Optimization method for ship formation cooperative air electromagnetic countermeasure

Abstract

The application discloses a ship formation cooperative air electromagnetic countermeasure optimization method, comprising: identifying intercepted air radiation source signals to obtain a working mode sequence; calculating mode filtering probability by using a hidden Markov model, and then inferring enemy tactical intention by fusing motion characteristics through a Bayesian network; establishing a cooperative jamming optimization model with the minimum comprehensive detection probability as a target, and imposing electromagnetic compatibility and radiation safety constraints; adopting a federal security reinforcement learning framework, each ship locally generates a jamming action and executes the jamming action after being corrected by a security projection layer, and a global strategy is obtained through federal aggregation; and implementing jamming according to the strategy and evaluating the effect to form a closed loop. The application organically fuses intention reasoning and security reinforcement learning, can realize predictive and adaptive optimization of formation cooperative jamming resources under the premise of ensuring electromagnetic compatibility safety, and significantly improves countermeasure efficiency.

Description

Technical Field

[0001] This invention relates to the field of electronic warfare technology, specifically to an optimized method for coordinated air-to-air electromagnetic warfare in naval formations. Background Technology

[0002] In modern naval warfare, the aerial threats faced by naval formations are becoming increasingly complex. Enemy forces often utilize multiple radiation sources, such as early warning radar, fire control radar, and terminal guidance radar, to work in concert to complete tactical missions including search, tracking, attack guidance, and terminal strikes. To enhance survivability, formations need to implement coordinated electromagnetic interference to disrupt or weaken the enemy's radar network's detection and locking capabilities.

[0003] Existing cooperative jamming methods are mostly based on preset rules or centralized optimization scheduling. Preset rules are difficult to adapt to dynamically changing threat patterns; centralized optimization relies on reliable central nodes and global information, which have single-point vulnerabilities in strong adversarial environments and are limited in real-time performance; some reinforcement learning-based methods can learn online, but most do not consider the strict constraints of electromagnetic compatibility between ships, which may lead to mutual interference or self-interference and cannot guarantee the safety of actual deployment; at the same time, these methods lack deep reasoning about the enemy's tactical intentions, and the strategies are often just passive responses, lacking foresight and resulting in low utilization of jamming resources.

[0004] Therefore, there is an urgent need for an optimized method for coordinated air-to-air electromagnetic countermeasures of naval formations that can both infer enemy intentions in real time and drive collaborative decision-making, while ensuring electromagnetic compatibility and safety. Summary of the Invention

[0005] To address this, the present invention provides an optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations, thereby resolving the problems in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An optimized method for cooperative air-to-air electromagnetic countermeasures in naval formations includes the following steps:

[0008] Step 1: Sort and identify the airborne radiation source signals intercepted by each shipborne electronic support system to obtain the radiation source type, working mode and mode switching sequence;

[0009] Step 2: Use a hidden Markov model to model the operating mode sequence of each radiation source and calculate the filtering probability of each radiation source being in a different operating mode at the current time.

[0010] Step 3: Using a Bayesian network, the filtered probability is used as soft evidence, and combined with the kinematic characteristics of the radiation source platform, the posterior probability distribution of the enemy platform's tactical intent is inferred. The tactical intent includes at least search, tracking, guided attack, and terminal guidance.

[0011] Step 4: Using the posterior probability distribution of the enemy platform's tactical intent as the suppression weight of each radiation source, and with the goal of minimizing the comprehensive detection probability of the enemy's multi-sensor network on the formation, establish a distributed cooperative interference optimization model.

[0012] Step 5: Solve the cooperative interference optimization model using the federated security reinforcement learning framework and generate security actions;

[0013] Step 6: Each ship generates and transmits jamming signals according to the aforementioned safety actions;

[0014] Step 7: Evaluate the interference effect in real time. If the overall detection probability drops below the threshold or the threat is eliminated, the process ends; otherwise, return to step 2 to continue reasoning and optimization.

[0015] Furthermore: the decision variables of the cooperative interference optimization model include the interference beam direction, interference power, interference pattern and time slot allocation of each ship; the constraints of the cooperative interference optimization model include the upper limit of interference resources of each ship, the protection frequency band of friendly frequency equipment, the electromagnetic radiation safety margin between ships and the priority suppression constraint based on intent probability.

[0016] Furthermore, the nodes of the Bayesian network include the enemy platform's intent, the operating mode status of each radiation source, the trend of route shortcut changes, and the platform's altitude / speed changes. The conditional probability table is constructed based on tactical rules, and the inference uses a joint tree algorithm.

[0017] Furthermore: a deep deterministic policy gradient algorithm is adopted, with each ship maintaining an Actor network and a Critic network; during training, the Critic network acquires global observations and joint actions, while only local observations are needed during execution; to satisfy constraints, a safety constraint projection layer is inserted between the Actor network output and execution.

[0018] Furthermore: the security constraint projection layer performs at least one of the following modifications:

[0019] Frequency constraints: If the interference frequency band overlaps with the protected frequency band, switch the interference mode or tune to a safe frequency band.

[0020] The equivalent interference strength constraint is used to estimate the equivalent received interference strength of the interference signal at the receiving antenna of the adjacent ship using the following formula. ;

[0021] ;

[0022] in, This refers to the jammer's transmission power. d represents the gain of the jamming antenna; d represents the distance between ships. Increase the receiving antenna gain of the neighboring ship;

[0023] like Exceeding the preset equivalent interference strength safety threshold Then adjust the transmission power. Perform geometric reduction;

[0024] Beam interference constraint: if the angle between the direction of the interfering beam and the direction of the friendly ship is less than the protection angle, then rotate the beam to outside the protection angle.

[0025] Furthermore: the local observations include the ship's distance and bearing relative to the radiation source, the radiation source's probability of intent, the location of neighboring ships, and the ship's available jamming resources.

[0026] Furthermore, the state set of the Hidden Markov Model includes search, scan-while-track, single-target tracking, continuous wave illumination, and silence. The observation matrix incorporates ESM identification confidence, and the forward algorithm is used for online filtering.

[0027] This invention has the following advantages: It combines Hidden Markov Models with Bayesian Networks to infer the enemy's tactical intentions online from the working mode sequence of radiation sources and the motion characteristics of platforms, making interference decisions predictable and significantly improving resource utilization efficiency; It adopts a federated security reinforcement learning framework, which not only realizes multi-ship distributed collaborative decision-making and avoids the bottleneck of the central node, but also eliminates interference actions that violate electromagnetic compatibility and radiation safety through a security constraint projection layer, ensuring that the method can be safely deployed in real formations.

[0028] Other features and advantages of the present invention will be set forth in the following description. Attached Figure Description

[0029] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0030] Figure 1 The flowchart illustrates the implementation of the optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations, as provided in this application embodiment.

[0031] Figure 2 This is a flowchart illustrating the online reasoning of enemy tactical intentions in the optimized method for coordinated air-to-air electromagnetic countermeasures of naval formations provided in this application embodiment. Detailed Implementation

[0032] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these embodiments are merely for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make some non-essential improvements and adjustments to the present invention based on the above-described content.

[0033] Please see Figure 1 The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations includes the following steps:

[0034] S1: Electromagnetic environment awareness; each ship's ESM intercepts airborne electromagnetic signals.

[0035] S2: Signal sorting and identification; signal sorting, modulation identification and fingerprint feature extraction to obtain radiation source type, working mode and mode sequence;

[0036] S3: Formation situation fusion; by sharing identification results through the formation data link, a unified radiation source situation map is constructed, including azimuth, distance, movement trajectory and working status;

[0037] S4: Online reasoning of tactical intentions, outputting the probability distribution of intentions;

[0038] See Figure 2 The online reasoning of enemy tactical intent in step S4 uses the working mode switching sequence of airborne radiation sources to infer the current tactical intent of the enemy platform; its reasoning framework consists of a bottom-level Hidden Markov Model (HMM) and an upper-level Bayesian Network (BN).

[0039] Among them, the Hidden Markov Model (HMM) models the operating mode sequence of a single radar; specifically:

[0040] State set Corresponding modes include search, scan-while-tracking, single-target tracking, continuous wave illumination, and silent operation; observation set Pattern recognition labels output in real time by ESM; model parameters Among them, the state transition matrix A is calibrated offline by the radiation source working logic; the observation probability matrix B represents the identification confidence level; and the initial distribution... Take a uniform or prior value; use a forward algorithm to recursively calculate the probability of the partial observation sequence at time t, and obtain the filtering probability of the current working mode. .

[0041] The upper-level Bayesian network can fuse the HMM state probabilities of multiple radars, platform kinematic characteristics, and tactical rules to infer the overall platform intent; its nodes include:

[0042] Enemy platform intentions Radar A mode status (HMM output probability distribution); Radar B mode status Target flight path shortcut change trend C (shrinking / stabilizing / increasing); altitude / speed change M; missile launch characteristic signal L (if any).

[0043] The Conditional Probability Table (CPT) is constructed based on expert knowledge. For example, if the early warning radar is scanning while tracking, the fire control radar is single-target tracking, and the flight path shortcut is narrowed, then the probability of "guided attack" is extremely high. The discrete probability values ​​output by the input Hidden Markov Model (HMM) are used as soft evidence, and the posterior probability of intent is calculated using the belief propagation algorithm. .

[0044] S5: Establish a distributed cooperative interference optimization model with the goal of minimizing the overall detection probability, and impose resource, frequency avoidance and electromagnetic compatibility constraints.

[0045] S6: Distributed policy solving and security action generation based on federated security reinforcement learning;

[0046] S7: Implement coordinated jamming and transmit jamming signals;

[0047] S8: Real-time evaluation of the interference effect, calculation of the decrease in detection probability as a reward, end if the threat is eliminated, otherwise return to S3.

[0048] In this embodiment, the specific steps for online reasoning of the enemy's tactical intentions are as follows:

[0049] (1) Offline preparation: For each type of known radiation source, construct the A and B matrices of HMM through intelligence or simulation; construct the Bayesian network topology and CPT, and solidify them into the inference engine.

[0050] (2) Online input acquisition: the current working mode labels of each radar identified by ESM, along with confidence level; target kinematic parameters (route shortcuts, altitude, speed changes, etc.) provided by the fusion center.

[0051] (3) HMM recursion: Perform the forward algorithm for each radar to calculate the current filtering probability vector.

[0052] (4) BN inference: The HMM probability vectors of each radar are used as soft evidence. After discretization of the kinematic features, they are input into BN and the joint tree algorithm is called to calculate the posterior distribution of intent.

[0053] (5) Output: Select the maximum a posteriori probability intention as the reasoning result, and output the probability distribution for subsequent model optimization as weights.

[0054] To explain the implementation of step S4 in more detail, we now assume that the enemy has an early warning aircraft and several fighter jets, with the early warning radar operating in the S-band and the fighter jet fire control radar operating in the X-band.

[0055] First, a Hidden Markov Model (HMM) is constructed offline for each type of radar; the state set includes typical modes such as search, scan-while-track, single-target tracking, and continuous wave illumination; the state transition matrix is ​​calibrated from intelligence data or simulation; and the observation matrix incorporates ESM to identify confidence levels.

[0056] During the online phase, for each mode label output by the ESM, the forward algorithm is used to update the mode filtering probabilities of each radar at the current time. .

[0057] Next, a Bayesian network (BN) is constructed; its parent node is the enemy platform intent, and the child nodes include the status of the early warning radar, the status of the fire control radar, the trend of changes in the flight shortcut (shrinking / stabilizing / increasing), and the characteristics of changes in altitude / speed.

[0058] The conditional probability table is set according to the tactical rules; for example, when the early warning radar is scanning and tracking simultaneously and the fire control radar is tracking a single target, and the flight path shortcut is narrowed, the probability of guiding the attack intention is set to 0.9.

[0059] During inference, the filtered probability output by the HMM is used as soft evidence input, and continuous variables such as route shortcuts are discretized and used as hard evidence. The joint tree inference algorithm is used to calculate the posterior distribution of intent. Finally, the maximum probability intent and the complete probability vector are output. This probability vector is directly used in step S5 to give higher suppression weights to high-threat intent targets.

[0060] In the process of solving and executing the federal security reinforcement learning in this embodiment, in order to explain steps S5 and S6 in more detail, it is assumed that each ship in the formation is defined as an intelligent agent, whose local observations include the distance, azimuth, hostile map probability, position of neighboring ships and available jamming resources of the ship relative to each radiation source; the action space is a four-dimensional continuous / discrete hybrid vector, namely beam azimuth angle, elevation angle, transmit power, jamming pattern number and time slot allocation.

[0061] The DDPG (Deep Deterministic Policy Gradient) algorithm is used to maintain the Actor network for each ship. and Critic Network During training, the Critic can obtain global observations and joint actions, while only local observations are needed during execution. To satisfy constraints, a safety constraint projection layer is inserted between the Actor's output and execution.

[0062] The projection rules include: frequency constraints, equivalent interference intensity constraints, and beam interference constraints.

[0063] Among them, frequency constraint: check whether the center frequency and bandwidth of the selected interference pattern conflict with the formation protection frequency list. If there is a conflict, automatically select the alternative pattern with a frequency separation greater than the preset value, or perform real-time frequency conversion control on the ship's interference exciter.

[0064] The equivalent interference strength constraint is used to estimate the equivalent received interference strength of the interference signal at the receiving antenna of the adjacent ship using the following formula. ;

[0065] ;

[0066] in, This refers to the jammer's transmission power. d represents the gain of the jamming antenna; d represents the distance between ships. Increase the receiving antenna gain of the neighboring ship;

[0067] like Exceeding the preset equivalent interference strength safety threshold Then adjust the transmission power. Reduce proportionally.

[0068] Beam interference constraint: Calculate the angle between the pointing vector of the interfering beam and the relative vectors of all friendly ships. If the minimum angle is less than the protection angle... , If so, adjust the beam azimuth or elevation angle to outside the nearest protection boundary.

[0069] The safety action is obtained after projection. Then it is handed over to the electronic attack system for execution;

[0070] The environmental feedback reward is defined as: the decrease in the sum of the instantaneous detection probabilities of any ship in the formation by multiple enemy sensors, as assessed by the formation radar detection system after the jamming is implemented; each ship stores experience tuples and performs local network updates;

[0071] Every K rounds, each ship encrypts and uploads its Actor network parameters to the formation command node, which then calculates the federated average. And issued, Where M represents the number of ships participating in the federation; Let be the local Actor network parameters for the i-th ship; thus enabling each ship to quickly learn local strategies and share collaborative knowledge through aggregation.

[0072] During online operation, once the enemy switches radar mode or changes its intent, the intent reasoning module outputs a new probability distribution, the state of the reinforcement learning module is updated accordingly, and the policy network is continuously adjusted online until the enemy detection probability is reduced to below the safe threshold.

[0073] This invention uses the tactical intent implied by the enemy's radiation source operating mode sequence as a priori driver for interference decision-making, transforming passive response into proactive anticipation. This makes the collaborative strategy (safe actions) both targeted and predictive, significantly improving the rationality of resource utilization compared to existing rule-based or simple reinforcement learning methods. By embedding an electromagnetic compatibility and radiation safety constraint layer within the distributed reinforcement learning framework, the risk of mutual interference and self-interference in multi-ship collaboration is eliminated from a mechanistic perspective, providing a practical guarantee for the safe deployment of online learning algorithms in real formation environments.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations, characterized in that, Includes the following steps: Step 1: Sort and identify the airborne radiation source signals intercepted by each shipborne electronic support system to obtain the radiation source type, working mode and mode switching sequence; Step 2: Use a hidden Markov model to model the operating mode sequence of each radiation source and calculate the filtering probability of each radiation source being in a different operating mode at the current time. Step 3: Using a Bayesian network, the filtered probability is used as soft evidence, and combined with the kinematic characteristics of the radiation source platform, the posterior probability distribution of the enemy platform's tactical intent is inferred. The tactical intent includes search, tracking, guided attack, and terminal guidance. Step 4: Using the posterior probability distribution of the enemy platform's tactical intent as the suppression weight of each radiation source, and with the goal of minimizing the comprehensive detection probability of the enemy's multi-sensor network on the formation, establish a distributed cooperative interference optimization model. Step 5: Solve the cooperative interference optimization model using the federated security reinforcement learning framework and generate security actions; Step 6: Each ship generates and transmits jamming signals according to the aforementioned safety actions; Step 7: Evaluate the interference effect in real time. If the overall detection probability drops below the threshold or the threat is eliminated, the process ends; otherwise, return to step 2 to continue reasoning and optimization.

2. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 1, characterized in that, The decision variables of the cooperative interference optimization model include the direction of each ship's interference beam, interference power, interference pattern, and time slot allocation; the constraints of the cooperative interference optimization model include the upper limit of each ship's interference resources, the protection frequency band of friendly frequency equipment, the electromagnetic radiation safety margin between ships, and the priority suppression constraint based on intent probability.

3. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 1, characterized in that, The nodes of the Bayesian network include the enemy platform's intent, the operating mode status of each radiation source, the trend of changes in flight shortcuts, and the changes in platform altitude / speed. The conditional probability table is constructed based on tactical rules, and the inference uses a joint tree algorithm.

4. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 1, characterized in that, A deep deterministic policy gradient algorithm is adopted, and each ship maintains an Actor network and a Critic network. During training, the Critic network acquires global observations and joint actions, while only local observations are needed during execution. To satisfy constraints, a safety constraint projection layer is inserted between the Actor network output and execution.

5. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 4, characterized in that, The security constraint projection layer performs at least one of the following corrections: Frequency constraints: If the interference frequency band overlaps with the protected frequency band, switch the interference mode or tune to a safe frequency band. The equivalent interference strength constraint is used to estimate the equivalent received interference strength of the interference signal at the receiving antenna of the adjacent ship using the following formula. ; ； in, This refers to the jammer's transmission power. d represents the gain of the jamming antenna; d represents the distance between ships. Increase the receiving antenna gain of the neighboring ship; like Exceeding the preset equivalent interference strength safety threshold Then adjust the transmission power. Perform geometric reduction; Beam interference constraint: if the angle between the direction of the interfering beam and the direction of the friendly ship is less than the protection angle, then rotate the beam to outside the protection angle.

6. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 4, characterized in that, The local observations include the ship's distance and bearing relative to the radiation source, the probability of the radiation source's intent, the location of neighboring ships, and the available jamming resources for the ship.

7. The optimized method for coordinated air-to-air electromagnetic countermeasures in naval formations according to claim 1, characterized in that, The state set of the Hidden Markov Model includes search, scan-while-track, single-target tracking, continuous wave illumination, and silence. The observation matrix incorporates the ESM identification confidence level, and the forward algorithm is used for online filtering.

Images