calculation speed of weight W. Figure 2 A diagram illustrating the working mode of a thread after adding a system clock.

[0046] Step 3: Weight calculation.

[0047] The specific formula for canceling weights is as follows:

[0048]

[0049] Using the least squares method Calculate weights

[0050] In the formula, Indicates gain. Indicates the target signal. Indicates interference signal. Indicates the number of interfering signals. Indicates the auxiliary path interference gain. Indicates the echo analog signal. This indicates the interference signal received by the auxiliary circuit. Indicates the weight.

[0051] Step 4: Interference projection and elimination.

[0052] By multiplying the weight matrix and the interference matrix, the interference can be projected onto the main lobe beam direction. Subtracting the interference projection value from the received echo signal in the main lobe direction allows for sidelobe cancellation calculation.

[0053] To better illustrate the algorithm efficiency of this invention, a simulation experiment is used to verify it, taking a specific set of parameters as an example:

[0054] First, the echo signal is stored in a two-dimensional matrix, where the rows represent the number of echo points and the columns represent the number of pulses. Parallel processing is implemented in the GPU kernel function. Then, weights are calculated using N interference sample points to perform sidelobe cancellation. Finally, the data is transferred from GPU memory to the CPU. The Visual Studio development environment is configured, CUDA 11.2 is installed, and radar echo signals are generated using integrated development within the Visual Studio software. These radar echo signals include target signals and interference signals. The target signals include linear frequency modulated (LFM) signals and pulse Doppler signals, both with a sampling rate set to 10. The number of targets is 2, and the sampling distance is 2. The radar carrier frequency is 5.5. The number of sampling points is 200,000, the number of sampling start points is 1,974, and the number of pulses for pulse Doppler is... The LFM single-pulse signal generated based on the above parameter settings is as follows: Figure 3 As shown, the generated PD multi-pulse signal is as follows: Figure 7As shown. Interference signal generation includes deception interference and suppression interference. Deception interference is generated using false target signals, while suppression interference is further divided into direct-play interference and repeater-type interference. For direct-play interference, a baseband random signal is generated and then combined with parameters (amplitude modulation coefficient, phase modulation coefficient, frequency modulation coefficient) to generate the corresponding interference signal. The interference signal is set to have 2 interference signals, a sampling rate of 10MHz, a baseband noise bandwidth of 1MHz, an LFM pulse width of 64e-6s, a PD pulse width of 1e-5s, a pulse repetition period of 5e-5, and a modulation bandwidth B of 5e6Hz. The interference signal is generated based on the linear frequency modulation interference parameters. Figure 4 Interference signals are generated based on pulse Doppler interference parameters. Figure 8 Interference signals will Figure 3 The generated linear frequency modulated signal and Figure 4 The generated interference signals superimposed form a single-pulse radar echo signal, such as Figure 5 As shown. Figure 7 The generated pulse Doppler signal and Figure 8 The generated interference signals superimposed to form multipulse radar echo signals, such as Figure 9 As shown. The radar signal after sidelobe cancellation of the monopulse radar echo signal is as follows. Figure 6 As shown, the radar signal after sidelobe cancellation of the multipulse radar echo signal is as follows: Figure 10 As shown in Table 1 below, the number of interferences was set to 1, 2, 3, ..., 5. The average time taken to calculate sidelobe cancellation was calculated for each loop of 100 iterations. When the number of interferences is greater than 2, the calculation speed is significantly improved when using this experimental method.

[0055] The above description is only a specific example of the present invention. For those skilled in the art, understanding the computational principle of the present invention can significantly reduce and improve the computational complexity of the algorithm.

Claims

1. A GPU-accelerated sidelobe cancellation calculation method, characterized in that, Includes the following steps: 1) Generate a simulated radar echo signal, wherein the simulated radar echo signal includes a target signal and a jamming signal, wherein the target signal includes a linear frequency modulated signal and a pulse Doppler signal, and the jamming signal includes deception jamming and suppression jamming; 2) Obtain interference sample points: The simulated radar echo signal generated in step 1) is transferred from the CPU memory to the GPU memory; a system clock is added as a delay in the GPU kernel function so that the calculation results of each thread in parallel computing form an execution time difference; the output of the threads after the delay is accumulated to determine the number of currently executing threads, and the first N sample points containing interference are selected by a preset decision threshold. 3) Calculate the weights: In the GPU, only the first N sample points containing interference selected in step 2) are used to calculate the sidelobe cancellation weights; 4) Perform interference projection and elimination: In the GPU, the weight matrix is ​​multiplied by the interference matrix to project the interference onto the main lobe beam direction, and the interference projection value is subtracted from the echo signal received in the main lobe direction to obtain the radar echo signal after sidelobe cancellation. 5) The radar echo signal after sidelobe cancellation is transferred from GPU memory to CPU memory.

2. The method according to claim 1, characterized in that, The parameters for the linear frequency modulated signal and pulse Doppler signal in step 1) are set as follows: The sampling rate for both radar signals is set to 10. The sampling distance is 2 The radar carrier frequency is 5.

5. The number of sampling points is 200,000, and the number of pulses for pulse Doppler is... .

3. The method according to claim 1, characterized in that, In step 2), N is obtained by accumulating the number of threads output by the delayer, and N is 16.

4. The method according to claim 1, characterized in that, In step 3), the least squares method is used to calculate the sidelobe cancellation weights based on the first N sample points containing interference that have been selected.

5. The method according to claim 4, characterized in that, The formula for calculating the weights in step 3) is as follows: The weights are calculated using the least squares method. In the formula, Indicates gain. Indicates the target signal. Indicates interference signal. Indicates the number of interfering signals. Indicates the auxiliary path interference gain. Indicates the echo analog signal. This indicates the interference signal received by the auxiliary circuit.

Images