A shield pulse design method based on MAB algorithm

By introducing a multi-arm gambling machine algorithm into the radar system, the pulse width of the cover signal is optimized, and the problem of degradation of anti-interference effect in the existing technology is solved, and efficient anti-interference and detection capabilities are achieved in dynamic environments.

CN118886340BActive Publication Date: 2024-11-29JIANGNAN ELECTROMECHANICAL DESIGN INST
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411368842.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-29
Publication Date
2024-11-29
Estimated Expiration
2044-09-29

AI Technical Summary

Technical Problem

When the existing RF masking technology faces interference from instantaneous frequency measurement technology, it is impossible to adjust the pulse width of the masking signal in time, resulting in a decrease in the anti-interference effect and it is difficult to adapt to changes in the interference side and environment.

Method used

Using the mask pulse design method based on multi-arm gambling machine algorithm (MAB), the pulse width of the mask signal is optimized cyclically through the MAB decision model and the return function, and a gradually optimized mask signal transmission strategy is constructed to reduce the probability of being intercepted and maximize the detection probability. .

Benefits of technology

It realizes the possibility of radar intercepting in complex electromagnetic environments, improves detection success rate, adapts to dynamic interference environments, and enhances the robustness and anti-interference ability of radar.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118886340B_ABST
    Figure CN118886340B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for designing a cover pulse based on a MAB algorithm, comprising the following steps: determining the frequency range of a detection signal and a cover signal; determining input parameters; loading a MAB decision model and a benefit function, wherein the MAB decision model is used to cyclically call a distribution function to obtain a pulse width, which is used as a cover signal pulse width for transmission, to obtain a signal detection result, and to optimize the distribution parameters of the distribution function according to the signal detection result; calling the MAB decision model, cyclically obtaining the pulse width, optimizing the distribution parameters of the distribution function, and forming an optimized MAB decision model. According to the above technical solution, a MAB decision model is constructed, and in each process of generating a cover signal pulse width for transmission and interacting with the environment, the model self-learning is continuously performed, the model parameters are optimized, and a gradually optimized cover signal pulse width transmission strategy is constructed, so that the radar can reduce the probability of being intercepted in the process of confronting an IFM jammer, while maximizing the detection probability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of radar anti-interference, and in particular to a cover pulse design method based on a MAB algorithm. Background Art

[0002] In radar anti-interference technology, frequency agile radar changes the carrier frequency between adjacent pulses, making it difficult for the jammer to align the frequency. Compared with traditional fixed-carrier pulse Doppler radar, it has the performance advantage of resisting active suppression interference. However, with the development of instantaneous frequency measurement (IFM) technology, the jammer can quickly measure the carrier frequency of the received radar signal, realize intra-pulse suppression interference, and make the radar frequency agility invalid.

[0003] To address this problem, RF masking technology is used to counter IFM jammers. RF masking technology transmits a masking signal with a large frequency difference from the detection signal. At this time, after the jammer samples the pulse front edge and measures the frequency, the measured signal frequency is the masking signal frequency of the front edge, and the real signal frequency is not measured, thereby shielding the radar detection signal. This technology is simple to implement and has excellent anti-interference performance. RF masking technology has been widely used in the field of radar anti-interference. In actual applications, typical reconnaissance aircraft intercept radar signals, sort and measure the signals, determine the pulse width, carrier frequency, arrival angle, arrival time, pulse repetition interval, etc. of the pulse signal, and generate RF masking signals around the measurement of these parameters to counter reconnaissance receivers.

[0004] However, most existing RF shielding technologies use a fixed shielding signal pulse width. When the interference party's strategy and environment change, if the shielding signal pulse width cannot be adjusted in time, the RF shielding anti-interference effect will be greatly reduced. Therefore, a shielding pulse design scheme is needed that can adjust the shielding signal pulse width in time according to the actual situation and generate an effective RF shielding signal. Summary of the invention

[0005] To achieve the above object, the present application provides a shield pulse design method based on the MAB algorithm, comprising the following steps:

[0006] Determine the frequency range of detection and cover signals;

[0007] Determine the input parameters including: the number of selectable pulse widths, detection frequency and cover frequency;

[0008] Load the MAB decision model and profit function. The MAB decision model is used to cyclically call the distribution function to obtain the pulse width. , with the pulse width As a cover signal pulse width transmission, a signal detection result is obtained, and the distribution parameter of the distribution function is optimized according to the signal detection result; the benefit function is used to calculate the pulse width The corresponding income;

[0009] Call the MAB decision model and loop to obtain the pulse width , optimize the distribution parameters of the distribution function and construct an optimized MAB decision model.

[0010] Among them, the frequency range of the detection signal and the cover signal is defined as: , where F is the center frequency of the radar detection signal, is the center frequency of the shielding pulse, and B is the bandwidth of the interference signal.

[0011] Before loading the MAB decision model, constructing the MAB decision model includes:

[0012] Define the radar action set, expressed as, ,in, is the maximum pulse width of the pulse, is the pulse width, For each action sequence number, is the maximum number of actions;

[0013] Define a distribution function, which is used to generate random numbers and is expressed as: ,in, , is the distribution parameter of the distribution function, indicating the number of pulses in the tth pulse repetition period. Parameters of an action pulse;

[0014] The distribution function is optimized by transmitting the mask pulse formed by the maximum random number among the random numbers.

[0015] Among them, when loading the MAB decision model, the distribution parameters of the initialization distribution function are expressed as:

[0016]

[0017] Furthermore, the optimization of the distribution function according to the maximum random number in the random numbers includes:

[0018] Step S123: Obtaining a random number corresponding to each action through a distribution function;

[0019] Step S124: extracting the maximum random number from the random numbers corresponding to each action;

[0020] Step S125: Extract the maximum random number pair as the pulse width , as the cover pulse width and emitted, expressed as: ,in, is the random number corresponding to the i-th action in step S124;

[0021] Step S126: receiving the echo signal and determining whether a target is detected;

[0022] Step S127: construct a revenue calculation system based on the target detection results, and obtain the revenue value from the revenue calculation system , to achieve distribution function optimization;

[0023] Determine whether to terminate the optimization. If not, re-execute step S123.

[0024] Among them, the construction of the profit calculation system is realized through the profit function, which is expressed as:

[0025] ,in, Gain revenue for radar target detection results.

[0026] Furthermore, the method for optimizing the distribution function is:

[0027] If the return value is 1, the method of updating the distribution parameter of the distribution function is:

[0028] , ,in, , For the In the pulse repetition period Parameters of an action pulse;

[0029] Otherwise, the method to update the distribution parameters of the distribution function is: , ,in, is the discount factor, and: .

[0030] Furthermore, when defining the radar action set, the pulse width is controlled and the detection signal pulse width The sum is a constant, expressed as:

[0031] ,in, To protect the maximum value that the signal pulse width can reach and the detection signal pulse width can reach.

[0032] According to the present invention, a multi-armed bandit model is added to the radar system as a MAB decision model. The MAB decision model continuously performs model self-learning and optimizes model parameters in the process of generating a cover signal pulse width for transmission and interacting with the environment each time, and constructs a gradually optimized cover signal pulse width transmission strategy, so that the radar can reduce the probability of being intercepted in the process of countering IFM interference while maximizing the detection probability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a step diagram of a shield pulse design method provided in an embodiment of the present invention;

[0034] Figure 2 is a simulation logic schematic diagram provided according to an embodiment of the present invention;

[0035] Figure 3 is the cumulative benefit value of the radar under the balanced environment fixed strategy provided by an embodiment of the present invention;

[0036] Figure 4 The cumulative benefit value of the radar in a non-stationary environment provided by an embodiment of the present invention;

[0037] Figure 5 The cumulative revenue value of the radar in a non-stationary environment provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention provides a cover pulse design method based on the MAB algorithm (multi-armed bandit algorithm). A multi-armed bandit model is added to a radar system as a MAB decision model. The MAB decision model continuously performs model self-learning and optimizes model parameters in each process of generating a cover signal pulse width for transmission and interacting with the environment, and constructs a gradually optimized cover signal pulse width transmission strategy, so that the radar can reduce the probability of being intercepted and maximize the detection probability in the process of countering IFM interference.

[0039] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings.

[0040] The implementation process of the shield pulse design method based on the MAB algorithm proposed in the present invention is as follows: Figure 1 As shown, the following steps are included:

[0041] Step S100: determining the frequency range of the detection signal and the cover signal;

[0042] Due to the principle of RF shielding pulse technology, anti-interference can be achieved when the frequency of the transmitted shielding pulse is significantly different from that of the detection signal. In this step, the frequency band distribution of the shielding signal and the real signal is defined as: , where F is the center frequency of the radar detection signal, is the center frequency of the shielding pulse, and B is the bandwidth of the interference signal.

[0043] Step S110: The input parameters determined include: the number of selectable pulse widths, the detection frequency and the cover frequency; wherein the detection frequency and the cover frequency are used to send signals to the interferometer in the MAB decision model, and the number of selectable pulse widths can affect the convergence speed of the MAB decision model.

[0044] Step S120: Load the MAB decision model and the benefit function. In the MAB decision model, call the distribution function cyclically to obtain the pulse width. , with the pulse width As a cover signal pulse width transmission, a signal detection result is obtained, and a distribution parameter of the distribution function is optimized according to the signal detection result to achieve MAB decision model optimization;

[0045] The gain function is used to calculate the pulse width Corresponding profit value, determine the optimization strategy of the distribution parameters of the distribution function.

[0046] In a specific application, before loading the MAB decision model, the MAB decision model is constructed.

[0047] Step S121: constructing the MAB decision model, including:

[0048] 1) Define the radar action set, expressed as, ,in, is the maximum pulse width of the pulse, is the pulse width, For each action sequence number, is the maximum number of actions.

[0049] In order to ensure the normal operation of the radar, it is necessary to constrain the maximum pulse width of the cover pulse to ensure that the sum of the cover signal pulse width and the detection signal pulse width is a certain value. Therefore, when defining the radar action set, it is necessary to control the pulse width. and the detection signal pulse width The sum is the fixed value of the inherent parameters of the radar system, expressed as:

[0050] ,in, To protect the maximum value that the signal pulse width can reach and the detection signal pulse width can reach.

[0051] 2) Define the distribution function, expressed as: ,in, , is the distribution parameter of the distribution function, indicating the tth pulse repetition period Parameters of an action pulse;

[0052] A random number can be generated according to the distribution function of each action, and the action corresponding to the maximum random number is selected as a cover pulse; and subsequently, a cover pulse formed by the maximum random number among the random numbers is transmitted, and the transmission result is identified to achieve the optimization of the distribution function.

[0053] To protect the pulse , according to the parameters of the action from the distribution Randomly sample the corresponding random number ; Select the maximum random number from the random numbers generated by all actions, and extract the corresponding action as the pulse width of the above-mentioned guard pulse, which can be expressed as: .

[0054] Sampling from the parameter distribution When selecting a pulse width, Thompson sampling (TS) is used. Based on Bayesian thinking, the uncertainty of pulse width selection is expressed by the probability distribution of actions, and random numbers are obtained by sampling from each distribution.

[0055] In subsequent applications, determine the action corresponding to the maximum random number To protect the pulse width.

[0056] The method uses the prior knowledge of the i-1th interaction to gradually explore the distribution function of the i-th interaction of the action, thereby improving the computational efficiency while avoiding the Matthew effect.

[0057] Step S122: Initialize the distribution parameters of the distribution function when loading the MAB decision model and , that is, the parameters of the first learning, expressed as:

[0058] Later, in the process of calling the MAB decision model, the pulse width is obtained cyclically , optimize the distribution parameters of the distribution function.

[0059] The process of optimizing the distribution function includes:

[0060] Step S123: Obtaining a random number corresponding to each action through a distribution function;

[0061] According to each action Parameters , , from the distribution Sample a random number from ;

[0062] Step S124: extracting the maximum random number from the random numbers corresponding to each action;

[0063] Step S125: Extract the maximum random number pair as the pulse width , as the cover pulse width, combined with the detection frequency and the cover frequency for transmission;

[0064] The guard pulse width is expressed as: ;

[0065] Step S126: receiving the echo signal and determining whether a target is detected;

[0066] In this step, down-conversion, pulse compression, constant false alarm detection and other processing operations are performed on the received echo signal to obtain a signal detection result.

[0067] Step S127: constructing a revenue calculation system according to the target detection results and optimizing the distribution function;

[0068] Among them, the construction of the profit calculation system is realized through the profit function, which is expressed as:

[0069] ,in, Gain revenue for radar target detection results.

[0070] That is, if the target is detected, the benefit value of the action is is 1, otherwise it is 0;

[0071] Distribution function optimization based on return value Completed, specifically including: If the return value is 1, adjust the distribution parameter of the distribution function, which is reflected as Incremental, Decrease, increase the probability of being selected in the next round of calls, the calculation method is:

[0072] , ,in, , For the In the pulse repetition period Parameters of an action pulse;

[0073] If the return value is 0, that is , at this time, the distribution parameters of the distribution function are adjusted, which is reflected as , Adjust by discount factor to reduce the probability of being selected in the next round of calls. The calculation method is:

[0074] , ,in, is the discount factor, and: .

[0075] In the model, every movement of the radar It is equivalent to an arm in the multi-armed bandit model. The initial Beta distribution parameters of each action are in the interaction process between the radar and the jammer. Since the reconnaissance window length of the jammer may change, the optimal strategy will change. Therefore, the discount factor is introduced. , reduce the weight of past experience and improve the radar's adaptability.

[0076] According to the self-training process of the MAB decision model, after step S127 is completed, step S123 is executed again. In this process, the distribution parameters of the distribution function are continuously optimized, and the pulse width is obtained in each cycle. As a cover pulse width, it is transmitted; in each round of transmission, a cover pulse width with a benefit value of 0 may be generated, which is an acceptable defect in the application process; when a defect occurs, the distribution parameter of the distribution function is reduced, thereby reducing the probability of the cover pulse width being selected in the next round of call, and increasing the probability of the advantage distribution parameter being selected in the next round of call, and the cover signal pulse width transmission strategy can be obtained through online learning, so that the radar can reduce the probability of being intercepted in the process of confronting the IFM jammer and maximize the detection probability.

[0077] During each iteration, it is determined whether to terminate the optimization; the determination can be based on the number of pulse repetition cycles t, or other conditions, such as the timing of the iteration, etc. If the optimization is not terminated, step S123 is re-executed.

[0078] Step S130: In the process from step S123 to step S127, the distribution parameters of the distribution function are continuously optimized to form a gradually optimized MAB decision model; in the shielding pulse design method provided by the present invention, the success probability of shielding is improved by continuously using and optimizing the MAB decision model in the application environment, and the robust detection capability of radar targets in complex electromagnetic environments is enhanced.

[0079] The revenue value generated during the execution of steps S123 to S127 within a period of time The sum of the cumulative benefit value constitutes the cumulative benefit value, and the trend of the cumulative benefit value can intuitively reflect the optimization result of the MAB decision-making model.

[0080] The present invention provides a specific simulation test, the specific logic is as follows Figure 2 As shown in the figure, the simulation test content includes:

[0081] (1) Determine simulation conditions: Assume that the inherent parameters of the radar system are constant. 10.5 , that is, the protection signal pulse width and the detection signal pulse width The sum is 10.5 The maximum pulse width of the shield pulse is 10 The number of selectable cover signal pulse widths is 20; the false alarm probabilities of radar and jammer are both When the radar is not interfered by the jammer, the signal-to-noise ratio of the radar receiving signal is , when the radar is jammed by the jammer, the signal-to-noise ratio is 0dB. Greater than the protection signal pulse width When the jammer receives the radar detection signal, the signal-to-noise ratio , when the jammer reconnaissance window length Smaller than the protection signal pulse width When the jammer is in operation, it cannot detect the radar's detection signal.

[0082] In subsequent experiments, the cumulative profitability of the cover signal pulse width fixed strategy (Fix for short), the cover signal pulse width random selection strategy (Rand for short), the classic multi-armed gambling machine algorithm UCB1 algorithm (UCB for short), the EXP3 algorithm (EXP3 for short) and the method of the present invention (DTS for short) were compared and analyzed.

[0083] (2) Simulation content:

[0084] Simulation experiment 1: Analyze the performance of the cover signal pulse width decision algorithm in a simulated stable environment:

[0085] Assume the jammer's reconnaissance window length is For 5 , in the fixed strategy 2.5 respectively and 7.5 , the discount factor of the method of the present invention is .

[0086] Simulation experiment 2: Assuming the detection window length of the jamming system It increases gradually over time, expressed as:

[0087] , the period of the reconnaissance window change is 10000 pulses, and other parameters are the same as those in simulation experiment 1.

[0088] Simulation experiment 3: Analyze the performance of the cover signal pulse width decision algorithm in a simulated non-stationary environment:

[0089] Assume the jammer's reconnaissance window length is , the period of the detection window change is 10000 pulses. Other parameters are the same as those in the simulation experiment 1.

[0090] (3) Analysis of simulation results:

[0091] Figure 3The cumulative benefit values ​​of radars using different anti-interference methods under a fixed strategy in a balanced environment are provided, wherein 201 is the cumulative benefit value of the radar using the anti-interference method provided by the present invention, namely DTS; 202 is the cumulative benefit value of the radar using the EXP3 algorithm; 203 is the cumulative benefit value of the radar using the UCB algorithm; 204 is the cumulative benefit value of the radar using a fixed strategy; 205 is the cumulative benefit value of the radar using the Rand algorithm; 206 is the cumulative benefit value of the radar using a fixed strategy;

[0092] like Figure 3 As shown, when the strategy is fixed to When, due to , the jammer can detect the radar's detection signal and interfere with it, causing the RF cover to fail; when the strategy is fixed to When, due to , the jammer cannot detect the radar's detection signal, thus avoiding interference. However, due to the shield pulse width If the value is too large, the detection signal energy will be too low, thus affecting the detection probability of the radar. By comparing the various methods, it can be seen that the method of the present invention has the highest benefit value, which can effectively reduce the probability of the radar being interfered with, while ensuring the detection probability of the radar to the target.

[0093] like Figure 4 The cumulative benefit values ​​of radar using different anti-interference measures in a non-stationary environment are provided:

[0094] Among them, 301 is the cumulative benefit value of the radar using the anti-interference method (DTS) provided by the present invention; 302 is the cumulative benefit value of the radar using the EXP3 algorithm; 303 is the cumulative benefit value of the radar using the UCB algorithm; 304 is the cumulative benefit value of the radar using a fixed strategy; 305 is the cumulative benefit value of the radar using a fixed strategy; 306 is the cumulative benefit value of the radar using the Rand algorithm.

[0095] It can be seen that fixed strategies are difficult to adapt to changes in the environment, resulting in reduced adversarial performance.

[0096] By comparing various methods, it can be seen that the method of the present invention has better adaptability to non-stationary interference environments and can effectively improve the performance of radar against IFM jammers.

[0097] Figure 5Provided are the cumulative benefit values ​​of radars using different anti-interference measures in a non-stationary environment, wherein 401 is the cumulative benefit value of the radar using the anti-interference method provided by the present invention, namely DTS; 402 is the cumulative benefit value of the radar using the EXP3 algorithm; 403 is the cumulative benefit value of the radar using the UCB algorithm; 404 is the cumulative benefit value of the radar using a fixed strategy; 405 is the cumulative benefit value of the radar using the Rand algorithm; and 406 is the cumulative benefit value of the radar using a fixed strategy.

[0098] It can be seen that when the reconnaissance window length of the jammer changes randomly, the method of the present invention can still adapt well to the enemy interference environment and has a higher benefit value. Therefore, the method of the present invention has a stronger adaptability to the dynamic interference environment and can adjust the cover signal pulse width in real time to make the radar system work in the best state as much as possible.

[0099] The above simulation experiments prove that the method provided by the present invention can learn the change strategy of the jammer by continuously interacting with the environment, determine the emission strategy of the cover pulse at this moment, and expand the application scenario. Furthermore, by determining the cover signal pulse width through selection probability based on the MAB algorithm, the randomness of the cover signal can be increased, and the anti-interference effect can be enhanced.

[0100] The above disclosures are only several specific embodiments of the present invention; however, the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A shield pulse design method based on MAB algorithm, characterized in that: The following steps are involved: Determine the frequency range of detection and cover signals; Determine the input parameters including: the number of selectable pulse widths, detection frequency and cover frequency; Load the MAB decision model and the profit function. The MAB decision model is used to cyclically call the distribution function to obtain the pulse width t c , with the pulse width t c As a cover signal pulse width transmission, a signal detection result is obtained, and the distribution parameter of the distribution function is optimized according to the signal detection result; the benefit function is used to calculate the pulse width t c The corresponding income; Call the MAB decision model and loop to obtain the pulse width t c , optimize the distribution parameters of the distribution function and construct an optimized MAB decision model; Before loading the MAB decision model, constructing the MAB decision model includes: Define the radar action set, expressed as, i=0,1,2,…,N-1,where T c is the maximum pulse width of the pulse, is the pulse width, i is the sequence number of each action, and N is the maximum number of actions; Define a distribution function, which is used to generate random numbers and is expressed as: in, is the distribution parameter of the distribution function, representing the parameter of the i-th action pulse in the t-th pulse repetition period; Optimizing the distribution function according to the result of transmitting the cover pulse formed by the maximum random number among the random numbers; The optimization of the distribution function is implemented according to the maximum random number among the random numbers, including: Step S123: Obtaining a random number corresponding to each action through the distribution function; Step S124: extracting the maximum random number from the random numbers corresponding to each action; Step S125: extract the maximum random number pair as the pulse width t c , as the cover pulse width and emitted, expressed as: Among them, θ i is the random number corresponding to the i-th action in step S124; Step S126: receiving the echo signal and determining whether a target is detected; Step S127: construct a revenue calculation system based on the target detection result, and obtain the revenue value r from the revenue calculation system t , to achieve distribution function optimization; Determine whether to terminate the optimization. If not, re-execute step S123.

2. The shield pulse design method according to claim 1, characterized in that: The frequency range of the detection signal and the cover signal is defined as: |F-F'|>B, where F is the center frequency of the radar detection signal, F' is the center frequency of the cover pulse, and B is the bandwidth of the interference signal.

3. The shield pulse design method according to claim 1, characterized in that: When loading the MAB decision model, the distribution parameters of the distribution function are initialized, which are expressed as:

4. The shield pulse design method according to claim 3, characterized in that: The construction of the profit calculation system is realized by a profit function, and the profit function is expressed as: Among them, r t Gain revenue for radar target detection results.

5. The shield pulse design method according to claim 4, characterized in that: The method for optimizing the distribution function is: If the return value is 1, the method of updating the distribution parameter of the distribution function is: in, is the parameter of the i-th action pulse in the t-th pulse repetition period; Otherwise, the method to update the distribution parameters of the distribution function is: Where γ is the discount factor and: γ<1.

6. The shield pulse design method according to claim 1, characterized in that: When defining the radar action set, the pulse width t is controlled c and the detection signal pulse width t d The sum is a constant, expressed as: t c +t d =t total , where t total To protect the maximum value that the signal pulse width can reach and the detection signal pulse width can reach.

Citation Information

Patent Citations

  • MAB model-based FAR anti-active suppression interference strategy generation method

    CN115586496A

  • Radar anti-interference method based on multi-arm bandit

    CN117129951A