Unmanned aerial vehicle countering system and method based on multi-sensor fusion

Abstract

The invention provides an unmanned aerial vehicle countering system and method based on multi-sensor fusion, and relates to the technical field of control systems, and the system comprises a detection sensor, a countering device, an excitation signal library, a white list data set and a processor. The processor comprises a fusion sensing unit, a hypothesis management unit and a decision control unit. And the fusion sensing unit performs time synchronization, space registration and multi-frame time sequence fusion on radio frequency, radar, electro-optical / infrared and acoustic data to generate a target track with confidence weight. And the hypothesis management unit establishes at least two parallel initial hypotheses based on the target trajectory, performs credibility updating in combination with behavior response data after the detection excitation signal acts, and outputs verified hypotheses. The decision control unit selects a detective excitation signal to form a cognitive-excitation-observation closed-loop query, generates a terminal countering strategy in combination with the environmental situation information and the white list data set when a verified hypothesis is obtained, and controls the countering device to perform soft killing or hard killing treatment.

Description

Technical Field

[0001] This invention relates to the field of control system technology, specifically to a UAV countermeasure system and method based on multi-sensor fusion. Background Technology

[0002] The widespread adoption of small, low-altitude, and low-cost drones has significantly increased the risk of them being illegally used for cross-border reconnaissance, interfering with civilian activities, disrupting public order, or even carrying dangerous payloads. To address these risks, existing drone countermeasures primarily rely on a passive detection and direct suppression approach. This involves using radar or electro-optical imaging to detect suspicious targets, and then directly triggering high-power jamming, forced landings, or physical destruction once the target enters a pre-designated no-fly zone or approaches a sensitive area. This approach suffers from the following significant shortcomings.

[0003] Existing technologies typically classify targets based on static or passive information such as shape, flight path, and positional relationships, lacking direct means of detecting the target's control logic. For example, conventional consumer-grade civilian drones, friendly drones operating on authorized routes, and high-threat platforms modified to resist interference may be similar in appearance, flight altitude, or speed. Relying solely on passive observation can easily lead to misjudgment. Such misjudgment could not only result in over-strikes against friendly, duty, or civilian targets, but also in delayed responses to genuine high-threat targets.

[0004] Current countermeasures are often "single-step" decisions. That is, once a threat is identified, a rapid switch to high-intensity, irreversible suppression measures is made, including high-power jamming, directed energy beam irradiation, or physical interception. This approach lacks an intermediate, controlled "questioning and confirmation" step, failing to prove malicious intent and failing to reflect the principle of proportionality in escalation. In scenarios such as civilian airspace, densely populated areas, and security zones for major events, this simplistic and brutal rapid hard-interception strategy carries significant collateral risks, such as falling debris, interference with unrelated communication links, and accidental damage to authorized drones, which are unacceptable from both legal and security perspectives.

[0005] In summary, existing technologies lack the following system capabilities: the ability to fuse multiple sensor outputs to obtain target trajectories with confidence weights; the ability to proactively apply controllable probing stimulus signals to targets and perform inference verification based on the target's behavioral responses; the ability to gradually converge to a high-confidence threat conclusion during multiple rounds of interrogation-response; the ability to implement differentiated, traceable, and compliant countermeasure strategies by integrating environmental situation information and whitelist data sets during the terminal countermeasure phase; and the ability to perform priority scheduling in multi-target concurrent scenarios. Summary of the Invention

[0006] Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a UAV countermeasure system and method based on multi-sensor fusion, which solves the problems of existing technologies.

[0008] Technical solution

[0009] To achieve the above objectives, the present invention provides the following technical solution: a UAV countermeasure system based on multi-sensor fusion, comprising:

[0010] Detection sensors, including those that collect multimodal detection data of the target airspace;

[0011] Countermeasures include transmitting probing excitation signals to the target drone and executing terminal countermeasures;

[0012] Excitation signal library, used to store a variety of preset probing excitation signals;

[0013] The whitelist data set is used to store the identity fingerprint information of authorized or friendly drones;

[0014] The processor is connected to the detection sensor, the countermeasure device, the excitation signal library, and the whitelist data set.

[0015] The processor includes a fusion perception unit, which performs time synchronization, feature alignment, filtering, and target association processing on multimodal detection data from the detection sensors to generate the spatial position, speed, heading trend, maneuvering mode, and corresponding confidence weight of the target UAV, and forms the target trajectory of the target UAV based on these data.

[0016] The assumption management unit includes establishing at least two parallel initial hypotheses for the attributes of the target UAV based on the target trajectory, the attributes including control logic, identity attributes and threat level, and updating the confidence, posterior probability or trust function of each hypothesis based on the behavioral response data of the target UAV to the probing stimulus signal, thereby outputting a verified hypothesis.

[0017] The decision control unit includes: selecting a probing stimulus signal from the stimulus signal library to distinguish the at least two parallel initial hypotheses and controlling the countermeasure device to directionally transmit the probing stimulus signal to the target UAV; acquiring and organizing behavioral response data for the hypothesis management unit to update the hypothesis; determining a terminal countermeasure strategy based on verified hypotheses, environmental situation information, and security constraint rules, and controlling the countermeasure device to execute the terminal countermeasure strategy; and implementing friendly exemption constraints on the terminal countermeasure strategy based on a whitelist data set to suppress irreversible hard-kill measures when an authorized UAV is identified.

[0018] Preferably, the detection sensor includes:

[0019] Radio frequency sensing components are used to detect and analyze the remote control link spectrum characteristics, telemetry signal protocols, and frequency agility behavior of target UAVs.

[0020] A radar sensing component, including at least one of an active radar structure or a passive radar, is used to output range, azimuth, pitch and radial velocity information of a target UAV.

[0021] Electro-optic / infrared imaging components are used to acquire the attitude, body geometry, and visible or infrared radiation characteristics of the target UAV.

[0022] Acoustic acquisition components are used to collect propeller noise from a target UAV and infer its propulsion status.

[0023] The fusion sensing unit of the control system includes performing multi-frame temporal fusion on the above-mentioned types of detection data to generate a fusion state quantity with confidence weights. The fusion state quantity is used to reduce the probability of tracking loss when any type of detection sensor is occluded, saturated, or deceived.

[0024] Preferably, the excitation signal library includes:

[0025] Low-power probing jamming signals targeting specific communication frequency bands are used to observe the link robustness and anti-jamming strategies of target UAVs.

[0026] Weak navigation deception signals that guide target drones to yaw, decelerate, hover, or make forced landings;

[0027] A limited-intensity directional laser irradiation signal that triggers the obstacle avoidance or optoelectronic protection response of the target UAV.

[0028] When selecting the probe excitation signal for transmission, the decision control unit of the control system considers the following:

[0029] (i) Discriminant power to distinguish different parallel initial hypotheses,

[0030] (ii) The collateral costs to civil airspace and surrounding communication systems.

[0031] (iii) Compliance constraint rules.

[0032] Preferably, the behavioral response data includes one or more of the following:

[0033] The abrupt change in the kinematic state of the target UAV within a preset time window, including a sharp change in attitude angular rate, sudden deceleration, track deviation, or a significant increase in descent rate;

[0034] Changes in the radio frequency signal characteristics of the target drone, including increased frequency agility, switching of uplink or downlink communication protocols, interruption of data link heartbeat, or increased encryption level of telemetry messages;

[0035] Rapid adjustment of the target drone's body attitude or sensor orientation, including gimbal orientation change, obstacle avoidance activation, or optical masking action;

[0036] The target drone's proactive response or countermeasures include broadcasting identification declaration signals or transmitting jamming pulses;

[0037] The hypothesis management unit of the control system processes the behavioral response data using Bayesian inference, evidence synthesis methods based on evidence theory, and / or decision criteria based on minimizing information entropy, and introduces a forgetting factor that decreases over time, thereby updating the posterior probability, trust function, or confidence interval of the at least two parallel initial hypotheses.

[0038] Preferably, the decision control unit further includes:

[0039] When the hypothesis management unit fails to output a verified hypothesis that meets the preset confidence threshold, it re-evaluates the stimulus signal library based on the expected information gain, selects a different probe stimulus signal than before, and controls the countermeasure device to emit again, so as to form a closed-loop interrogation mechanism of cognition-stimulation-observation.

[0040] The countermeasures against multiple candidate terminals are scored based on a joint optimization objective function. The candidate terminal countermeasures include soft-kill strategies and hard-kill strategies. The soft-kill strategies include continuous directional jamming, navigation deception, link takeover, or forced landing guidance. The hard-kill strategies include high-directional beam illumination, kinetic interception, or deployment of net-and-tow capture structures.

[0041] The joint optimization objective function of the control system simultaneously considers: (i) the threat level corresponding to the verified assumptions, (ii) the estimated collateral damage costs, (iii) regulatory and security boundary constraints, and (iv) the resource consumption costs.

[0042] Furthermore, when generating the terminal countermeasure strategy, the strategy is constrained by environmental situation information. Specifically, when the predicted flight path of the target UAV points to a sensitive area on the ground, the priority of rapid disabling measures is increased. When the target UAV is near a civilian flight path, physical interception that would generate high collateral risks is suppressed, and soft kill measures that would not cause debris to scatter are given priority.

[0043] When the whitelist data set indicates that the target drone is flying on a friendly or authorized route, it automatically prevents irreversible hard kill measures and only outputs alarms or soft restriction operation instructions.

[0044] A preferred method for countering drones based on multi-sensor fusion includes the following steps:

[0045] Sp1: Receives multimodal detection data from the detection sensor, detects, locates and tracks the target UAV, and generates the target trajectory of the target UAV;

[0046] Sp2: Based on the target trajectory, establish at least two parallel initial hypotheses about the attributes of the target UAV, the attributes including at least control logic, identity attributes, and threat level;

[0047] Sp3: Select a probing excitation signal from a library of excitation signals, designed to distinguish the at least two parallel initial hypotheses;

[0048] Sp4: Send a control command to a countermeasure device, causing the countermeasure device to transmit the detection excitation signal in a directional manner toward the target UAV;

[0049] Sp5: Collect behavioral response data of the target UAV to the probing excitation signal, wherein the behavioral response data includes at least one of kinematic response, radio frequency communication response and / or attitude optical response;

[0050] Sp6: Based on the behavioral response data, perform multi-hypothesis testing on the at least two parallel initial hypotheses, and update the confidence, posterior probability or confidence function of each hypothesis to obtain the verified hypothesis with the highest confidence.

[0051] Sp7: Based on the verified assumptions, combined with environmental situation information and security constraints, determine a terminal countermeasure strategy, and control the countermeasure device to execute the terminal countermeasure strategy, wherein irreversible hard-kill measures are suppressed against target UAVs identified as friendly or authorized flight routes.

[0052] Preferably, Sp1 includes:

[0053] It integrates radio frequency sensing data, radar sensing data, electro-optic / infrared imaging data, and acoustic acquisition data;

[0054] The data is subjected to time synchronization, coordinate alignment, filtering, target association, and consistency verification.

[0055] The system tracks and outputs the target UAV's three-dimensional position, velocity, heading trend, and maneuvering mode through multi-frame time-series tracking, and assigns confidence weights to the above outputs to reduce the probability of tracking loss due to the failure of a single detection source.

[0056] Preferably, the Sp3 includes:

[0057] Calculate the expected information gain metric for the at least two parallel initial hypotheses;

[0058] Based on the expected information gain index, a probing excitation signal that can maximize the hypothetical discriminative power under the expected minimum excitation power is selected from the excitation signal library;

[0059] The probing excitation signal is compared with compliance constraints, civil airspace impact restrictions, and surrounding communication system interference restrictions to eliminate excitation signals that do not meet the restrictions.

[0060] Preferably, the Sp6 includes:

[0061] Using Bayesian inference methods and / or evidence synthesis methods based on evidence theory, the behavioral response data obtained from Sp5 is used as new evidence input to update the posterior probability, trust function, or confidence interval of the at least two parallel initial hypotheses;

[0062] Introduce a time-decreasing forgetting factor to reduce the influence of historical evidence;

[0063] When the confidence level of at least one hypothesis is higher than a preset threshold, the hypothesis is output as the verified hypothesis.

[0064] When the confidence level of all hypotheses fails to reach the preset threshold, return to execute Sp3, select a different exploratory stimulus signal than the previous one, and execute Sp4 and Sp5 again to form a closed-loop inquiry process of cognition-stimulation-observation.

[0065] Preferably, the Sp7 includes:

[0066] Multiple candidate terminal countermeasure strategies are scored and ranked based on a joint optimization objective function. The candidate terminal countermeasure strategies include soft-kill strategies and hard-kill strategies. The soft-kill strategies include continuous directional jamming, navigation deception, link takeover, or forced landing guidance. The hard-kill strategies include high-directional beam illumination, kinetic interception, or deployment of net-and-tether capture structures.

[0067] The joint optimization objective function simultaneously considers: (i) the threat level corresponding to the verified hypothesis, (ii) the estimated collateral damage cost, (iii) regulatory and security boundary constraints, and (iv) the resource consumption cost;

[0068] The terminal countermeasure strategy is constrained by combining environmental situation information. When the current or predicted flight path of the target UAV points to a sensitive area on the ground, the priority of rapid disabling measures is increased. When the target UAV is near a civilian flight path, physical interception that will cause high collateral danger is suppressed and soft kill measures that will not cause debris to scatter are given priority.

[0069] The identity fingerprint information of the target drone is compared with the whitelist data set. When the target drone is instructed to fly on a friendly or authorized route, irreversible hard kill measures are prohibited, and only risk warnings or soft restriction operation instructions are output.

[0070] Beneficial effects

[0071] This invention provides a UAV countermeasure system and method based on multi-sensor fusion. It has the following beneficial effects:

[0072] 1. This invention integrates data from radio frequency sensing, radar sensing, electro-optic / infrared imaging, and acoustic acquisition, and employs time synchronization, spatial registration, and multi-frame temporal fusion to output a target trajectory with confidence weights. This significantly improves the continuous tracking stability of small, low-altitude, and maneuverable UAVs and reduces the probability of loss when a single sensor is blocked, saturated, or deceived.

[0073] 2. This invention introduces a parallel initial hypothesis and a closed-loop interrogation-based identification mechanism. The decision control unit selects probing stimulus signals (low-power communication interference, slight navigation bias, directional light / laser amplitude limiting stimulation, etc.) from a stimulus signal library to induce distinguishable behavioral responses from the target UAV. The hypothesis management unit updates the credibility of each hypothesis based on the behavioral response data using Bayesian inference, evidence theory, or the information entropy minimization criterion, combined with a time-decreasing forgetting factor, until a verified hypothesis is output. This achieves highly reliable identification of control logic, identity attributes, and threat levels, reducing the probability of misjudging friendly or low-threat targets.

[0074] 3. This invention incorporates environmental situational information and a whitelist dataset when generating terminal countermeasure strategies. The environmental situational information includes the distribution of civilian flight paths, sensitive ground areas, prohibited debris landing zones, and no-fire zones, used to suppress actions that may cause collateral damage or violate airspace constraints. The whitelist dataset is used to identify authorized or friendly drones and automatically prohibits irreversible hard-kill actions, allowing only warnings, forced landing guidance, or soft-restriction operations. This decision-making model ensures that the countermeasure process complies with regulatory requirements and the principle of proportionality, reduces collateral risks, and improves traceability.

[0075] 4. This invention supports priority scheduling in multi-target concurrent scenarios. The decision control unit can sort multiple UAVs based on factors such as the degree of proximity of the target to the sensitive area, anti-interference / anti-guidance behavior characteristics, and potential payload hazards. High-priority targets are given priority to enter the closed-loop inquiry and handling process, while low-priority targets are kept in a monitoring state and their incentive resource usage is restricted, thereby achieving orderly control of group intrusion scenarios. Attached Figure Description

[0076] Figure 1 This is a system architecture diagram of the present invention;

[0077] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Example 1:

[0080] The UAV countermeasure system based on multi-sensor fusion of the present invention includes detection sensors, countermeasure devices, an excitation signal library, a whitelist data set, and a processor. The processor includes a fusion perception unit, a hypothesis management unit, and a decision control unit. The detection sensors are responsible for multimodal detection of target UAVs in the airspace. The fusion perception unit is responsible for time synchronization, spatial registration, temporal fusion, and target tracking of the detection data. The hypothesis management unit is responsible for generating and maintaining parallel initial hypotheses about the target UAV and updating confidence levels based on behavioral response data. The decision control unit is responsible for selecting and issuing probing excitation signals, driving the closed-loop interrogation process, and generating a terminal countermeasure strategy in the final stage.

[0081] To describe the signal, data, and control relationships between the various units within the system, this invention marks the key data flows during system operation as several progressive data sets. First, the raw multimodal detection data (denoted as D1) represents the radio frequency observation data, radar observation data, electro-optic or infrared image observation data, and acoustic observation data directly acquired by the detection sensors. Second, the unified observation data (denoted as D2) represents the time-stamped target observation set obtained by the fusion sensing unit after performing unified timestamp alignment, pose compensation, coordinate registration, angle registration, and association with the same target on the raw multimodal detection data. Third, the target trajectory information (denoted as D3) represents the target UAV state estimate obtained by the fusion sensing unit after performing multi-frame temporal fusion, filtering prediction, maneuver pattern recognition, and confidence calculation based on the unified observation data. The target trajectory information includes spatial position, velocity vector, heading trend, attitude maneuver pattern, and corresponding confidence weights. Fourth, the parallel initial hypothesis information (denoted as D4) represents at least two parallel initial hypotheses given by the hypothesis management unit based on the target trajectory information, along with the initial confidence level of each hypothesis. Fifth, the probe excitation signal command (denoted as D5) represents the control command, including the type, direction, power or amplitude limit, and duration of the probe excitation signal selected by the decision control unit from the excitation signal library based on the parallel initial hypothesis information and the previous round of inference. Sixth, the excitation emission behavior (denoted as D6) represents the probe excitation signal actually radiated by the countermeasure device towards the target UAV according to the probe excitation signal command. Seventh, the behavior response data (denoted as D7) represents the response characteristics of the target UAV, re-collected by the detection sensor and processed by the fusion perception unit after the probe excitation signal is applied. This includes kinematic change characteristics, radio frequency link behavior change characteristics, attitude or optical evasion characteristics, and possible active response behavior characteristics. Eighth, the hypothesis update result (denoted as D8) represents the result after the hypothesis management unit updates the credibility of each parallel initial hypothesis using the behavior response data. This includes the latest credibility distribution of each hypothesis, the time decay processing result, and whether there are any verified hypotheses with credibility exceeding the threshold. Ninth, the terminal countermeasure strategy command (denoted as D9) indicates that after receiving the verified hypothesis, the decision control unit, in conjunction with environmental situation information and the comparison results of the whitelist data set, calculates the terminal countermeasure strategy through multi-objective constraint optimization, and sends it to the countermeasure device in the form of execution parameters. The above data flow forms a closed loop in time, that is, the original multimodal detection data is sequentially transformed into unified observation data, target trajectory information, and parallel initial hypothesis information. Then, the behavioral response data of the target UAV is induced by the probe excitation signal command and the excitation launch behavior. The hypothesis update result is then used to determine whether a credible verified hypothesis has been obtained. If not, it returns to select a new probe excitation signal command to form a new interrogation round. If a hypothesis has been obtained, the terminal countermeasure strategy command is generated to complete the countermeasure decision.

[0082] The detection sensors preferably include radio frequency (RF) sensing components, radar sensing components, electro-optic or infrared imaging components, and acoustic acquisition components. The RF sensing components output information including, but not limited to, the operating frequency band of the control link, carrier occupancy characteristics, modulation or coding fingerprints, frequency agility behavior, and the presence of telemetry or remote control heartbeat data. This information reflects the target UAV's communication link type, anti-jamming capability, and control source characteristics. The radar sensing components output information including, but not limited to, range, azimuth, pitch, radial velocity, and echo intensity. Multi-frame measurements can be used to deduce the UAV's three-dimensional spatial position, velocity vector, and maneuvering trends. The radar sensing components can be active or passive radar. The electro-optic or infrared imaging components output information including, but not limited to, the target UAV's image frame, attitude angle estimates (e.g., yaw, pitch, roll), thermal radiation characteristic distribution, and gimbal or sensor orientation. This information reflects whether the UAV is performing obstacle avoidance, whether it has a payload, or whether it has performed optical obstruction maneuvers. The acoustic acquisition components output, but are not limited to, propeller noise spectrum characteristics, rotor quantity indication characteristics, and acoustic envelope information corresponding to thrust status. This information can be used to determine the propulsion condition of the UAV, such as high-thrust ascent, hovering, or rapid descent. Before processing, the above-mentioned observation information constitutes the raw multimodal detection data, i.e., the aforementioned raw multimodal detection data, which has characteristics such as multi-source heterogeneity, time asynchrony, and different coordinate systems.

[0083] The fusion sensing unit first performs standardization processing on the raw multimodal detection data to form unified observation data. This processing includes at least unified timestamps, coordinate and angle registration, and data alignment correlation. Unified timestamps refer to the fusion sensing unit using a unified system clock to synchronize the observation results from the radio frequency sensing component, radar sensing component, electro-optic or infrared imaging component, and acoustic acquisition component. This ensures that different observations correspond within the same time slice, avoiding situations where radar observations precede photoelectric observations, making parallel analysis impossible. Coordinate and angle registration refers to the fusion sensing unit using the installation pose information of each detection sensor within the system to map the polar coordinate information output by the radar, the line-of-sight angle output by the electro-optic or infrared imaging component, the beam pointing angle of the acoustic acquisition component, and the direction-finding angle of the radio frequency sensing component to the same spatial reference coordinate system. This allows it to determine whether multiple observations point to the same physical target. Data alignment correlation refers to the fusion sensing unit creating a target entry for each candidate airborne target, merging multiple observation records pointing to similar azimuths, distances, or angles within the same time slice into this target entry. After the above processing, the unified observation data can provide a spatial orientation estimate, initial velocity value, initial attitude value, initial RF link characteristic value, and initial acoustic fingerprint value for a single target UAV in a single time slice.

[0084] After acquiring unified observation data across consecutive time points, the fusion sensing unit performs multi-frame temporal fusion and target tracking for each target entry to output target trajectory information. First, the fusion sensing unit models the motion state of the target UAV. The state of the target UAV can be represented as a vector X_t = (x, y, z, v_x, v_y, v_z, ψ, dψ / dt, κ), where x, y, and z are spatial coordinates, v_x, v_y, and v_z are spatial velocity components, ψ is the heading angle, dψ / dt is the rate of change of the heading angle, and κ is the maneuver mode label. The maneuver mode label represents typical flight maneuver modes, including smooth cruise, rapid evasion, dive descent, hovering landing, etc. Second, the fusion sensing unit uses extended Kalman filtering, unscented Kalman filtering, particle filtering, or trajectory prediction methods based on deep temporal networks to fuse multi-source measurements in the unified observation data, completing noise suppression, missing measurement compensation, and short-term state prediction. Third, the fusion perception unit determines the maneuver mode label based on features such as velocity abrupt changes, peak acceleration, abrupt attitude changes, and acoustic thrust abrupt changes, to identify behaviors such as high thrust ramp-up, anti-interception evasion, and forced landing trends. Fourth, the fusion perception unit calculates confidence weights for the current state estimate. Confidence weights can be understood as a reliability score or covariance measure of the state estimate, used to measure whether the current trajectory estimate is stable and reliable. The target trajectory information includes the target ID, three-dimensional position, velocity vector, heading trend, maneuver mode label, and the aforementioned confidence weights. The target trajectory information will be transmitted to the hypothesis management unit as the basis for subsequent threat modeling and hypothesis generation.

[0085] The assumption management unit is used to answer questions such as what type of target drone it is, who controls it, what its intentions are, and what its threat level is. This unit takes target trajectory information as input, outputs parallel initial hypothesis information, and then iteratively updates it by combining behavioral response data, finally outputting the updated hypothesis result. This updated hypothesis result may include verified hypotheses with a confidence level exceeding a threshold.

[0086] After receiving target trajectory information, the assumption management unit generates at least two parallel initial hypotheses based on spatial location, speed, heading trend, maneuver mode labels, and corresponding confidence weights. The parallel initial hypothesis information includes a set of candidate identities and threat profiles for the current target UAV, as well as the initial confidence level of each candidate hypothesis. For example, one hypothesis might correspond to a compliant, registered civilian aerial photography UAV with a conventional remote control link, flying along a predetermined route within an authorized area and employing a conservative safety strategy; another hypothesis might correspond to a consumer-grade or quasi-industrial-grade UAV with basic anti-jamming capabilities but a moderate overall threat level; yet another hypothesis might correspond to a highly maneuverable modified platform or a high-threat illegal payload platform with an anti-jamming link, strong anti-guidance behavior, and a tendency to intrude into sensitive areas. Each hypothesis is bound to a control logic assumption during its establishment, such as whether to prioritize maintaining the mission objective over hazard avoidance, whether to actively hop frequencies when encountering external disturbances, and whether to exhibit forced landing or evasive behavior when guided by navigation. The parallel initial hypothesis information is output by the hypothesis management unit to the decision control unit to guide the selection of subsequent probing stimulus signals.

[0087] Behavioral response data is generated by the detection sensors after the application of a probing excitation signal and processed by the fusion sensing unit. This data includes the target UAV's flight maneuvers, radio frequency behavior, attitude or sensor actions, and possible proactive response behaviors after the excitation signal. Specifically, under low-power communication interference-type probing excitation signals, ordinary consumer-grade UAVs often enter conservative modes such as hovering or returning to home, exhibiting characteristics such as decreased trajectory speed, maintaining altitude, and reduced rate of change of heading angle; while high-threat targets often rapidly hop frequencies, switch communication protocols, briefly interrupt existing heartbeat signals, and accompany off-axis maneuvers or accelerated escape. Under mild navigation deception or slightly biased trajectory guidance-type probing excitation signals, ordinary consumer-grade UAVs often cooperate with forced landing guidance or deceleration descent, showing an increased descent rate and a forced landing trend; while high-threat targets may forcibly maintain their original intrusion route or even increase thrust to enter a breakthrough attitude, manifested as a transition in maneuver mode label from stable cruise to rapid evasion or high-speed intrusion. Under probing excitation signals such as directional light or laser-limited stimuli, ordinary drones often trigger obstacle avoidance or sensor protection logic, resulting in nose-up, circling, or brief hovering, along with noticeable gimbal orientation adjustments. High-threat targets, however, may ignore visual protection, prioritizing maintaining their offensive target trajectory, or even initiating anti-detection maneuvers such as rapid ground-hugging flight or optical obstruction. Differences in behavioral response data are considered the "typical responses" of different hypotheses to intentional stimuli. Once fed back to the hypothesis management unit, this behavioral response data serves as evidence-based differentiation.

[0088] The management unit matches behavioral response data with parallel initial hypothesis information to determine the explanatory power of each hypothesis for the behavioral response, thereby updating the credibility of each hypothesis and outputting the hypothesis update results. This update employs a probabilistic update method based on Bayesian inference. Specifically, for each hypothesis H_i, the updated credibility Bel_new(H_i) is calculated based on the prior credibility Bel_old(H_i) and the likelihood L(E|H_i) of the behavioral response data E. This can be expressed as Bel_new(H_i) = [Bel_old(H_i) × L(E|H_i)] / Σ_j [Bel_old(H_j) × L(E|H_j)]. An evidence synthesis method based on evidence theory is also used, treating behavioral response data as a new source of evidence. A basic trust assignment function is constructed and Dempster-Shafer synthesis is performed to obtain the trust function for each hypothesis. Furthermore, a decision criterion based on minimizing information entropy is employed, treating behavioral response data as evidence to reduce uncertainty and prioritizing the increase in credibility of hypotheses that can reduce overall uncertainty to the greatest extent. To avoid the long-term suppression of higher-risk behavioral characteristics by early observations, the management unit introduces a time-decreasing forgetting factor λ (0 < λ < 1) into the credibility update, meaning that the influence of recent behavioral responses on credibility is greater, while the influence of earlier behavioral responses on credibility gradually decreases. The form Bel_t(H_i) ← λ × Bel_(t-1)(H_i) + (1-λ) × Bel*(H_i) is adopted, where Bel*(H_i) is the updated value obtained based on the current behavioral response data. Through the above update, the hypothesis update result will include the latest credibility of each hypothesis, the difference between the credibility of each hypothesis, and whether there are any verified hypotheses with credibility exceeding a preset threshold. If there is a verified hypothesis with credibility exceeding the threshold, it will be submitted to the decision control unit for subsequent terminal countermeasure strategy generation. If there is no verified hypothesis with credibility exceeding the threshold, the hypothesis management unit will indicate to the decision control unit that the current conclusion has not yet converged.

[0089] The decision control unit undertakes two core responsibilities. First, during the identification phase, based on parallel initial hypothesis information and hypothesis update results, the decision control unit proactively selects probing stimulus signals and issues commands to the countermeasure device to induce the target UAV to generate distinctive behavioral response data, helping the hypothesis management unit converge to credible, verified hypotheses. Second, during the response phase, after receiving verified hypotheses, the decision control unit, in conjunction with environmental situation information and a whitelist dataset, generates terminal countermeasure strategy commands through multi-objective constraint optimization and delivers them to the countermeasure device for execution.

[0090] In each round of questioning, the decision control unit selects a probe excitation signal from the excitation signal library to maximize information gain, that is, to maximize the difference in response exhibited by different hypotheses under this excitation. To illustrate the mechanism and discriminative role of probe excitation signals, three typical signal categories are described below. The first category is low-power communication interference probe excitation signals. Their mechanism is to apply a slight disturbance to the suspected control link or data link under short-term, low-power, and narrow-band conditions, rather than high-power suppression. The typical response of ordinary consumer-grade or compliant drones is to enter a hovering, return-to-home, and slow-down conservative mode, manifested as a decrease in speed, altitude hold, and a reduced rate of change of heading angle, indicating that their control logic tends to be safe and conservative; while high-threat drones may immediately hop frequencies, change communication protocols, briefly interrupt their original heartbeats, and accelerate off-axis escape, indicating that they have anti-interference links and evasion tendencies. The second type is navigation deception or mild track offset probing excitation signals. Their mechanism involves injecting a slight amplitude of virtual navigation reference, causing the target drone to perceive a guided waypoint or altitude deviation, but the amplitude is controlled to avoid immediate crashes or severe veergence. Ordinary consumer drones typically comply with this guidance, exhibiting a forced landing tendency, increased descent rate, or a trajectory away from sensitive areas; high-threat drones may reject this guidance and maintain their original intrusion direction, or even increase thrust to intrude into sensitive areas at high speed. The third type is directional light or laser-limited amplitude stimulation probing excitation signals. Their mechanism involves briefly stimulating targets suspected of relying on visual obstacle avoidance or optical sensors with controlled radiation intensity, guiding them to trigger self-protective behaviors such as obstacle avoidance, head-up flight, gimbal turning, and brief hovering. Compliant or civilian drones often exhibit these protective responses, while high-threat drones may prioritize maintaining their intrusion trajectory or even engage in counter-detection, optical sensor blocking, or low-altitude penetration. The common feature of the three types of signals is that they elicit observable and categorizable behavioral response data through mild stimulation, amplifying the differences in response patterns among different hypotheses, thereby improving the hypotheses' discriminative power.

[0091] After selecting a probing stimulus signal, the decision control unit generates a probing stimulus signal command and issues it to the countermeasure device. The countermeasure device then uses this command to directionally emit signals towards the target UAV; the actual launch is the induced launch behavior. The detection sensor then observes the target again, and the fusion perception unit extracts changes in the target's flight maneuvers, link behavior, attitude avoidance behavior, and response behavior from these new observations, forming new behavioral response data. The hypothesis management unit uses this behavioral response data to update the confidence level of each hypothesis. If no hypothesis reaches the confidence threshold at this point, the decision control unit re-evaluates the stimulus signal library based on the current confidence distribution and the differences between the remaining hypotheses, selects a new probing stimulus signal, and issues the probing stimulus signal command again, repeating the collection of behavioral response data and confidence level updates. This cycle constitutes a closed-loop questioning process for cognitive stimulus observation. Essentially, it involves posing probing, low-collateral-impact questions round by round, observing the target's responses, and gradually approaching a credible, verified hypothesis, rather than taking irreversible, highly destructive measures at the identification stage.

[0092] When the hypothesis update result output by the hypothesis management unit indicates that there is a verified hypothesis with a credibility exceeding a preset threshold, the decision control unit enters the handling phase. First, the decision control unit reads the threat level, control logic characteristics, and current behavioral tendencies from the verified hypothesis, such as whether the target is actively approaching a sensitive area or exhibiting anti-interference and anti-guidance behavior. Second, the decision control unit introduces environmental situation information as a constraint. This environmental situation information can come from system presets or external situation interfaces, including information on the distribution of civilian flight routes, the location of sensitive ground areas, the distribution of no-landing and no-fire zones, and the setting of temporary control zones. For example, when the target is near a civilian flight route or above a densely populated area, the decision control unit automatically lowers the score of high-collateral-hazard hard-kill strategies, prioritizing soft-kill strategies or guided forced landing strategies to avoid generating a large amount of debris or high-energy fall risks. When the target is rapidly approaching a highly sensitive protected object or restricted area, the decision control unit increases the priority of rapid disabling measures to prevent the target from causing high-value harm in a short period. Simultaneously, the decision control unit compares the target UAV's identity fingerprint information with a whitelist database. The whitelist database stores the identity fingerprint information of authorized or friendly drones. This fingerprint information may include specific identification codes or registration numbers, telemetry heartbeat message structure characteristics or encrypted handshake characteristics, authorized flight corridor information, and specific response behavior patterns. If the comparison result shows that the target drone matches the whitelist or is flying on an authorized route, the decision control unit will forcibly suppress irreversible hard-kill measures, allowing only soft restriction operation commands such as alarms, track restrictions, forced landing guidance, or continuous directional interference to be output. This process ensures that the system takes minimal necessary intervention against friendly or authorized targets. If the comparison result shows that the target drone does not match the whitelist and its verified assumption is that it is a high threat, then high-intensity measures are allowed to enter the candidate strategy set.

[0093] In the specific strategy generation process, the decision control unit performs multi-objective optimization and scoring on multiple candidate terminal countermeasure strategies. A decision function F(strategy) = w1 × R_threat − w2 × C_collateral − w3 × C_regulatory − w4 × C_resource can be constructed, where R_threat represents the timeliness and effectiveness of the strategy in suppressing the current threat, C_collateral represents the potential collateral damage risk, C_regulatory represents the compliance cost under the current airspace or regulatory boundaries, C_resource represents the resource consumption and sustainability of subsequent protection by the strategy, and w1, w2, w3, and w4 are weighting coefficients that can be configured according to the deployment scenario, mission level, and regulatory requirements. Based on this function, the decision control unit scores and ranks soft-kill and hard-kill strategies separately, selects the strategy with the highest score that does not violate environmental situation constraints and whitelist exemption constraints, forms the terminal countermeasure strategy instruction, and issues it to the countermeasure device for execution.

[0094] The execution chain of a soft-kill strategy typically includes persistent directional jamming, navigation deception-based forced landing guidance, link takeover, or soft limitation operations. The goal of persistent directional jamming is to limit the effective control link of the target UAV, forcing it into a safe, conservative mode, such as hovering or returning to base, rather than immediately disintegrating. The goal of navigation deception-based forced landing guidance is to gradually remove the target from sensitive airspace and land it in a safe area with low kinetic energy. The goal of link takeover or forced landing guidance is to guide the target to a controlled landing. The decision control unit can specify parameters such as target identification, orientation angle, output power or amplitude limit, duration of action, and desired flight attitude changes in the terminal countermeasure strategy command, thereby enabling the countermeasure device to implement soft-kill in a controllable manner with low collateral damage.

[0095] The execution chain of a hard-kill strategy includes high-direction beam illumination, kinetic interception, and deployment of net-and-tether capture structures. High-direction beam illumination can disable critical components of the target UAV in a very short time; kinetic interception or net-and-tether capture structures can directly limit the target's flight capabilities. Because hard-kill attacks may cause the target to stall and crash, generating debris or collateral damage, the decision control unit only allows a hard-kill strategy to be a candidate strategy and to dominate the scoring when the following conditions are met: the target has been identified as a high threat; the target is approaching or intruding into a highly sensitive area or a critical protected object; environmental situational information indicates that the airspace allows for rapid disabling measures; it is not an area where debris is strictly prohibited from falling; and the whitelist data set comparison confirms that the target is not a friendly or authorized flight path target. For hard-kill strategies, the terminal countermeasure command will include more precise execution parameters, such as interception timing, directional window, effective altitude range, maximum permissible spillover power density, and net-and-tether deployment altitude limit, to ensure that the countermeasure device complies with area security requirements during execution.

[0096] Environmental situation information is explicitly incorporated into the generation of terminal countermeasure strategies. This environmental situation information can be pre-set by the system, distributed in real-time by the operations management platform, or dynamically provided by the on-site security control system. The information includes at least the distribution areas of civil aviation routes, sensitive ground areas, prohibited debris landing areas, prohibited high-power radiation areas, and temporary control zones. When generating terminal countermeasure strategy instructions, the decision control unit filters candidate strategies based on this environmental situation information, using methods including risk weighting and hard constraint elimination. Risk weighting refers to increasing the scoring weight of rapid disabling measures near high-risk sensitive areas, while decreasing the scoring weight of high-collateral-risk strategies above civil aviation routes or densely populated areas. Hard constraint elimination refers to directly eliminating physical interception schemes that would generate a large amount of detached debris in prohibited debris landing areas, and eliminating excessively high-intensity wide-area interference schemes that would significantly affect unrelated radio links near civil aviation routes. The decision control unit also records the decision-making process for adopting the above constraints to form subsequent proof of legality and rationality. The whitelist exemption mechanism is also recorded. The whitelist database contains the identity fingerprint information of authorized drones. This information may include a specific registration number, a unique heartbeat message structure, encrypted handshake sequence characteristics, authorized flight path definitions, and agreed-upon response behavior patterns. Before generating terminal countermeasure instructions, the decision control unit compares the target drone's current identity fingerprint characteristics with the whitelist database. If the target is determined to be a whitelist match or flying on an authorized flight path, irreversible hard-kill methods are automatically excluded when generating terminal countermeasure instructions, allowing only soft-kill, forced landing guidance, or alarm-type actions. This comparison and its conclusion are archived to prove that the invention has performed a "friendly exemption" verification process before taking any mandatory countermeasure actions.

[0097] The following is a typical time-series operational example to illustrate how the various units of this invention work collaboratively and demonstrate the transmission relationship between the original multimodal detection data, unified observation data, target trajectory information, parallel initial hypothesis information, probing excitation signal commands, excitation launch behavior, behavior response data, hypothesis update results, and terminal countermeasure strategy commands. First, after the system starts operating, the detection sensors continuously collect airspace information. The fusion perception unit performs unified timestamp, spatial registration, and target association processing on this multimodal information, gradually forming unified observation data and further calculating the target trajectory information. Second, the hypothesis management unit generates at least two parallel initial hypotheses based on the target trajectory information and provides corresponding initial confidence levels, serving as parallel initial hypothesis information. Third, the decision control unit evaluates the excitation signal library based on the parallel initial hypothesis information, selects the most discriminative probing excitation signal, and forms a probing excitation signal command. The countermeasure device then implements the excitation launch behavior accordingly. Fourth, the detection sensors observe the target UAV again, and the fusion perception unit extracts the flight actions, communication behavior changes, attitude avoidance behavior, or active response behavior after this round of excitation as behavior response data. Fifth, the hypothesis management unit matches the behavioral response data with the parallel initial hypotheses, updates the credibility of each hypothesis using Bayesian inference, evidence theory synthesis, or information entropy minimization criteria, and applies a time-decreasing forgetting factor to obtain the hypothesis update results. In the hypothesis update results, if there is a verified hypothesis with a credibility exceeding a preset threshold, this verified hypothesis is passed to the decision control unit. If there is no verified hypothesis with sufficiently high credibility, the decision control unit will continue to select new exploratory stimulus signals and repeat the aforementioned process, thereby forming a closed-loop inquiry of cognitive stimulus observation. In a preferred embodiment, the hypothesis management unit uses the behavioral response data as new evidence to update the credibility of the parallel initial hypotheses and decides whether to trigger the next round of exploratory stimulus accordingly. One or a combination of the following three complementary inference paradigms is used:

[0098] 1. Update based on Bayesian inference, assuming a parallel initial hypothesis set of {H_i} and a prior confidence level of . The latest behavioral response data is Define a likelihood model for each hypothesis. Then the posterior:

[0099]

[0100] To suppress the "long tail" effect of early evidence, a time-decreasing forgetting factor is introduced. Exponential smoothing over a number field:

[0101]

[0102] Normalization This is the result of this round of updates. The likelihood model can be learned from offline samples, such as conditional density modeling (probability output of Gaussian mixture, kernel density, or discriminant calibration) of behavioral features such as "frequency hopping rate increase", "protocol switching", "track offset angle", "descent rate", and "gimbal attitude change amplitude".

[0103] 2. Synthesis based on evidence theory (DS) maps behavioral response data to basic trust assignments. Regarding the results of the previous synthesis With the present Perform Dempster synthesis to obtain To reflect the forgetting factor, one can... First, make a reduction: The remaining amount will be added to the entire series. Ultimately, it will... Convert the confidence level of each hypothesis or use pignistic transformation to obtain comparable decision probabilities.

[0104] 3. Update based on minimizing information entropy with the current distribution entropy As a measure of uncertainty, a discrimination score is used to evaluate new evidence. Forming a soft posterior ,in The scaling factor is selected after smoothing with the forgetting factor. The solution with the largest decrease achieves the goal of "accelerating convergence". 5.x.4 Threshold Determination and Triggering of Continued Queries: Set a confidence threshold. If it exists If the hypothesis is verified, the process proceeds to the action phase; otherwise, the process moves to the next round of inquiry. The selection of the next round of exploratory stimulus signals can be based on the expected information gain (mutual information) criterion.

[0105]

[0106] in This is a subset of the available stimulus signal library (with compliance and security constraints taken into account). The results are predicted by the response model under each assumption. In practice, approximate Monte Carlo methods or offline table lookups can be used to accelerate the process.

[0107] Sixth, when a verified hypothesis is found in the hypothesis update results, the decision control unit combines the threat level of the verified hypothesis, environmental situation information, and the comparison results of the whitelist data set, calculates the scores of multiple candidate terminal countermeasure schemes based on the multi-objective optimization function, selects the highest-scoring scheme that meets environmental constraints, security constraints, and friendly exemption constraints, and forms the terminal countermeasure strategy instruction.

[0108] Terminal countermeasure strategy optimization and instruction generation: After establishing verified hypotheses, the decision control unit performs constrained multi-objective optimization on candidate terminal countermeasure schemes. Let the candidate set be denoted as... For each scheme Defining Objectives and Costs: Threat Suppression Benefits in the Logarithmic Domain The assessment is based on a comprehensive evaluation of the threat level of validated hypotheses, the movement trends of the target in relatively sensitive areas, and the prior knowledge of historical response effectiveness; and includes the cost of collateral damage in the logarithmic domain. Based on the current ground environment, population density, and debris / energy spillover risk model estimation; logistic domain compliance costs. Rule costs triggered by environmental constraints (civil aviation corridors, no-landing / no-fire zones, etc.) are considered unfeasible if violated; this also affects the consumption of resources across multiple domains. The device's available power, time slots, pointing occupancy, and continuous protection capabilities all have an impact. A weighted summation feasibility maximization model can be used.

[0109]

[0110] constraint: (Hard constraints of environmental conditions and hard constraints of whitelist exemptions) Infeasible solutions are directly eliminated. Weighting It can be preset at the task level or adapted online (e.g., improved during critical sensitive periods). Improve during peak passenger flow ).

[0111] To improve transparency, a two-stage decision-making process can be adopted: first, conduct feasibility screening and compliance elimination, and then, based on the feasible set... Sort the results to find the optimal one. Use the Pareto front to assist in careful selection if necessary.

[0112] The generated terminal countermeasure strategy instructions include: target identifier and pointing parameters, strategy category (soft kill / hard kill / alarm only), action window (time / space), intensity limit, termination conditions and fallback strategy, monitoring indicators and stop rules (such as reaching the escape distance, descent rate, disability flag, etc.).

[0113] Seventh, the countermeasure device executes soft kill, hard kill, or alarm-only actions according to the terminal countermeasure strategy instructions, and returns execution status information to the system to complete the closed loop of the response to the target;

[0114] After receiving the countermeasure strategy command from the terminal, the countermeasure device enters a strictly controlled execution state and sends the status back to the system, forming a closed-loop response. Key implementation points are as follows:

[0115] Verify the command signature and validity period, and check the hard constraints of the environmental situation and the whitelist exemption bits; if not satisfied, refuse execution and provide the reason. Load the pointing angle, power / amplitude limit, effective height / azimuth window, maximum duration, emergency stop conditions, etc.

[0116] Soft-kill attacks (continuous directional interference, navigation guidance, link takeover, soft restriction operations) are controlled and output according to the set closed-loop mechanism; hard-kill attacks (directed energy, kinetic energy interception, netting capture) are triggered only within the feasible window, with real-time verification of "direction / altitude / spillover energy" to ensure they do not exceed limits. During execution, the system continuously subscribes to the target trajectory updates from the fusion sensing unit for fine-tuning of pointing and timing.

[0117] Set success criteria and termination conditions, such as the target leaving the sensitive area, the descent rate reaching the threshold and entering the safe recovery area, and the appearance of incapacitation characteristics; if an anomaly occurs (the target suddenly enters the civil aviation corridor, or the ground risk suddenly increases), immediately downgrade or terminate the strategy and revert to a low-risk strategy.

[0118] The system periodically transmits execution status (timestamp, output level, pointing status, whether the threshold was triggered, abnormal events), target response summary (from the fusion sensing unit), and final result (drive-away / forced landing / disabling / failure and reason). This information is simultaneously written into the electronic evidence chain for compliance auditing and post-event review.

[0119] In practical applications, multiple UAVs may appear simultaneously in the airspace. The decision control unit of this invention supports a multi-target concurrent handling strategy, achieving threat priority ranking and polling-style interrogation control through resource scheduling. First, the decision control unit calculates a priority score for each target trajectory maintained by the fusion perception unit. This score can consider factors such as whether the target is rapidly approaching a sensitive area or critical protected object, whether the target exhibits high-maneuverability, high-thrust penetration, or high-threat behaviors such as anti-guidance and anti-jamming, the rate of change of distance between the target and sensitive areas or densely populated areas on the ground, and whether the target is suspected of carrying high-risk payloads. Second, the system sorts targets according to priority. The highest priority target is immediately included in the closed-loop interrogation process, that is, a probing excitation signal command is generated for the target and an excitation launch is performed, its behavioral response data is collected, and the hypothesis credibility is updated. Lower priority targets may only be passively monitored at this stage, without immediately using the probing excitation signal, to save resources and avoid unnecessary intervention on low-risk targets. Third, the decision control unit can limit the number of targets being interrogated simultaneously to avoid multiple high-power or highly directional actions conflicting at the same time, preventing resource contention or superimposed effects on the airspace. Fourth, the priority score is dynamically updated over time. If a previously low-priority target suddenly exhibits high-risk characteristics (such as high-speed dives into sensitive areas or strong anti-guided behavior), that target will be promoted to a high-priority target in the next scheduling cycle and enter the closed-loop inquiry process. Through this scheduling logic, this invention can maintain high-frequency proactive inquiry and rapid verification of high-threat targets under multi-target concurrent conditions, while maintaining continuous monitoring of low-threat targets without excessive intervention, thereby achieving overall airspace security control under limited resources.

[0120] The system structure and method of this invention can be applied to fixed deployment scenarios, vehicle-mounted mobile scenarios, and temporary deployment scenarios. In fixed deployment scenarios, the detection sensors can be arranged in a fixed omnidirectional or hemispherical coverage layout to achieve routine protection of key sensitive areas. In vehicle-mounted mobile scenarios, the system is installed on a mobile carrier platform and platform self-motion compensation is introduced into the fusion sensing unit, so that radar and electro-optical sensing data can still be correctly registered in a mobile state, suitable for mobile patrols or temporary escorts. In temporary deployment scenarios, the system can use a lightweight detection sensor group and a library of detection excitation signals with limited power levels to quickly establish temporary no-fly zones or temporary security airspace. Under different deployment forms, the types of detection excitation signals that can be used, the upper limit of output power, the set of permitted terminal countermeasure strategies, and the range of whitelist entries can be tailored according to regulations, mission level, or activity nature, but the data flow relationship, closed-loop query mechanism, dynamic update mechanism of hypothesis credibility, and terminal countermeasure strategy generation mechanism of this invention remain unchanged.

[0121] Through the aforementioned structure and mechanism, this invention achieves the following technical effects. First, by generating target trajectory information containing confidence weights through multi-sensor fusion and multi-frame temporal tracking, this invention enables robust locking of high-speed maneuvering, small, low-radar-cross-section, and even partially adversarial UAV targets, thus overcoming the problem of single-sensor target loss under conditions of obstruction, camouflage, and low signal-to-noise ratio. Second, through an active interrogation mechanism using probing excitation signals, this invention extends the identification process from traditional passive observation to active probing, allowing the system to observe the autonomous response patterns of target UAVs under mild stimuli and use these patterns as key evidence to distinguish different hypotheses. Third, through a parallel initial hypothesis and confidence iteration update mechanism in the hypothesis management unit, combined with Bayesian inference, evidence theory synthesis, information entropy minimization criteria, and time-decreasing forgetting factors, this invention achieves dynamic determination and rapid convergence of UAV identity attributes, control logic, and threat level. Fourth, this invention utilizes a multi-objective constraint optimization strategy in the decision control unit to comprehensively consider threat suppression capabilities, collateral damage risks, regulatory constraints, resource consumption, environmental situation information, and whitelist exemption logic to generate the final terminal countermeasure strategy instruction. This ensures that while meeting the security disposal objective, it avoids excessive attacks, accidental harm to friendly forces, or secondary risks in restricted airspace. Fifth, this invention supports a resource scheduling mechanism for multi-target concurrent scenarios, enabling high-threat targets to receive high-priority inquiries and timely handling, while maintaining moderate monitoring of low-threat targets without excessive intervention, thereby ensuring overall airspace security in resource-constrained environments.

[0122] To meet the requirements of safety compliance, accountability, and objective evidence collection, the system of this invention can form an electronic evidence chain during operation to record all key decisions and evidence from the initial detection to the final disposal. First, in the perception phase, the system records when, where, and with what confidence level the detection sensors observed the target UAV, and records the target trajectory information generated by the fusion perception unit, including the target position, speed, maneuver mode label, and confidence weight. Second, in the hypothesis building and updating phase, the system records parallel initial hypothesis information, including the meaning of each hypothesis, the corresponding control logic assumption, and the initial confidence level; records the type, power limit, pointing angle, and duration of each round of exploratory excitation signal; records the behavioral response data of the target UAV after each round of excitation; records the reasoning method (e.g., Bayesian inference, evidence theory synthesis, or information entropy minimization criterion) used by the hypothesis management unit to update the confidence level of each hypothesis based on the behavioral response data, as well as the set value of the time-decreasing forgetting factor; and records the hypothesis update results, including whether any verified hypotheses with confidence levels exceeding a threshold have appeared. Third, during the closed-loop inquiry phase, the system records the order of each round of inquiries, the correspondence between each round of probing stimulus signal commands and stimulus launch behaviors, and the content of each round of behavioral response data, thus forming a complete "inquiry-response" history. This demonstrates that the system does not indiscriminately suppress, but rather verifies and upgrades gradually. Fourth, during the disposal phase, the system records the use of environmental situation information by the decision control unit, including the constraints of civilian flight path distribution, sensitive ground areas, prohibited debris landing areas, prohibited high-power radiation areas, and temporary control zones; it records the conclusions of comparing the target UAV's identity fingerprint information with the whitelist data set, indicating whether the system has confirmed that the target is a non-friendly or unauthorized flight path target before taking disposal actions; and it records the key input factors of the decision function F(strategy) = w1 × R_threat − w2 × C_collateral − w3 × C_regulatory − w4 × C_resource, which explains why a specific terminal countermeasure strategy command was chosen instead of other alternative strategies. Finally, the system also records the execution feedback of the countermeasures, including whether soft or hard kill was carried out according to the specified parameters, whether the target drone was disabled, forced to land, or escaped from the sensitive area, whether collateral damage occurred, and whether further alarms or reporting to regulatory agencies are required. Through the above records, this invention not only achieves closed-loop control from identification to disposal in terms of technical path, but also forms traceable and interpretable disposal evidence at the evidentiary level to support compliance review, post-event liability determination, and security review.

[0123] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A drone countermeasure system based on multi-sensor fusion, characterized in that, include: Detection sensors, including those that collect multimodal detection data of the target airspace; Countermeasures include transmitting probing excitation signals to the target drone and executing terminal countermeasures; Excitation signal library, used to store a variety of preset probing excitation signals; The whitelist data set is used to store the identity fingerprint information of authorized or friendly drones; The processor is connected to the detection sensor, the countermeasure device, the excitation signal library, and the whitelist data set. The processor includes: a fusion sensing unit, which forms the target trajectory of the target UAV; The management unit assumes that it establishes at least two parallel initial hypotheses regarding the attributes of the target UAV based on the target trajectory. The decision control unit includes: selecting a probing stimulus signal from the stimulus signal library to distinguish the at least two parallel initial hypotheses and controlling the countermeasure device to directionally launch the probing stimulus signal toward the target UAV; and acquiring and organizing behavioral response data for the hypothesis management unit to update the hypothesis.

2. The UAV countermeasure system based on multi-sensor fusion according to claim 1, characterized in that, The detection sensor includes: Radio frequency sensing components are used to detect and analyze the remote control link spectrum characteristics, telemetry signal protocols, and frequency agility behavior of target UAVs. A radar sensing component, including at least one of an active radar structure or a passive radar, is used to output range, azimuth, pitch and radial velocity information of a target UAV. Electro-optic / infrared imaging components are used to acquire the attitude, body geometry, and visible or infrared radiation characteristics of the target UAV. Acoustic acquisition components are used to collect propeller noise from a target UAV and infer its propulsion status. The fusion sensing unit of the control system includes performing multi-frame temporal fusion on the above-mentioned types of detection data to generate a fusion state quantity with confidence weights. The fusion state quantity is used to reduce the probability of tracking loss when any type of detection sensor is occluded, saturated, or deceived.

3. The UAV countermeasure system based on multi-sensor fusion according to claim 1, characterized in that, The excitation signal library includes: Low-power probing jamming signals targeting specific communication frequency bands are used to observe the link robustness and anti-jamming strategies of target UAVs. Weak navigation deception signals that guide target drones to yaw, decelerate, hover, or make forced landings; A limited-intensity directional laser irradiation signal that triggers the obstacle avoidance or optoelectronic protection response of the target UAV.

4. The UAV countermeasure system based on multi-sensor fusion according to claim 1, characterized in that, The behavioral response data includes one or more of the following: The abrupt change in the kinematic state of the target UAV within a preset time window, including a sharp change in attitude angular rate, sudden deceleration, track deviation, or a significant increase in descent rate; Changes in the radio frequency signal characteristics of the target drone, including increased frequency agility, switching of uplink or downlink communication protocols, interruption of data link heartbeat, or increased encryption level of telemetry messages; Rapid adjustment of the target drone's body attitude or sensor orientation, including gimbal orientation change, obstacle avoidance activation, or optical masking action; The target drone's proactive response or countermeasures include broadcasting identification declaration signals or transmitting jamming pulses; The hypothesis management unit of the control system processes the behavioral response data using Bayesian inference, evidence synthesis methods based on evidence theory, and / or decision criteria based on minimizing information entropy, and introduces a forgetting factor that decreases over time, thereby updating the posterior probability, trust function, or confidence interval of the at least two parallel initial hypotheses.

5. The UAV countermeasure system based on multi-sensor fusion according to claim 1, characterized in that, The decision control unit also includes: When the hypothesis management unit fails to output a verified hypothesis that meets the preset confidence threshold, it re-evaluates the stimulus signal library based on the expected information gain, selects a different probe stimulus signal than before, and controls the countermeasure device to emit again, so as to form a closed-loop interrogation mechanism of cognition-stimulation-observation. The countermeasures against multiple candidate terminals are scored based on a joint optimization objective function. The candidate terminal countermeasures include soft-kill strategies and hard-kill strategies. The soft-kill strategies include continuous directional jamming, navigation deception, link takeover, or forced landing guidance. The hard-kill strategies include high-directional beam illumination, kinetic interception, or deployment of net-and-tow capture structures. When generating the terminal countermeasure strategy, the strategy is constrained by environmental situation information. When the predicted flight path of the target UAV points to a sensitive area on the ground, the priority of rapid disabling measures is increased. When the target UAV is near a civilian flight path, physical interception that will cause high collateral risks is suppressed and soft kill measures that will not cause debris to scatter are given priority. When the whitelist data set indicates that the target drone is flying on a friendly or authorized route, it automatically prevents irreversible hard kill measures and only outputs alarms or soft restriction operation instructions.

6. A method for countering unmanned aerial vehicles (UAVs) based on multi-sensor fusion, characterized in that: Includes the following steps: Sp1: Receives multimodal detection data from the detection sensor, detects, locates and tracks the target UAV, and generates the target trajectory of the target UAV; Sp2: Based on the target trajectory, establish at least two parallel initial hypotheses about the attributes of the target UAV, the attributes including at least control logic, identity attributes, and threat level; Sp3: Select a probing excitation signal from a library of excitation signals, designed to distinguish the at least two parallel initial hypotheses; Sp4: Send a control command to a countermeasure device, causing the countermeasure device to transmit the detection excitation signal in a directional manner toward the target UAV; Sp5: Collect behavioral response data of the target UAV to the probing excitation signal, wherein the behavioral response data includes at least one of kinematic response, radio frequency communication response and / or attitude optical response; Sp6: Based on the behavioral response data, perform multi-hypothesis testing on the at least two parallel initial hypotheses, and update the confidence, posterior probability or confidence function of each hypothesis to obtain the verified hypothesis with the highest confidence. Sp7: Based on the verified assumptions, combined with environmental situation information and security constraints, determine a terminal countermeasure strategy, and control the countermeasure device to execute the terminal countermeasure strategy, wherein irreversible hard-kill measures are suppressed against target UAVs identified as friendly or authorized flight routes.

7. The UAV countermeasure method based on multi-sensor fusion according to claim 6, characterized in that, Sp1 includes: It integrates radio frequency sensing data, radar sensing data, electro-optic / infrared imaging data, and acoustic acquisition data; The data is subjected to time synchronization, coordinate alignment, filtering, target association, and consistency verification. The system outputs the target UAV's three-dimensional position, velocity, heading trend, and maneuvering mode through multi-frame time-series tracking, and assigns different information weights to reduce the probability of tracking loss due to the failure of a single detection source.

8. The UAV countermeasure method based on multi-sensor fusion according to claim 6, characterized in that, The Sp3 includes: Calculate the expected information gain metric for the at least two parallel initial hypotheses; Based on the expected information gain index, a probing excitation signal that can maximize the hypothetical discriminative power under the expected minimum excitation power is selected from the excitation signal library; The probing excitation signal is compared with compliance constraints, civil airspace impact restrictions, and surrounding communication system interference restrictions to eliminate excitation signals that do not meet the restrictions.

9. The UAV countermeasure method based on multi-sensor fusion according to claim 6, characterized in that, The Sp6 includes: Using Bayesian inference methods and / or evidence synthesis methods based on evidence theory, the behavioral response data obtained from Sp5 is used as new evidence input to update the posterior probability, trust function, or confidence interval of the at least two parallel initial hypotheses; Introduce a time-decreasing forgetting factor to reduce the influence of historical evidence; When the confidence level of at least one hypothesis is higher than a preset threshold, the hypothesis is output as the verified hypothesis. When the confidence level of all hypotheses fails to reach the preset threshold, return to execute Sp3, select a different exploratory stimulus signal than the previous one, and execute Sp4 and Sp5 again to form a closed-loop inquiry process of cognition-stimulation-observation.

10. The UAV countermeasure method based on multi-sensor fusion according to claim 6, characterized in that, The Sp7 includes: Multiple candidate terminal countermeasure strategies are scored and ranked based on a joint optimization objective function. The candidate terminal countermeasure strategies include soft-kill strategies and hard-kill strategies. The soft-kill strategies include continuous directional jamming, navigation deception, link takeover, or forced landing guidance. The hard-kill strategies include high-directional beam illumination, kinetic interception, or deployment of net-and-tether capture structures. The terminal countermeasure strategy is constrained by combining environmental situation information. When the current or predicted flight path of the target UAV points to a sensitive area on the ground, the priority of rapid disabling measures is increased. When the target UAV is near a civilian flight path, physical interception that will cause high collateral danger is suppressed and soft kill measures that will not cause debris to scatter are given priority. The stored identity fingerprint information of authorized or friendly drones constitutes a whitelist data set; the identity fingerprint information of the target drone is compared with the whitelist data set, and when the target drone is instructed to fly a friendly or authorized route, irreversible hard kill measures are prohibited, and only risk warnings or soft restriction operation instructions are output.

Images