An unmanned aerial vehicle countermeasure plan generation method based on artificial intelligence technology

Abstract

The application relates to the technical field of data analysis, in particular to a UAV countermeasure plan generation method based on artificial intelligence technology, which comprises the following steps: a database with multiple countermeasure success cases is established; a UAV threat model is established; the threat model is coupled with a current use environment to obtain a coupling model; the matching degree of countermeasures is calculated; an artificial intelligence module is used to train the threat model and the coupling model according to the multiple countermeasure success cases; according to the determined threat model, coupling model and matching degree of UAV behavior countermeasures, the highest corresponding countermeasures are selected and output; the model focuses on the key features of speed, acceleration, mass and external load, the generation time of the plan is shortened, and the response speed is improved; through subsequent environment coupling, the subsequent strategy can be adjusted according to the specific environment; the unit cost, response time and adaptation rate of the countermeasures are calculated, and the optimal allocation of resources is realized.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and specifically to a method for generating drone countermeasure plans based on artificial intelligence technology. Background Technology

[0002] With technological advancements, drones, manned aircraft, and wearable drones are becoming increasingly common, leading to a wider range of aerial applications. While the widespread use of drones supports the rapid development of my country's low-altitude economy, it also brings numerous challenges. As the influence of drones grows, attracting widespread attention, they are increasingly being used for other activities. Therefore, countermeasures against drones are necessary.

[0003] Traditional contingency plans are mostly based on "if-then" rule bases or fixed procedures based on simple threat levels. Once the drone's behavior exceeds the preset rules, the system cannot respond effectively. This "reactive" approach is always a step behind when facing high-speed, highly maneuverable drones. The selection of countermeasures (such as electromagnetic jammers, net-trapping drones, and laser weapons) lacks quantitative basis and relies heavily on experience-based judgment. For example, using high-power laser weapons to counter low-altitude, slow-moving, and non-threatening small drones would result in energy waste and potential light pollution, and some drones might be shot down without needing to be shot down. In other words, existing countermeasure plans cannot achieve optimal matching between the actual situation and the countermeasures; furthermore, they cannot select countermeasures based on the environment. For example, in urban areas, using high-power electromagnetic interference could easily cause collateral damage to other equipment. To address the problems of existing technologies, this application provides a drone countermeasure plan generation method based on artificial intelligence technology that can generate contingency plans by matching drone threats, environment, and countermeasures. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a drone countermeasure plan generation method based on artificial intelligence technology that can generate a plan by matching drone threats, environment, and countermeasures.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A method for generating drone countermeasure plans based on artificial intelligence technology includes:

[0007] Establish a database that stores the shape and weight of drones, the shape and risk level of drone external payloads, and multiple successful countermeasure cases.

[0008] Establish a drone threat model The flight speed v and acceleration a of the drone are obtained, and the weight m and risk factor L of the identified drone are obtained by combining the identification device with the drone information database. = ×( )+ ×( )+ ×L, where and The maximum allowable drone weight and maximum flight speed in the current operating environment;

[0009] The threat model is coupled with the current usage environment to obtain a coupled model T, where T = ×(1+ ×ρ)×(1+ ×α)×(1+ × ), where ρ is the obstacle density of the current operating environment, and α is the electromagnetic environment attenuation coefficient;

[0010] Calculate the matching degree of countermeasures , = ,in, To determine the environmental adaptability of the i-th type of countermeasure, the adaptability of the i-th type of countermeasure to the current usage environment is calibrated based on the current usage environment information. ; Let be the unit cost of the i-th type of countermeasure; Response time of the i-th type of countermeasure;

[0011] The artificial intelligence module takes v, a, m, L, ρ, and α from multiple successful countermeasure cases as inputs to the input layer, and T and M. i The output layer serves as the output of the output layer; the temporal layer uses a bidirectional LSTM to capture the temporal correlation of drone behavior, which consists of flight speed v and acceleration a; the causal attention layer determines the weights of key causal links through an attention mechanism. - ;

[0012] According to the determined - and drone behavior computing Select the highest few Corresponding countermeasures will be implemented.

[0013] Preferably, based on the determined - and drone behavior computing From large to small Sort the results and select the first three. Corresponding countermeasures will be implemented.

[0014] Preferably, the selected countermeasure from the three countermeasures obtained is stored as a successful countermeasure case.

[0015] Preferably, the AI ​​module updates when the number of successful countermeasures reaches a preset value. - .

[0016] Preferably, the identification device is one or more of radar, camera, infrared sensor, multispectral sensor and laser rangefinder.

[0017] Preferably, the identification device includes an image recognition module. After identifying the shape of the UAV and the shape of the UAV's external payload through the image recognition module, the module compares the data with a database and outputs the weight m and risk factor L of the closest UAV.

[0018] Preferably, the flight speed v and acceleration a of the UAV are obtained through the identification device.

[0019] Preferably, after determining the number of obstacles by using a 3D map matched to the current environment, the density of environmental obstacles is then determined by manual verification.

[0020] Preferably, α is obtained through an electromagnetic monitoring device.

[0021] Preferably, ρ and α are updated periodically.

[0022] The beneficial effects of this invention are as follows: Threats are quantified through the dynamic attributes of the drone itself, such as mass, speed, acceleration, and external payload. The identification device can be AI recognition or image recognition, which can quickly determine the shape and risk factor, and thus the mass, thereby enhancing the credibility of the threat model. Particular attention is paid to external payloads. Generally, drones do not have external payloads, but the purpose of external payloads is to enhance certain functions. Traditional drones' built-in cameras are sufficient, while external payloads are likely devices not present on the drone itself, such as explosives, jammers, wired fiber optic cables, detection equipment, and other dangerous or anti-jamming devices. Therefore, the presence of these external devices significantly increases the threat level (T). The model significantly increases efficiency by focusing on key features such as speed, acceleration, mass, and external payload, shortening contingency plan generation time and improving response speed. Furthermore, through subsequent environmental coupling, it can adjust strategies based on specific environments, preventing countermeasures from causing larger incidents. For example, the threat level of the same drone model in residential areas and open areas is completely different under calculation T, while existing technologies consider it the same threat level. Moreover, by calculating the unit cost, response time, and adaptability of countermeasures, it can select one or more solutions with the best cost, response time, and adaptability from among numerous measures, achieving optimal resource allocation and reducing waste. Detailed Implementation

[0023] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments.

[0024] A method for generating drone countermeasure plans based on artificial intelligence technology includes:

[0025] Establish a database that stores the shape and weight of drones, the shape and risk level of drone external payloads, and multiple successful countermeasure cases.

[0026] Establish a drone threat model The flight speed v and acceleration a of the drone are obtained, and the weight m and risk factor L of the identified drone are obtained by combining the identification device with the drone information database. = ×( )+ ×( )+ ×L, where and The maximum allowable drone weight and maximum flight speed in the current operating environment;

[0027] The threat model is coupled with the current usage environment to obtain a coupled model T, where T = ×(1+ ×ρ)×(1+ ×α)×(1+ × ), where ρ is the obstacle density of the current operating environment, and α is the electromagnetic environment attenuation coefficient;

[0028] Calculate the matching degree of countermeasures , = ,in, To determine the environmental adaptability of the i-th type of countermeasure, the adaptability of the i-th type of countermeasure to the current usage environment is calibrated based on the current usage environment information. ; Let be the unit cost of the i-th type of countermeasure; Response time of the i-th type of countermeasure;

[0029] The artificial intelligence module takes v, a, m, L, ρ, and α from multiple successful countermeasure cases as inputs to the input layer, and T and M. i The output layer serves as the output of the output layer; the temporal layer uses a bidirectional LSTM to capture the temporal correlation of drone behavior; the causal attention layer determines the weights of key causal links through an attention mechanism. - ;

[0030] According to the determined - and drone behavior computing Select the highest few Corresponding countermeasures will be implemented.

[0031] As described above, threats can be quantified by analyzing the drone's dynamic attributes, such as mass, speed, acceleration, and external payload. The identification device can be AI-based or image-based, quickly determining the drone's shape and hazard level, thereby determining its mass and enhancing the credibility of the threat model. Particular attention is paid to external payloads. While drones typically lack external payloads, their purpose is to enhance certain functions. Traditional drones' built-in cameras are sufficient, but external payloads are likely devices not present on the drone itself, such as explosives, jammers, wired fiber optic cables, detection equipment, or other dangerous or anti-jamming devices. Therefore, the presence of such external devices significantly increases the threat level (T). Furthermore, by focusing on speed, acceleration, mass, and external payload, the model concentrates on these key characteristics, shortening the time for generating contingency plans and improving response speed. Through subsequent environmental coupling, it can adjust subsequent strategies according to the specific environment, avoiding countermeasures from causing larger accidents. For example, the threat level of the same type of drone in residential areas and open areas is completely different under the T calculation, while existing technology considers it to be the same threat level. Moreover, by calculating the unit cost, response time, and adaptability of countermeasures, it can select one or several solutions with the best cost, response time, and adaptability from among many measures, achieving optimal resource allocation and reducing waste.

[0032] Furthermore, based on the determined - and drone behavior computing From large to small Sort the results and select the first three. Corresponding countermeasures will be implemented.

[0033] As can be seen from the above description, by providing a variety of countermeasures, staff can make more appropriate or accurate judgments based on their experience.

[0034] Furthermore, the selected countermeasure from the three countermeasures obtained is stored as a successful countermeasure case.

[0035] As can be seen from the above description, closed-loop learning enables the output of the contingency plan to be continuously optimized as cases accumulate, thereby improving the success rate of countermeasures.

[0036] Furthermore, once the number of successful countermeasures reaches a preset threshold, the AI ​​module updates. - .

[0037] Furthermore, the identification device is one or more of radar, camera, infrared sensor, multispectral sensor and laser rangefinder.

[0038] Furthermore, the identification device includes an image recognition module. After identifying the shape of the drone and the shape of the drone's external payload through the image recognition module, the module compares the data with a database and outputs the weight m and risk factor L of the drone that is closest to the identification.

[0039] Furthermore, the flight speed v and acceleration a of the drone are obtained through the identification device.

[0040] Furthermore, after determining the number of obstacles by matching the current environment with a 3D map, the density of environmental obstacles is then manually verified.

[0041] As can be seen from the above description, the accuracy of the environmental obstacle density ρ value is ensured by matching the map and then manually verifying it.

[0042] Furthermore, α is obtained through electromagnetic monitoring equipment.

[0043] Furthermore, ρ and α are updated periodically.

[0044] As can be seen from the above description, regular updates can ensure that the model can adapt to changes in the environment, thereby improving accuracy.

[0045] Example 1

[0046] A method for generating drone countermeasure plans based on artificial intelligence technology includes:

[0047] Establish a database that stores the shape and weight of drones, the shape and risk level of drone external payloads, and multiple successful countermeasure cases.

[0048] Establish a drone threat model The flight speed v and acceleration a of the drone are obtained, and the weight m and risk factor L of the identified drone are obtained by combining the identification device with the drone information database. = ×( )+ ×( )+ ×L, where and The maximum allowable drone weight and maximum flight speed in the current operating environment;

[0049] The threat model is coupled with the current usage environment to obtain a coupled model T, where T = ×(1+ ×ρ)×(1+ ×α)×(1+ × ), where ρ is the obstacle density of the current operating environment, and α is the electromagnetic environment attenuation coefficient;

[0050] Calculate the matching degree of countermeasures , = ,in, To determine the environmental adaptability of the i-th type of countermeasure, the adaptability of the i-th type of countermeasure to the current usage environment is calibrated based on the current usage environment information. ; Let be the unit cost of the i-th type of countermeasure; Response time of the i-th type of countermeasure;

[0051] The artificial intelligence module takes v, a, m, L, ρ, and α from multiple successful countermeasure cases as inputs to the input layer, and T and M. i As the output of the output layer; the temporal layer uses a bidirectional LSTM to capture the temporal correlation of drone behavior; drone behavior consists of flight speed v and acceleration a; drone behavior is not isolated, that is, there is a temporal correlation between flight speed v and acceleration a (vt=v0+at). Based on the temporal correlation of successful countermeasures, the next trajectory of the drone can be predicted (such as continuously flying towards the target while the acceleration rapidly increases in a short period of time, inferring its impact intention), avoiding countermeasure lag; combined with the subsequent causal attention layer, the change in acceleration (temporal correlation feature) is regarded as a change in threat level, and the dynamic causal relationship is quantified, making the decision more in line with the real-time behavior logic of the drone. The causal attention layer determines the weight of key causal links through an attention mechanism. - ;

[0052] According to the determined - and drone behavior computing Select the highest few Corresponding countermeasures are then output. The threat level is determined based on the T value: T < 0.5 is low threat, 0.5 ≤ T < 1.2 is medium threat, and T ≥ 1.2 is high threat. Different levels correspond to basic contingency plan frameworks, such as low threat: early warning and expulsion; high threat: forced landing, shooting down.

[0053] According to the determined - and drone behavior computing From large to small Sort the results and select the first three. The corresponding countermeasures are output. The selected countermeasure from the three output countermeasures is stored as a successful countermeasure case. When the number of successful countermeasure cases accumulates to a preset value, the artificial intelligence module updates. - .

[0054] The identification device is one or more of radar, camera, infrared sensor, multispectral sensor, and laser rangefinder. The identification device includes an image recognition module, which identifies the shape of the UAV and its external payload, compares the results to a database, and outputs the weight m and risk factor L of the closest matching UAV.

[0055] The flight speed v and acceleration a of the UAV are obtained through the identification device.

[0056] The number of obstacles is determined by ρ through a 3D map matching the current environment, and then the density of environmental obstacles is determined by manual verification.

[0057] The α value is obtained through electromagnetic monitoring equipment.

[0058] The ρ and α are updated periodically.

[0059] A concert was held at a city sports center, and the area around the sports center was designated as a no-fly zone for drones, which is the current operating environment. Radar or cameras detected a DJI Mini 3 Pro drone (consumer-grade, unregistered) flying into the no-fly zone at an altitude of 80m and a speed of 4m / s. It was initially determined that it was a photography enthusiast who had mistakenly entered the no-fly zone, and a low-threat level countermeasure plan needs to be generated.

[0060] Radar / electro-optical data: speed v=4m / s, heading angle 315°, weight m=0.24kg (obtained by comparison with database), load identification L=0 (no dangerous load, obtained by comparison with database), acceleration Δv / Δt=0 (uniform flight);

[0061] Environmental data: wind speed 2m / s, rainfall 0, obstacle density ρ = 800 obstacles / km² calculated by GIS and real-time image fusion (dense audience area, density calculated after camera capture, obtained after manual verification), electromagnetic attenuation coefficient α = 0.9 (no obvious electromagnetic interference, obtained through electromagnetic monitoring equipment).

[0062] Countermeasures data: Three types of measures are deployed in the area: A (portable electromagnetic jammer, located 1.2km from the drone, power 10W, response time t=2s, cost C=50 yuan / time), B (early warning and deterrence broadcast, located 0.8km from the drone, response time t=1s, cost C=1 yuan / time), and C (netting the drone, located 3km from the drone, response time t=8s, cost C=500 yuan / time).

[0063] Threat level T0 = 0.3 × (0.24 / 50) + 0.4 × (4 / 50) + 0.3 × 0 = 0.3 × 0.0048 + 0.4 × 0.08 + 0 = 0.03344 (close to 0, inherent threat is extremely low); and The maximum weight and maximum flight speed of a drone allowed in the current operating environment are defined by M, which is set to 50 kg, representing the upper limit of the weight of a large industrial-grade drone. Take 50 m / s, referring to the Civil Aviation Administration's drone management standards; , , The weighting coefficients are set to 0.3, 0.4, and 0.3 respectively (these values ​​are set arbitrarily for ease of understanding; the same values ​​are used in Example 2 for the same purpose).

[0064] Calculate the environmental coupling threat level T: T=0.03344×(1+0.5×800 / 10000)×(1-0.4×0.9)×(1+0.6×0)=0.03344×1.04×0.64×1≈0.022 (T<0.5, judged as low threat); , , The weighting coefficients are set to 0.5, 0.4, and 0.6 respectively (these values ​​are set arbitrarily for ease of understanding; the same values ​​are used in Example 2 for the same purpose).

[0065] Calculate the matching degree M of the measures i :

[0066] Warning broadcast B: Environmental adaptation efficiency η=1 (no electromagnetic dependence). =(0.022×1) / (1×1)=0.022;

[0067] Electromagnetic interference device A: η=0.88 (α=0.9, high adaptability). =(0.022×0.88) / (50×2)=0.0001936;

[0068] Net-catching drone C: η=0.9 (high adaptability to open areas). =(0.022×0.9) / (500×8)=0.00000495;

[0069] Matching degree ranking: B>A>C, the optimal measure is early warning and drive-away broadcast, the other two are alternative measures.

[0070] Example 2

[0071] A method for generating drone countermeasure plans based on artificial intelligence technology, the similarities with Embodiment 1 will not be repeated here, the difference being:

[0072] Within the safety isolation zone (radius 1km) of a certain substation, i.e. the current operating environment, an unidentified drone (black fuselage, carrying a rectangular payload) was detected. It was flying at an altitude of 150m and a speed of 12m / s, continuously flying towards the center of the substation. It suddenly accelerated to 5m / s² (suspected dive). It was determined to be a high-threat target and required mandatory handling.

[0073] Radar / electro-optical data: velocity v=12m / s, weight m=8kg (compared to the database, no matching drone model was found, the most similar drone shape was selected and the corresponding weight was selected), payload identification L=1 (suspected explosive), acceleration Δv / Δt=5m / s²;

[0074] Environmental data: wind speed 5 m / s, rainfall 0, obstacle density ρ = 120 obstacles / km² (industrial area, sparse buildings, obtained from GIS map, manually determined during system layout), electromagnetic attenuation coefficient α = 0.7 (weak electromagnetic shielding around the substation).

[0075] Countermeasures: D (high-power electromagnetic jammer, 0.6km from the drone, 1500W power, t=1.5s, C=300 yuan / time), E (laser interception system, 1.2km from the drone, 2000W power, t=3s, C=2 yuan / time), F (incendiary-explosive countermeasure device, 0.8km from the drone, t=2s, C=800 yuan / time).

[0076] Threat level T0 = 0.3 × (8 / 50) + 0.4 × (12 / 50) + 0.3 × 1 = 0.3 × 0.16 + 0.4 × 0.24 + 0.3 = 0.444;

[0077] Calculate the environmental coupling threat level T: T = 0.444 × (1 + 0.5 × 120 / 10000) × (1 - 0.4 × 0.7) × (1 + 0.6 × 5) = 0.444 × 1.006 × 0.72 × 4 ≈ 1.28 (T ≥ 1.2, judged as high threat);

[0078] Calculate the matching degree M of the measures i :

[0079] Electromagnetic interference device D: η=0.75 (α=0.7, moderate adaptability). =(1.28×0.75) / (300×1.5)=0.96 / 450≈0.00213;

[0080] Laser system E: η=0.95 (no electromagnetic dependence, high adaptability). =(1.28×0.95) / (2×3)=1.216 / 6≈0.202;

[0081] Explosion-prone device F: η=0, because it violates the rule that "flammable and explosive materials are prohibited from being stored in the area surrounding the substation", so it is eliminated in advance by taking 0;

[0082] Matching degree ranking: E>D>F, the optimal resource is laser system E, and the alternative measure is electromagnetic interference device D.

[0083] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for generating drone countermeasure plans based on artificial intelligence technology, characterized in that, include: Establish a database that stores the shape and weight of drones, the shape and risk level of drone external payloads, and multiple successful countermeasure cases. Establish a drone threat model The flight speed v and acceleration a of the drone are obtained, and the weight m and risk factor L of the identified drone are obtained by combining the identification device with the drone information database. The identification device has an image recognition module. The image recognition module identifies the shape of the UAV and the shape of the UAV's external load and compares them with the database. It then selects the weight m and the risk factor L of the closest UAV and outputs them. When the comparison with the database shows that there is no dangerous load, L=0. When a load is identified, L=1. = ×( )+ ×( )+ ×L, where and The maximum allowable drone weight and maximum flight speed in the current operating environment; The threat model is coupled with the current usage environment to obtain a coupled model T, where T = ×(1+ ×ρ)×(1+ ×α)×(1+ × ), where ρ is the obstacle density of the current operating environment, and α is the electromagnetic environment attenuation coefficient; Calculate the matching degree of countermeasures , = ,in, To determine the environmental adaptability of the i-th type of countermeasure, the adaptability of the i-th type of countermeasure to the current usage environment is calibrated based on the current usage environment information. ; Let be the unit cost of the i-th type of countermeasure; Response time of the i-th type of countermeasure; The artificial intelligence module takes v, a, m, L, ρ, and α from multiple successful countermeasure cases as inputs to the input layer, and T and M. i The output layer serves as the output of the output layer; the temporal layer uses a bidirectional LSTM to capture the temporal correlation of drone behavior, which consists of flight speed v and acceleration a; the causal attention layer determines the weights of key causal links through an attention mechanism. - ; According to the determined - and drone behavior computing Select the highest number of Corresponding countermeasures will be implemented.

2. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, According to the determined - and drone behavior computing From large to small Sort the results and select the first three. Corresponding countermeasures will be implemented.

3. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 2, characterized in that, The selected countermeasure from the three countermeasures output is stored as a successful countermeasure case.

4. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 3, characterized in that, Once the number of successful countermeasures reaches a preset threshold, the AI ​​module will update. - .

5. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, The identification device is one or more of radar, camera, infrared sensor, multispectral sensor and laser rangefinder.

6. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, The flight speed v and acceleration a of the UAV are obtained through the identification device.

7. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, The number of obstacles is determined by ρ through a 3D map matching the current environment, and then the density of environmental obstacles is determined by manual verification.

8. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, The α value is obtained through electromagnetic monitoring equipment.

9. The method for generating drone countermeasure plans based on artificial intelligence technology according to claim 1, characterized in that, The ρ and α are updated periodically.

Images