A hierarchical adaptive compression representation of spatiotemporal conflict graphs oriented towards conflict prediction error constraints
By constructing a multi-layered spatiotemporal conflict map and establishing a cross-layer index, distinguishing between stable regions and conflict critical regions, and adopting differentiated compression granularity and verification mechanisms, the conflict prediction problem of the internal trajectory association structure of the compression server in the high-density low-altitude UAV operation area was solved, achieving high-efficiency conflict prediction and compression.
Patent Information
- Application Number
- CN202611618111.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-20
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies struggle to compress the internal trajectory association structure of servers based on differences in conflict criticality in high-density low-altitude UAV operating areas, leading to the loss of key conflict relationships. Furthermore, the compression granularity is disconnected from the conflict sensitivity, making it impossible to balance the compression of association structures with the preservation of conflict prediction information.
Spatiotemporal conflict maps of track layer, region layer and functional airspace layer are constructed, and cross-layer indexes are established. Stable areas and critical conflict areas are distinguished based on node state similarity, adjacent structure change, predicted nearest interval and conflict prediction sensitivity. Differentiated compression granularity is adopted, and the compression results are verified by conflict recall rate, conflict probability error and key event retention rate. The affected local subgraphs are reconstructed only when the track or conflict relationship changes.
It reduces the risk of key conflict evolutions being merged, lowers the overhead of redundant storage and computation in the rolling prediction process, and balances the efficiency of spatiotemporal conflict graph compression with the reliability of conflict prediction.
Smart Images

Figure CN123115072A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of low-altitude traffic management, and in particular to a hierarchical adaptive compressed representation method for spatiotemporal conflict graphs oriented towards conflict prediction error constraints. Background Technology
[0002] Within urban high-density low-altitude drone operation areas, regional low-altitude traffic management servers need to simultaneously process the monitoring tracks, flight plans, and airspace structure of logistics delivery drones, inspection drones, and take-off and landing drones. During continuous rolling prediction, they must calculate the future positions and safe intervals between different drones. For route intersection areas, take-off and landing zones, and different cruising altitude layers, a large number of drones form dense track states and conflict relationships during continuous prediction. If track nodes and their inter-node connections are fully saved at each prediction moment, the number of nodes and connections within the server will rapidly accumulate as the prediction window grows or the number of drones increases. This generates a large number of repetitive or similar track states between adjacent prediction times. Existing compression processing usually uses fixed time windows, fixed spatial scales, or single-node similarity relationships to merge data. However, stable segments far from the conflict boundary and critical conflict segments with the predicted closest interval close to the safety interval have significantly different sensitivities to compression errors. When using a uniform compression granularity, if the compression scale is too large, it will cause the generation or disappearance of conflict edges or the merging of key changes such as the predicted closest interval crossing the safety interval boundary. On the other hand, if the compression scale is too small, it is difficult to reduce the node storage of the server and the computational load of conflict prediction. Therefore, it is difficult to balance the compression of associated structures with the preservation of conflict prediction information.
[0003] CN120803042A discloses a multi-UAV dynamic path planning system and method. Its technical focus is on multi-UAV path planning and spatiotemporal occupancy management. However, it does not establish a hierarchical compression relationship between the track layer, region layer, and functional airspace layer for the continuously generated conflict association structure, nor does it adjust the compression granularity of different regions based on differences in track node status, changes in adjacency structure, predicted nearest interval, and conflict prediction sensitivity. Therefore, in the process of continuous conflict prediction of high-density low-altitude UAVs, this technology still has difficulty compressing the track association structure inside the server according to the difference in the degree of conflict criticality and avoiding information loss in the key conflict evolution during the compression process.
[0004] CN119046486A discloses a graph compression method based on the relationship between nodes and edges. Although this technology uses the connection relationship between nodes and their neighborhoods to compress the associated structure, its compression is mainly based on the topological relationship between nodes and the query dependency relationship. It does not involve the physical relationship between the actual spatial location of the UAV, its flight status, the predicted nearest interval, and the change in the probability of conflict, nor does it distinguish between stable flight segments and critical conflict flight segments. Furthermore, it does not re-input the candidate compression results into the conflict prediction process after compression to inversely limit the compression scale with conflict recall rate, conflict probability error, and the degree of retention of key events. Therefore, when this method is directly applied to a regional-level low-altitude traffic management server, the problem of losing key predicted conflict relationships after merging structurally similar nodes may still occur, making it difficult to adaptively constrain the compression process based on the conflict prediction error.
[0005] Given the problems of existing multi-UAV spatiotemporal occupancy management technologies and general associated structure compression technologies, such as the disconnect between compression granularity and conflict sensitivity, lack of compression protection for key conflict events, lack of reverse verification of prediction errors in compression results, and difficulty in accurately defining the local impact range after track changes, this invention proposes a spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints. By constructing a track layer, a region layer, and a functional airspace layer and establishing a cross-layer index, the method distinguishes between stable regions and conflict critical regions based on node state similarity, adjacent structure changes, predicted nearest interval, and conflict prediction sensitivity. Different temporal merging windows and spatial aggregation scales are used for different regions. The candidate compression results are then verified by conflict recall rate, conflict probability error, and key event retention rate. When the track or predicted conflict edge changes, the affected local subgraph is limited based on the cross-layer index, thereby resolving the contradiction between the growth of associated structure scale and the retention of key conflict information in the high-density low-altitude UAV rolling conflict prediction process. Summary of the Invention
[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.
[0007] In view of the aforementioned existing problems, the present invention is proposed.
[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a spatiotemporal conflict graph hierarchical adaptive compression representation method oriented to conflict prediction error constraints, comprising: establishing a spatiotemporal conflict graph including a track layer, a region layer and a functional airspace layer based on the UAV track status and airspace structure, and establishing cross-layer indexes between nodes of each layer; Based on the node state similarity, adjacency structure change, predicted nearest interval and conflict prediction sensitivity of the track layer, the stable region and the conflict critical region are determined, and the temporal merging window and spatial aggregation scale are determined respectively to form a candidate compressed map. Conflict prediction is performed on the candidate compressed map, and the compression result is checked based on the conflict recall rate, conflict probability error and key event retention rate. If the constraints are not met, the time merging window or spatial aggregation scale is reduced to form a hierarchical compressed map. Based on newly added tracks, deleted tracks, abnormal maneuvers, or predicted changes in conflict edges, the affected local subgraphs of the hierarchical compressed map are located, the affected local subgraphs are reconstructed, and the cross-layer index is updated.
[0009] In a second aspect, the present invention provides a computer device, comprising: One or more processors; The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints as described above.
[0010] Thirdly, the present invention provides a computer-readable medium for storing software, the software including instructions executable by one or more computers, the instructions causing the one or more computers to perform operations, the operations including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints as described above.
[0011] The beneficial effects of this invention are as follows: By constructing a spatiotemporal conflict map at the track layer, region layer, and functional airspace layer and establishing a cross-layer index, this invention continuously maps the UAV track status, region occupancy relationship, and functional airspace relationship. Based on node state similarity, adjacent structure changes, predicted nearest interval, and conflict prediction sensitivity, it distinguishes between stable regions and conflict critical regions, allowing for differentiated compression granularity across different conflict-sensitive regions, reducing the risk of key conflict evolutions being merged. Furthermore, it uses conflict recall rate, conflict probability error, and key event retention rate to back-verify the compression results, ensuring that the compression scale is constrained by conflict prediction accuracy. When track or conflict relationship changes, only the affected local subgraph is reconstructed, reducing the redundant storage and computational overhead during the rolling prediction process, thus balancing the spatiotemporal conflict map compression efficiency with conflict prediction reliability. Attached Figure Description
[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as shown in this invention. Detailed Implementation
[0013] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0014] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.
[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0016] This embodiment uses a regional-level low-altitude traffic management server in a high-density urban low-altitude UAV operating area as a specific application scenario. The horizontal range of this low-altitude operating area is 5km × 5km, and it includes 300 UAVs, including logistics delivery UAVs, inspection UAVs, and take-off and landing operation UAVs. The airspace structure includes fixed flight corridors, take-off and landing areas, and cruise layers divided according to flight altitude. The regional-level low-altitude traffic management server receives the monitoring tracks and flight plans of each UAV and saves the three-dimensional spatial boundaries of the flight corridors, take-off and landing areas, and cruise layers. The objects processed in this embodiment are the track nodes, node connection relationships, area occupancy relationships, and functional airspace relationships formed within the regional-level low-altitude traffic management server based on the future track status of the UAVs.
[0017] In this embodiment, the start time of each rolling prediction is used as the unified prediction time starting point, and the prediction range is 300s. As a preferred implementation, the prediction range is discretized according to a basic time interval of 1s. Each UAV forms about 300 track nodes within a prediction range, and 300 UAVs correspond to about 90,000 original track nodes. The UAV track status includes three-dimensional position, three-dimensional velocity, and heading. The three-dimensional position adopts the lateral coordinate, longitudinal coordinate, and altitude under a unified spatial coordinate reference, and the three-dimensional velocity adopts three velocity components under the same coordinate reference.
[0018] The regional low-altitude traffic management server is configured with a conflict edge distance threshold corresponding to the current operating airspace. The conflict edge distance threshold is determined based on the UAV safety interval configuration parameters used in the current airspace, rather than being arbitrarily given by this method. As a preferred implementation, the distance judgment boundary for entering the conflict critical processing can be set to 1.5 times the corresponding safety interval, thereby preserving a finer track state before the UAV enters the actual safety interval.
[0019] The regional-level low-altitude traffic management server is also equipped with a conflict prediction model. As a preferred implementation, the conflict prediction model adopts a logistic regression model. Its inputs include the predicted nearest interval, the remaining time to reach the predicted nearest interval, and the relative speed of the UAVs. Its output is the conflict probability of the corresponding UAV pair. The model training samples are from historical UAV operation records. The conflict labels are historical confirmed conflict events, simulation confirmed conflict events, or conflict records determined by the low-altitude traffic management system. The model parameters are determined through the training samples, and the output stability is checked using independent validation samples. The conflict prediction sensitivity calculation, uncompressed spatiotemporal conflict map prediction, and candidate compressed map prediction described later all use the same model structure and the same model parameters, so that the difference in prediction results before and after compression comes from the spatiotemporal conflict map compression relationship, rather than the change in the prediction model.
[0020] Under the above operating conditions, the multi-layered spatiotemporal conflict relationship formed within the regional low-altitude traffic management server is taken as the processing object, and the following processing is carried out in sequence.
[0021] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a hierarchical adaptive compression representation method for spatiotemporal conflict graphs oriented towards conflict prediction error constraints, which specifically includes the following steps: S1. Based on the UAV flight path status and airspace structure, establish a spatiotemporal conflict map including the flight path layer, regional layer and functional airspace layer, and establish cross-layer indexes between nodes of each layer.
[0022] S1.1 Unify the time and spatial coordinate references for monitoring flight paths, flight plans, and airspace structures.
[0023] The current rolling forecast start time is determined as the forecast time starting point, and the monitoring time in the monitoring track and the planned arrival time in the flight plan are converted into the forecast time relative to the forecast time starting point; the longitude, latitude and altitude in the monitoring track, flight plan and airspace structure are converted to the local northeastern rectangular coordinate system established based on the current low-altitude operating area.
[0024] For the UAV track status after the current monitoring time, as a preferred implementation method, segmented motion extrapolation is performed based on the current monitoring position, the current three-dimensional velocity, and adjacent planned waypoints in the flight plan; when the predicted flight segment corresponds to two adjacent waypoints in the flight plan, the motion direction and average velocity of the predicted flight segment are determined based on the positions of the two waypoints and the planned arrival time, and the three-dimensional position at each prediction time is calculated according to a 1-second basic time interval; when there is a deviation between the current monitoring status and the starting point of the flight plan flight segment, the current monitoring position is used as the starting position of the current predicted flight segment, so that the predicted track is continuous with the actual monitoring status.
[0025] S1.2 Establish track nodes and continuous edges for the same aircraft status.
[0026] For each drone, the drone identifier, prediction time, 3D position, 3D velocity, and heading at the same prediction time are combined to form a track node.
[0027] The track nodes of the same UAV are arranged in the order of the predicted time, and the two track nodes corresponding to adjacent predicted times are connected to form a continuous edge of the same state; the continuous edge of the same state records the track node identifiers at both ends and the corresponding prediction time interval; thus forming the track node chain corresponding to each UAV.
[0028] S1.3 Calculate the predicted nearest interval between track nodes and establish the predicted conflict edge for different aircraft.
[0029] In order to ensure that the predicted nearest interval has a clear occurrence time, this embodiment does not calculate a fixed global minimum distance for the 300s prediction range, but sets a conflict determination look-ahead interval for each prediction time.
[0030] As a preferred embodiment, in the first The calculation range is the conflict determination look-ahead time from the predicted time. The conflict determination look-ahead time can be determined based on the current conflict prediction lead of the regional low-altitude traffic management server, for example, 30 seconds. When the remaining prediction range is less than the conflict determination look-ahead time, the remaining prediction range is used as the actual calculation range.
[0031] For drones and drones At any prediction time within the conflict determination forward look-ahead interval The spatial distance between the two is:
[0032] in, Indicates drone With drones At the predicted time Spatial distance; Indicates drone At the predicted time The three-dimensional position; Indicates drone At the predicted time The three-dimensional position.
[0033] Select the minimum spatial distance from the conflict determination forward look-ahead interval corresponding to the current prediction time as the nearest prediction interval for that prediction time:
[0034] in, Indicates the first The nearest prediction interval corresponding to each prediction time; Indicates the first The conflict determination forward interval corresponding to each predicted time.
[0035] when When the distance is not greater than the conflict edge distance threshold, in the first... The drone corresponding to the predicted time track nodes and drones Establish conflict prediction edges between track nodes.
[0036] When there is no inter-machine prediction conflict edge in the previous prediction time but there is an inter-machine prediction conflict edge in the next prediction time, the next prediction time is determined as the time when the inter-machine prediction conflict edge is generated; when there is an inter-machine prediction conflict edge in the previous prediction time but there is no inter-machine prediction conflict edge in the next prediction time, the next prediction time is determined as the time when the inter-machine prediction conflict edge disappears.
[0037] The track layer is formed by track nodes, continuous edges of the same aircraft status, and predicted conflict edges of different aircraft.
[0038] S1.4 Establish regional layers.
[0039] As a preferred implementation, the 5km×5km low-altitude operating area is divided into basic horizontal spatial units with a horizontal dimension of 100m×100m, and basic height units are divided into basic height units with a height interval of 20m. The basic horizontal spatial units and basic height units are combined to form a three-dimensional basic spatial unit, and each three-dimensional basic spatial unit is equipped with a unique area unit identifier.
[0040] The three-dimensional basic space unit where each track node is located is determined based on its three-dimensional position, and the corresponding track node is mapped to the region-occupied node corresponding to that three-dimensional basic space unit.
[0041] The region occupancy node records the region unit identifier, time coverage interval, and track node identifier mapped to the region occupancy node; the region layer is formed by establishing the connection relationship within the region layer based on the spatial adjacency relationship and the traversal relationship of the UAV track node from one three-dimensional basic spatial unit to another three-dimensional basic spatial unit.
[0042] S1.5 Establish functional spatial domain layers and cross-layer indexes.
[0043] Based on the three-dimensional spatial boundaries of airway corridors, take-off and landing areas, and cruise layers in the airspace structure, the three-dimensional basic spatial units corresponding to the nodes occupied by each area are spatially contained or spatially intersected with the three-dimensional spatial boundaries.
[0044] When an area occupancy node is located within the airway corridor space, it is mapped to the corresponding airway corridor node; when it is located within the take-off and landing area space, it is mapped to the corresponding take-off and landing area node; when its altitude range falls within the corresponding cruise layer, it is mapped to the corresponding cruise layer node; airway corridor nodes, take-off and landing area nodes, and cruise layer nodes together constitute the functional airspace layer.
[0045] For a track node, record the correspondence between the track node identifier, the area occupied node identifier, and the functional airspace node identifier to form a cross-layer index; when the same area occupied node intersects with multiple functional airspace ranges at the same time, record the corresponding cross-layer indexes respectively.
[0046] The spatiotemporal conflict graph formed by the above processing is a node-connection data structure inside the regional low-altitude traffic management server, consisting of track nodes, area occupancy nodes, functional airspace nodes, continuous edges of the same aircraft status, predicted conflict edges of different aircraft, intra-layer connection relationships, and cross-layer indexes. Among them, track nodes correspond to the actual movement status of UAVs, and area occupancy nodes and functional airspace nodes correspond to the actual low-altitude space range.
[0047] It should be noted that this step, through a layer-by-layer mapping between the UAV's motion state, spatial occupancy, and functional airspace, ensures that subsequent compression processing maintains the correspondence between the UAV's actual flight state and the actual airspace structure.
[0048] S2. Based on the node state similarity, adjacent structure changes, predicted nearest interval, and conflict prediction sensitivity of the track layer, determine the stable region and the conflict critical region, and determine the time merging window and spatial aggregation scale respectively to form a candidate compressed map.
[0049] S2.1 Form a set of track nodes to be merged.
[0050] Only consecutive track nodes are selected from the track node chain corresponding to the same UAV, and track nodes between different UAVs are not merged.
[0051] Within a candidate time merging window, track nodes of the same UAV arranged consecutively according to the predicted time are added to the set of track nodes to be merged; track nodes that exceed the current candidate time merging window are added to the next set of track nodes to be merged.
[0052] S2.2 Determine the similarity of node states.
[0053] For track nodes to be merged and track nodes The three-dimensional position difference, three-dimensional velocity difference, and heading difference between the two are calculated respectively.
[0054] The three types of differences are transformed to the range of 0 to 1 according to the normalization scale of the corresponding state quantity; the normalization scale can be determined by the range of changes of the corresponding state quantity in the historical normal flight segment, or by the normal motion range of the current aircraft type.
[0055] Node state similarity is determined by the following formula:
[0056] in, Indicates the similarity of node states; Indicates the difference in normalized positions; Indicates the difference in normalization speed; Indicates normalized course differences; , and These represent the weights of positional differences, speed differences, and heading differences, respectively, with the sum of the three being 1.
[0057] S2.3 Determine the similarity of adjacent structures.
[0058] Obtain track nodes respectively and track nodes The connected inter-machine predicted conflict edges are used to form a set of adjacent nodes with the corresponding inter-machine drone identifier. and the set of adjacent nodes .
[0059] The similarity of adjacent structures is determined by the following formula:
[0060] in, This indicates the similarity of adjacent structures.
[0061] when and When all sets are empty, the similarity of the adjacency structure is set to 1.
[0062] Therefore, when the local motion state of a continuous track node changes little but its conflicting object has changed, the similarity of the corresponding adjacent structure decreases.
[0063] S2.4 Determine the continuity of time.
[0064] Obtain track nodes To track node The predicted time interval between two track nodes is obtained by summing the time intervals of the consecutive edges of the same state traversed in the chain of track nodes. .
[0065] when Not greater than the current candidate time merging window The continuity of time is determined by the following formula:
[0066] in, Indicates temporal continuity; This represents the prediction time interval between two track nodes; This indicates the current candidate time merging window.
[0067] when If the time window exceeds the current candidate time merging window, the two will no longer be considered as nodes to be compared within the same set of track nodes to be merged.
[0068] S2.5 Determine the sensitivity of conflict prediction.
[0069] Small positive and negative perturbations are applied to the three-dimensional position, three-dimensional velocity, and heading of the track nodes to be merged. The perturbation amplitude is determined according to the measurement error range of the corresponding state variables by the current monitoring system, so that the perturbation is within the actual observed change range of the UAV's current track state.
[0070] First, the undisturbed flight path state and the corresponding adjacent flight path state of other aircraft are input into the conflict prediction model to obtain the baseline conflict probability. .
[0071] Subsequently, only one perturbation is applied to each state variable at a time, and the conflict probability is recalculated while keeping other state variables unchanged, to obtain the first... Conflict probability under various perturbation conditions .
[0072] The sensitivity to conflict prediction is determined by the following formula:
[0073] in, Indicates sensitivity to conflict prediction; Indicates the baseline conflict probability; Indicates the first The probability of conflict under various perturbation conditions.
[0074] When a small change in flight path status causes a significant change in the probability of conflict, the sensitivity of conflict prediction increases accordingly.
[0075] S2.6 Determine the merging similarity.
[0076] Node state similarity, adjacency structure similarity, and temporal continuity have a positive effect on the tendency of track node merging, while conflict prediction sensitivity has a negative effect on the tendency of track node merging.
[0077] Merging similarity is determined according to the following formula:
[0078] in, Represents track nodes With track nodes Merging similarity; Indicates the weight of node state similarity; Indicates the similarity weight of the adjacency structure; Indicates the weight of time continuity; This represents the sensitivity weight for conflict prediction.
[0079] Each weight and merging threshold is calibrated using historical conflict scenarios. As a preferred implementation, historical operational samples including smooth segments, route intersections, takeoff and landing merging, and near-conflict segments are divided into parameter calibration samples and independent verification samples. Compression is performed under multiple merging thresholds, and the conflict recall rate, conflict probability error, and critical event retention rate are used as verification conditions. Among the candidate merging thresholds that meet the verification conditions, the merging threshold with the smaller overall compression scale is selected.
[0080] S2.7 Determine the stable region and the critical conflict region, and the corresponding compression particle size.
[0081] When the predicted nearest interval corresponding to a track node is not greater than the conflict edge distance threshold, the track node is classified into the conflict critical zone.
[0082] When the predicted nearest interval is greater than the conflict edge distance threshold, and the merging similarity between consecutive track nodes is not lower than the merging threshold, the corresponding consecutive track nodes are assigned to the stable region.
[0083] All other track nodes are classified as being in the conflict critical zone.
[0084] A stable zone consists of track nodes that meet the conditions for a stable zone and corresponding area-occupied nodes; a conflict critical zone consists of conflict critical track nodes and corresponding area-occupied nodes; when the same area-occupied node corresponds to both a stable zone track node and a conflict critical zone track node, it is treated as a conflict critical zone.
[0085] As a preferred implementation, the initial time merging window for the stable zone is 10s, and the initial time merging window for the conflict critical zone is 2s. The stable zone is spatially aggregated according to two adjacent basic horizontal spatial units located in the same basic height unit, while the conflict critical zone maintains a single 100m×100m×20m basic spatial unit.
[0086] S2.8. Form candidate compressed graphs.
[0087] Within the stable zone, consecutive track nodes that are located in the same time window and belong to the same UAV are merged to form compressed track nodes.
[0088] The compressed track node saves the original track node's starting 3D position, ending 3D position, starting 3D velocity, ending 3D velocity, starting heading, ending heading, time coverage interval, and the range of the original track node's identifier, so that the corresponding time interval can be directly processed based on the endpoint status of the compressed track node.
[0089] Within the critical conflict zone, the process is handled using a smaller time merging window. When the time merging window spans the time when the predicted conflict edge of the different aircraft is generated, the time when the predicted conflict edge of the different aircraft disappears, or the time when the predicted closest interval crosses the distance threshold of the conflict edge, the track node merging is not performed across the event time, and the track nodes before and after the event are retained respectively.
[0090] For the regional occupancy nodes in the stable zone, merge the spatially adjacent regional occupancy nodes according to the spatial aggregation scale, and record the spatial range and time coverage interval after aggregation; for the conflict critical zone, maintain the regional occupancy nodes corresponding to the basic spatial unit.
[0091] Update the corresponding nodes and intra-layer connections in the regional layer and functional airspace layer based on the merged or retained track nodes, and maintain the cross-layer indexes between track nodes, regional occupancy nodes and functional airspace nodes to form a candidate compressed map.
[0092] It should be noted that this step reduces repetitive nodes in stable flight segments by using different temporal and spatial processing granularities for stable flight segments and conflict-sensitive flight segments, while maintaining a fine-grained structure of locations where inter-aircraft conflict relationships change.
[0093] S3. Perform conflict prediction on the candidate compressed map, and verify the compression results based on the conflict recall rate, conflict probability error and key event retention rate. If the constraints are not met, reduce the time merging window or spatial aggregation scale to form a hierarchical compressed map.
[0094] S3.1. Generate reference conflict prediction results for an uncompressed spatiotemporal conflict map.
[0095] Input the uncompressed spatiotemporal conflict graph established by S1 into the conflict prediction model.
[0096] For each UAV pair, the predicted nearest interval, the remaining time to reach the predicted nearest interval, and the relative speed are determined within the corresponding conflict determination forward look-ahead interval, and then input into the conflict prediction model to obtain the reference conflict probability.
[0097] As a preferred implementation method, the logistic regression conflict prediction model is expressed as:
[0098] in, Indicates drone With drones The probability of conflict; Indicates the nearest interval of the normalized prediction; This represents the remaining time to reach the nearest predicted interval after normalization. Represents normalized relative velocity; , , and This represents the model parameters determined based on historical conflict samples.
[0099] The reference conflict prediction result is determined by the reference conflict probability, the conflict edge generation event predicted by the different machines, the conflict edge disappearance event predicted by the different machines, and the distance threshold event of the predicted nearest interval crossing the conflict edge.
[0100] S3.2 Calculate the candidate conflict prediction results directly based on the candidate compression map.
[0101] For two compressed track nodes with overlapping time coverage intervals, a piecewise linear relationship between the UAV position and the prediction time is established within the overlapping time range based on the starting track state, ending track state, and time coverage interval saved by each compressed track node.
[0102] Based on the relative position and relative velocity of the two compressed track nodes, the nearest predicted interval and corresponding occurrence time within the overlapping time range are directly determined, without first restoring all 1-second original track nodes within the time coverage interval.
[0103] The conflict prediction model, which is the same type as the uncompressed spatiotemporal conflict map, is formed by inputting the predicted nearest interval, the remaining time to reach the predicted nearest interval, and the relative velocity, and the conflict probability corresponding to the candidate compressed map is obtained.
[0104] Therefore, the candidate compressed graph can directly participate in conflict prediction calculations while maintaining fewer nodes and connections.
[0105] S3.3 Establish the correspondence between key events before and after compression.
[0106] Event identifiers are established for key events in the uncompressed spatiotemporal conflict graph and the candidate compressed graph.
[0107] The event identifier includes at least the drone pair identifier, the event direction, and the predicted time of occurrence, wherein the event direction includes the distance thresholds for entering the conflict edge and leaving the conflict edge.
[0108] When the predicted conflict edge generation event and the entry conflict edge distance threshold event occur in the same UAV pair and are predicted at the same time, they are considered as the same composite critical event. Similarly, when the predicted conflict edge disappearance event and the exit conflict edge distance threshold event occur in the same UAV pair and are predicted at the same time, they are also considered as the same composite critical event, thus avoiding the double counting of the same physical change.
[0109] When the candidate compressed map event and the reference event have the same UAV pair, the same event direction, and the predicted occurrence time deviation does not exceed one basic time interval, the two are determined as matching events; in this embodiment, the basic time interval is 1 second, so a 1 second event time deviation is preferably allowed.
[0110] S3.4 Calculate the conflict recall rate.
[0111] Using the set of reference conflict events formed by the uncompressed spatiotemporal conflict graph as a reference, the number of reference conflict events that are successfully matched in the candidate compressed graph is counted.
[0112] The conflict recall rate is determined by the following formula:
[0113] in, Indicates the conflict recall rate; Indicates the number of reference conflict events; This indicates the number of reference conflict events that were successfully retained in the candidate compressed graph.
[0114] When there are no reference conflict events in the current evaluation scope, the conflict recall rate of the evaluation scope is set to 1, and the conflict probability error and critical event retention rate verification are still performed.
[0115] S3.5 Calculate the collision probability error.
[0116] For the same UAV pair, the maximum conflict probability output by the uncompressed spatiotemporal conflict map and the candidate compressed map within the same prediction range is obtained, and the two are used as the corresponding prediction results.
[0117] The collision probability error is expressed as mean absolute error:
[0118] in, Indicates the probability error of conflict; Indicates the number of drone pairs participating in the comparison; Indicates the first The maximum collision probability of each drone pair obtained from the candidate compressed graph; This represents the maximum collision probability of the same UAV pair obtained from the uncompressed spatiotemporal collision map.
[0119] By using the same UAV to compare the maximum collision probability across the entire prediction range, the candidate compressed map can be evaluated for error without recovering all the base prediction times.
[0120] S3.6 Calculate the critical event retention rate.
[0121] The set of key events is the union of the composite key events formed by the inter-machine predicted conflict edge generation event, the inter-machine predicted conflict edge disappearance event, and the predicted nearest interval crossing conflict edge distance threshold event according to S3.3.
[0122] Critical event retention rate is determined by the following formula:
[0123] in, Indicates critical event retention rate; This indicates the number of reference critical events in the uncompressed spatiotemporal conflict graph; This indicates the number of reference key events that were successfully matched in the candidate compressed graph.
[0124] When there are no reference critical events within the current evaluation scope, the critical event retention rate is set to 1.
[0125] S3.7 Verify the candidate compressed graph and perform compression granularity rollback.
[0126] As a preferred implementation, the lower limit of the conflict recall rate is set at 0.99, the upper limit of the conflict probability error is set at 0.02, and the critical event retention rate requirement is 1.
[0127] The time merging window can be selected from the candidate set consisting of 10s, 5s, 2s and 1s; the spatial aggregation scale can be selected from the candidate set consisting of horizontally adjacent 2×2 basic spatial units, horizontally adjacent 2×1 basic spatial units and single basic spatial units.
[0128] Multiple candidate compressed maps are generated corresponding to combinations of time merging windows and spatial aggregation scales.
[0129] A candidate compressed graph is added to the qualified candidate set when it meets the following conditions: the conflict recall rate is not lower than the lower limit of recall, the conflict probability error is not higher than the upper limit of error, and the critical event retention rate is 1.
[0130] If the current candidate compressed map does not meet any constraint, reduce the time merging window or spatial aggregation scale in descending order and re-perform the verification.
[0131] When the time merging window has been reduced to a basic time interval of 1 second, and the spatial aggregation scale has been reduced to a single basic spatial unit but still cannot meet the constraints, track node merging and regional occupancy node aggregation are not performed on the corresponding range, and the corresponding structure in the uncompressed spatiotemporal conflict map is directly retained.
[0132] This sets a defined termination boundary for the compression rollback process.
[0133] S3.8. Generate a hierarchical compressed graph and compressed representation results from the qualified candidate set.
[0134] Calculate the overall size for each qualified candidate compressed graph:
[0135] in, Indicates the overall size of the candidate compressed graph; This indicates the number of nodes in the candidate compressed graph; Indicates the number of connected edges in the candidate compressed graph; This represents the edge size weight.
[0136] The edge size weight is determined based on the average number of bytes stored after node objects and connected edge objects in the regional low-altitude traffic management server are serialized using the same serialization method. Specifically, it is the ratio of the average number of bytes stored for a single connected edge to the average number of bytes stored for a single node, so that the overall size corresponds to the actual storage size inside the server.
[0137] Select the candidate compressed graph with the smallest overall size from the qualified candidate set as the hierarchical compressed graph.
[0138] A hierarchical compression code is generated based on the track layer, region layer, and functional airspace layer of the hierarchical compression map, and a cross-layer mapping table, time coverage interval, and upper bound of collision probability error associated with the hierarchical compression code are also generated.
[0139] The upper bound of the conflict probability error represents the maximum allowable conflict probability error used in the S3 check; the conflict probability error calculated from the actual candidate compressed map still adopts... This indicates that the two have different meanings.
[0140] Compression ratio is determined by the following formula:
[0141] in, Indicates the compression ratio; This indicates the number of nodes in the spatiotemporal conflict graph before compression; This indicates the number of connected edges in the spatiotemporal conflict graph before compression.
[0142] The final compressed representation includes hierarchical compression coding, cross-layer mapping tables, time coverage intervals, upper bound of collision probability error, and compression ratio.
[0143] As an example, under the conditions of a 5km×5km low-altitude operating area, 300 UAVs, and a 300s prediction range, the original track layer contains approximately 90,000 track nodes. When a 10s time merging window is used in the stable region and a 2s time merging window is used in the conflict critical region, the example node compression rate is 82%, the conflict edge recall rate is 99.1%, and the mean absolute error of the conflict probability is 0.018. The above values are used as example data to illustrate the relationship processing of this method and are not intended to limit the actual project operation performance.
[0144] Preferably, this step incorporates both the spatiotemporal conflict map structure size and the compressed conflict prediction error into the screening criteria, ensuring that the reduction of nodes and connections does not come at the cost of losing key conflict relationships; and progressively reduces the compression granularity of candidate compression results that do not meet the conflict prediction constraints, so that the final hierarchical compressed map retains the track states that have a significant impact on the actual UAV conflict relationships.
[0145] S4. Based on the changes in newly added tracks, deleted tracks, abnormal maneuvers, or predicted conflict edges, locate the affected local subgraphs of the hierarchical compressed map, reconstruct the affected local subgraphs, and update the cross-layer index.
[0146] S4.1 Determine the initial affected track nodes.
[0147] The regional low-altitude traffic management server acquires new surveillance tracks in each rolling forecast cycle. As a preferred implementation, the rolling forecast cycle can be consistent with a 1-second base time interval.
[0148] When a drone identifier that did not exist in the previous rolling cycle appears in the current rolling cycle, the newly generated track node corresponding to that drone is identified as the initial affected track node corresponding to the newly added track.
[0149] When a drone that existed in the previous rolling cycle exits the current prediction area or its corresponding flight plan is canceled in the current rolling cycle, the hierarchical compression map of the previous rolling cycle is used as the positioning reference to obtain the original track node of the drone and the track nodes of other drones that have a conflict edge with it in the prediction. These track nodes are determined as the initial affected track nodes corresponding to the deleted track.
[0150] For drones that still exist in the current cycle, compare the current monitored 3D position, 3D velocity, and heading with the predicted state formed in the previous rolling cycle for the current moment.
[0151] As a preferred implementation, the position prediction residual, velocity prediction residual, and heading prediction residual are calculated using historical normal operating tracks under the same prediction lead condition. The position abnormal maneuver threshold, velocity abnormal maneuver threshold, and heading abnormal maneuver threshold are determined based on independent verification samples. When any state residual exceeds the corresponding abnormal maneuver threshold, the track node of the UAV in the current and subsequent affected time range is determined as the initial affected track node corresponding to the abnormal maneuver.
[0152] The set of predicted conflict edges for the same UAV pair in the current rolling cycle and the previous rolling cycle is compared. When a predicted conflict edge for a different UAV is added or deleted, the track nodes at both ends of the predicted conflict edge are determined as the initial affected track nodes corresponding to the change in the predicted conflict edge.
[0153] S4.2 Extend the range of affected track nodes along the connection relationship.
[0154] Establish a set of visited track nodes and add the initially affected track nodes to the set to be expanded.
[0155] Select track nodes sequentially from the set to be expanded, add adjacent track nodes of the same UAV that are within the predicted state change time range along the continuous edge of the same UAV state, and add other UAV track nodes that have conflict relationships with the track node along the predicted conflict edge of different UAVs.
[0156] Whenever a new track node is added, it is marked as visited; track nodes that already exist in the set of visited track nodes are not added to the set to be expanded again.
[0157] For newly added track nodes, compare the UAV track status and the set of predicted conflict edges between different UAVs in the current rolling period with the previous rolling period.
[0158] When the difference in the drone track status of a boundary track node does not exceed the corresponding abnormal maneuver judgment threshold, and the set of predicted conflict edges of the different drones of the boundary track node has not been added or deleted compared with the previous rolling cycle, the expansion will no longer start from the boundary track node.
[0159] Stop expanding the track layer range when the set to be expanded is empty.
[0160] S4.3. Form affected local subgraphs based on cross-level indexes.
[0161] The cross-layer index retrieves all affected track nodes and their corresponding area occupancy nodes, and further retrieves the route corridor nodes, take-off and landing area nodes, or cruise layer nodes corresponding to the area occupancy nodes.
[0162] The affected track nodes, affected area occupied nodes, affected functional airspace nodes, as well as the continuous edges of the same aircraft status, the conflict edges of different aircraft predictions, the intra-layer connection relationships, and the cross-layer indexes between them are collectively determined as the affected local subgraph.
[0163] Track nodes, area-occupied nodes, functional airspace nodes, and their internal connections that have not entered the affected area will remain in the state of the previous rolling cycle's hierarchical compression map.
[0164] S4.4 Reconstruct the affected local subgraph.
[0165] Based on the current rolling cycle of monitoring tracks and flight plans, reconstruct the track nodes and continuous edges of the same aircraft status within the affected range, and recalculate the predicted nearest interval and predicted conflict edges between different aircraft according to S1.3.
[0166] Based on node state similarity, adjacency structure similarity, temporal continuity, and conflict prediction sensitivity, the stable zone and conflict critical zone within the affected area are re-determined, and track node merging and regional occupancy node aggregation are re-executed according to the corresponding time merging window and spatial aggregation scale.
[0167] The reconstructed affected local candidate compressed map is used to perform conflict prediction according to the same conflict prediction model as the uncompressed local structure, and the conflict recall, conflict probability error and critical event retention rate are recalculated.
[0168] If the corresponding constraints are not met, reduce the time merging window or spatial aggregation scale within the affected range; if the constraints are still not met after reaching the basic time interval and basic spatial unit, retain the uncompressed structure of that local range.
[0169] S4.5 Connect the affected local subgraphs after reconstruction and update the cross-layer index.
[0170] For track nodes within the affected local subgraph after reconstruction, the recalculated continuous edges of the same aircraft state and the predicted conflict edges of different aircraft are used.
[0171] For the boundary track nodes of the affected local subgraph, the same-aircraft state continuity relationship and the different-aircraft prediction conflict relationship between them and the unaffected track nodes are recalculated, and the reconstructed affected local subgraph and the unaffected part are connected by the recalculated boundary connecting edge.
[0172] The nodes and connections within the unaffected parts remain unchanged.
[0173] Subsequently, the track layer, region layer, and functional airspace layer are updated sequentially, and cross-layer indexes involving the affected range are re-established based on the updated three-layer node relationships; cross-layer indexes related to unaffected nodes and whose mapping relationships have not changed are kept in their original state.
[0174] The corresponding local range in the original hierarchical compressed map is replaced by the reconstructed affected local submap to form the updated hierarchical compressed map for the current rolling prediction period.
[0175] As example data for the technical solution, in the scenario of 300 drones mentioned above, the local subgraph rebuilt in each rolling update cycle accounts for approximately 7% of the complete spatiotemporal conflict graph. The values are only used as examples to illustrate the local incremental update relationship.
[0176] As an alternative implementation, the regional low-altitude traffic management server can also re-perform global conflict prediction error verification on the complete layered compressed map according to the calibration interval; the calibration interval is determined based on the server's current processing load and the density of UAV operations.
[0177] It should be noted that this step, based on the track layer connection relationship and the cross-layer index between the three layers of nodes, limits the recalculation of new tracks, deleted tracks, abnormal maneuvers, and changes in predicted conflict edges to the range of actually affected nodes, while maintaining the remaining nodes and connection relationships. This reduces the server computation overhead caused by the repeated creation and verification of the complete three-layer spatiotemporal conflict map during the rolling prediction process.
[0178] In summary, this embodiment first establishes a track layer based on the actual spatial motion state of the UAV, and then forms a regional layer, a functional airspace layer, and a cross-layer index by using the regional occupancy location and functional airspace boundaries. Based on this, it distinguishes between stable regions and critical conflict regions according to the continuity of track states, conflict adjacency relationships, and the impact of small track state changes on conflict probability, and employs differentiated time merging windows and spatial aggregation scales. Furthermore, it uses an uncompressed spatiotemporal conflict map as a conflict prediction reference, and verifies the conflict recall rate, conflict probability error, and key event retention rate of the candidate compressed map. When the compression result fails to meet the prediction constraints, the compression granularity is gradually reduced. When track additions, track deletions, abnormal maneuvers, or changes in predicted conflict edges occur during the rolling operation, the affected local subgraphs are defined along the track layer connection relationships and cross-layer indexes, and local reconstruction is completed. Thus, while preserving the critical conflict track states and key conflict relationships, the scale of track nodes and connection relationships continuously stored and repeatedly processed by the regional-level low-altitude traffic management server is reduced.
[0179] Other aspects disclosed in the embodiments of the present invention also provide a computer device including one or more processors and a memory.
[0180] The memory is used to store operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints as described in the foregoing embodiments, especially... Figure 1 The flowchart of the method is shown.
[0181] Other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium for storing software including instructions executable by one or more computers, which, upon execution, cause the one or more computers to perform operations including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints of the foregoing embodiments, particularly... Figure 1 The flowchart of the method is shown.
[0182] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.
[0183] The method can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program in the computer program, wherein the storage medium is configured such that the computer operates in a specific and predefined manner.
[0184] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if necessary, the program can be implemented in assembly or machine language.
[0185] In any case, the language can be either compiled or interpreted.
[0186] Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit.
[0187] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.
[0188] Furthermore, the method can be implemented in any suitable computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices.
[0189] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.
[0190] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.
[0191] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.
[0192] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A hierarchical adaptive compression representation method for spatiotemporal conflict graphs oriented towards conflict prediction error constraints, characterized in that, include: Based on the UAV flight path status and airspace structure, a spatiotemporal conflict map including the flight path layer, the area layer, and the functional airspace layer is established, and cross-layer indexes between nodes of each layer are established. Based on the node state similarity, adjacency structure change, predicted nearest interval and conflict prediction sensitivity of the track layer, the stable region and the conflict critical region are determined, and the temporal merging window and spatial aggregation scale are determined respectively to form a candidate compressed map. Conflict prediction is performed on the candidate compressed map, and the compression result is checked based on the conflict recall rate, conflict probability error and key event retention rate. If the constraints are not met, the time merging window or spatial aggregation scale is reduced to form a hierarchical compressed map. Based on newly added tracks, deleted tracks, abnormal maneuvers, or predicted changes in conflict edges, the affected local subgraphs of the hierarchical compressed map are located, the affected local subgraphs are reconstructed, and the cross-layer index is updated.
2. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as described in claim 1, characterized in that, The establishment of the spatiotemporal conflict graph includes: Unify surveillance tracks, flight plans, and airspace structure to the same time and space coordinate reference; Based on the monitoring track and flight plan, track nodes are established, and track nodes of the same UAV at adjacent monitoring times are connected as continuous edges of the same UAV status. When the predicted nearest interval between track nodes of different UAVs is not greater than the conflict edge distance threshold, the corresponding track nodes are connected as inter-UAV predicted conflict edges to form the track layer. Based on the spatial grid or airspace partition, the track nodes are mapped to area occupancy nodes to form the area layer, and the area occupancy nodes are mapped to the route corridor, take-off and landing area and cruise layer to form the functional airspace layer. The cross-layer index between the track nodes, area occupancy nodes and functional airspace nodes is recorded.
3. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as described in claim 2, characterized in that, The determination of the stable region and the critical conflict region includes: The node state similarity is determined based on the differences in the UAV track states of the track nodes to be merged, the adjacency structure similarity is determined based on the differences in the set of associated edges of the track nodes to be merged, and the temporal continuity is determined based on the time interval of the continuous edges of the same state. The conflict prediction sensitivity is determined based on the change in conflict prediction probability before and after the disturbance of the track node state, and the merging similarity is determined based on the node state similarity, adjacency structure similarity, temporal continuity and conflict prediction sensitivity. Track nodes whose predicted nearest interval is not greater than the conflict edge distance threshold are assigned to the conflict critical region; consecutive track nodes whose predicted nearest interval is greater than the conflict edge distance threshold and whose merging similarity is not lower than the merging threshold are assigned to the stable region; and the remaining track nodes are assigned to the conflict critical region.
4. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as described in claim 3, characterized in that, The formation of the candidate compressed map includes: Within the stable zone, consecutive track nodes of the same UAV are merged according to the time merging window, and the area occupancy nodes corresponding to the merged track nodes are merged according to the spatial aggregation scale. Within the conflict critical zone, the time merging window is reduced. When the time merging window crosses the time of generation of the predicted conflict edge of the different aircraft, the time of disappearance of the predicted conflict edge of the different aircraft, or the time when the predicted closest interval crosses the distance threshold of the conflict edge, the corresponding track node is retained. The corresponding nodes and intra-layer connections in the regional layer and functional airspace layer are updated based on the merged or retained track nodes, and the candidate compressed map is formed according to the cross-layer index.
5. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as described in claim 3, characterized in that, The determination of the merged similarity includes: The node state similarity, adjacency structure similarity, temporal continuity, and conflict prediction sensitivity are determined respectively. The node state similarity, adjacency structure similarity, and temporal continuity are positively weighted according to their respective weights, and the conflict prediction sensitivity is negatively weighted according to their respective weights. The weighted results are combined to form the merged similarity. The merge similarity is compared with the merge threshold determined by historical conflict scenarios, and the continuous track nodes with a merge similarity not lower than the merge threshold are determined as track nodes to be merged.
6. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints according to claim 1, characterized in that, The formation of the layered compressed map includes: The candidate compressed maps are formed by combining multiple time merging windows and spatial aggregation scales, and each candidate compressed map is input into the conflict prediction model in the low-altitude traffic management server. Based on the conflict prediction results of the uncompressed spatiotemporal conflict graph, the conflict recall rate and conflict probability error of each candidate compressed graph are calculated, and the key event retention rate is determined according to the number of events generated by conflict edges predicted by different machines, events disappeared by conflict edges predicted by different machines, and events that cross the nearest interval conflict edge distance threshold. Candidate compressed graphs are selected based on a conflict recall rate not lower than the lower limit of recall, a conflict probability error not higher than the upper limit of error, and a critical event retention rate of 1. The candidate compressed graph with the smallest sum of the number of nodes and the number of edges calculated by edge size weight is selected as the hierarchical compressed graph. If no candidate compressed graph meets the conditions, the time merging window or spatial aggregation scale is reduced and the graph is re-verified.
7. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints according to claim 6, characterized in that, The compressed representation of the hierarchical compressed graph includes: A hierarchical compression code is generated based on the nodes and intra-layer connections in the track layer, region layer, and functional airspace layer, and a cross-layer mapping table, time coverage interval, and upper bound of collision probability error associated with the hierarchical compression code are generated. The original comprehensive scale is determined by the sum of the number of nodes in the spatiotemporal conflict graph before compression and the number of edges calculated according to the edge size weight. The compressed comprehensive scale is determined by the sum of the number of nodes in the layered compressed graph and the number of edges calculated according to the same edge size weight. The compression rate is determined by the ratio of the difference between the original comprehensive scale and the compressed comprehensive scale to the original comprehensive scale. The resulting compressed representation includes the layered compression coding, cross-layer mapping table, time coverage interval, upper bound of collision probability error, and compression ratio.
8. The spatiotemporal conflict graph hierarchical adaptive compression representation method oriented towards conflict prediction error constraints as described in claim 2, characterized in that, The incremental reconstruction of the affected local subgraph includes: The affected track nodes are determined based on changes in the predicted conflict edges caused by new track additions, deleted tracks, abnormal maneuvers, or different aircraft. The affected area is expanded along the continuous edge of the same aircraft status and the predicted conflict edge of different aircraft connected to the affected track node, and the corresponding area occupancy node and functional airspace node are added according to the cross-layer index to determine the affected local subgraph; The affected local subgraph is re-executed with track node merging, inter-layer mapping, and conflict prediction error verification. The reconstructed affected local subgraph is then combined with the unaffected portion of the hierarchical compressed map, and the cross-layer index is updated.
9. A computer device, characterized in that, include: One or more processors; The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform an operation, the operation including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints as described in any one of claims 1 to 8.
10. A computer-readable medium for storing software, characterized in that: The software includes instructions executable by one or more computers, which cause the one or more computers to perform operations, including the flow of the spatiotemporal conflict graph hierarchical adaptive compression representation method for conflict prediction error constraints as described in any one of claims 1 to 8.
Citation Information
Patent Citations
Method for carrying out graph compression based on relation between nodes and edges
CN119046486A
Multi-unmanned aerial vehicle dynamic path planning system and method
CN120803042A
