Intelligent management and control method and system for low-altitude aircraft

Abstract

The application provides a wisdom management and control method and system for low-altitude aircraft, and relates to the technical field of low-altitude aircraft management and control. First, flight state message streams periodically broadcast by low-altitude aircraft are received. Then, airspace projection gridding processing is performed on the message streams to obtain a single-grid occupation marker set and a grid occupation aircraft identifier mapping relationship. Then, occupation state continuation analysis processing is performed on the single-grid occupation marker set to generate grid occupation continuation mode records. Then, cross-comparison structure is used to perform cross-comparison processing on the mode records of adjacent grid units. Finally, according to the cross-comparison results, a flight track intersection situation is determined, flight avoidance scheduling instructions are generated, and the corresponding aircraft control terminal is distributed, so that the wisdom management and control of low-altitude aircraft are realized.

Description

Technical Field

[0001] This invention relates to the field of low-altitude aircraft control technology, and more specifically, to an intelligent control method and system for low-altitude aircraft. Background Technology

[0002] With the increasing frequency of low-altitude flight activities, the safety management of low-altitude airspace faces unprecedented challenges. As drones and other low-altitude aircraft are widely used in logistics, agricultural plant protection, surveying, and exploration, the number of aircraft in low-altitude airspace has increased dramatically, and their flight paths have become more complex and diverse. Existing methods for managing low-altitude aircraft largely rely on traditional radar monitoring and manual command and dispatch. However, radar monitoring suffers from high costs, limited coverage, and insufficient detection accuracy for small low-altitude aircraft, making it difficult to comprehensively and accurately grasp the real-time status of all aircraft in low-altitude airspace. Meanwhile, manual command and dispatch is not only inefficient and susceptible to human interference, but also struggles to quickly and accurately assess the overlap of flight paths between aircraft in large-scale low-altitude flight scenarios, making it impossible to issue timely and effective avoidance commands. This increases the risk of low-altitude collisions and seriously threatens the safety of low-altitude flights. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an intelligent control method for low-altitude aircraft, the method comprising:

[0004] Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field.

[0005] The flight status message stream is subjected to airspace projection rasterization processing. Based on the instantaneous position coordinate field, each flight status message is mapped to the corresponding grid cell in the preset airspace control grid system to obtain the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp.

[0006] For each grid cell, the single grid occupancy marker set is processed by time sequence occupancy status change analysis to generate a grid occupancy change pattern record for that grid cell under multiple consecutive acquisition time stamps. The grid occupancy change pattern record includes an occupancy duration sequence and an occupancy aircraft switching sequence record.

[0007] The pre-defined multi-grid occupancy change pattern cross-comparison structure is used to perform cross-comparison processing on the grid occupancy change pattern records of adjacent grid cells, generating occupancy pattern cross-comparison difference records between adjacent grid cells;

[0008] Based on the cross-comparison difference records of the occupancy patterns, the aircraft track interleaving situation category and track interleaving spatiotemporal distribution record between the adjacent grid cells are determined. Based on the aircraft track interleaving situation category and the track interleaving spatiotemporal distribution record, a flight avoidance scheduling instruction carrying a grid coordinate chain is generated and the flight avoidance scheduling instruction is distributed to the corresponding aircraft control terminal.

[0009] Furthermore, embodiments of the present invention also provide an intelligent control system for low-altitude aircraft, comprising:

[0010] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the aforementioned intelligent control method for low-altitude aircraft by executing the machine-executable instructions.

[0011] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of an intelligent control system for low-altitude aircraft reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the intelligent control system for low-altitude aircraft to execute the aforementioned intelligent control method for low-altitude aircraft.

[0012] Based on the above, by receiving the periodically broadcast flight status message streams from low-altitude aircraft, key information such as the instantaneous position, speed, and heading of the aircraft is comprehensively and accurately obtained. The flight status message streams are then subjected to airspace projection rasterization processing. Occupancy status continuity analysis is performed on the single-grid occupancy marker set of each grid cell, generating grid occupancy continuity pattern records. This allows for in-depth exploration of the dynamic changes in aircraft occupancy within each grid cell. By utilizing a multi-grid occupancy continuity pattern cross-comparison structure to cross-compare adjacent grid cells, the overlapping of aircraft tracks between adjacent grid cells can be accurately detected, generating detailed records of cross-comparison differences in occupancy patterns. Based on these cross-comparison difference records, the types and spatiotemporal distribution of aircraft track overlap situations are determined, and flight avoidance scheduling instructions carrying grid coordinate chains are generated. This enables precise avoidance scheduling of low-altitude aircraft, effectively reducing the risk of low-altitude collisions and significantly improving the safety and flight efficiency of low-altitude airspace. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the execution flow of the intelligent control method for low-altitude aircraft provided in an embodiment of the present invention.

[0014] Figure 2This is a schematic diagram of exemplary hardware and software components of an intelligent control system for low-altitude aircraft provided in an embodiment of the present invention. Detailed Implementation

[0015] Figure 1 This is a flowchart illustrating an intelligent control method for low-altitude aircraft provided in one embodiment of the present invention, which will be described in detail below.

[0016] Step S110: Receive the flight status message stream periodically broadcast by low-altitude aircraft in the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition time stamp field.

[0017] In a scenario applying to urban low-altitude drone traffic management, the airspace to be managed is the low-altitude airspace below 120 meters above the city center. Step S110 receives flight status message streams broadcast by low-altitude aircraft within this airspace via a 5G network. This flight status message stream contains a sequence of messages with an aircraft identifier. Each message carries instantaneous position coordinates (longitude, latitude, altitude), instantaneous velocity vector fields (northward velocity, eastward velocity, vertical velocity), heading azimuth field (0 to 360 degrees), and a data collection timestamp field (millisecond precision). Reception of all messages has been authorized by the aircraft operator and complies with data security and privacy protection regulations.

[0018] Step S120: Perform airspace projection rasterization processing on the flight status message stream, and map each flight status message to the corresponding grid cell in the preset airspace control grid system based on the instantaneous position coordinate field, to obtain the single grid occupancy mark set and grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp.

[0019] Step S121: Extract the instantaneous position coordinate field from the flight status message stream one by one, and obtain the longitude coordinate component, latitude coordinate component and altitude coordinate component of the instantaneous position coordinate field in the three-dimensional airspace coordinate system.

[0020] For each received flight status message, the data offset defined in the message format is parsed. The longitude coordinate component is read from the specified byte offset position in the message, using degrees in the WGS84 coordinate system, with a value range of -180 to 180. The latitude coordinate component is read from the next byte offset position, with a value range of -90 to 90. The altitude coordinate component, representing the altitude in meters, is read from the next byte offset position.

[0021] Step S122: Retrieve the preset airspace control grid system. The airspace control grid system divides the airspace to be controlled in the three-dimensional airspace coordinate system according to a uniform grid side length, forming longitude layer grid units, latitude layer grid units and altitude layer grid units with unique grid numbers.

[0022] The pre-defined airspace control grid system is a three-dimensional grid structure. In the longitude direction, the longitude range of the airspace to be controlled, from the minimum longitude Lmin to the maximum longitude Lmax, is divided into multiple longitude-layer grid units with a fixed longitude step size dLon. Each longitude-layer grid unit is assigned a unique longitude grid number Ig. In the latitude direction, the latitude range, from the minimum latitude Bmin to the maximum latitude Bmax, is divided into multiple latitude-layer grid units with a fixed latitude step size dLat. Each latitude-layer grid unit is assigned a unique latitude grid number Jg. In the altitude direction, the altitude range, from the minimum altitude Hmin to the maximum altitude Hmax, is divided into multiple altitude-layer grid units with a fixed altitude step size dH. Each altitude-layer grid unit is assigned a unique altitude grid number Kg. Each grid unit is uniquely determined by its longitude-layer number Ig, latitude-layer number Jg, and altitude-layer number Kg. The three-dimensional grid number is represented as (Ig, Jg, Kg).

[0023] Step S123: Perform coordinate assignment matching processing based on the longitude coordinate components of the instantaneous position coordinate field and the boundary coordinate range of each longitude layer grid unit in the airspace control grid system to determine the target longitude layer grid number to which the instantaneous position coordinate field falls.

[0024] The formula Ig is used to calculate the longitude grid number corresponding to the longitude coordinate component Lo: Ig = floor((Lo - Lmin) / dLon). If Lo equals Lmax, then Ig takes the maximum value. This formula maps the longitude coordinate to the corresponding longitude grid cell. For example, for a longitude range of 116° to 117° E, with a step size dLon of 0.01°, the longitude coordinate 116.035° falls into the 3rd grid cell, and Ig is 3.

[0025] Step S124: Perform coordinate attribution matching processing based on the latitudinal coordinate components of the instantaneous position coordinate field and the boundary coordinate range of each latitudinal layer grid unit in the airspace control grid system to determine the target latitudinal layer grid number to which the instantaneous position coordinate field falls.

[0026] Calculate the latitude layer grid number corresponding to the latitude coordinate component La: Jg = floor((La - Bmin) / dLat). If La equals Bmax, then Jg takes the maximum value. This formula maps the latitude coordinates to the corresponding latitude layer grid cell.

[0027] Step S125: Perform coordinate assignment matching processing based on the height coordinate component of the instantaneous position coordinate field and the boundary coordinate range of each height layer grid unit in the airspace control grid system to determine the target height layer grid number to which the instantaneous position coordinate field falls.

[0028] Calculate the height layer grid number corresponding to the height coordinate component Al: Kg = floor((Al - Hmin) / dH). If Al equals Hmax, then Kg takes the maximum value. This formula maps the height coordinates to the corresponding height layer grid cell.

[0029] Step S126: Combine the target longitude layer grid number, the target latitude layer grid number, and the target altitude layer grid number to generate the three-dimensional grid number of the target airspace control grid unit corresponding to the flight status message.

[0030] The longitude layer grid number Ig, latitude layer grid number Jg, and altitude layer grid number Kg obtained in steps S123, S124, and S125 are combined into a triple (Ig, Jg, Kg) as the three-dimensional grid number of the target airspace control grid unit mapped to the flight status message.

[0031] Step S127: Establish an association record between the aircraft identifier field of the current flight status message and the three-dimensional grid number, and store it in the grid occupancy temporary mapping table entry corresponding to the acquisition timestamp.

[0032] Create a hash table MapT with the acquisition timestamp as the key. For the current acquisition timestamp Tm, look up the value corresponding to key Tm in hash table MapT. If it does not exist, create an empty list as the value. Store the aircraft identifier Id of the flight status message and its corresponding 3D grid number (Ig, Jg, Kg) as an association pair in the list corresponding to the acquisition timestamp. Each element in the list is (Id, Ig, Jg, Kg).

[0033] Step S128: Traverse all flight status messages with the same acquisition timestamp field in the flight status message stream, perform coordinate attribution matching processing one by one, and generate a set of multiple association records between the multi-dimensional grid number and the aircraft identifier field under the acquisition timestamp.

[0034] For each acquisition time stamp Tm, process all received flight status messages at that time, repeating steps S121 to S127 until all messages at that time have been processed. Obtain the set of associated records SetR={(Id1, Ig1, Jg1, Kg1), (Id2, Ig2, Jg2, Kg2), ...} for all aircraft identifiers and their respective grid numbers under that acquisition time stamp Tm.

[0035] Step S129: The set of multiple associated records of the multi-dimensional grid number and the aircraft identifier field are grouped and aggregated according to the three-dimensional grid number. All aircraft identifier fields belonging to the same three-dimensional grid number are summarized to form a list of single-grid occupied aircraft identifiers for the three-dimensional grid number at the time stamp of the acquisition.

[0036] Iterate through the associated record set SetR generated in step S128, using the 3D grid number (Ig, Jg, Kg) as the grouping key, and collect all aircraft identifiers Id with the same 3D grid number into a list. For each grid cell (Ig, Jg, Kg), obtain the list of occupying aircraft identifiers at that moment, Lid(Ig, Jg, Kg) = [Id_a, Id_b, Id_c, ...].

[0037] Step S1210: Based on the existence status of the entries in the aircraft identifier field of the single-grid occupancy aircraft identifier list, generate a single-grid occupancy mark for the 3D grid number under the acquisition time stamp. Arrange the single-grid occupancy marks of all 3D grid numbers in the order of the acquisition time stamp field to generate a set of single-grid occupancy marks for each grid cell under each acquisition time stamp. Establish a mapping relationship between each 3D grid number and the corresponding single-grid occupancy aircraft identifier list under the acquisition time stamp as the grid occupancy aircraft identifier mapping relationship.

[0038] For each grid cell, check the occupied aircraft identifier list Lid generated in step S129. If Lid is not empty (containing at least one aircraft identifier), set the single-grid occ tag Oc for that grid cell at the acquisition time stamp Tm to 1, indicating that it is occupied; if Lid is empty, set Oc to 0, indicating that it is not occupied. Organize the single-grid occ tags of all grid cells into a three-dimensional array Occ[Ig][Jg][Kg] with the same grid architecture as the grid structure, as the set of single-grid occupancy tags at the acquisition time stamp Tm. At the same time, establish a mapping relationship from the three-dimensional grid number (Ig, Jg, Kg) to the occupied aircraft identifier list Lid at that time, as the grid occupied aircraft identifier mapping relationship.

[0039] Step S130: Perform occupancy status continuity analysis on the single grid occupancy mark set of each grid cell according to the time sequence to generate a grid occupancy continuity pattern record for the grid cell under multiple continuous acquisition time stamps. The grid occupancy continuity pattern record includes an occupancy duration sequence and an occupancy aircraft switching sequence record.

[0040] Step S131: Traverse the single-grid occupancy marker set of the target grid cell in the chronological order of the acquisition timestamps, identify the acquisition timestamp when the single-grid occupancy marker changes from the occupied state to the unoccupied state as the occupancy termination timestamp, and identify the acquisition timestamp when the single-grid occupancy marker changes from the unoccupied state to the occupied state as the occupancy start timestamp.

[0041] Obtain the sequence of single-grid occupancy marker values [Oc1, Oc2, Oc3,..., Ocn] corresponding to the target grid cell (Ig, Jg, Kg) at consecutive acquisition timestamps [T1, T2, T3,..., Tn]. Starting from i = 1 and traversing to i = n - 1, if Oci is equal to 0 and Oc(i + 1) is equal to 1, then record the moment T(i + 1) as an occupancy start timestamp Ts. If Oci is equal to 1 and Oc(i + 1) is equal to 0, then record the moment T(i + 1) as an occupancy termination timestamp Te.

[0042] Step S132: From the occupancy start timestamp to the occupancy termination timestamp, mark the corresponding consecutive acquisition timestamp interval as an occupancy event of the target grid cell, extract the occupancy start timestamp and occupancy termination timestamp corresponding to this occupancy event, calculate the time span between the occupancy termination timestamp and the occupancy start timestamp, and generate an occupancy duration segment of this occupancy event.

[0043] Pair the identified Ts and Te in Step S131 (Ts < Te), and mark the time interval [Ts, Te] as an occupancy event Ev. Calculate the occupancy duration Du = Te - Ts (in milliseconds). Record this occupancy event as a triple (Ts, Te, Du), which is called an occupancy duration segment.

[0044] Step S133: For all occupancy events identified for the target grid cell during all consecutive monitoring periods, calculate the occupancy duration segment of each occupancy event one by one, and arrange the occupancy duration segments in the chronological order of the occurrence of the occupancy events to form an occupancy duration sequence of the target grid cell.

[0045] Repeat Step S131 and Step S132 to process all occupancy events of the target grid cell during the entire monitoring period, and obtain an occupancy duration segment sequence [(Ts1, Te1, Du1), (Ts2, Te2, Du2), (Ts3, Te3, Du3),...]. This sequence is called an occupancy duration sequence SeqDur.

[0046] Step S134: Between two adjacent occupancy events, retrieve the list of single-grid occupied aircraft identifiers corresponding to the previous occupancy event and the list of single-grid occupied aircraft identifiers corresponding to the subsequent occupancy event through the grid occupancy aircraft identifier mapping relationship. Compare the list of single-grid occupied aircraft identifiers of the previous occupancy event with the list of single-grid occupied aircraft identifiers of the subsequent occupancy event, and extract the aircraft identifier fields that exist in the previous list but not in the subsequent list as the switching-out aircraft identifiers, and extract the aircraft identifier fields that do not exist in the previous list but exist in the subsequent list as the switching-in aircraft identifiers.

[0047] For two adjacent occupancy events Ev1 and Ev2 (Te1 of Ev1 < Ts2 of Ev2), obtain the list L1 in the grid occupancy aircraft identifier mapping relationship corresponding to the end time Te1 of Ev1. Obtain the list L2 in the grid occupancy aircraft identifier mapping relationship corresponding to the start time Ts2 of Ev2. Calculate the difference set L1\L2 of L1 with respect to L2 to obtain the set of switching-out aircraft identifiers IdLeave. Calculate the difference set L2\L1 of L2 with respect to L1 to obtain the set of switching-in aircraft identifiers IdEnter. Usually, due to the small size of the grid cells, at most one aircraft occupies a grid cell at the same time. Therefore, IdLeave and IdEnter each contain at most one aircraft identifier.

[0048] Step S135: Pair the switching-out aircraft identifiers and the switching-in aircraft identifiers to form a record of an aircraft identity switching event between these adjacent occupancy events. The record of the aircraft identity switching event includes a switching-before aircraft identifier field and a switching-after aircraft identifier field.

[0049] Use the aircraft identifier in IdLeave extracted in Step S134 as the switching-before aircraft identifier IdPrev, and use the aircraft identifier in IdEnter as the switching-after aircraft identifier IdNext. If IdLeave or IdEnter is empty, the corresponding identifier is set to a null value. Pair IdPrev and IdNext to form a record of an aircraft identity switching event SwRec = (IdPrev, IdNext). This record of the aircraft identity switching event occurs at the boundary time between the end of Ev1 and the start of Ev2.

[0050] Step S136: Arrange all the records of the aircraft identity switching events in chronological order of the occupancy events to generate a record of the occupancy aircraft switching sequence of the target grid cell under multiple consecutive acquisition timestamps. Synchronously align and combine the occupancy duration sequence and the record of the occupancy aircraft switching sequence in the time dimension to generate a record of the grid occupancy continuous change pattern of the target grid cell.

[0051] Arrange all aircraft identity switching event records SwRec in the order of their corresponding occupancy events to obtain the occupancy aircraft switching sequence record SeqSw=[SwRec1, SwRec2, SwRec3, ...]. Assign a one-to-one correspondence between the occupancy duration sequence SeqDur and the occupancy aircraft switching sequence record SeqSw according to their indices. Each index position i corresponds to a triple (Ev_i, SwRec_i), where Ev_i contains (Du_i). The combination of all index positions forms the grid occupancy continuity pattern record PatOcc.

[0052] Step S140: Use the preset multi-grid occupancy variation pattern cross-comparison structure to perform cross-comparison processing on the grid occupancy variation pattern records of adjacent grid cells, and generate occupancy pattern cross-comparison difference records between adjacent grid cells.

[0053] Step S141: In the airspace control grid system, using the three-dimensional grid number of each target grid cell as an index, locate the adjacent grid cells that are directly adjacent to the target grid cell in the adjacent longitude direction, adjacent latitude direction, and adjacent altitude direction, and establish a one-to-one pairing relationship between the target grid cell and each adjacent grid cell.

[0054] For a target mesh element Cu with the 3D mesh number (Ig, Jg, Kg), its adjacent mesh elements in the direction of increasing longitude are Cu_e=(Ig+1, Jg, Kg) and in the direction of decreasing longitude are Cu_w=(Ig-1, Jg, Kg). Adjacent mesh elements in the direction of increasing latitude are Cu_n=(Ig, Jg+1, Kg) and in the direction of decreasing latitude are Cu_s=(Ig, Jg-1, Kg). Adjacent mesh elements in the direction of increasing height are Cu_u=(Ig, Jg, Kg+1) and in the direction of decreasing height are Cu_d=(Ig, Jg, Kg-1). The target mesh element Cu is paired with each of these adjacent mesh elements to form a pairing list: Pairs=[(Cu, Cu_e), (Cu, Cu_w), (Cu, Cu_n), (Cu, Cu_s), (Cu, Cu_u), (Cu, Cu_d)].

[0055] Step S142: For each pairing relationship, extract the occupancy duration sequence of the target grid cell, and simultaneously extract the occupancy duration sequences of adjacent grid cells. Subtract the occupancy duration segments within the same time window from each pair of occupancy duration sequences using a time window subtraction process to generate a time window difference record sequence for that pairing relationship. Note that since the occupancy events of different grid cells may not be perfectly aligned in time, a time window alignment algorithm is needed to match the two sequences along the time axis.

[0056] For two grid cells A and B in a pairing relationship, their occupancy duration sequences SeqDurA and SeqDurB are obtained respectively. Since the event times of the two sequences may not be aligned, a sliding time window alignment method is adopted: a fixed time window length Lw is set, dividing the entire monitoring period into consecutive time windows [W1, W2, W3, ...]. Within each time window Wi, the total occupancy duration of A within that window, sumA_i, is calculated, and the total occupancy duration of B within that window, sumB_i, is calculated. The difference, dif_i, is calculated as sumA_i - sumB_i. The dif_i of all windows constitutes the window-by-window difference value recording sequence SeqDif.

[0057] Step S143: Extract the occupancy aircraft switching sequence record of the target grid cell, and simultaneously extract the occupancy aircraft switching sequence record of the adjacent grid cell. Compare the aircraft identity switching event records in the same time window in the two sets of occupancy aircraft switching sequence records before and after the switching time, calculate the time misalignment between the switching time of the target grid cell and the switching time of the adjacent grid cell, and generate a switching event time misalignment record sequence for the pairing relationship.

[0058] For two grid cells A and B in a pairing relationship, their occupied aircraft switching sequence records SeqSwA and SeqSwB are obtained respectively. Each switching event record contains the time tsw when the switching occurs (equal to the end time of the corresponding occupied event). Within each time window Wi, all switching times tswA_list of A occurring within that window are collected, and all switching times tswB_list of B occurring within that window are collected. If both lists are not empty, the minimum time misalignment is calculated as dif_t = min(|tswA - tswB|)overalltswAintswA_list, tswBintswB_list. If only one list is not empty, dif_t takes the preset maximum misalignment value. If both lists are empty, dif_t is 0. The dif_t of all windows constitutes the switching event time misalignment record sequence SeqMis.

[0059] Step S144: Each difference value in the window-by-window difference value record sequence of the duration sequence is compared and graded one by one with a preset difference level grading benchmark sequence to obtain the duration dimension occupancy difference grading record under this pairing relationship. The difference level grading benchmark sequence defines the mapping relationship between difference value ranges and grading levels. For example, an absolute difference value between 0 and 0.1 times the window length is level 0, between 0.1 and 0.3 times the window length is level 1, and so on. Each dif_i in SeqDif is mapped to the corresponding grading level grade_i to obtain the duration dimension occupancy difference grading record SeqGrd.

[0060] Step S145: Compare and classify each time misalignment in the switching event time misalignment record sequence with a preset misalignment degree classification benchmark sequence to obtain the switching dimension difference classification record under this pairing relationship. The misalignment degree classification benchmark sequence defines the mapping relationship between the time misalignment interval and the classification level. For example, a time misalignment between 0 and 1 second is level 0, between 1 and 3 seconds is level 1, and so on. Map each dif_t_i in SeqMis to the corresponding classification level gt_i to obtain the switching dimension difference classification record SeqGt.

[0061] Step S146: Merge and combine the historical dimension occupancy difference classification record and the switching dimension difference classification record window by window in the same time window to generate occupancy pattern cross-comparison difference record entries for each time window between adjacent grid cells under the pairing relationship. Arrange all occupancy pattern cross-comparison difference record entries in chronological order to generate occupancy pattern cross-comparison difference records between adjacent grid cells under the pairing relationship.

[0062] For each time window Wi, the epochal dimension occupancy difference grade level grade_i obtained in step S144 and the switching dimension difference grade level gt_i obtained in step S145 are merged into a tuple (grade_i, gt_i). This tuple serves as the occupancy pattern cross-comparison difference record entry for that time window Wi. All entries for all time windows are arranged in chronological order to form the sequence SeqEntry=[(grade_1, gt_1), (grade_2, gt_2), (grade_3, gt_3), ...], which is the occupancy pattern cross-comparison difference record RecDiff between adjacent grid cells.

[0063] Step S150: Determine the aircraft track crossing situation category and track crossing spatiotemporal distribution record between adjacent grid cells based on the cross-comparison difference record of the occupancy mode, and generate a flight avoidance scheduling instruction carrying a grid coordinate chain based on the aircraft track crossing situation category and the track crossing spatiotemporal distribution record, and distribute the flight avoidance scheduling instruction to the corresponding aircraft control terminal.

[0064] Step S151: Extract the occupancy pattern cross-comparison difference record entries for each time window from the occupancy pattern cross-comparison difference records one by one, and separate the occupancy difference classification records of the historical dimension and the switching dimension difference classification records.

[0065] Iterate through each entry (grade_i, gt_i) in the RecDiff record of occupancy pattern cross-comparison differences, and use grade_i as the occupancy difference classification record of the epochal dimension of the i-th time window, and use gt_i as the switching dimension difference classification record of the i-th time window.

[0066] Step S152: Input the duration dimension occupancy difference classification record into a preset track interleaving duration classification judgment logic structure. The track interleaving duration classification judgment logic structure maintains a mapping table between duration difference classification values ​​and track interleaving duration intensity level codes, and outputs the track interleaving duration intensity level codes by looking up the table; and input the switching dimension difference classification record into a preset track interleaving switching classification judgment logic structure. The track interleaving switching classification judgment logic structure maintains a mapping table between switching difference classification values ​​and track interleaving switching intensity level codes, and outputs the track interleaving switching intensity level codes by looking up the table.

[0067] The preset logic structure for determining the duration of track interleaving is a function mapping F_hist(grade_i). For the input grade_i, the mapping table is looked up: grade_i = 0 corresponds to duration intensity code 0 (no interleaving), grade_i = 1 corresponds to duration intensity code 1 (slight interleaving), grade_i = 2 corresponds to duration intensity code 2 (moderate interleaving), and grade_i greater than or equal to 3 corresponds to duration intensity code 3 (severe interleaving). The duration intensity code C_hist is output. Similarly, the logic structure for determining the switching duration of track interleaving is F_sw(gt_i): gt_i = 0 corresponds to switching intensity code 0, gt_i = 1 corresponds to switching intensity code 1, gt_i = 2 corresponds to switching intensity code 2, and gt_i greater than or equal to 3 corresponds to switching intensity code 3. The switching intensity code C_sw is output.

[0068] Step S153: Combine the track interleaving duration intensity level code and the track interleaving switching intensity level code to form a hierarchical code combination. Input the hierarchical code combination into a pre-constructed track interleaving situation category determination logic structure. The track interleaving situation category determination logic structure maintains a mapping table between the hierarchical code combination and the track interleaving situation category code. Output the aircraft track interleaving situation category code corresponding to the pairing relationship of the adjacent grid cells by looking up the table.

[0069] Combine C_hist and C_sw into an ordered pair (C_hist, C_sw). The preset logic structure for determining the situation category of track interleaving includes a two-dimensional mapping table, with row index C_hist and column index C_sw. The situation category code St is obtained by looking up the table. For example, (0,0) corresponds to the non-interleaving situation code 0; (1,1) corresponds to the crossing situation code 1; (2,2) corresponds to the same-direction chasing situation code 2; and (3,3) corresponds to the head-on encounter situation code 3. Output the situation category code St.

[0070] Step S154: Extract the set of time windows containing the difference records from the cross-comparison difference records of the occupancy patterns, extract the start time stamp and end time stamp of each time window in the set of time windows, and generate a track cross-spatiotemporal distribution record containing the cross-start time, cross-duration duration and cross-grid number pairs.

[0071] In steps S151 to S153, identify time windows Wi where St is not 0, and record the start time T_start_i and end time T_end_i of these windows. The staggered duration Dur_i = T_end_i - T_start_i. The staggered grid number pair is (GridA, GridB). Generate the staggered spatiotemporal distribution record RecSpace = (T_start_i, Dur_i, GridA, GridB).

[0072] Step S155: Based on the aircraft track interleaving situation category code, retrieve the matching avoidance strategy rule entry in the preset flight avoidance strategy rule set, and extract the avoidance direction command and avoidance priority parameter contained in the avoidance strategy rule entry.

[0073] The preset flight avoidance strategy rule set is a mapping table, where the key is the situation category code St, and the value is the avoidance strategy rule entry. Each entry contains an avoidance direction instruction Dir (e.g., "ascend", "descend", "turn left", "turn right", "hover") and an avoidance priority parameter Pri (the higher the value, the higher the priority). Based on the St output in step S153, the corresponding Dir and Pri are looked up in the mapping table.

[0074] Step S156: Extract the interlaced grid number pair from the interlaced spatiotemporal distribution record of the flight path, and in the airspace control grid system, take the interlaced grid number pair as the reference, expand the adjacent grid cells outward along the spatial adjacency direction, arrange the grid numbers on the expansion path in spatial connection order, and generate the avoidance path grid coordinate chain.

[0075] Extract the staggered grid number pair (GridA, GridB) from RecSpace in step S154. In the airspace control grid system, with GridA as the starting point and GridB as the ending point, calculate the Manhattan path between them. Move step-by-step along the longitude, latitude, or altitude directions, moving to an adjacent grid cell at each step and recording the 3D grid number of each grid cell passed. Arrange the recorded grid number sequence sequentially to form the avoidance path grid coordinate chain Chain=[GridA, GridX1, GridX2, ..., GridB].

[0076] Step S157: Encapsulate and combine the avoidance direction command, the avoidance priority parameter, and the avoidance path grid coordinate chain to generate a flight avoidance scheduling command carrying the grid coordinate chain. Based on the mapping relationship between the staggered grid number and the associated grid-occupied aircraft identifier, extract the set of aircraft identifier fields involved in the staggered grid. Distribute the flight avoidance scheduling command to the aircraft control terminal corresponding to the set of aircraft identifier fields through the communication link.

[0077] The avoidance direction command Dir from step S155, the avoidance priority parameter Pri, and the avoidance path grid coordinate chain Chain from step S156 are encapsulated into a data packet CmdAvoid. From the mapping relationship of the grid occupancy aircraft identifiers corresponding to the staggered grid number pairs (GridA, GridB), the occupancy aircraft identifier IdA corresponding to GridA and the occupancy aircraft identifier IdB corresponding to GridB are obtained. IdA and IdB are combined into a set {IdA, IdB}. CmdAvoid is sent to the aircraft control terminal corresponding to IdA and IdB via the 5G communication link, triggering the aircraft to perform an avoidance maneuver.

[0078] Step S210: After distributing the flight avoidance scheduling command to the corresponding aircraft control terminal, retrieve the staggered grid number pair and avoidance path grid coordinate chain corresponding to the flight avoidance scheduling command in the distributed record.

[0079] Read the staggered grid number pair (Ga, Gb) and the avoidance path grid coordinate chain ChainP corresponding to the flight avoidance scheduling command sent in step S150 from the distributed command record database.

[0080] Step S211: Using the grid number contained in the staggered grid number pair as an index, retrieve all flight status messages falling into the staggered grid number pair in the subsequently collected flight status message stream, extract the instantaneous velocity vector field and heading azimuth field of the flight status message, and simultaneously extract the aircraft identifier field of the flight status message.

[0081] Using Ga and Gb as query keys, search for all flight status messages mapped to grid numbers Ga or Gb in the flight status message stream of subsequent acquisition timestamps. For each found message, extract its instantaneous velocity vector field (Vx, Vy, Vz), heading and azimuth angle field Hd, and aircraft identifier field Id.

[0082] Step S212: Extract the avoidance direction instruction from the flight avoidance scheduling instruction, parse the reference heading azimuth value indicated by the avoidance direction instruction, perform difference calculation processing between the heading azimuth field of the retrieved flight status message and the reference heading azimuth value, and generate the heading deviation angle value corresponding to the flight status message.

[0083] Extract the avoidance direction instruction Dir from the flight avoidance dispatch instructions, and determine the reference heading azimuth value Hr based on the avoidance direction instruction Dir. For example, when the avoidance direction instruction is "turn left 90 degrees", the reference heading azimuth value Hr is equal to the aircraft's original heading minus 90 degrees. Calculate the heading deviation angle value Da = |Hd - Hr|. If Da is greater than 180 degrees, then Da = 360 - Da.

[0084] Step S213: Perform difference calculation on the velocity magnitude component in the instantaneous velocity vector field of the retrieved flight status message and the preset avoidance reference velocity value to generate the velocity deviation value corresponding to the flight status message. Assign a heading response classification label to the flight status message according to the deviation angle range into which the heading deviation angle value falls.

[0085] Calculate the magnitude of the instantaneous velocity vector field: Sp = sqrt(Vx^2 + Vy^2 + Vz^2). Preset avoidance reference velocity value Sr. Calculate the velocity deviation value Dv = |Sp - Sr|. Preset a list of deviation angle ranges, for example, [0 degrees, 10 degrees) is level 0, [10 degrees, 30 degrees) is level 1, [30 degrees, 90 degrees) is level 2, and [90 degrees, 180 degrees] is level 3. Assign a heading response classification label Lh based on the range into which Da falls.

[0086] Step S214: Based on the deviation range into which the speed deviation value falls, assign a speed response classification mark to the flight status message, combine the heading response classification mark and speed response classification mark corresponding to the same aircraft identifier field, and generate the avoidance response execution feedback record corresponding to the aircraft identifier field.

[0087] A list of preset deviation ranges is provided, for example, [0 m / s, 1 m / s) is level 0, [1 m / s, 3 m / s) is level 1, [3 m / s, 5 m / s) is level 2, and [5 m / s, ∞) is level 3. Based on the range into which Dv falls, a velocity response classification label Lv is assigned. (Lh, Lv) are combined into a tuple, which serves as the avoidance response execution feedback record Rfb=(Id, Lh, Lv) corresponding to the aircraft identifier Id.

[0088] Step S215: Associate the avoidance response execution feedback record with the aircraft identifier field and store it in the aircraft avoidance history feedback record set. When the low-altitude aircraft corresponding to the aircraft identifier field is again the object of flight avoidance scheduling instruction distribution, extract the historical avoidance response execution feedback record corresponding to the aircraft identifier field from the aircraft avoidance history feedback record set, and adjust the avoidance priority parameter of the low-altitude aircraft based on the historical avoidance response execution feedback record.

[0089] The Rfb is stored in the historical feedback record database HistDb, with Id as the primary key. When a flight avoidance scheduling command is sent again, the historical record corresponding to Id in HistDb is queried, and the average values ​​of historical Lh and Lv are extracted. An adjustment factor is calculated based on the average Lh and Lv, for example, the adjustment factor coef = 1 + (Lh_avg + Lv_avg) / 10. The original avoidance priority parameter Pri is multiplied by coef to obtain the adjusted priority parameter Pri_new.

[0090] Step S310: After generating the grid occupancy continuous change mode record, extract the aircraft identity switching event record one by one from the occupancy aircraft switching sequence record of each grid cell, and obtain the aircraft identifier field before switching and the aircraft identifier field after switching in each aircraft identity switching event record.

[0091] From the mesh occupancy continuation mode record PatOcc generated in step S136, iterate through the aircraft identity switching event record SwRec=(IdPrev, IdNext) corresponding to each occupancy event. Extract IdPrev and IdNext.

[0092] Step S311: Using the pre-switch aircraft identifier field as the query basis, retrieve the pre-built aircraft basic parameter filing record library, and extract the cruise rate parameter and turning rate parameter of the preceding aircraft corresponding to the pre-switch aircraft identifier field; and using the post-switch aircraft identifier field as the query basis, retrieve the aircraft basic parameter filing record library, and extract the cruise rate parameter and turning rate parameter of the following aircraft corresponding to the post-switch aircraft identifier field.

[0093] The aircraft basic parameter record database is a dictionary with the aircraft identifier Id as the key. A query retrieves the cruise rate parameter Vc_prev and the turning rate parameter Rc_prev of the preceding aircraft corresponding to IdPrev. A query retrieves the cruise rate parameter Vc_next and the turning rate parameter Rc_next of the following aircraft corresponding to IdNext.

[0094] Step S312: From the occupancy duration sequence of the grid cell, locate the occupancy duration segment of the previous occupancy event and the occupancy duration segment of the next occupancy event adjacent to the current aircraft identity switching event record.

[0095] From the occupancy duration sequence SeqDur generated in step S133, find the occupancy event index i corresponding to the current switching event SwRec. Obtain the duration Du_prev of the (i-1)th occupancy event and the duration Du_next of the (i+1)th occupancy event.

[0096] Step S313: Calculate the cruise rate difference between the cruise rate parameter of the preceding aircraft and the cruise rate parameter of the following aircraft, and associate the cruise rate difference with the duration difference between the duration of the previous occupation event and the duration of the occupation event to generate an association record of cruise rate difference and occupation duration change; and calculate the turning angle rate difference between the turning angle rate parameter of the preceding aircraft and the turning angle rate parameter of the following aircraft, and associate the turning angle rate difference with the duration difference between the duration of the previous occupation event and the duration of the occupation event to generate an association record of turning angle rate difference and occupation duration change.

[0097] Calculate the cruise speed difference dVc = |Vc_prev - Vc_next|. Calculate the duration difference dDu = Du_next - Du_prev. Use (dVc, dDu) as the associated record R1. Calculate the steering angle rate difference dRc = |Rc_prev - Rc_next|. Use (dRc, dDu) as the associated record R2.

[0098] Step S314: Aggregate the correlation records of cruise rate differences and occupancy duration changes corresponding to all aircraft identity switching event records within the same grid cell, extract the corresponding patterns of recurring cruise rate difference intervals and occupancy duration change directions in the grid cell, and generate the cruise rate-related occupancy change pattern record for the grid cell; and aggregate the correlation records of turning angle rate differences and occupancy duration changes corresponding to all aircraft identity switching event records within the same grid cell, extract the corresponding patterns of recurring turning angle rate difference intervals and occupancy duration change directions in the grid cell, and generate the turning angle rate-related occupancy change pattern record for the grid cell.

[0099] For the same grid cell, collect all associated records R1 and statistically analyze the sign distribution of dDu when dVc falls within different intervals. If dDu is mostly positive when dVc is in interval A, then record pattern A: dVc belongs to interval A, corresponding to an increase in occupancy time. Similarly, process R2 to obtain the steering angle rate associated occupancy change pattern records.

[0100] Step S315: Merge the cruise rate-related occupancy change mode record and the steering angle rate-related occupancy change mode record to generate the aircraft performance-related occupancy change comprehensive mode record for the grid cell. Attach the aircraft performance-related occupancy change comprehensive mode record to the grid occupancy change mode record for the grid cell to form an expanded grid occupancy change mode record. Replace the original grid occupancy change mode record with the expanded grid occupancy change mode record for cross-comparison processing of multiple grid occupancy change modes.

[0101] The cruise rate-related occupancy change pattern record and the steering angle rate-related occupancy change pattern record are merged into a single integrated pattern record, PatPerf. PatPerf is then appended to PatOcc generated in step S136 to obtain the expanded grid occupancy change pattern record, PatOccNew. In subsequent step S140, PatOccNew replaces the original PatOcc.

[0102] Step S410: After obtaining the set of single grid occupancy markers for each grid cell under each acquisition time stamp, perform spatial statistical processing on the single grid occupancy markers of all grid cells under the same acquisition time stamp, count the number of aircraft in the occupied state of each grid cell under the acquisition time stamp, generate the distribution record of the number of grid occupancy in the entire airspace under the acquisition time stamp, arrange the distribution records of the number of grid occupancy in the entire airspace under multiple consecutive acquisition time stamps in chronological order, and generate a time-series change record sequence of the number of grid occupancy in the entire airspace.

[0103] For each acquisition time stamp Tm, iterate through all grid cells and count the length of the occupied aircraft identifier list Lid for each grid cell, denoted as the occupancy quantity Nocc(Ig, Jg, Kg). Organize Nocc(Ig, Jg, Kg) into a three-dimensional array OccNum[Tm]. Arrange the OccNum values ​​at all times in order of Tm to obtain the temporal change sequence of the total airspace grid occupancy quantity SeqDens.

[0104] Step S411: Extract the time-series curve of the occupancy quantity of each grid cell from the time-series record of the change in the occupancy quantity of the full-space grid, identify the continuous time segments in the time-series curve where the slope of the occupancy quantity increases exceeds a preset slope threshold, and mark the continuous time segments as density surge time windows.

[0105] For each grid cell, extract the occupancy sequence Nocc(t) from SeqDens. Calculate the difference in occupancy between adjacent time steps ΔN = Nocc(t+1) - Nocc(t). Calculate the rising slope k = ΔN / Δt. Set a preset slope threshold Kt. Find the time interval where k > Kt for multiple consecutive time steps, and mark it as the density surge time window Wden = [Ts, Te].

[0106] Step S412: Extract the three-dimensional grid number of the corresponding grid cell within the density surge time window to form a density surge grid number set. From the grid-occupied aircraft identifier mapping relationship corresponding to the density surge grid number set, extract all aircraft identifier fields that exist in the grid cell within the density surge time window to form a density surge associated aircraft identifier set.

[0107] Add the grid cell numbers (Ig, Jg, Kg) corresponding to the density surge time window Wden to the set GridDense. For each grid cell in GridDense, find all aircraft identifiers Id that appear in that grid cell within the time window Wden from the grid-occupied aircraft identifier mapping relationship, and add them to the set SetDenId.

[0108] Step S413: For each aircraft identifier field in the density surge associated aircraft identifier set, retrieve its corresponding instantaneous velocity vector field and heading azimuth field. Based on the instantaneous velocity vector field and the heading azimuth field, extrapolate the expected arrival grid number of the aircraft identifier at subsequent acquisition timestamps along the time axis. Compare the expected arrival grid number of each aircraft identifier with the density surge grid number set. When the expected arrival grid number of an aircraft identifier belongs to the density surge grid number set, mark the aircraft identifier as an aircraft flowing into the surge area.

[0109] For each Id in SetDenId, obtain its velocity vector V and heading angle Hd, and calculate its motion direction unit vector. Set the prediction time step Δt_pred, and the prediction position P_pred = P_cur + V * Δt_pred. Map P_pred to the grid system to obtain the expected grid number G_pred. If G_pred belongs to GridDense, then mark Id as the aircraft Id_in flowing into the surge region.

[0110] Step S414: Count the aircraft identifier fields of all aircraft marked as aircraft entering the surge area, generate a list of aircraft entering the surge area, and mark the expected inflow timestamp for each aircraft entering the surge area. Based on the list of aircraft entering the surge area and the expected inflow timestamp, generate an airspace density continuous surge warning notification. The airspace density continuous surge warning notification includes the surge grid number, warning time window and the list of aircraft entering the surge area identifiers. Send the airspace density continuous surge warning notification to the aircraft control terminal corresponding to each aircraft identifier field in the list of aircraft entering the surge area identifiers.

[0111] Collect all Id_in values ​​into a list ListIn, where the estimated inflow time for each Id_in is T_cur + Δt_pred. Generate a warning notification WarnDen = (GridDense, Wden, ListIn). Send WarnDen to the aircraft control terminal corresponding to each Id_in in ListIn.

[0112] Step S510: After generating the grid occupancy continuity pattern record, extract the occupancy duration sequence of the target grid cell within a preset long-period monitoring period, perform occupancy period segmentation processing on the occupancy duration sequence, detect the occupancy duration peaks that repeat at fixed time intervals in the occupancy duration sequence, record the time span between every two adjacent occupancy duration peaks as an occupancy period segment, and extract the moment when the occupancy duration peak appears in each occupancy period segment as the period peak moment.

[0113] Extract the occupancy duration sequence of the target grid cell from the SeqDur generated in step S133, taking only the duration value Du sequence. Perform autocorrelation analysis on this sequence, and find the lag order with the largest autocorrelation coefficient as the period T_period. Detect peak points in the sequence (Du greater than the preceding and following neighbors), and use the period T_period as an interval to divide the interval between adjacent peak points into an occupancy period segment Cyc_i. The time corresponding to the peak point within each segment is taken as the periodic peak time T_peak_i.

[0114] Step S511: The average time span of multiple occupancy period segments is calculated to obtain the average occupancy period duration of the target grid cell. The average occupancy period duration is compared and matched with a preset set of period type determination benchmark durations to determine the spatial occupancy period type label of the target grid cell.

[0115] Calculate the average duration T_avg of all occupied period segments. Establish a preset set of period type determination benchmark durations; for example, T < 10 seconds is a short period type, 10 seconds ≤ T < 30 seconds is a medium period type, and T ≥ 30 seconds is a long period type. Determine the spatial occupancy period type label TypeCyc based on the interval where T_avg falls.

[0116] Step S512: Extract the peak duration of occupancy within each occupancy period segment, compare and match each peak duration with a preset set of peak intensity classification benchmarks, assign a periodic peak intensity classification marker to each occupancy period segment, arrange the spatial occupancy period type marker and the periodic peak intensity classification marker in chronological order of the occupancy period segment, and generate a periodic occupancy pattern record for the target grid cell.

[0117] Obtain the peak value Du_peak_i within each occupied period segment Cyc_i. A preset set of peak intensity classification benchmarks is established; for example, Du < 5 seconds is a weak peak, 5 seconds ≤ Du < 15 seconds is a medium peak, and Du ≥ 15 seconds is a strong peak. Based on the interval in which Du_peak_i falls, assign a peak intensity classification label Lev_i. Arrange (TypeCyc, Lev_i) in i-order to generate a periodic occupancy pattern record PatCyc.

[0118] Step S513: Generate corresponding periodic occupancy pattern records for all grid cells in the airspace control grid system to form a full airspace periodic occupancy pattern record set. Perform periodic phase comparison processing on the periodic occupancy pattern records of spatially adjacent grid cells in the full airspace periodic occupancy pattern record set, and calculate the difference in average occupancy period duration and the time offset of the periodic peak time between adjacent grid cells.

[0119] For each grid cell, perform steps S510 to S512 to generate PatCyc_all. For two spatially adjacent grid cells A and B, extract their average occupied cycle durations T_avg_A and T_avg_B, as well as their cycle peak time sequences T_peak_A_list and T_peak_B_list. Calculate the cycle duration difference dT_avg = |T_avg_A - T_avg_B|. Perform dynamic time warping and alignment on the two peak time sequences, and calculate the average time offset dT_peak = average(|T_peak_A_i - T_peak_B_i|).

[0120] Step S514: Generate an occupancy cycle phase difference record between adjacent grid cells based on the difference in the average occupancy cycle duration and the time offset of the peak time of the cycle. Perform difference fusion processing on the same grid cell pairing relationship between the occupancy cycle phase difference record and the occupancy pattern cross-comparison difference record. Incorporate the difference information in the occupancy cycle phase difference record into the occupancy pattern cross-comparison difference record of the pairing relationship to form an enhanced occupancy pattern cross-comparison difference record.

[0121] (dT_avg, dT_peak) is used as the periodic phase difference record RecPhase. In the occupancy pattern cross-matching difference record RecDiff generated in step S146, the record corresponding to the grid cell pairing is found. The information in RecPhase is added as a new field to RecDiff to obtain the enhanced record RecDiffEnh.

[0122] For example, the method may further include: step S610: after distributing the flight avoidance scheduling instruction to the corresponding aircraft control terminal, extracting the avoidance path grid coordinate chain from the flight avoidance scheduling instruction, obtaining all grid number sequences contained in the avoidance path grid coordinate chain, using each grid number in the grid number sequence as an index, retrieving all flight status messages falling into the grid number in the subsequently continuously collected flight status message stream, and extracting the instantaneous position coordinate field and aircraft identifier field of the flight status message.

[0123] Extract the avoidance path grid coordinate chain ChainP=[G1, G2, G3, ..., Gm] from the flight avoidance scheduling command CmdAvoid. For each grid number Gi in ChainP, find all flight status messages mapped to grid number Gi in the subsequent flight status message stream, and extract the instantaneous position coordinate P(Id) and aircraft identifier Id from each message.

[0124] Step S611: Connect the instantaneous position coordinate fields belonging to the same aircraft identifier field according to the order of the acquisition timestamp fields to generate the actual avoidance track line of the low-altitude aircraft corresponding to the aircraft identifier field in the avoidance path grid coordinate chain; and extract the avoidance direction instruction from the flight avoidance scheduling instruction, and generate a reference avoidance track line extending along the avoidance direction from the staggered grid number pair according to the avoidance direction instruction.

[0125] For each aircraft identifier Id, all its instantaneous position coordinates P within the avoidance path grid coordinate chain are sorted by the acquisition timestamp and connected sequentially to form a polyline, thus obtaining the actual avoidance track trackReal. The avoidance direction command Dir is extracted from CmdAvoid, and a straight line is generated along the specified direction using the staggered grid number pair (Ga, Gb) as a reference, as the reference avoidance track trackRef.

[0126] Step S612: Perform point-by-point deviation distance calculation on the actual avoidance track and the reference avoidance track to generate the avoidance track deviation distance sequence of the low-altitude aircraft. Compare and classify each deviation distance value in the avoidance track deviation distance sequence with a preset deviation distance classification benchmark sequence to obtain the deviation distance classification record sequence of the low-altitude aircraft.

[0127] For each point P_real on TrackReal, find the corresponding point P_ref on TrackRef with the same timestamp, and calculate the Euclidean distance D_dist = |P_real - P_ref| between the two points. All points' D_dists form the sequence SeqDist. A default deviation distance classification benchmark sequence is used, for example, [0 m, 5 m) is level 0, [5 m, 15 m) is level 1, [15 m, 30 m) is level 2, and [30 m, ∞) is level 3. Each D_dist in SeqDist is mapped to a classification level Ld_i, resulting in the deviation distance classification record sequence SeqLd.

[0128] Step S613: Extract the deviation direction corresponding to the time segment in which the deviation distance classification record exceeds the preset classification tolerance level from the deviation distance classification record sequence. Calculate the angle between the deviation direction and the extension direction of the reference avoidance track to generate a deviation direction angle sequence. Compare each angle value in the deviation direction angle sequence with the preset deviation direction classification benchmark sequence to obtain the deviation direction classification record sequence of the low-altitude aircraft.

[0129] Find the time segments in SeqLd where Ld_i≥2 occur consecutively, and record the deviation direction within these time segments. Calculate the angle Da_dir between the deviation direction and the extension direction of TrackRef. All angles form the sequence SeqAng. Predetermine a baseline sequence for deviation direction classification, for example, [0 degrees, 15 degrees) is level 0, [15 degrees, 45 degrees) is level 1, [45 degrees, 90 degrees) is level 2, and [90 degrees, 180 degrees] is level 3. Map each Da_dir in SeqAng to a classification level La_i, obtaining the deviation direction classification record sequence SeqLa.

[0130] Step S614: Combine the deviation distance classification record sequence and the deviation direction classification record sequence time-by-time to generate the execution accuracy classification record of the low-altitude aircraft's current avoidance response. Associate the execution accuracy classification record with the aircraft identifier field of the low-altitude aircraft and store it in the aircraft avoidance execution accuracy history record library. When the low-altitude aircraft corresponding to the aircraft identifier field is again identified as the object of the flight avoidance scheduling instruction, extract the historical execution accuracy classification record corresponding to the aircraft identifier field from the aircraft avoidance execution accuracy history record library, and adjust the avoidance priority parameter corresponding to the aircraft identifier field based on the historical execution accuracy classification record.

[0131] Combine (Ld_i, La_i) at the same time position in SeqLd and SeqLa into a tuple to obtain the execution precision classification record RecPre. Store RecPre in the execution precision history database AccuDb, using the aircraft identifier Id as the primary key. When distributing instructions again, query the historical RecPre corresponding to Id from AccuDb, and calculate the average Ld_avg and average La_avg. Adjust the priority parameter Pri_new = Pri / (1 + Ld_avg + La_avg) based on the average values.

[0132] Step S710: After generating the grid occupancy change pattern record, construct the grid occupancy change pattern records of all grid cells in the airspace to be controlled within a preset historical time period into an airspace occupancy pattern historical sample set. Each sample in the airspace occupancy pattern historical sample set corresponds to a grid cell's grid occupancy change pattern record within a historical time period.

[0133] Extract all grid cell records for the historical period (e.g., the past 7 days) from the grid occupancy pattern record PatOcc generated in step S136. Each grid cell's PatOcc for each hour (or each preset time period) is treated as an independent sample. Collect all samples to form a historical sample set of spatial occupancy patterns, S_hist={s1, s2, s3, ..., sN}, where N is the total number of samples.

[0134] Step S711: Perform occupancy pattern decomposition processing on each sample in the historical sample set of airspace occupancy patterns, extract the duration distribution statistical features of the occupancy duration sequence and the switching frequency statistical features of the occupancy aircraft switching sequence record in each sample, concatenate the duration distribution statistical features and the switching frequency statistical features to form the occupancy pattern statistical representation vector of the sample, perform vector clustering processing on the occupancy pattern statistical representation vectors of all samples, and group samples with similar duration distribution statistical features and similar switching frequency statistical features into the same occupancy pattern cluster category to form multiple occupancy pattern cluster categories, and assign a unique occupancy pattern category label to each occupancy pattern cluster category.

[0135] For each sample PatOcc, extract the occupancy duration sequence SeqDur=[Du1, Du2, Du3, ...]. Calculate the statistical characteristics of SeqDur: mean μ=(ΣDu_i) / n, standard deviation σ=sqrt(Σ(Du_i-μ)^2 / n), maximum value max=max(Du_i), minimum value min=min(Du_i). Extract the switching frequency statistical characteristics from the occupancy aircraft switching sequence record SeqSw: number of switching events per unit time f=(total number of switching events) / (total duration). Concatenate (μ, σ, max, min, f) into a 5-dimensional statistical representation vector V_stat. Cluster the V_stat of all samples in S_hist using the K-means clustering algorithm. K-means clustering minimizes the sum of squares within a cluster by iteratively updating the cluster centers. Initialize K random centers, assign each sample to the nearest center, then recalculate the centers, and repeat until convergence. After clustering is completed, each cluster corresponds to an occupancy pattern, and a unique category label L_occ is assigned to each cluster, such as "high frequency short time type" or "low frequency long time type".

[0136] Step S712: Arrange the occupancy pattern category labels corresponding to different time periods within the historical period of each grid cell in chronological order to generate a time sequence of occupancy pattern category labels for each grid cell.

[0137] For each grid cell, the occupancy pattern category label L_occ obtained by clustering the cell across different historical time periods (e.g., time windows divided by hours) is arranged in chronological order. For example, for grid cell GridA, the sequence SeqL_GridA=[L1, L2, L3, ..., Lm] is obtained, where L1 corresponds to the first hour, L2 corresponds to the second hour, and so on.

[0138] Step S713: Taking each grid cell as the target grid cell, extract the occupancy pattern category label time sequence of the target grid cell, and at the same time extract the occupancy pattern category label time sequence of all spatially adjacent grid cells of the target grid cell. Pair the occupancy pattern category label time sequence of the target grid cell with the occupancy pattern category label time sequence of each adjacent grid cell to construct a set of adjacent grid cell occupancy pattern category label time sequence sample pairs.

[0139] For the target grid cell GridT, obtain its sequence SeqL_T. Obtain the adjacent grid cells GridA (directly adjacent cells in the longitude, latitude, and altitude directions) of GridT, and obtain the sequence SeqL_A of GridA. Add (SeqL_T, SeqL_A) as a sample pair to the training sample pair set Pairs. Traverse all grid cells and all adjacent directions to construct a complete time-series sample pair set.

[0140] Step S714: Retrieve the pre-constructed occupancy pattern spatiotemporal correlation prediction network. The occupancy pattern spatiotemporal correlation prediction network includes a spatial adjacency encoding module, a temporal dependency encoding module, and an occupancy pattern category label prediction module. The spatial adjacency encoding module is used to convert the spatial relative positional relationship of adjacent grid cells into a spatial adjacency encoding vector. The temporal dependency encoding module is used to convert the temporal dependency relationship in the temporal sequence into a temporal dependency encoding vector. The occupancy pattern category label prediction module outputs the predicted probability distribution of the occupancy pattern category label combination of adjacent grid cells in the future time window based on the spatial adjacency encoding vector and the temporal dependency encoding vector.

[0141] The specific architecture of the occupancy pattern spatiotemporal association prediction network is as follows: The spatial adjacency encoding module is an embedding layer. The input is the categories of spatial relative positions (6 categories in total: East, West, South, North, Up, Down), and the output is a 64-dimensional dense vector. The weights of this embedding layer are learned during training. The temporal dependency encoding module uses a Long Short-Term Memory (LSTM) network, consisting of two layers, each with 128 hidden units. The input is a sequence of occupancy pattern category labels (the labels are first converted into 32-dimensional vectors by another embedding layer). The LSTM outputs the hidden state of the last time step as the temporal dependency encoding vector (128 dimensions). The occupancy pattern category label prediction module is a multilayer perceptron, containing two fully connected layers: the first layer has an input dimension of 64 + 128 = 192 and an output dimension of 128, with the ReLU activation function; the second layer has an input dimension of 128 and an output dimension of K × K, where K is the total number of occupancy pattern categories. The output is converted into a probability distribution using a Softmax function. The forward propagation process of the network is as follows: spatial relationship categories are input into the embedding layer to obtain V_pos (64 dimensions), sequences SeqL_T and SeqL_A are input into LSTM to obtain V_seq_T and V_seq_A (128 dimensions each), V_seq_T and V_seq_A are concatenated to obtain V_seq (256 dimensions), and V_pos and V_seq are concatenated to obtain V_joint (occupancy pattern spatiotemporal association prediction network). The training process is as follows: sample pairs in Pairs are used, and the actual occupancy pattern category labels of the target grid cell and adjacent grid cells in the future time window are used as supervision signals. The classification cross-entropy loss function is adopted, the optimizer is Adam, the initial learning rate is 0.001, the batch size is 64, the training epochs are 50, and the early stopping condition is that the validation set loss does not decrease for 5 consecutive epochs.

[0142] Step S715: Input the set of temporal sample pairs of adjacent grid cell occupancy pattern category labels into the occupancy pattern spatiotemporal correlation prediction network, and use the spatial adjacency relationship encoding module to encode the spatial relative positional relationship between adjacent grid cells and target grid cells in each sample pair to generate the spatial adjacency relationship encoding vector of the sample pair.

[0143] For each sequence pair (SeqL_T, SeqL_A) in the training sample pairs, determine the spatial positional relationship category Pos (values ​​from 0 to 5, representing east, west, south, north, up, and down, respectively) relative to the target grid cell. Input Pos into the embedding layer of the spatial adjacency encoding module. The embedding layer finds the corresponding 64-dimensional row vector from the weight matrix to obtain the spatial adjacency encoding vector V_pos.

[0144] Step S716: The temporal dependency encoding module is used to perform temporal encoding on the occupancy pattern category label temporal sequence of the target grid cell and the adjacent grid cell in each pair of samples, generating the target grid cell temporal dependency encoding vector and the adjacent grid cell temporal dependency encoding vector. The target grid cell temporal dependency encoding vector and the adjacent grid cell temporal dependency encoding vector are then fused to generate the joint temporal dependency encoding vector of the sample pair.

[0145] Each category label in SeqL_T is first converted into a 32-dimensional vector through an embedding layer to obtain a sequence vector representation. This sequence is then input into a Long Short-Term Memory (LSTM) network, which contains three gated units: a forget gate, an input gate, and an output gate. The forget gate determines how much of the cell state from the previous time step is retained in the current time step, the input gate determines how much of the current input is entered into the cell state, and the output gate determines how much of the current cell state is output to the hidden state. After LSTM processing, the hidden state of the last time step is taken as the target grid cell temporal dependency encoding vector V_seq_T (128 dimensions). Similarly, SeqL_A is processed to obtain V_seq_A (128 dimensions). V_seq_T and V_seq_A are concatenated to obtain a 256-dimensional joint temporal dependency encoding vector V_seq_joint.

[0146] Step S717: Input the spatial adjacency relationship encoding vector and the joint temporal dependency encoding vector of each sample pair into the occupancy pattern category label prediction module, output the predicted probability distribution of the occupancy pattern category label of the adjacent grid cells in the sample pair relative to the target grid cell in the future time window, and generate the future occupancy pattern category label prediction result based on the combination of the occupancy pattern category labels with the highest probability in the predicted probability distribution.

[0147] The V_pos (64-dimensional) and V_seq_joint (256-dimensional) vectors are concatenated to form a 320-dimensional vector, which is then input into the multilayer perceptron of the occupancy pattern category label prediction module. The multilayer perceptron passes through a first fully connected layer (320->128) and ReLU activation, followed by a second fully connected layer (128->K×K), outputting a K×K dimensional raw score vector. The Softmax function is applied to this vector to convert it into a probability distribution, where each element represents the probability that the target grid cell is predicted as category i and its adjacent grid cells are predicted as category j. The element with the highest probability is found, and its corresponding (i, j) is the future occupancy pattern category label prediction result (L_pred_T, L_pred_A).

[0148] Step S718: When the combination of the occupancy pattern category labels of the adjacent grid cell and the target grid cell in the future occupancy pattern category label prediction result conforms to the preset conflict category label combination rule, the target grid cell and the adjacent grid cell are marked as a future conflict grid pair, and a preemptive track conflict warning instruction is generated based on the spatial location information of the future conflict grid pair and the time information of the future time window.

[0149] A pre-defined conflict category label combination rule table defines which label combinations indicate high-risk conflicts. For example, if both L_pred_T and L_pred_A are "high-frequency short-term," it means that two adjacent grid cells will be occupied frequently in the future, posing a risk of track conflict. When the prediction result matches the conflict rule, the target grid cell GridT and the adjacent grid cell GridA are marked as a future conflict grid pair. The center time or start time T_future of the future time window targeted by the prediction is obtained. A pre-emptive track conflict warning command Warn_conflict is generated, which includes the position coordinates of the conflict grid pair, the predicted conflict time T_future, and the conflict type. This warning command is sent to the airspace control center dispatch terminal and the control terminals of aircraft that may be involved in the area.

[0150] Step S810: After generating the grid occupancy continuity mode record, extract the collection timestamp corresponding to each aircraft identity switching event record from the occupancy aircraft switching sequence record in the grid occupancy continuity mode record as the switching occurrence timestamp. Using the switching occurrence timestamp as the time retrieval basis, extract the flight status message corresponding to the switching occurrence timestamp from the flight status message stream, and obtain the instantaneous velocity vector field and heading azimuth angle field of the aircraft identifier field before the switch at the switching occurrence timestamp recorded in the flight status message. At the same time, obtain the instantaneous velocity vector field and heading azimuth angle field of the aircraft identifier field after the switch at the switching occurrence timestamp.

[0151] The SeqSw record in the PatOcc grid occupancy continuation mode record contains the aircraft identity switching event record SwRec=(IdPrev, IdNext, Ts), where Ts is the switching time (i.e., the time when the preceding aircraft leaves the grid and the following aircraft enters the grid). Using Ts as the query key, messages with a collection timestamp equal to Ts are retrieved from the flight status message stream. From the retrieved messages, the instantaneous velocity vector (V_prev_x, V_prev_y, V_prev_z) and heading / azimuth angle H_prev corresponding to IdPrev are extracted. Similarly, the velocity vector (V_next_x, V_next_y, V_next_z) and heading / azimuth angle H_next corresponding to IdNext are extracted.

[0152] Step S811: Construct the departure trajectory direction vector of the aircraft before the switch based on the instantaneous velocity vector field and heading azimuth field of the aircraft identifier field before the switch; construct the entry trajectory direction vector of the aircraft after the switch based on the instantaneous velocity vector field and heading azimuth field of the aircraft identifier field after the switch; calculate the trajectory direction angle between the departure trajectory direction vector and the entry trajectory direction vector; compare the trajectory direction angle with a preset trajectory direction angle classification threshold set to determine the trajectory direction change type marker of this aircraft identity switch event; the trajectory direction change type marker includes same-direction continuation type, cross-direction reversal type, and opposite-direction rendezvous type.

[0153] Before the switch, the departure direction vector of the aircraft is taken as the horizontal velocity direction vector D_prev=(V_prev_x, V_prev_y), and normalized. After the switch, the entry direction vector of the aircraft is taken as D_next=(V_next_x, V_next_y), and normalized. The angle θ between the two direction vectors is calculated as θ=arccos((D_prev·D_next) / (|D_prev|·|D_next|)), and the result is between 0 and 180 degrees. Preset classification thresholds for the trajectory direction angle are: when θ<30 degrees, it is marked as a continuation in the same direction, indicating that the subsequent aircraft enters the grid in the same direction as the preceding aircraft; when 30 degrees≤θ≤150 degrees, it is marked as a cross-direction reversal type, indicating that the flight paths intersect; when θ>150 degrees, it is marked as a meeting in opposite directions, indicating that the two aircraft are traveling towards each other.

[0154] Step S812: Extract the duration of occupation of the grid cell at multiple consecutive acquisition timestamps before and after the switching timetamp, count the change in duration of occupation and the duration of the occupation interruption gap before and after the switching timetamp, combine the track direction change type marker, the change in duration of occupation and the duration of the occupation interruption gap into a multi-dimensional feature record of the switching event for this aircraft identity switching event, collect the multi-dimensional feature records of the switching events of all aircraft identity switching events of all grid cells in the airspace to be controlled within a preset historical time period, and construct a multi-dimensional feature sample set of switching events.

[0155] From the SeqDur generated in step S133, find the duration of occupancy before and after the occupancy event containing the handover time Ts. Let the duration of the occupancy event before the handover be D_before, and the duration of the occupancy event after the handover be D_after. Calculate the change in occupancy duration ΔD = D_after - D_before. Calculate the occupancy interruption interval ΔG, which is the time interval between the end of the previous occupancy event and the start of the next occupancy event. Combine (TypeDir, ΔD, ΔG) into a three-dimensional feature vector V_feat = [TypeDir encoding, ΔD, ΔG]. Wherein, TypeDir encoding adopts one-hot encoding: the same-direction continuation type is encoded as [1, 0, 0], the crossover type is encoded as [0, 1, 0], and the opposite rendezvous type is encoded as [0, 0, 1]. Perform the above extraction on all switching events of all grid cells within the historical time period to obtain the feature sample set S_feat={V_feat1, V_feat2, ..., V_featM}.

[0156] Step S813: Retrieve the pre-constructed handover event security level determination network. The handover event security level determination network includes a multi-dimensional feature input layer, a feature cross-combination layer, a deep feature transformation layer, and a security level classification output layer. The multi-dimensional feature input layer is used to receive the type encoding vector, the change in occupation duration, and the occupation interruption gap duration corresponding to the track direction change type marker in the multi-dimensional feature record of the handover event. The feature cross-combination layer is used to perform pairwise feature cross-combinations on the type encoding vector, the change in occupation duration, and the occupation interruption gap duration to generate a cross-combination feature vector. The deep feature transformation layer is used to perform multi-level nonlinear transformations on the cross-combination feature vector to generate a deep abstract feature vector. The security level classification output layer is used to output the security level classification result of the aircraft identity handover event based on the deep abstract feature vector.

[0157] The event security level determination network is a multilayer perceptron classifier. The input layer (multidimensional feature input layer) receives a 5-dimensional input vector: a type encoding vector (3D), ΔD (1D), and ΔG (1D). The feature cross-combination layer combines the inputs pairwise: it calculates the outer product of the type encoding and ΔD (3D vector multiplied by ΔD, resulting in 3D), the outer product of the type encoding and ΔG (3D vector multiplied by ΔG, resulting in 3D), and the product of ΔD and ΔG (1D). These three cross-combination results are concatenated to obtain a 7-dimensional cross-combination feature vector. The deep feature transformation layer contains three fully connected layers with 64, 32, and 16 neurons respectively. Each fully connected layer is followed by a batch normalization layer and a ReLU activation function, with dropout regularization (dropout rate 0.3). The security level classification output layer is a fully connected layer that maps the 16-dimensional features to 4 dimensions (corresponding to four security levels: Level 1, Level 2, Level 3, and Level 4), followed by a Softmax function to output the probability distribution. Network training process: The feature vector in S_feat is used as input, and the manually labeled safety level labels (0-3 integers) are used as supervision signals. The classification cross-entropy loss function is adopted, the optimizer is Adam, the initial learning rate is 0.0005, the batch size is 32, the training epochs are 40, and the early stopping epochs are 10.

[0158] Step S814: Input each multi-dimensional feature record of the switching event in the multi-dimensional feature sample set of the switching event into the switching event security level determination network. The multi-dimensional feature input layer converts the track direction change type mark into a type encoding vector, and concatenates the type encoding vector with the change in occupation duration and the occupation interruption interval duration to form an input feature vector.

[0159] For each switching event multidimensional feature record, its trajectory direction change type is converted into a 3-dimensional one-hot encoded vector C_type. C_type, ΔD, and ΔG are concatenated into a 5-dimensional input feature vector X_in=[C_type0, C_type1, C_type2, ΔD, ΔG].

[0160] Step S815: The feature cross-combination layer performs an outer product operation on the type encoding vector and the change in occupation duration in the input feature vector to generate a first cross feature vector, performs an outer product operation on the type encoding vector and the occupation interruption gap duration to generate a second cross feature vector, performs an outer product operation on the change in occupation duration and the occupation interruption gap duration to generate a third cross feature vector, and concatenates the first cross feature vector, the second cross feature vector and the third cross feature vector into a cross-combination feature vector.

[0161] Extract the type encoding vector C_type (3D) and ΔD (1D) from X_in. Calculate their outer product to obtain the 3D first cross feature V_cross1 = C_type * ΔD. Calculate the outer product of C_type and ΔG to obtain the 3D second cross feature V_cross2 = C_type * ΔG. Calculate the product of ΔD and ΔG to obtain the 1D third cross feature V_cross3 = ΔD * ΔG. Concatenate V_cross1, V_cross2, and V_cross3 to form a 7D cross-combination feature vector X_cross = [V_cross1, V_cross2, V_cross3].

[0162] Step S816: The deep feature transformation layer performs layer-by-layer feature transformation processing on the cross-combined feature vector through multiple fully connected network layers. Each fully connected network layer performs linear transformation and non-linear activation processing on the input features and then passes the output to the next layer. After processing by all fully connected network layers, a deep abstract feature vector is obtained.

[0163] The X_cross vector is input into the first fully connected layer (7-dimensional input, 64-dimensional output), and subjected to a linear transformation Y1 = W1·X_cross + b1. After batch normalization and ReLU activation, H1 is obtained. H1 is then input into the second fully connected layer (64-dimensional -> 32-dimensional), and H2 is obtained similarly. H2 is then input into the third fully connected layer (32-dimensional -> 16-dimensional), and after ReLU activation, the deep abstract feature vector H_deep (16-dimensional) is obtained.

[0164] Step S817: The security level classification output layer calculates the probability distribution of the aircraft identity switching event belonging to each preset security level category based on the deep abstract feature vector, and outputs the security level category with the highest probability as the switching security level determination result of the aircraft identity switching event.

[0165] Input H_deep into the security level classification output layer (16D -> 4D), perform a linear transformation to obtain the original score Z = W_out·H_deep + b_out. Apply the Softmax function to calculate the probability distribution P = softmax(Z), where P_k = exp(Z_k) / Σexp(Z_j) for j = 1..4. Take the security level (0, 1, 2, 3) corresponding to argmax(P) as the judgment result L_safety, where 0 corresponds to level 1 (safe), 1 corresponds to level 2 (low risk), 2 corresponds to level 3 (medium risk), and 3 corresponds to level 4 (high risk).

[0166] Step S818: When the handover security level determination result belongs to the preset high-risk security level category, extract the grid cell number, handover timestamp, pre-handover aircraft identifier field and post-handover aircraft identifier field corresponding to the aircraft identity handover event, generate a handover security risk warning instruction carrying the grid location and the aircraft identifier involved, and distribute the handover security risk warning instruction to the aircraft control terminal corresponding to the pre-handover aircraft identifier field and the post-handover aircraft identifier field.

[0167] If L_safety equals 3 (Level 4 High Risk), then extract the grid cell number G_switch, the time of the switchover T_switch, the aircraft identifier Id_prev before the switchover, and the aircraft identifier Id_next after the switchover corresponding to the switchover event. Generate a switchover safety risk warning command Warn_risk={“type”:“switch_risk”,“grid”:G_switch,“time”:T_switch,“aircrafts”:[Id_prev,Id_next],“risk_level”:“high”}. Send this warning command through the communication link to the aircraft control terminal corresponding to Id_prev and Id_next, indicating the existence of a high-risk switchover event.

[0168] Step S910: After generating the grid occupancy continuity mode record, perform switching flow statistics processing on the occupancy aircraft switching sequence record of each grid cell in the airspace to be controlled under continuous acquisition timestamps, extract the pre-switching aircraft identifier field and the post-switching aircraft identifier field of the aircraft identity switching event record in each grid cell, count the occurrence frequency of each target grid cell after each aircraft identifier field leaves from a grid cell, and generate the inter-grid flow frequency distribution record corresponding to each aircraft identifier field.

[0169] For each grid cell, iterate through each transition record (Id_prev, Id_next, Ts) in its SeqSw. Record one flow direction: after aircraft Id_prev leaves the grid cell, aircraft Id_next enters the grid cell. Using Id_prev as the key, record the number of times the aircraft enters the grid cell (i.e., the current grid) after leaving the grid cell. At the same time, using Id_prev as the key, record the pairings of leaving and entering the grid cells. After statistics, obtain the flow direction frequency distribution record FlowStat={aircraft identifier:{(leaving grid, entering grid): frequency of occurrence}}.

[0170] Step S911: Extract the main flow direction from the inter-grid flow frequency distribution record corresponding to each aircraft identifier field, and extract the flow direction combination from the preceding grid cell number to the following grid cell number with the highest frequency as the normal flow path record of the aircraft identifier field.

[0171] For each aircraft ID, find the most frequent (leaving the grid, entering the grid) pairings from its FlowStat, denoted as the normal flow path Path_norm=(Grid_leave, Grid_enter). This normal flow path indicates that the most common flight direction for this aircraft is from Grid_leave to Grid_enter.

[0172] Step S912: Arrange the normal flow path records of each aircraft identifier field in multiple consecutive time periods in chronological order to generate the normal flow path time sequence of the aircraft identifier field.

[0173] Divide time into continuous periods (e.g., every hour). For each period, calculate the normal flow path Path_norm(t) for the aircraft. Arrange the Path_norm of all periods in chronological order to obtain the normal flow path time series Seq_flow=[Path_norm1, Path_norm2, Path_norm3, ...].

[0174] Step S913: Retrieve the pre-constructed aircraft flow direction anomaly deviation detection network. The aircraft flow direction anomaly deviation detection network includes a flow direction path embedding layer, a temporal context encoding layer, a flow direction path reconstruction layer, and a deviation degree evaluation layer. The flow direction path embedding layer is used to convert the grid cell numbers in the normal flow direction path record into dense embedding vector representations. The temporal context encoding layer is used to encode the temporal correlation between the preceding and following paths in the temporal sequence of the normal flow direction path to generate a temporal context-aware vector. The flow direction path reconstruction layer is used to reconstruct the expected flow direction path embedding representation at the current moment based on the temporal context-aware vector. The deviation degree evaluation layer is used to calculate the deviation degree quantization value between the actual flow direction path embedding representation and the expected flow direction path embedding representation.

[0175] The aircraft flow direction deviation detection network employs an autoencoder architecture. The flow path embedding layer is an embedding matrix that maps grid numbers (L grids in total) to 32-dimensional vectors. The temporal context encoding layer uses gated recurrent units, taking the embedding vector sequence as input and outputting the hidden state (64-dimensional) of the last time step. The flow path reconstruction layer is a fully connected network that maps the 64-dimensional context encoding vector back to 32-dimensional embedding vectors. The deviation evaluation layer calculates the Euclidean distance between two 32-dimensional vectors. Network training is performed using a Seq_flow sequence under normal flight conditions, trained in a self-supervised manner. The input is the first t-1 path embedding vectors, and the output is the reconstructed t-th path embedding vector. The loss function is mean squared error. The optimizer uses Adam with an initial learning rate of 0.001.

[0176] Step S914: Using the flow path embedding layer, each normal flow path record in the normal flow path time sequence of each aircraft identifier field is embedded. The preceding grid cell number and the following grid cell number in the normal flow path record are mapped to the preceding grid embedding vector and the following grid embedding vector, respectively. The preceding grid embedding vector and the following grid embedding vector are concatenated to form the path embedding vector of the normal flow path record.

[0177] For each normal flow path record Path_norm=(Grid_l, Grid_e), query the embedding matrix of the flow path embedding layer to obtain the 32-dimensional embedding vector Emb_l of Grid_l and the 32-dimensional embedding vector Emb_e of Grid_e. Concatenate Emb_l and Emb_e to form a 64-dimensional path embedding vector Emb_path.

[0178] Step S915: The temporal context coding layer is used to perform temporal encoding processing on the path embedding vector temporal sequence of each aircraft identifier field in chronological order. A temporal attention weighting mechanism is applied to multiple path embedding vectors before the current moment. Temporal attention weights are assigned according to the relevance of the path embedding vectors at each historical moment to the context at the current moment. The path embedding vectors at each historical moment are weighted and summed based on the temporal attention weights to generate the temporal context-aware vector at the current moment.

[0179] The historical path embedding vector sequence [Emb1, Emb2, ..., Emb_{T-1}] of the aircraft ID is input into the gated recurrent unit. At each time step, the gated recurrent unit calculates the update gate and the reset gate. The update gate determines how much of the hidden state from the previous time step is retained in the current time step, and the reset gate determines how much of the current input is combined with the hidden state from the previous time step. After T-1 time steps, the hidden state sequence [h1, h2, ..., h_{T-1}] is obtained. An attention mechanism is used to calculate the context vector at the current time step: the relevance score s_i = h_i·h_{T-1} between each historical hidden state h_i and the current query vector (taking the last hidden state) is calculated. The attention weight α_i = exp(s_i) / Σexp(s_j) is obtained through Softmax normalization, and then a weighted sum is taken to obtain the context vector Ctx = Σα_i*h_i.

[0180] Step S916: The current time-series context-aware vector is input into the path reconstruction network using the flow path reconstruction layer. The path reconstruction network generates the expected flow path embedding vector for the current time based on the historical flow path patterns and time-series evolution rules contained in the time-series context-aware vector.

[0181] The context vector Ctx is input into the fully connected network of the flow path reconstruction layer (64-dimensional input, 32-dimensional output). After linear transformation and ReLU activation, a 32-dimensional expected flow path embedding vector Emb_pred is obtained.

[0182] Step S917: Calculate the actual flow path embedding vector of the aircraft identifier field actually observed through the flight status message stream at the current moment using the deviation assessment layer, and perform vector space distance calculation processing on the actual flow path embedding vector and the expected flow path embedding vector to generate a flow path deviation quantification value.

[0183] Obtain the actual observed aircraft flow path at the current time: Path_actual = (Grid_l_act, Grid_e_act). Convert it into an actual embedding vector Emb_actual (32-dimensional) using the same flow path embedding layer. Calculate the Euclidean distance between Emb_actual and Emb_pred: D_dev = ||Emb_actual - Emb_pred||2.

[0184] Step S918: Compare the quantified value of the deviation degree of the flow path with the preset deviation degree classification judgment benchmark sequence. When the quantified value of the deviation degree of the flow path falls into the preset abnormal deviation degree range, it is determined that the flow path of the aircraft identifier field has deviated abnormally. Extract the current grid cell number, the current collection timestamp, and the target grid cell number pointed to by the actual flow path when the aircraft identifier field deviates abnormally. Generate an aircraft flow abnormal deviation alarm command. The aircraft flow abnormal deviation alarm command includes the abnormal aircraft identifier field, the abnormal occurrence grid number, the abnormal occurrence time, and the abnormal deviation target grid number. Distribute the aircraft flow abnormal deviation alarm command to the aircraft control terminal and the airspace control center dispatch terminal corresponding to the aircraft identifier field.

[0185] A preset deviation severity grading benchmark sequence is established: D_dev < 1.0 is normal, 1.0 ≤ D_dev < 2.0 is slight deviation, 2.0 ≤ D_dev < 3.0 is moderate deviation, and D_dev ≥ 3.0 is severe deviation. When D_dev ≥ 2.0, it is determined to be an abnormal deviation of the flow path. The current grid number Grid_cur, the current data collection time stamp T_cur, and the target grid number Grid_target of the actual flow path for the aircraft are extracted. An abnormal flow path deviation alarm command is generated: Warn_flow = {"type":"flow_anomaly", "aircraft_id":Id", "current_grid":Grid_cur", "time":T_cur", "target_grid":Grid_target", "deviation":D_dev}. This alarm command is sent to the control terminal corresponding to the aircraft ID and the airspace control center dispatch terminal.

[0186] In one exemplary embodiment, an intelligent control system for low-altitude aircraft is provided. This intelligent control system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2As shown, the intelligent control system for low-altitude aircraft includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements an intelligent control method for low-altitude aircraft. The display unit is used to generate a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of the intelligent control system for low-altitude aircraft, or an external keyboard, touchpad, or mouse, etc.

[0187] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A smart control method for low-altitude aircraft, characterized in that, The method includes: Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field. The flight status message stream is subjected to airspace projection rasterization processing. Based on the instantaneous position coordinate field, each flight status message is mapped to the corresponding grid cell in the preset airspace control grid system to obtain the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp. For each grid cell, the single grid occupancy marker set is processed by time sequence occupancy status change analysis to generate a grid occupancy change pattern record for that grid cell under multiple consecutive acquisition time stamps. The grid occupancy change pattern record includes an occupancy duration sequence and an occupancy aircraft switching sequence record. The pre-defined multi-grid occupancy change pattern cross-comparison structure is used to perform cross-comparison processing on the grid occupancy change pattern records of adjacent grid cells, generating occupancy pattern cross-comparison difference records between adjacent grid cells; Based on the cross-comparison difference records of the occupancy patterns, the aircraft track interleaving situation category and track interleaving spatiotemporal distribution record between the adjacent grid cells are determined. Based on the aircraft track interleaving situation category and the track interleaving spatiotemporal distribution record, a flight avoidance scheduling instruction carrying a grid coordinate chain is generated and the flight avoidance scheduling instruction is distributed to the corresponding aircraft control terminal.

2. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The process of performing airspace projection rasterization on the flight status message stream maps each flight status message to a corresponding grid cell in a preset airspace control grid system based on the instantaneous position coordinate field, obtaining the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship for each grid cell at each acquisition time stamp, including: Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field. Extract the instantaneous position coordinate field from the flight status message stream one by one, and obtain the longitude coordinate components, latitude coordinate components and altitude coordinate components of the instantaneous position coordinate field in the three-dimensional airspace coordinate system; The preset airspace control grid system is retrieved. The airspace control grid system divides the airspace to be controlled in a three-dimensional airspace coordinate system according to a uniform grid side length, forming longitude layer grid units, latitude layer grid units and altitude layer grid units with unique grid numbers. Based on the longitude coordinate components of the instantaneous position coordinate field and the boundary coordinate range of each longitude layer grid unit in the airspace control grid system, coordinate assignment matching is performed to determine the target longitude layer grid number to which the instantaneous position coordinate field falls. Based on the latitudinal coordinate components of the instantaneous position coordinate field and the boundary coordinate range of each latitudinal layer grid unit in the airspace control grid system, coordinate attribution matching is performed to determine the target latitudinal layer grid number to which the instantaneous position coordinate field falls. Based on the height coordinate component of the instantaneous position coordinate field and the boundary coordinate range of each height layer grid unit in the airspace control grid system, coordinate assignment matching is performed to determine the target height layer grid number to which the instantaneous position coordinate field falls; The target longitude layer grid number, the target latitude layer grid number, and the target altitude layer grid number are combined to generate the three-dimensional grid number of the target airspace control grid unit corresponding to the flight status message; Establish an association record between the aircraft identifier field of the current flight status message and the three-dimensional grid number, and store it in the grid occupancy temporary mapping table entry corresponding to the acquisition timestamp; Traverse all flight status messages with the same acquisition timestamp field in the flight status message stream, perform coordinate attribution matching processing one by one, and generate a set of multiple association records between the multi-dimensional mesh number and the aircraft identifier field under the acquisition timestamp; The set of multiple associated records of the multi-dimensional grid number and the aircraft identifier field are grouped and aggregated according to the three-dimensional grid number. All aircraft identifier fields belonging to the same three-dimensional grid number are summarized to form a list of single-grid aircraft identifiers for that three-dimensional grid number at the time of acquisition. Based on the existence status of the entries in the aircraft identifier field of the single-grid occupancy aircraft identifier list, a single-grid occupancy mark for the 3D grid number under the acquisition time stamp is generated. The single-grid occupancy marks of all 3D grid numbers are arranged in the order of the acquisition time stamp field to generate a set of single-grid occupancy marks for each grid cell under each acquisition time stamp. A mapping relationship between each 3D grid number and the single-grid occupancy aircraft identifier list under the corresponding acquisition time stamp is established as the grid occupancy aircraft identifier mapping relationship.

3. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The process of performing temporal occupancy status continuity analysis on the single-grid occupancy marker set for each grid cell generates a record of the grid occupancy continuity pattern for that grid cell under multiple consecutive acquisition time stamps, including: Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field. The flight status message stream is subjected to airspace projection rasterization processing. Based on the instantaneous position coordinate field, each flight status message is mapped to the corresponding grid cell in the preset airspace control grid system to obtain the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp. The single grid occupancy marker set of the target grid cell is traversed according to the time sequence of the acquisition timestamp field. The acquisition timestamp when the single grid occupancy marker changes from an occupied state to an unoccupied state is identified as the occupancy end timestamp, and the acquisition timestamp when the single grid occupancy marker changes from an unoccupied state to an occupied state is identified as the occupancy start timestamp. From the start time of occupation to the end time of occupation, the corresponding continuous collection time stamp interval is marked as an occupation event of the target grid cell. The start time of occupation and the end time of occupation corresponding to the occupation event are extracted. The time span between the end time of occupation and the start time of occupation is calculated to generate the occupation duration segment of the occupation event. For all occupancy events identified in the target grid cell during the entire continuous monitoring period, the occupancy duration segment of each occupancy event is calculated one by one, and the occupancy duration segments are arranged in the order of the occurrence of the occupancy events to form the occupancy duration sequence of the target grid cell. Between two adjacent occupancy events, the single-grid occupancy aircraft identifier list corresponding to the previous occupancy event and the single-grid occupancy aircraft identifier list corresponding to the next occupancy event are retrieved through the grid occupancy aircraft identifier mapping relationship. The single-grid occupancy aircraft identifier list of the previous occupancy event is compared with the single-grid occupancy aircraft identifier list of the next occupancy event. The aircraft identifier field that exists in the previous list but does not exist in the next list is extracted as the switching departure aircraft identifier, and the aircraft identifier field that does not exist in the previous list but exists in the next list is extracted as the switching entry aircraft identifier. The switching departure aircraft identifier and the switching entry aircraft identifier are paired to form an aircraft identity switching event record between two adjacent occupancy events. The aircraft identity switching event record includes an aircraft identifier field before switching and an aircraft identifier field after switching. Arrange all aircraft identity switching event records according to the chronological order of the occupancy events to generate an occupancy aircraft switching sequence record for the target grid cell under multiple consecutive acquisition timestamps. Synchronize and align the occupancy duration sequence and the occupancy aircraft switching sequence record in the time dimension to generate a grid occupancy continuity change mode record for the target grid cell.

4. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The process of cross-comparing the grid occupancy change pattern records of adjacent grid cells using a preset multi-grid occupancy change pattern cross-comparison structure to generate occupancy pattern cross-comparison difference records between adjacent grid cells includes: Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field. The flight status message stream is subjected to airspace projection rasterization processing. Based on the instantaneous position coordinate field, each flight status message is mapped to the corresponding grid cell in the preset airspace control grid system to obtain the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp. For each grid cell, the single grid occupancy marker set is processed by time sequence occupancy status change analysis to generate a grid occupancy change pattern record for that grid cell under multiple consecutive acquisition time stamps. The grid occupancy change pattern record includes an occupancy duration sequence and an occupancy aircraft switching sequence record. In the airspace control grid system, the three-dimensional grid number of each target grid cell is used as an index to locate the adjacent grid cells that are directly adjacent to the target grid cell in the longitude, latitude, and altitude directions, respectively, and to establish a one-to-one pairing relationship between the target grid cell and each adjacent grid cell. For each pair of relationships, the occupancy duration sequence of the target grid cell is extracted, and the occupancy duration sequence of the adjacent grid cells is also extracted. The occupancy duration segments in the same time window in the two sets of occupancy duration sequences are subtracted one time window at a time to generate a window-by-window difference record sequence of the duration sequence of the pairing relationship. Extract the aircraft switching sequence record of the target grid cell, and simultaneously extract the aircraft switching sequence record of the adjacent grid cell. Compare the aircraft identity switching event records in the same time window in the two sets of aircraft switching sequence records before and after the switching time, calculate the time misalignment between the switching time of the target grid cell and the switching time of the adjacent grid cell, and generate a sequence of switching event time misalignment records for this pairing relationship. Each difference value in the window-by-window difference value record sequence of the duration sequence is compared and graded with the preset difference degree grading benchmark sequence to obtain the difference grading record of the duration dimension under the pairing relationship. Each time misalignment in the sequence of switching event time misalignment is compared and graded with a preset misalignment degree grading benchmark sequence to obtain the switching dimension difference grading record under the pairing relationship. The historical dimension occupancy difference classification record and the switching dimension difference classification record are merged and combined window by window in the same time window to generate occupancy pattern cross-comparison difference record entries for each time window between adjacent grid cells under the pairing relationship. The occupancy pattern cross-comparison difference record entries of all time windows are arranged in chronological order to generate occupancy pattern cross-comparison difference records between adjacent grid cells under the pairing relationship.

5. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The process of determining the aircraft track overlap situation category and track overlap spatiotemporal distribution record between adjacent grid cells based on the cross-comparison difference record of the occupancy pattern, generating a flight avoidance scheduling command carrying a grid coordinate chain based on the aircraft track overlap situation category and track overlap spatiotemporal distribution record, and distributing the flight avoidance scheduling command to the corresponding aircraft control terminal includes: Receive flight status message streams periodically broadcast by low-altitude aircraft within the airspace to be controlled. The flight status message stream contains a message sequence with an aircraft identifier. Each message carries an instantaneous position coordinate field, an instantaneous velocity vector field, a heading azimuth field, and a data acquisition timestamp field. The flight status message stream is subjected to airspace projection rasterization processing. Based on the instantaneous position coordinate field, each flight status message is mapped to the corresponding grid cell in the preset airspace control grid system to obtain the single grid occupancy mark set and the grid occupancy aircraft identifier mapping relationship of each grid cell under each collection time stamp. For each grid cell, the single grid occupancy marker set is processed by time sequence occupancy status change analysis to generate a grid occupancy change pattern record for that grid cell under multiple consecutive acquisition time stamps. The grid occupancy change pattern record includes an occupancy duration sequence and an occupancy aircraft switching sequence record. Using a preset multi-grid occupancy change pattern cross-comparison structure, the grid occupancy change pattern records of adjacent grid cells are cross-compared to generate occupancy pattern cross-comparison difference records between adjacent grid cells. From the occupancy pattern cross-comparison difference records, the occupancy pattern cross-comparison difference record entries for each time window are extracted one by one, and the historical dimension occupancy difference classification records and the switching dimension difference classification records are separated. The time-dimension occupancy difference classification records are input into a preset track interleaving time-dimension classification judgment logic structure. The track interleaving time-dimension classification judgment logic structure maintains a mapping table between the time-dimension difference classification values ​​and the track interleaving time-dimension intensity level codes, and outputs the track interleaving time-dimension intensity level codes by looking up the table. Additionally, the switching dimension difference classification records are input into a preset track interleaving switching classification judgment logic structure. The track interleaving switching classification judgment logic structure maintains a mapping table between the switching difference classification values ​​and the track interleaving switching intensity level codes, and outputs the track interleaving switching intensity level codes by looking up the table. The track interleaving duration intensity level code and the track interleaving switching intensity level code are combined to form a hierarchical code combination. The hierarchical code combination is input into a pre-constructed track interleaving situation category determination logic structure. The track interleaving situation category determination logic structure maintains a mapping table between the hierarchical code combination and the track interleaving situation category code. The aircraft track interleaving situation category code corresponding to the pairing relationship of the adjacent grid cells is output by looking up the table. Extract the time window set containing the difference records of the cross-comparison of the occupancy patterns, extract the start collection time stamp and end collection time stamp of each time window in the time window set, and generate a track cross-spatiotemporal distribution record containing the cross-start time, cross-duration duration and cross-grid number pair; Based on the aircraft track interleaving situation category code, a matching avoidance strategy rule entry is retrieved from the preset flight avoidance strategy rule set, and the avoidance direction command and avoidance priority parameter contained in the avoidance strategy rule entry are extracted. Extract the interlaced grid number pair from the interlaced spatiotemporal distribution record of the flight path, and in the airspace control grid system, take the interlaced grid number pair as the reference, expand the adjacent grid cells outward along the spatial adjacency direction, arrange the grid numbers on the expansion path in spatial connection order, and generate the avoidance path grid coordinate chain. The avoidance direction command, the avoidance priority parameter, and the avoidance path grid coordinate chain are encapsulated and combined to generate a flight avoidance scheduling command carrying the grid coordinate chain. Based on the mapping relationship between the staggered grid number and the associated grid-occupied aircraft identifier, the set of aircraft identifier fields involved in the staggered grid is extracted. The flight avoidance scheduling command is then distributed to the aircraft control terminal corresponding to the set of aircraft identifier fields through a communication link.

6. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The method further includes: After distributing the flight avoidance scheduling command to the corresponding aircraft control terminal, retrieve the staggered grid number pair and the avoidance path grid coordinate chain corresponding to the flight avoidance scheduling command in the distributed record; Using the grid numbers contained in the staggered grid number pair as an index, all flight status messages falling into the staggered grid number pair are retrieved in the subsequently collected flight status message stream. The instantaneous velocity vector field and heading azimuth field of the flight status message are extracted, and the aircraft identifier field of the flight status message is also extracted. Extract the avoidance direction instruction from the flight avoidance scheduling instruction, parse the reference heading azimuth value indicated by the avoidance direction instruction, perform difference calculation processing between the heading azimuth field of the retrieved flight status message and the reference heading azimuth value, and generate the heading deviation angle value corresponding to the flight status message. The speed magnitude component in the instantaneous speed vector field of the retrieved flight status message is calculated with the preset avoidance reference speed value to generate the speed deviation value corresponding to the flight status message. Based on the deviation angle range into which the heading deviation angle value falls, a heading response classification label is assigned to the flight status message. Based on the deviation range into which the speed deviation value falls, a speed response classification flag is assigned to the flight status message. The heading response classification flag and speed response classification flag corresponding to the same aircraft identifier field are combined to generate the avoidance response execution feedback record corresponding to the aircraft identifier field. The avoidance response execution feedback record is associated with the aircraft identifier field and stored in the aircraft avoidance history feedback record set. When the low-altitude aircraft corresponding to the aircraft identifier field is again the object of flight avoidance scheduling instruction distribution, the historical avoidance response execution feedback record corresponding to the aircraft identifier field is extracted from the aircraft avoidance history feedback record set, and the avoidance priority parameter of the low-altitude aircraft is adjusted based on the historical avoidance response execution feedback record.

7. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The method further includes: After generating the grid occupancy continuous change mode record, extract the aircraft identity switching event record one by one from the occupancy aircraft switching sequence record of each grid cell, and obtain the aircraft identifier field before switching and the aircraft identifier field after switching in each aircraft identity switching event record. Using the pre-switch aircraft identifier field as the query basis, the pre-built aircraft basic parameter filing record library is retrieved to extract the cruise rate parameter and the turning rate parameter of the preceding aircraft corresponding to the pre-switch aircraft identifier field; and using the post-switch aircraft identifier field as the query basis, the aircraft basic parameter filing record library is retrieved to extract the cruise rate parameter and the turning rate parameter of the following aircraft corresponding to the post-switch aircraft identifier field. From the occupancy duration sequence of this grid cell, locate the occupancy duration segments of the preceding and following occupancy events adjacent to the current aircraft identity switching event record; The method calculates the cruise rate difference between the cruise rate parameters of the preceding and following aircraft, and correlates this cruise rate difference with the duration difference between the duration of the previous and following occupancy events to generate a correlation record between cruise rate difference and occupancy duration change; and calculates the turn rate difference between the turn rate parameters of the preceding and following aircraft, and correlates this turn rate difference with the duration difference between the duration of the previous and following occupancy events to generate a correlation record between turn rate difference and occupancy duration change. Aggregate the cruise rate difference and occupation duration change association records corresponding to all aircraft identity switching event records within the same grid cell, extract the corresponding patterns of recurring cruise rate difference intervals and occupation duration change directions in the grid cell, and generate the cruise rate associated occupation change pattern record for the grid cell; and aggregate the turning angle rate difference and occupation duration change association records corresponding to all aircraft identity switching event records within the same grid cell, extract the corresponding patterns of recurring turning angle rate difference intervals and occupation duration change directions in the grid cell, and generate the turning angle rate associated occupation change pattern record for the grid cell. The cruise rate-related occupancy change pattern record and the steering angle rate-related occupancy change pattern record are merged and combined to generate the aircraft performance-related occupancy change comprehensive pattern record for the grid cell. The aircraft performance-related occupancy change comprehensive pattern record is then appended to the grid occupancy change pattern record for the grid cell to form an expanded grid occupancy change pattern record. This expanded grid occupancy change pattern record replaces the original grid occupancy change pattern record for cross-comparison processing of multiple grid occupancy change patterns.

8. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The method further includes: After obtaining the set of single-grid occupancy markers for each grid cell at each acquisition time stamp, spatial statistical processing is performed on the single-grid occupancy markers of all grid cells at the same acquisition time stamp. The number of aircraft in the occupied state of each grid cell at the acquisition time stamp is counted, and a record of the distribution of the number of grid occupancy in the entire airspace at the acquisition time stamp is generated. The records of the distribution of the number of grid occupancy in the entire airspace at multiple consecutive acquisition time stamps are arranged in chronological order to generate a sequence of time-series changes in the number of grid occupancy in the entire airspace. Extract the time-series curve of the occupancy quantity of each grid cell from the time-series record of the change in the occupancy quantity over time, identify the continuous time segments in the time-series curve where the slope of the occupancy quantity increases exceeds a preset slope threshold, and mark the continuous time segments as density surge time windows. Extract the 3D grid number of the corresponding grid cell within the density surge time window to form a density surge grid number set. From the grid-occupied aircraft identifier mapping relationship corresponding to the density surge grid number set, extract all aircraft identifier fields that exist in the grid cell within the density surge time window to form a density surge associated aircraft identifier set. For each aircraft identifier field in the density surge associated aircraft identifier set, retrieve its corresponding instantaneous velocity vector field and heading azimuth field. Based on the instantaneous velocity vector field and the heading azimuth field, extrapolate the expected arrival grid number of the aircraft identifier at the subsequent acquisition timestamp along the time axis. Compare the expected arrival grid number of each aircraft identifier with the density surge grid number set. When the expected arrival grid number of a certain aircraft identifier belongs to the density surge grid number set, mark the aircraft identifier as an aircraft flowing into the surge area. The aircraft identifier fields of all aircraft marked as aircraft entering the surge area are counted to generate a list of aircraft entering the surge area. An expected inflow time stamp is marked for each aircraft entering the surge area. Based on the list of aircraft entering the surge area and the expected inflow time stamp, an airspace density continuous surge warning notification is generated. The airspace density continuous surge warning notification includes the surge grid number, the warning time window and the list of aircraft entering the surge area identifier. The airspace density continuous surge warning notification is sent to the aircraft control terminal corresponding to each aircraft identifier field in the list of aircraft entering the surge area identifier.

9. The intelligent control method for low-altitude aircraft according to claim 1, characterized in that, The method further includes: After generating the grid occupancy change pattern record, the occupancy duration sequence of the target grid cell within a preset long-term monitoring period is extracted. The occupancy duration sequence is segmented into occupancy periods. The peak occupancy duration that repeats at fixed time intervals in the occupancy duration sequence is detected. The time span between two adjacent peak occupancy durations is recorded as an occupancy period segment. The moment when the peak occupancy duration occurs in each occupancy period segment is extracted as the period peak moment. The average occupancy period of the target grid cell is obtained by averaging the time spans of multiple occupancy period segments. The average occupancy period is then compared and matched with a preset set of period type determination benchmark durations to determine the spatial occupancy period type label of the target grid cell. Extract the peak duration of occupation within each occupation period segment, compare and match each peak duration with a preset set of peak intensity classification benchmarks, assign a periodic peak intensity classification marker to each occupation period segment, and arrange the spatial occupation period type marker and the periodic peak intensity classification marker in chronological order of the occupation period segments to generate a periodic occupation pattern record for the target grid cell. For all grid cells in the airspace control grid system, corresponding periodic occupancy pattern records are generated to form a full airspace periodic occupancy pattern record set. Periodic phase comparison processing is performed on the periodic occupancy pattern records of spatially adjacent grid cells in the full airspace periodic occupancy pattern record set to calculate the difference in average occupancy period duration and the time offset of the periodic peak time between adjacent grid cells. Based on the difference in the average occupancy period duration and the time offset of the peak moment of the period, an occupancy period phase difference record is generated between adjacent grid cells. The occupancy period phase difference record and the occupancy pattern cross-comparison difference record are then subjected to difference fusion processing on the same grid cell pairing relationship. The difference information in the occupancy period phase difference record is then incorporated into the occupancy pattern cross-comparison difference record of the pairing relationship to form an enhanced occupancy pattern cross-comparison difference record.

10. A smart control system for low-altitude aircraft, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the intelligent control method for low-altitude aircraft as described in any one of claims 1 to 9 by executing the machine-executable instructions.

Images