A method for integrated airspace surveillance of UAVs and manned aircraft based on ADS-B

Abstract

This invention discloses a method for fusion airspace surveillance of unmanned and manned aircraft based on ADS-B, belonging to the field of air traffic control technology. This invention acquires manned aircraft attitude vectors through dual channels: a ground-based ADS-B receiver and an unmanned aircraft broadcast. It then combines this with relative measurement vectors from the unmanned aircraft's self-sensing sensors and converts them into absolute attitude vectors in a global coordinate system. Based on the signal reliability weighted fusion of the three acquisition channels, an accurate attitude vector is obtained. Sequence features of position, velocity, and heading angle changes are extracted to calculate the probabilities of various flight intentions. Based on the intention probabilities, the mixed future position and covariance are predicted. Finally, a spatial kernel function is constructed and a directional amplification factor is introduced to generate an intention risk field. This invention improves the accuracy of attitude vectors through multi-source information fusion and optimizes modeling by combining intention probability and directional risk amplification mechanisms, significantly improving the accuracy of fusion airspace surveillance and the reliability of risk assessment.

Description

Technical Field

[0001] This invention relates to the field of air traffic control technology, specifically to an ADS-B-based method for integrated airspace surveillance of unmanned aerial vehicles and manned aircraft. Background Technology

[0002] With the accelerated opening of low-altitude airspace, the demand for collaborative operation between unmanned and manned aircraft in the same airspace is constantly increasing, leading to a significant increase in airspace traffic density and placing higher demands on the real-time performance and security of airspace surveillance systems. Existing surveillance systems primarily rely on ground-based ADS-B receivers to acquire the position and motion status of manned aircraft, supplemented in some areas by self-sensing sensors carried by unmanned aerial vehicles (UAVs) to compensate for ADS-B signal obstruction and attenuation in low-altitude environments. While multi-source surveillance technology has improved monitoring coverage to some extent, the various data sources differ significantly in terms of accuracy, latency, and noise characteristics, making data fusion still quite challenging.

[0003] In airspace safety management, predicting the future flight status of manned aircraft and generating spatial risk distributions is crucial for supporting collision avoidance and airspace control for unmanned aerial vehicles (UAVs). However, existing risk assessment methods typically construct static or semi-static risk areas based solely on current position, speed, or heading information. This makes it difficult to effectively utilize the continuous flight status changes of manned aircraft and lacks the ability to characterize multimodal future motion trends corresponding to different flight intentions. Because the probability and directional characteristics of intentions are not incorporated into risk calculations, the risk fields generated by existing technologies have limited accuracy and cannot accurately reflect the risk distribution in complex airspace. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides an ADS-B-based method for fusion airspace surveillance of unmanned and manned aircraft, which solves the problem of low accuracy in the generated risk fields in existing technologies.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: a method for integrated airspace surveillance combining unmanned aerial vehicles (UAVs) and manned aircraft based on ADS-B, comprising the following steps:

[0006] S1. Collect manned aircraft attitude vectors through ground ADS-B receiver and UAV broadcast, and at the same time collect relative measurement vectors using UAV self-sensing sensors, and convert the relative measurement vectors into manned aircraft absolute attitude vectors in global coordinate system.

[0007] S2. Obtain the signal confidence level from the ground ADS-B receiver, the UAV, and the self-sensing sensor on the UAV. Based on the confidence level of the three, weight the attitude vectors of the three to obtain the accurate attitude vector of the manned aircraft.

[0008] S3. Subtract the accurate attitude vectors of the manned aircraft from adjacent time points to obtain the position change sequence, velocity change sequence, and heading angle change sequence. Extract features from the three change sequences and obtain the intent probability for each intent category based on the features.

[0009] S4. Predict the future location for each intent category, and based on the intent probability for each intent category, obtain the mixed predicted location and the mixed predicted covariance.

[0010] S5. Based on the mixed prediction location and mixed prediction covariance, construct a kernel function for any point in space and introduce a directional amplification factor to generate the intentional risk field.

[0011] Furthermore, the attitude vector of the manned aircraft collected by the ground ADS-B receiver in S1 includes: the manned aircraft position decoded by the ground, the manned aircraft velocity decoded by the ground, and the heading angle decoded by the ground.

[0012] The attitude vector of the manned aircraft collected by the UAV includes: the manned aircraft position decoded by the UAV, the manned aircraft speed decoded by the UAV, and the heading angle decoded by the UAV.

[0013] The relative measurement vectors collected by the self-sensing sensors on the drone include: the relative distance between the drone and the manned aircraft, the azimuth angle, and the pitch angle.

[0014] Furthermore, S2 includes the following sub-steps:

[0015] S21. Obtain the signal reliability from the ground ADS-B receiver, the drone, and the self-sensing sensor on the drone.

[0016] S22. Add the confidence levels of the three signals to obtain the total signal confidence level;

[0017] S23. The ratio of the confidence level of each signal to the total confidence level of the signals is taken as the confidence level ratio of each signal.

[0018] S24. Multiply each signal confidence ratio by the corresponding attitude vector to obtain a local attitude vector. Add the three local attitude vectors to obtain the manned aircraft accurate attitude vector.

[0019] Furthermore, the formula for obtaining the signal reliability is:

[0020]

[0021] in, Let i be the confidence level of the i-th signal. Let i be the reliability level of the i-th signal source. The value range is 0~1. For the continuity of the i-th signal, The value range is 0~1. Let be the normalized value of the reciprocal of the update time interval for the i-th signal. Assigning weights to the reliability level of the signal source. For signal continuity weights, To update the time interval weight, the value of i ranges from 1 to 3. When i equals 1, it corresponds to the signal of data acquired by the ground ADS-B receiver. When i equals 2, it corresponds to the signal of data acquired by the UAV. When i equals 3, it corresponds to the signal of data acquired by the self-sensing sensor carried by the UAV.

[0022] Furthermore, S3 includes the following sub-steps:

[0023] S31. Subtract the accurate attitude vectors of the manned aircraft from adjacent time points to obtain the accurate attitude change sequence of the manned aircraft.

[0024] S32. Separate the data belonging to position, velocity and heading angle from the accurate attitude change sequence of the manned aircraft to form a position change sequence, a velocity change sequence and a heading angle change sequence;

[0025] S33. Using a sliding window, slide it over the position change sequence, velocity change sequence, and heading angle change sequence respectively, and extract the mean and standard deviation under each sliding window to obtain the position change feature matrix, velocity change feature matrix, and heading angle change feature matrix. The first row of the matrix is ​​the mean, and the second row is the standard deviation.

[0026] S34. Configure corresponding weights and biases for the intent category set, and weight the three feature matrices respectively to obtain the linear discrimination score for each intent category;

[0027] S35. Use the softmax function to process the linear discrimination score to obtain the intent probability of each intent category at each time step.

[0028] Furthermore, S4 includes the following sub-steps:

[0029] S41. Predict the future position of straight-line travel using a uniform linear model;

[0030] S42. Predict the future position of the turn using a coordinated turning model;

[0031] S43. Use a vertical velocity model to predict the future position of ascent / descent;

[0032] S44. Multiply the intent probability of each intent category by the future position of the corresponding intent category, and add all the multiplication results to obtain the mixed predicted position;

[0033] S45. Obtain the mixed prediction covariance based on the difference between the future location of each intent category and the mixed prediction location.

[0034] Furthermore, the formula for the mixed prediction covariance in S45 is:

[0035]

[0036] in, For the first Mixed forecast covariance at the forecast time For the first Predicted time number The prediction covariance matrix for each intent category, For the first Hybrid prediction location at the predicted time For the first Predicted time number Future location of each intent category For the first Time of the first The intent probability of each intent category, where I is the set of intent categories. Numbering for future moments. As a number for historical moments, The number representing the intent category. For transpose operation, for The transpose of .

[0037] Furthermore, S5 includes the following sub-steps:

[0038] S51. Construct a kernel function for any point in space based on the mixed prediction location and the mixed prediction covariance;

[0039] S52. When turning, set the directional magnification factor;

[0040] S53. Multiply the directional amplification factor by the kernel function to obtain the risk of any point in space in the future;

[0041] S54. Take the average risk of all points in any grid in the three-dimensional spatial grid in the future to obtain the risk of that grid, thus forming the intentional risk field.

[0042] Furthermore, the formula for the kernel function is:

[0043]

[0044] in, For any point in space In the Kernel function for predicting time, For the first Hybrid prediction location at the predicted time For the first Mixed forecast covariance at the forecast time Let be the determinant of the covariance matrix. Let be the coordinates of any point in space. It is an exponential function. Numbering for future moments. This refers to the numbering of historical moments.

[0045] Furthermore, the formula for the directional magnification factor is:

[0046]

[0047] in, This is the directional magnification factor during turning. Forward weighting factor, For indicator functions, Let be the coordinates of any point in space. For the first Hybrid prediction location at the predicted time This is the unit direction vector corresponding to the heading angle of the manned aircraft. Let be the angle between the vector from any point in space to the manned aircraft and the heading direction, and max is the maximum of the two.

[0048] The beneficial effects of this invention are as follows:

[0049] 1. In step S2, this invention calculates the signal reliability of three types of data: ground ADS-B receiver, UAV broadcast signal, and UAV self-sensing sensor. Based on the reliability, the data is weighted and fused, which effectively balances the advantages and disadvantages of data from different sources (such as the wide coverage of ADS-B and the low-altitude blind spot filling capability of the self-sensing sensor), reduces the fusion error caused by data heterogeneity, and finally outputs a high-precision manned aircraft attitude vector.

[0050] 2. This invention extracts the sequence features of position, velocity, and heading angle changes at adjacent moments in step S3, and calculates the probability of different flight intentions (such as straight flight, turning, climbing, etc.) based on continuous motion features, breaking through the limitation of "only relying on the current state"; step S4 further fuses the future position prediction results under multiple intentions based on the intention probability to obtain the mixed predicted position and mixed predicted covariance, which not only covers all possible motion trends, but also quantifies the prediction uncertainty, and is more in line with the actual scenario of dynamic flight of manned aircraft compared with static prediction methods.

[0051] 3. This invention is based on a hybrid predicted position and covariance, combined with a directional amplification factor to construct a kernel function. This allows the generated risk field to not only correlate with the probability distribution of the manned aircraft's future position but also reflect the influence of flight direction on the risk propagation range. Compared with existing technologies that only generate symmetrical, static risk regions, this invention can present a dynamically changing, directional risk distribution in the airspace, which is more consistent with the actual collision avoidance logic of UAVs and the airspace operation rules, thus improving the accuracy of the generated risk field.

[0052] 4. Because the risk field generated by this invention can reflect the possible future positions of manned aircraft and their uncertainties in real time, unmanned aerial vehicles (UAVs) can achieve more accurate path planning and risk avoidance based on this. This invention can significantly reduce the probability of aerial conflicts and has important application value for track management, conflict early warning, and collaborative scheduling in complex mixed airspace. Attached Figure Description

[0053] Figure 1 This is a flowchart of an ADS-B-based method for integrated airspace surveillance of unmanned and manned aircraft. Detailed Implementation

[0054] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0055] like Figure 1 As shown, a method for integrated airspace surveillance based on ADS-B for unmanned aerial vehicles and manned aircraft includes the following steps:

[0056] S1. Collect manned aircraft attitude vectors through ground ADS-B receiver and UAV broadcast, and at the same time collect relative measurement vectors using UAV self-sensing sensors, and convert the relative measurement vectors into manned aircraft absolute attitude vectors in global coordinate system.

[0057] S2. Obtain the signal confidence level from the ground ADS-B receiver, the UAV, and the self-sensing sensor on the UAV. Based on the confidence level of the three, weight the attitude vectors of the three to obtain the accurate attitude vector of the manned aircraft.

[0058] S3. Subtract the accurate attitude vectors of the manned aircraft from adjacent time points to obtain the position change sequence, velocity change sequence, and heading angle change sequence. Extract features from the three change sequences and obtain the intent probability for each intent category based on the features.

[0059] S4. Predict the future location for each intent category, and based on the intent probability for each intent category, obtain the mixed predicted location and the mixed predicted covariance.

[0060] S5. Based on the mixed prediction location and mixed prediction covariance, construct a kernel function for any point in space and introduce a directional amplification factor to generate the intentional risk field.

[0061] In this embodiment, the attitude vector of the manned aircraft collected by the ground ADS-B receiver in S1 includes: the manned aircraft position decoded by the ground, the speed of the manned aircraft decoded by the ground, and the heading angle decoded by the ground.

[0062] The attitude vector of the manned aircraft collected by the UAV includes: the manned aircraft position decoded by the UAV, the manned aircraft speed decoded by the UAV, and the heading angle decoded by the UAV.

[0063] The relative measurement vectors collected by the self-sensing sensors on the drone include: the relative distance between the drone and the manned aircraft, the azimuth angle, and the pitch angle.

[0064] S1 specifically involves: acquiring the manned aircraft's attitude vector via a ground-based ADS-B receiver, acquiring the manned aircraft's attitude vector via UAV broadcast, and acquiring the UAV's relative measurement vector via its onboard self-sensing sensors; then converting the relative measurement vector into the manned aircraft's absolute attitude vector in a global coordinate system. ,in, For the first The position of the manned aircraft in the absolute attitude vector at a given moment. It is a homogeneous transformation matrix. For the first The position and attitude of the drone in the global coordinate system at a given moment. To convert the relative measurement vector into a three-dimensional position vector in the UAV coordinate system based on the sensor model. , The relative distance between the drone and the manned aircraft. For pitch angle, It is the azimuth angle. The homogeneous transformation matrix for transforming the UAV coordinate system to the global coordinate system is calculated from the UAV's position and attitude in the global coordinate system.

[0065] The heading angle of the UAV in the absolute attitude vector of the manned aircraft is equal to The manned aircraft's velocity in its absolute attitude vector can be obtained from the position changes at adjacent time points.

[0066] Self-sensing sensors are modules that can measure relative distance, azimuth, and elevation angles, such as millimeter-wave radar.

[0067] In this embodiment, S2 includes the following sub-steps:

[0068] S21. Obtain the signal reliability from the ground ADS-B receiver, the drone, and the self-sensing sensor on the drone.

[0069] S22. Add the confidence levels of the three signals to obtain the total signal confidence level;

[0070] S23. The ratio of the confidence level of each signal to the total confidence level of the signals is taken as the confidence level ratio of each signal.

[0071] S24. Multiply each signal confidence ratio by the corresponding attitude vector to obtain a local attitude vector. Add the three local attitude vectors to obtain the manned aircraft accurate attitude vector.

[0072] In this embodiment, the formula for obtaining signal confidence level is:

[0073]

[0074] in, Let i be the confidence level of the i-th signal. Let i be the reliability level of the i-th signal source. The value range is 0~1. For the continuity of the i-th signal, The value range is 0~1. Let be the normalized value of the reciprocal of the update time interval for the i-th signal. Assigning weights to the reliability level of the signal source. For signal continuity weights, To update the time interval weight, the value of i ranges from 1 to 3. When i equals 1, it corresponds to the signal of data acquired by the ground ADS-B receiver. When i equals 2, it corresponds to the signal of data acquired by the UAV. When i equals 3, it corresponds to the signal of data acquired by the self-sensing sensor carried by the UAV.

[0075] In this embodiment, the source reliability level is: ground ADS-B receiver > UAV > self-sensing sensor carried by the UAV.

[0076] The normalization formula is:

[0077]

[0078] in, The reciprocal of the update time interval for the i-th signal. It is the minimum value of the reciprocal of the signal update time interval. It is the maximum value of the reciprocal of the signal update time interval.

[0079] This invention calculates credibility through a multi-dimensional weighted calculation of "signal source reliability level + continuity + update frequency" to avoid the one-sidedness of a single dimension (such as only looking at the source); at the same time, it combines the characteristics of ground ADS-B, UAV broadcast, and self-sensing sensors (such as ground ADS-B having a higher source level) to assign differentiated weights, so that the fused manned aircraft accurate attitude vector is closer to the real state.

[0080] In this embodiment, based on the data source characteristics of the low-altitude surveillance scenario, the example value range is as follows: ground ADS-B receiver: R1 is set to 0.8, UAV broadcast channel R2 is set to 0.6; UAV self-sensing sensor R3 is set to 0.4.

[0081] The rule for setting signal continuity is based on "successfully acquiring signals at multiple consecutive moments". An example value is: ≥5 consecutive signal acquisition moments: S i Set to 0.9 (stable and continuous signal); continuously acquire signals at 3-4 time points: S i Set to 0.6; continuously acquire signals for 1-2 time points: S i Set to 0.3; acquire only one time point signal: S i Set it to 0.1.

[0082] In this embodiment, the formula for obtaining the accurate attitude vector of the manned aircraft is:

[0083]

[0084] in, This is the accurate attitude vector of the manned aircraft. Let i be the i-th attitude vector. Let be the confidence ratio of the i-th signal.

[0085] In this embodiment, S3 includes the following sub-steps:

[0086] S31. Subtract the accurate attitude vectors of the manned aircraft from those of adjacent time points to obtain the accurate attitude change sequence of the manned aircraft, that is, subtract the accurate attitude vectors of the manned aircraft at time t+1 from those at time t.

[0087] S32. Separate the data belonging to position, velocity and heading angle from the accurate attitude change sequence of the manned aircraft to form a position change sequence, a velocity change sequence and a heading angle change sequence;

[0088] S33. Using a sliding window, slide it over the position change sequence, velocity change sequence, and heading angle change sequence respectively, and extract the mean and standard deviation under each sliding window to obtain the position change feature matrix, velocity change feature matrix, and heading angle change feature matrix. The first row of each matrix is ​​the mean of each window, and the second row is the standard deviation of each window.

[0089] S34. Assign corresponding weights and biases to the intent category set, and then weight the three feature matrices to obtain the linear discrimination score for each intent category:

[0090]

[0091] in, For the first The first moment Linear discriminant scores for each intent category, For the first The weight matrix of each intent category to the k-th feature matrix. For the first The bias matrix of each intent category with respect to the k-th feature matrix. k takes values ​​from 1 to 3, corresponding to the th, ... The feature matrices for position changes, velocity changes, and heading angle changes at any given moment are included. The intent categories in the intent category set include: straight line, turn, climb, and descent.

[0092] S35. The softmax function is used to process the linear discriminant scores to obtain the intent probability of each intent category at each time step:

[0093]

[0094] in, For the first Time of the first The probability of intent for each intent category. Let I be an exponential function, and let I be the set of intent categories.

[0095] This invention effectively captures the changing patterns of position, speed, and heading angle through feature engineering using a sliding window to extract the mean and standard deviation (e.g., the standard deviation of the heading angle increases significantly when turning). Combined with the logic of "multi-feature matrix weighting + softmax probability transformation", it can accurately map different motion features to intention categories, reduce the possibility of misjudgment of a single feature, and make the recognition results of intentions such as straight lines and turns more reliable.

[0096] In this embodiment, S4 includes the following sub-steps:

[0097] S41. Predict the future position of straight-line travel using a uniform linear model:

[0098]

[0099]

[0100] in, For the first Predict the future x, y positions of the line moving straight ahead at a given time. Let x and y be the positions of the manned aircraft's accurate attitude vector at time t. The magnitude of the horizontal velocity. Numbering for future moments. This is the unit direction vector corresponding to the heading angle of the manned aircraft. Let be the height in the manned aircraft's accurate attitude vector at time t. For the first Predict the future height and position of the line of sight at the predicted moment, and the future x, y position. and height Together, they constitute the future position for going straight;

[0101] S42. Predict the future position of the turn using a coordinated turning model:

[0102]

[0103]

[0104]

[0105] in, For the first The heading angle at the predicted time. For the first The heading angle at any moment, For turning angular velocity, For the first Predict the future x and y positions at the moment of the turn. For the first Predict the future height and position of the turn at the predicted moment, and the future x and y positions. and height Together, they constitute the future position of the turn;

[0106] S43. Predict future climbing / descent positions using a vertical velocity model:

[0107]

[0108] in, For the first Predict the future position of the ascent / descent at the current moment. For the first Vertical velocity at time t, It is vertical acceleration;

[0109] S44. Multiply the intent probability of each intent category by the future position of the corresponding intent category, and sum all the multiplication results to obtain the mixed predicted position:

[0110]

[0111] in, For the first Hybrid prediction location at the predicted time For the first Future location of each intent category For the first Time of the first The intent probability of each intent category, where I is the set of intent categories. Numbering for future moments. As a number for historical moments, The number representing the intent category;

[0112] S45. Obtain the mixed prediction covariance based on the difference between the future location of each intent category and the mixed prediction location.

[0113] In this embodiment, the formula for the mixed prediction covariance in S45 is:

[0114]

[0115] in, For the first Mixed forecast covariance at the forecast time For the first Predicted time number The prediction covariance matrix for each intent category, For the first Hybrid prediction location at the predicted time For the first Predicted time number Future location of each intent category For the first Time of the first The intent probability of each intent category, where I is the set of intent categories. Numbering for future moments. As a number for historical moments, The number representing the intent category. For transpose operation, for The transpose of .

[0116] For the straight line intention: covariance matrix This primarily covers horizontal velocity measurement errors and uncertainties in x and y position predictions; regarding turning intentions: covariance matrix. This primarily covers the uncertainty of angular velocity and the uncertainty of horizontal position (x, y) prediction; regarding climb / descent intentions: covariance matrix. It mainly covers vertical velocity measurement errors, acceleration prediction uncertainties, and height position prediction uncertainties.

[0117] This invention employs tailored professional models (uniform linear motion / coordinated turning / vertical velocity models) for three typical motion intentions: "straight ahead," "turning," and "climbing / descending." This allows for precise matching of the patterns of different motion states, avoiding the prediction bias of a single model for complex motions, and ensuring that the position prediction for each intention more closely reflects the actual motion characteristics. This invention obtains a hybrid predicted position through "intention probability weighted fusion," and simultaneously calculates the hybrid covariance by combining the deviation between the intended position and the hybrid position. This integrates the prediction information of all possible intentions and quantifies the uncertainty brought by different intentions, enabling the final prediction result to adapt to dynamic scenarios with "multi-intention switching" and reducing the risk of prediction failure due to misjudgment of a single intention.

[0118] In this embodiment, S5 includes the following sub-steps:

[0119] S51. Construct a kernel function for any point in space based on the mixed prediction location and the mixed prediction covariance;

[0120] S52. When turning, set the directional magnification factor;

[0121] S53. Multiply the directional amplification factor by the kernel function to obtain the risk of any point in space in the future;

[0122] S54. Take the average risk of all points in any grid in the three-dimensional spatial grid in the future to obtain the risk of that grid. Arrange the risks of each grid according to the grid position to form the intended risk field.

[0123] In this embodiment, the formula for the kernel function is:

[0124]

[0125] in, For any point in space In the Kernel function for predicting time, For the first Hybrid prediction location at the predicted time For the first Mixed forecast covariance at the forecast time Let be the determinant of the covariance matrix. Let be the coordinates of any point in space. It is an exponential function. Numbering for future moments. As a number for historical moments, To Take the reverse.

[0126] In this embodiment, the formula for the directional magnification factor is:

[0127]

[0128] in, This is the directional magnification factor during turning. Forward weighting factor, This is an indicator function; its value is 1 if the intention of the manned aircraft is to "turn," and 0 otherwise. Let be the coordinates of any point in space. For the first Hybrid prediction location at the predicted time This is the unit direction vector corresponding to the heading angle of the manned aircraft. Let be the angle between the vector from any point in space to the manned aircraft and the heading direction, and max is the maximum of the two.

[0129] This invention uses a Gaussian kernel function (S51) constructed with "hybrid predicted position + covariance" to quantify the spatial uncertainty of position prediction. The more dispersed the probability distribution of the UAV's future position (the larger the covariance), the higher the kernel function value (basic risk) of the corresponding area, which is consistent with the actual logic that "uncertain areas have higher risks". The directional amplification factor (S52-S53) during turning specifically strengthens the risk weight of the area ahead of the flight path, matching the actual scenario that "the direction ahead is the direction of the movement trend during turning, and the risk needs to be focused on", avoiding the bias of indiscriminate assessment of the entire area.

[0130] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for integrated airspace surveillance combining unmanned aerial vehicles (UAVs) and manned aircraft based on ADS-B, characterized in that, Includes the following steps: S1. Collect the manned aircraft attitude vector through the ground ADS-B receiver and the UAV broadcast through two channels. At the same time, use the UAV self-sensing sensor to collect the relative measurement vector and convert the relative measurement vector into the manned aircraft absolute attitude vector in the global coordinate system. The relative measurement vector includes: the relative distance between the UAV and the manned aircraft, the azimuth angle and the pitch angle. S2. Obtain the signal reliability from the ground ADS-B receiver, the drone broadcast and the self-sensing sensor on the drone respectively. Based on the reliability of the three, weight the attitude vectors of the three to obtain the accurate attitude vector of the manned aircraft. S3. Subtract the accurate attitude vectors of the manned aircraft from adjacent time points to obtain the position change sequence, velocity change sequence, and heading angle change sequence. Extract features from these three change sequences and obtain the intent probability for each intent category based on these features: Use a sliding window to slide over the position change sequence, velocity change sequence, and heading angle change sequence respectively, and extract the mean and standard deviation under each sliding window to obtain the position change feature matrix, velocity change feature matrix, and heading angle change feature matrix, where the first row of the matrix is ​​the mean and the second row is the standard deviation; Assign corresponding weights and biases to the intent category set, and weight the three feature matrices respectively to obtain the linear discrimination score for each intent category; Use the softmax function to process the linear discrimination score to obtain the intent probability for each intent category at each time point; S4. Predict the future location for each intent category, and based on the intent probability for each intent category, obtain the mixed predicted location and mixed predicted covariance: A uniform linear motion model is used to predict the future position for straight-line travel; a coordinated turning model is used to predict the future position for turning; a vertical velocity model is used to predict the future position for ascent / descent; the intent probability of each intent category is multiplied by the corresponding future position, and all multiplication results are summed to obtain the mixed predicted position; the mixed predicted covariance is obtained based on the difference between the future position of each intent category and the mixed predicted position. ； in, For the first Mixed forecast covariance at the forecast time For the first Predicted time number The prediction covariance matrix for each intent category, For the first Hybrid prediction location at the predicted time For the first Predicted time number Future location of each intent category For the first Time of the first The intent probability of each intent category, where I is the set of intent categories. Numbering for future moments. As a number for historical moments, The number representing the intent category. For transpose operation, for The transpose of the term; S5. Based on the mixed prediction location and mixed prediction covariance, construct a kernel function for any point in space, and introduce a directional amplification factor to generate the intentional risk field: Based on the mixed predicted location and mixed predicted covariance, a kernel function is constructed for any point in space; a directional magnification factor is set when turning; the directional magnification factor is multiplied by the kernel function to obtain the future risk of any point in space; the average future risk of all points in any grid in the three-dimensional spatial grid is taken to obtain the risk of that grid, thus forming the intended risk field: ； in, This is the directional magnification factor during turning. Forward weighting factor, For indicator functions, Let be the coordinates of any point in space. For the first Hybrid prediction location at the predicted time This is the unit direction vector corresponding to the heading angle of the manned aircraft. is the angle between "the vector from any point in space to the manned aircraft" and "the heading direction", and max is the maximum of the two.

2. The method for integrated airspace surveillance of UAVs and manned aircraft based on ADS-B according to claim 1, characterized in that, The attitude vector of the manned aircraft collected by the ground ADS-B receiver in S1 includes: the manned aircraft position, the manned aircraft speed, and the heading angle. The attitude vector of the manned aircraft collected by the UAV includes: the manned aircraft position decoded by the UAV, the manned aircraft speed decoded by the UAV, and the heading angle decoded by the UAV.

3. The method for integrated airspace surveillance of UAVs and manned aircraft based on ADS-B according to claim 1, characterized in that, S2 includes the following steps: S21. Obtain the signal reliability from the ground ADS-B receiver, the drone broadcast, and the self-sensing sensor on the drone. S22. Add the confidence levels of the three signals to obtain the total signal confidence level; S23. The ratio of the confidence level of each signal to the total confidence level of the signals is taken as the confidence level ratio of each signal. S24. Multiply each signal confidence ratio by the corresponding attitude vector to obtain a local attitude vector. Add the three local attitude vectors to obtain the manned aircraft accurate attitude vector.

4. The ADS-B-based UAV and manned aircraft integrated airspace surveillance method according to claim 1 or 3, characterized in that, The formula for obtaining signal reliability is: ； in, Let i be the confidence level of the i-th signal. Let i be the reliability level of the i-th signal source. The value range is 0~1. For the continuity of the i-th signal, The value range is 0~1. Let be the normalized value of the reciprocal of the update time interval for the i-th signal. Assigning weights to the reliability level of the signal source. For signal continuity weights, To update the time interval weight, the value of i ranges from 1 to 3. When i equals 1, it corresponds to the signal of data acquired by the ground ADS-B receiver. When i equals 2, it corresponds to the signal of data acquired by the UAV. When i equals 3, it corresponds to the signal of data acquired by the self-sensing sensor carried by the UAV.

5. The method for integrated airspace surveillance of UAVs and manned aircraft based on ADS-B according to claim 1, characterized in that, The formula for the kernel function is: ； in, For any point in space In the Kernel function for predicting time, For the first Hybrid prediction location at the predicted time For the first Mixed forecast covariance at the forecast time Let be the determinant of the covariance matrix. Let be the coordinates of any point in space. It is an exponential function. Numbering for future moments. This refers to the numbering of historical moments.

Images