Low-altitude air route corridor octree multi-dimensional modeling method

Abstract

The application discloses a kind of low-altitude air route corridor octree multidimensional modeling methods, belong to low-altitude traffic technical field. For the problem that existing low-altitude modeling is only with geometry occupation driving space discrete, cannot unifiedly fuse the timeliness and the reliable attribute of multi-source heterogeneous environment, the application obtains multi-source environment data and unifies space-time and attribute format benchmark, constructs octree space model and mapping generation includes five-dimensional environment vector of geometry occupation, dynamic barrier, meteorological disturbance, communication and navigation risk;With node environment change intensity and credibility double-driven node splitting, merging and state updating, when data anomaly, generate conservative degenerative occupation result, output low-altitude multidimensional environment model.The application significantly improves the delicacy, timeliness and reliability of low-altitude corridor modeling, is applicable to low-altitude path planning, traffic capacity assessment and the like scene.

Description

Technical Field

[0001] This invention belongs to the field of low-altitude transportation technology, specifically relating to a method for three-dimensional environmental modeling of low-altitude airway corridors. Background Technology

[0002] With the development of low-altitude logistics, urban patrol, emergency transport, and manned low-altitude transportation, low-altitude airway corridors are gradually shifting from discrete pilot airspaces to continuous, regularized, and computable operational spaces. For aircraft, an airway corridor is not merely a geometric passage, but a constrained space simultaneously influenced by various environmental factors such as buildings, poles, trees, temporary facilities, moving obstacles, wind fields, precipitation, communication coverage, and navigation quality. Common spatial representation methods in engineering include regular voxels, hierarchical grids, and octree structures. Among these, octrees, due to their ability to locally refine three-dimensional space through recursive subdivision, have become an important candidate structure for representing low-altitude environments.

[0003] However, traditional octree modeling mostly revolves around the single objective of geometric occupancy, with node states typically only recording whether they are occupied or not, or accompanied by simple category labels. In practical applications, this type of modeling usually assumes that node attributes are approximately static during the update cycle and that different data sources projected onto the same node can be treated with equal semantics. When these assumptions are not met, the simultaneous superposition of geometric boundaries, dynamic disturbances, and timeliness differences within a node can lead to semantic aliasing in the same state description, making it difficult for subsequent route planning or traffic assessment to distinguish between actual occupancy changes and the increased uncertainty caused by expired data. From the perspective of data organization and computer implementation, the difficulty in modeling low-altitude airway corridors lies not in a single three-dimensional coordinate mapping, but in the joint alignment of multi-source heterogeneous data across spatial, temporal, and reliability scales. Building outlines are typically derived from oblique photogrammetry, BIM, or GIS vector surfaces, offering high spatial accuracy but with update cycles measured in days or weeks. Temporary construction facilities and mobile obstacles may originate from visual perception, radar, or ground-based reporting, with update frequencies reaching seconds or even sub-seconds. Low-altitude meteorological fields and communication / navigation quality fields are usually represented by raster interpolation or regional statistics, exhibiting continuous spatial distribution but suffering from observation noise and sampling sparsity. If fixed-resolution voxels or static threshold-driven splitting are still used, two types of problems easily arise: one is that in rapidly changing regions, local gradients from different sources within a node are averaged, leading to blurred boundaries; the other is that in stable regions, excessively fine resolution results in numerous redundant nodes, increasing storage and update costs. Therefore, how to unify spatial occupancy representation, environmental change representation, and node timeliness reliability representation into a single octree node, and how to drive resolution adjustment by the intensity of environmental change rather than simple geometric density, becomes a key technical issue in low-altitude airway corridor environmental modeling.

[0004] To address specific problems in the processing, existing technologies have explored various approaches. Based on their technical approaches, they can be mainly divided into two categories: one focuses on hierarchical grid segmentation and path planning in low-altitude space, and the other focuses on three-dimensional track search under multi-scale grid encoding.

[0005] The first category is hierarchical grid segmentation and path planning. For example, Chinese patent CN202610071432.4 discloses a path planning and traffic capacity assessment system for low-altitude operations in communities. This system analyzes the traffic capacity of low-altitude operations in communities through airspace facility modeling, facility relationship determination, hierarchical airspace segmentation, dynamic obstacle marking, and multi-objective path planning. This type of technology typically first completes a structured representation of spatial facilities and obstacle areas, and then performs path search and bottleneck identification on this representation. Its spatial representation is more focused on obstacle marking and safe distance calculation for planning purposes. The decision timescale is usually within the path planning cycle, focusing on which areas are passable and which areas should be avoided, without maintaining the update time, reliability, and intensity of changes within nodes as a unified node state.

[0006] The second category is 3D trajectory search under multi-scale grid encoding. For example, Chinese patent CN202310860845.7 discloses a 3D trajectory planning method and system based on a DQG grid-based improved JPS algorithm. It starts with the underlying spatial representation, uses DQG-3D to construct a gridded airspace, segments it according to the complexity of obstacles, and completes trajectory planning using multi-scale grids and different step lengths. The core of this technology lies in improving search efficiency through multi-scale discrete grids, and its effective boundary is mainly reflected in the discrete representation of the airspace at the planning time. Even with multi-scale grids, it essentially still serves neighborhood expansion and jump point selection in trajectory search, without solving the problem of unified binding of multi-dimensional environmental attributes in octree nodes, nor explicitly establishing a node time-related decay and model reliability update mechanism.

[0007] While the aforementioned existing technologies have made some progress in path planning and multi-scale spatial discretization, they still have the following shortcomings in the multi-dimensional environment modeling scenario of low-altitude airway corridors.

[0008] First, the spatial discretization criteria are relatively singular, with node semantics primarily remaining at the geometric occupancy level. The first type of method mentioned above mainly focuses on facility relationships, safety distances, and obstacle markings, while the second type primarily uses obstacle complexity to drive grid scale switching. A common problem with both is that the dominant variable for resolution adjustment remains geometric structure or planning needs, rather than the intensity of changes in the multidimensional environmental state within a node. This leads to weather disturbances, communication fluctuations, mobile obstacle crossings, and temporary facility changes within the same spatial node being compressed into a single "accessible" or "inaccessible" semantic, making it difficult to accurately represent asynchronous environmental changes from different sources.

[0009] Secondly, multi-source data is typically only juxtaposed and mapped, lacking a unified state evolution mechanism for timeliness and reliability. The differences in low-altitude environmental data lie not only in the coordinate systems but also in their sampling frequencies, observation noise, and expiration rates. Most existing technologies assume that mapped data is valid within the current planning period, but fail to establish a continuous evolution chain between update time, validity period, and reliability decay for nodes. As a result, once node attributes are written, they are easily used by subsequent systems as equally reliable facts. This makes it impossible to distinguish between the continued existence of real obstacles and data that has not been updated but is still being used, and it also fails to prevent expired information from continuing to participate in subsequent airspace analysis in a highly reliable state.

[0010] Secondly, node splitting and merging lack a closed-loop driving mechanism oriented towards dynamic environments. Traditional hierarchical grid or multi-scale grid schemes typically complete scale division during the initialization or planning phase, and subsequent updates often employ relabeling or local recalculation methods, lacking a closed-loop mechanism that considers the combined effects of the intensity of node environmental changes, data timeliness decay, and reliability thresholds. In other words, existing schemes are more about modeling before use, rather than updating and reconstructing simultaneously. This leads to dynamically changing areas not being continuously densified, and stable areas not being merged in a timely manner, resulting in a long-term mismatch between spatial representation and the true complexity of the environment, thus affecting update efficiency and the fundamental accuracy of subsequent path planning, risk assessment, and corridor operation management.

[0011] Therefore, there is an urgent need in this field for a low-altitude airway corridor modeling method that can uniformly bind multi-source environmental states, sources of change, update time, validity period and model credibility to octree nodes, and drive node splitting, merging and state updates by the intensity of environmental change and the timeliness credibility of nodes. Summary of the Invention

[0012] The purpose of this invention is to provide a multi-dimensional octree modeling method for low-altitude airway corridors, which solves the problems in existing low-altitude airspace modeling techniques that only use geometric occupancy or obstacle complexity as the basis for spatial discretization and cannot uniformly bind heterogeneous information such as buildings and temporary obstacles, low-altitude meteorology, and communication and navigation quality to the same spatial node; at the same time, it solves the problems in existing schemes that lack node time-related decay and reliable update chains, and that it is difficult to distinguish between real environmental changes and the uncertainty growth of observation expiration; furthermore, it solves the problem that most existing hierarchical grids and multi-scale grids adopt static partitioning or passive recalculation, and it is difficult to actively refine or reclaim nodes according to local environmental changes.

[0013] To achieve the above objectives, this invention provides a multi-dimensional octree modeling method for low-altitude airway corridors, executed by a computer device. The computer device includes a processor, memory, a communication interface, and a data acquisition interface for accessing external data sources; the processor reads program instructions from the memory and executes the following steps. This method uses the intensity of node environmental changes and node reliability as dual driving forces for octree node evolution, controlling node splitting, merging, and state updates. Based on this, a unified mapping mechanism for multi-source environmental records and an expired degradation and conservative occupancy mechanism are employed to form an online-updable multi-dimensional environmental model of low-altitude airway corridors.

[0014] Step S1, Multi-source Environmental Data Acquisition and Parameter Dictionary Establishment: Acquire multi-source environmental data within the low-altitude airway corridor and write source type identifier, observation time, initial source confidence value, and failure duration for each data record. The multi-source environmental data includes at least building outlines, tower centerlines and outer dimensions, tree crown width and height, temporary construction facility boundaries, moving obstacle trajectory points, low-altitude meteorological grids, communication coverage intensity, navigation and positioning quality, and generation timestamps for each data source. By establishing a unified parameter dictionary, records from different sources have consistent data description fields before entering the subsequent node mapping and confidence evolution process.

[0015] Step S2, unified preprocessing of spatial and temporal attributes: The multi-source environmental data is unified in terms of spatial reference, temporal reference, and attribute format. The spatial data is converted to a unified three-dimensional coordinate frame. Records within the same update cycle are time-aligned according to the latest priority rule, and continuous numerical attributes are normalized into comparable risk quantities. Integer encoding can be used for discrete category data; for continuous numerical data, it is converted to a unified comparison scale according to its respective engineering value boundaries. For slow data that has not been updated within the update cycle, the previous version record is retained, but its impact is reduced in subsequent confidence calculations according to the time decay rule; records that have exceeded the failure time are no longer included in the regular risk fusion, but are only used as a reference for the conservative degradation process.

[0016] Step S3, Octree Spatial Model Construction and Node Field Initialization: Construct an octree spatial model of the target region, generate the root node and hierarchical index structure, and initialize fields such as spatial bounding box, hierarchy, occupancy status, multidimensional environment vector, change source vector, last update time, effective deadline, confidence level, and change intensity for each node. The multidimensional environment vector includes at least five dimensions: geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor, and navigation quality risk factor, used to uniformly express the multi-source environmental state within the node.

[0017] Step S4, Multi-source Environmental Data Mapping and Node Environmental Vector Generation: The multi-source environmental data is mapped to corresponding octree nodes, forming a five-dimensional environmental vector within each node that includes at least geometric occupancy factors, dynamic obstacle factors, meteorological disturbance factors, communication quality risk factors, and navigation quality risk factors. The node's most recent update time and effective deadline are then written into this vector. For buildings, poles, trees, and temporary facilities, the geometric occupancy contribution is determined by the spatial intersection between the entity's bounding body and the node's bounding box. For moving obstacle trajectories, the dynamic obstacle contribution is determined by the proportion of trajectory points relative to the node and the speed's influence range. For low-altitude meteorological data, communication quality data, and navigation quality data, aggregation is performed based on the node's internal sampling results. Subsequently, contributions of the same type are weighted and fused according to the current credibility weight of the corresponding data source to generate the node's five-dimensional environmental vector. Further, based on the proportion of the current change contribution of each environmental factor to the total change, a change source vector for the node is generated to mark which type of information primarily triggers the node's change.

[0018] Step S5, Calculation of Node Environmental Change Intensity: Based on the difference between the five-dimensional environmental vectors of the current period and the previous period, and the node's time-related disturbances, the intensity of node environmental change is calculated. Specifically, the five-dimensional environmental vectors of the current period and the previous period are read, and the changes in the geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor, and navigation quality risk factor are calculated respectively. Then, based on the current system time, the node's most recent update time, and the node's effective deadline, a time-related disturbance term characterizing the degree of uncertainty growth due to the node's lack of updates is calculated. Finally, the changes in each environmental factor and the time-related disturbance term are weighted accordingly, summed, and subjected to amplitude limiting to obtain the node's environmental change intensity. This allows for a mathematical distinction between the actual environmental transitions of a node and the accumulation of uncertainty caused solely by the long-term lack of data updates.

[0019] Step S6, Node Trustworthiness Calculation and Splitting, Merging, and State Update Control: Node trustworthiness is calculated based on the prior trustworthiness of the data source mapped to the node, the most recent action time, the failure duration, and the spatial contribution weight. The splitting, merging, and state updates of the octree nodes are controlled according to the intensity of the node's environmental changes and the node trustworthiness. Specifically, the set of data sources mapped to the node is determined. Each data source is weighted according to its spatial contribution weight within the node. Based on the initial source trustworthiness value of each data source and the time difference between the current system time and the most recent action time of that data source, exponential decay is performed according to the failure duration. The weakened trustworthiness contributions of each data source are normalized and fused to obtain the node trustworthiness. When a node is not mapped to a valid data source or the sum of the spatial contribution weights of each data source is lower than a preset lower limit, the node trustworthiness is set to a low trustworthiness state. Subsequently, when the intensity of node environment change is not lower than a preset splitting threshold and the node level is lower than a preset maximum level, the current node is split into child nodes; when the intensity of node environment change is not higher than a preset merging threshold, the node credibility is not lower than a preset credibility threshold, the variance of sibling node occupancy risk is not greater than a preset variance threshold, and the number of instantiated sibling nodes reaches a preset lower limit, the sibling nodes are merged into the parent node; when the node credibility is lower than a preset credibility threshold, the node structure is maintained and conservative degradation occupancy update is implemented; in other cases, the current level is maintained and the node environment vector, change source, occupancy risk value, and occupancy status are updated.

[0020] Step S7, Node Timeliness Trust Record Writing and Abnormal Degradation Processing: When a node has no valid data source, its trustworthiness is lower than a preset trust threshold, or the node has timeliness anomalies, a conservative degradation occupancy result is generated based on the parent node's occupancy risk, the average risk of the same-layer neighborhood, and historical occupancy status. Specifically, for each valid data source within the node, candidate cutoff times are determined based on its most recent action time and failure duration. Each candidate cutoff time is combined with its corresponding spatial contribution weight to form a time weight sequence, and sorted in ascending order of candidate cutoff times. When the cumulative weight ratio reaches a preset ratio, the corresponding minimum candidate cutoff time is taken as the node's valid cutoff time. When a node has no valid data source, the node's valid cutoff time is set to the node's most recent update time, and a timeliness anomaly marker is written. Subsequently, for abnormal nodes, the occupancy risk of its parent node, the occupancy risk of neighboring nodes at the same level as the abnormal node, and the historical occupancy risk of the abnormal node are obtained and fused according to a preset degradation weight to obtain a conservative degradation occupancy result. The same-layer neighborhood is the set of instantiated same-layer nodes that are adjacent to the abnormal node in three-dimensional space. Furthermore, when the proportion of abnormal nodes or the model credibility meets the preset abnormal conditions, the model is rolled back or rebuilt based on the local subtree where the abnormal node is located, according to the historical stable version.

[0021] Step S8, Model Organization and Output Publication: The output is a low-altitude multidimensional environment model containing node status information and hierarchical relationships. The node status information includes at least occupancy status, occupancy risk value, source of change, last update time, effective deadline, and credibility. The model can be organized into a hierarchical node library structure with parent-child relationship indexes, and model snapshots can be published to downstream path planning, conflict detection, accessibility assessment, or digital twin systems through a communication interface.

[0022] Furthermore, the process of mapping multi-source environmental data to corresponding octree nodes and forming a five-dimensional environmental vector within each node includes: performing spatial intersection calculations of entity bounding volumes and node bounding boxes on buildings, towers, trees, and temporary facilities to obtain geometric occupancy contributions; extracting trajectory point sequences from moving obstacle trajectories according to time windows and obtaining dynamic obstacle contributions based on the dwell ratio and speed influence range of trajectory points relative to nodes; aggregating low-altitude meteorological data, communication quality data, and navigation quality data according to the sampling results within nodes to obtain meteorological disturbance contributions, communication quality risk contributions, and navigation quality risk contributions, respectively; and weighting and fusing similar contributions according to the current credibility weights of the corresponding data sources to generate geometric occupancy factors, dynamic obstacle factors, meteorological disturbance factors, communication quality risk factors, and navigation quality risk factors for nodes.

[0023] Furthermore, the aggregation of low-altitude meteorological data, communication quality data, and navigation quality data based on the sampling results within each node includes: when the node side length is greater than a preset multiple of the original sampling step size of the corresponding continuous field, the average value of the sampling points within the node is used as the aggregation result of the continuous field for the node; when the node side length is not greater than the preset multiple of the original sampling step size of the corresponding continuous field, the maximum risk value of the sampling points within the node is used as the aggregation result of the continuous field for the node. By using average aggregation on coarse-scale nodes and peak-preserving aggregation on fine-scale nodes, continuous field noise can be suppressed while avoiding the average masking of local high-risk peaks.

[0024] Furthermore, when calculating the intensity of environmental changes at a node, in addition to the weighted changes of each environmental factor, a time-dependent disturbance term representing the degree of uncertainty growth in the node's unrefreshed state is introduced. A change source vector for the node is generated based on the proportion of the weighted changes of each environmental factor in the total change. This allows the node refinement conditions to directly respond to the combined state of real environmental changes and time-dependent distortions, rather than merely responding to static obstacle contours.

[0025] Furthermore, when writing the effective deadline of a node, the candidate deadlines and spatial contribution weights of the effective data sources within the node are sorted, and the minimum candidate deadline is selected as the effective deadline of the node when the cumulative weight ratio reaches a preset ratio. This avoids a single low-weight, short-term noise source from directly dominating the node's validity period, while maintaining the recalculation of the node's freshness boundary.

[0026] Furthermore, when generating the conservative degenerate occupancy result, the abnormal node is not directly set to full occupancy or cleared from occupancy. Instead, the occupancy risk of the parent node, the average risk of the same-layer neighborhood, and the historical occupancy risk are integrated to ensure that the abnormal node maintains spatial continuity and risk continuity even when the sensing source is interrupted, the record expires, or the credibility decreases.

[0027] Furthermore, the generation of the occupancy risk value includes: applying corresponding fusion weights to the node's geometric occupancy factor, dynamic obstacle factor, weather disturbance factor, communication quality risk factor, navigation quality risk factor, and low-confidence penalty term, and then summing them to obtain the node occupancy risk value; then comparing the node occupancy risk value with a preset occupancy threshold; if the node occupancy risk value is not lower than the preset occupancy threshold, the node is marked as occupied; if the node occupancy risk value is lower than the preset occupancy threshold, the node is marked as traversable. The low-confidence penalty term is determined by the inverse vector of node confidence and is used to explicitly project timeliness uncertainty into the occupancy risk.

[0028] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for multidimensional modeling of low-altitude airway corridors using an octree.

[0029] Correspondingly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described low-altitude airway corridor octree multidimensional modeling method.

[0030] Compared with the prior art, the present invention has the following beneficial effects.

[0031] Firstly, this invention unifies the binding of geometric occupancy factors, dynamic obstacle factors, meteorological disturbance factors, communication quality risk factors, and navigation quality risk factors to the same octree node, and further writes the source of change, update time, effective deadline, and credibility, so that the node semantics are no longer limited to a single geometric occupancy layer. In this way, downstream path planning, conflict detection, or risk assessment systems can directly read the multi-dimensional environmental information in the same node without having to repeatedly associate it between multiple heterogeneous data layers, thus enabling the node to directly carry the environmental state related to operational decisions.

[0032] Secondly, this invention drives node splitting and merging by the intensity of changes in the node environment, enabling spatial resolution adjustment to directly respond to multi-dimensional state transitions in the environment, rather than merely responding to obstacle geometric complexity or pre-defined planning needs. Regions experiencing drastic changes are continuously encrypted due to geometrical variations, dynamic obstacle traversal, enhanced weather disturbances, communication and navigation fluctuations, or increased timeliness distortion; stable regions reclaim resolution when the intensity of change is low, the credibility condition is met, and the risk of sibling nodes occupying the same area is consistent. This allows for finer boundary representation in highly disturbed regions and avoids uniform refinement across the entire domain in stable regions.

[0033] Third, this invention maintains the most recent update time, valid expiration time, and credibility at the node level, and incorporates the timeliness disturbances caused by non-refreshing into the change intensity calculation. Simultaneously, it triggers a conservative degradation occupancy mechanism when credibility is insufficient. This allows for a computational distinction between two scenarios: a truly unchanged environment and expired records. It prevents expired data from continuing to participate in spatial domain modeling with a high credibility state, and also avoids directly generating state blanks when sensing link anomalies. Through joint degradation of parent nodes, neighboring nodes at the same level, and historical states, the model maintains the continuity of the risk field during anomaly periods.

[0034] Fourth, this invention links the judgment of change intensity, credibility decay, splitting and merging rules, occupancy risk fusion, abnormal degradation mechanism, and local rollback or reconstruction under abnormal conditions into a single node update chain, transforming the octree from a static structure that is modeled first and then used into an online evolutionary structure that is updated and reconstructed simultaneously. This closed-loop update method enables dynamic regions to be continuously encrypted and stable regions to be merged in a timely manner, and makes the model output simultaneously possess spatial hierarchy, risk status, and timeliness credibility attributes, making it suitable for direct use by the low-altitude corridor operation and management system. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments are briefly described below. The drawings are only used to illustrate the core process of the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0036] Figure 1 This is a flowchart of the overall method in an embodiment of the present invention; Figure 2 This is a flowchart of the data acquisition and unified preprocessing logic unit in an embodiment of the present invention; Figure 3 This is a flowchart of the octree construction and node environment mapping logic unit in an embodiment of the present invention; Figure 4 This is a flowchart of the node change intensity calculation and splitting / merging control logic unit in an embodiment of the present invention; Figure 5This is a flowchart of the logic unit for generating time-sensitive and reliable records, handling anomaly degradation, and publishing models in an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Embodiment 1 is an embodiment of the overall method flow, while Embodiments 2 and 3 are embodiments of urban logistics scenarios and general aviation inspection scenarios, respectively.

[0038] Example 1: Overall Flow of an Octree Multidimensional Modeling Method for Low-Altitude Airway Corridors

[0039] like Figures 1 to 5 As shown in this embodiment, a multi-dimensional octree modeling method for low-altitude airway corridors includes eight collaborative steps: multi-source environmental data acquisition and parameter dictionary establishment, unified preprocessing of spatial and temporal attributes, octree spatial model construction, node environment mapping, node change intensity calculation, node splitting and merging and state updating, generation of reliable time-sensitive records and anomaly degradation, and model organization and release. This method is executed by a computer device, which includes a processor, memory, communication interface, and data acquisition interface.

[0040] Step S1: Multi-source environmental data acquisition and parameter dictionary establishment.

[0041] S101, the processor receives multi-source environmental data of the low-altitude airway corridor through a data acquisition interface. The data sources include at least building outlines, tower centerlines and outer dimensions, tree canopies and heights, temporary construction facility boundaries, moving obstacle trajectory points, low-altitude meteorological grids, communication coverage strength, navigation and positioning quality, and the generation timestamps of each data source. In this embodiment, the bounding box dimension of the airway corridor to be modeled is... The width dimension is The dimension in the height direction is The root node level of an octree is denoted as . The maximum subdivision level is The update cycle is The bounding box dimensions are determined based on the actual management scope of the target airway corridor. The maximum subdivision level is obtained offline based on spatial resolution requirements and the processor's computing power limit. The update cycle is determined based on real-time requirements.

[0042] S102. The processor writes a unified description field to each data record, including source type identifier, observation time, initial source reliability value, and failure duration. The source type identifier is used to distinguish geometric entities, trajectory data, and continuous field data; the observation time is used to establish a node update time chain; the initial source reliability value is used to characterize the prior reliability of the data source when it enters the system; and the failure duration is used to limit the duration for which this type of data can be directly used in engineering. In this embodiment, the failure duration of building vector data can be taken as... Construction facilities can be taken Moving obstacles are acceptable Low-altitude weather is favorable The communication and navigation quality is acceptable. These preset durations can be obtained through historical update frequency statistics and offline calibration of the service-side tolerance latency.

[0043] Step S2: Unified preprocessing of spatial and temporal attributes.

[0044] S201. The processor transforms all spatial data into a unified three-dimensional coordinate frame. In this embodiment, a local East-North-Sky coordinate system with the corridor origin as the origin is adopted. The geographic coordinates of buildings, towers, trees, and construction facilities are projected into this coordinate system; the trajectories of moving obstacles are spatially interpolated based on time-aligned timestamps; low-altitude meteorology and communication / navigation quality are mapped to the unified spatial frame based on the original sampling resolution. For continuous numerical attributes, the processor obtains the normalized risk quantity at a unified comparison scale by subtracting the engineering-defined lower limit from the original observation value and then dividing it by the sum of the difference between the upper and lower limits and the smoothing term. The corresponding basic normalization relationship is: , in, Represents continuous features after normalization; Represents the original continuous observations; and These represent the minimum and maximum values ​​of the feature under the engineering definition, respectively; This represents a smoothing term, used to prevent the denominator from being zero. In this embodiment, Desirable Its size is set offline based on the stability of floating-point operations. When a certain continuous feature satisfies When the value is zero, the processor immediately sets the normalization result to zero and writes an indistinguishable constant flag to prevent numerical anomalies.

[0045] S202. The processor performs time alignment based on the current system time. For newly arrived records within the current update cycle time window, the record with the latest timestamp is prioritized; when timestamps are the same, the record with the higher initial source confidence value is prioritized; for slow data that has not been refreshed in the current cycle, the previous version node record is retained, but its subsequent confidence decreases according to the time decay rule; when the missing time exceeds the corresponding failure duration, the record only participates in subsequent conservative degradation and no longer participates in regular risk fusion. After this processing, data from different sources are mapped to a unified processing scale, and subsequent node fusion can be performed within the same time base and numerical space.

[0046] Step S3: Octree spatial model construction and node field initialization.

[0047] S301, the processor generates a root node based on the corridor bounding box size and recursively builds an octree index structure. Each node contains at least the following fields: spatial bounding box, level, occupancy status, multidimensional environment vector, change source vector, last update time, valid deadline, confidence level, and change intensity. In this embodiment, the first... The edge lengths of the layer nodes in the three directions are the corresponding edge lengths of the bounding box divided by 1 / 2. For example, when When, the node side lengths can be respectively , and This scale allows for a fine representation of building edges, tower cladding, and construction boom areas, while avoiding a level lower than the reliable spatial resolution of most input sources.

[0048] S302, the processor defines a five-dimensional environment vector for each node. This environment vector represents, in sequence, the geometric occupancy factor, dynamic obstacle factor, weather disturbance factor, communication quality risk factor, and navigation quality risk factor. The corresponding five-dimensional environment vector form is: , in, Indicates the first A multidimensional environment vector for each node; Indicates the geometric occupancy factor; Indicates dynamic obstacle factor; Indicates meteorological disturbance factors; Indicates communication quality risk factors; Indicates navigation quality risk factor; superscript The transpose indicates that the vector is stored as a column vector. All five components take values ​​between zero and one, used to represent environmental information from different sources under a unified risk scale. If subsequent engineering projects require expansion to dimensions such as noise, electromagnetic interference, or no-fly zones, the environmental vector can be expanded without changing the dual-drive update mechanism; however, this embodiment is based on five dimensions.

[0049] Step S4: Multi-source environmental data mapping and node environment vector generation.

[0050] S401. The processor performs spatial intersection calculations between the entity bounding volume and the node bounding box for buildings, towers, trees, and construction facilities, obtaining the geometric occupancy contribution. This contribution is determined by the proportion of the entity-node overlap volume to the node volume, specifically: , in, Indicates data source For nodes Geometric occupancy contribution; This represents a volume operator used to calculate the volume of a space volume; Indicates source The corresponding spatial volume; Represents a node bounding box; symbol Indicates the space intersection operation; This indicates a smoothing term. When nodes are at the maximum level and their spatial dimensions are close to the lower limit of sensor resolution, if the intersection volume calculation is unstable, the bounding box overlap rate can be used to approximate the fine volume calculation. This process ensures that geometric entities can be projected as occupancy contributions in a uniform manner across nodes at different levels.

[0051] S402, the processor performs node aggregation for moving obstacles, low-altitude weather, communication quality, and navigation quality. For moving obstacle trajectories, it first extracts the trajectory point sequence within a preset time window, and then obtains the dynamic obstacle contribution based on the dwell ratio of trajectory points relative to nodes and the velocity expansion influence range. First, the velocity expansion influence radius is defined for each trajectory point: , in, Indicates trajectory source The The radius of influence of the velocity expansion of each trajectory point; This indicates the maximum outward radius, used to limit the unlimited expansion of the influence range under high-speed conditions; Indicates the basic safety radius; The proportionality coefficient representing the velocity's expansion into space, with its unit being seconds; Indicates the first The velocity of each trajectory point, in meters per second; function This indicates taking the smaller value. In this embodiment, Desirable , Desirable , Desirable These values ​​can be obtained through offline calibration using a mobile obstacle safety buffer test.

[0052] Subsequently, the processor determines whether each trajectory point falls within the node's bounding box or its velocity expansion influence range, and calculates the dwell time ratio accordingly. The dwell time ratio is then weighted and fused with the velocity influence normalization factor to obtain the node's dynamic obstacle contribution to the trajectory source. If there are no valid trajectory points within the time window, the dynamic contribution of the trajectory source to the node is directly recorded as zero. This reflects both the dwell time of moving obstacles within the node and the amplification effect of high-speed passage on node disturbance.

[0053] For continuous field data, the processor selects the aggregation method based on the relationship between the node side length and the original sampling step size. When the node side length is greater than a preset multiple of the original sampling step size of the corresponding continuous field, the average value of the sampling points within the node is used as the aggregation result for that continuous field. When the node side length is not greater than the preset multiple, the maximum risk value of the sampling points within the node is used as the aggregation result. In this embodiment, the preset multiple can be taken as... This value can be determined offline based on the sparsity of continuous field sampling and the requirement for peak preservation. This allows noise to be suppressed at coarse-scale nodes while preserving local peak values ​​at fine-scale nodes.

[0054] In one optional implementation, for meteorological disturbance factors, wind speed, vertical gusts, and rainfall intensity can be selected as basic raw quantities, and then normalized according to the engineering boundary before weighted fusion. For communication quality risk factors, received signal strength indication and signal-to-noise ratio can be selected as basic raw quantities, with higher risk for weaker signals and lower signal-to-noise ratios. For navigation quality risk factors, positioning error estimate and satellite geometric accuracy factor can be selected as basic raw quantities, with risk increasing synchronously as positioning error and geometric accuracy deteriorate. When a node is created for the first time and a certain component is missing, the corresponding component can be temporarily recorded as zero, and the contribution of the corresponding source can be reduced in the confidence calculation; when subsequent periods are missing, the value of the previous period is preferentially used to maintain field continuity.

[0055] S403. The processor weights and fuses contributions of the same type according to the current trusted weights of the corresponding data sources to generate a five-dimensional environment vector for the node. For example, the node's geometric occupancy factor can be obtained by normalizing and weighting the contributions of the data sources participating in the geometric occupancy fusion according to the current trusted weights; the fusion method for dynamic obstacles, meteorology, communication, and navigation factors is the same, only the corresponding components are replaced. When there is no valid data source of the same type acting on the node in the current period, the processor does not perform updates for that type of component, but retains the value of the previous period and applies time decay through the trusted module. When the total weight used for normalization is too small, below the preset lower limit, the value of the previous period is directly used to avoid amplifying noise with extremely small weights.

[0056] After generating the five-dimensional environment vector, the processor further constructs a change source vector to indicate whether node changes are primarily triggered by geometric occupancy factors, dynamic obstacle factors, meteorological disturbance factors, communication quality risk factors, or navigation quality risk factors. When the changes in each dimension are all zero and the time-related disturbances are also zero, the change source vector is set to a zero vector. This vector can be directly used by downstream systems to explain the sources of risk changes in subsequent outputs.

[0057] Step S5: Calculate the intensity of changes in the node environment.

[0058] S501, the processor reads the current cycle node's environment vector and the previous cycle node's environment vector, and calculates the changes in five dimensions. Simultaneously, it reads the node's most recent update time and effective deadline for subsequent calculation of the timeliness perturbation term. In this embodiment, all changes are calculated within the normalized space to ensure that variables with different dimensions can be compared using the same change intensity formula.

[0059] S502, the processor constructs the node environment change intensity based on multi-dimensional environmental changes and time-dependent disturbances. The corresponding change intensity calculation relationship is as follows: , in, Represents a node The intensity of environmental change; , , , , and These represent the weighting coefficients for changes in geometric occupancy, dynamic obstacles, weather disturbances, communication quality, navigation quality, and timeliness disturbances, respectively. , , , and These represent the absolute changes of the five corresponding environmental factors between the current period and the previous period; Represents the node time-related disturbance term; outer function This is used to limit the intensity of change to within one. In this embodiment, the six weighting coefficients can be respectively taken as... , , , , and Its value is obtained through offline grid search calibration in a digital twin environment, and satisfies that the sum of the six is ​​one.

[0060] The aforementioned time-sensitivity disturbance term is used to characterize the increased uncertainty of a node due to prolonged periods without being refreshed, and its calculation relationship is as follows: , in, Represents a node The time-dependent disturbance term; Indicates the current system time; Represents a node The time of the most recent update triggered by actual observation; Represents a node Current valid deadline; Represents the smoothing term; inner function To prevent exceptions caused by negative time differences, the outer function... Used to limit the result to one. When When an expired record is found to be abnormal or missing, the processor immediately sets the expiration disturbance term to one and marks the node as having an expiration anomaly. This design allows nodes to exhibit an increase in change intensity due to data expiration even when they are stationary on the surface of geometric and environmental observations, thus preventing expired nodes from being misclassified as stable nodes.

[0061] Step S6: Node credibility calculation and split, merge and state update control.

[0062] S601, the processor determines the set of data sources mapped to the nodes and calculates the node credibility based on spatial contribution weight, initial source credibility value, most recent action time, and failure duration. The corresponding credibility calculation relationship is as follows: , in, Represents a node Current credibility; Indicates mapping to nodes The complete collection of data sources; Indicates source For nodes The spatial contribution weight can be determined by the overlap volume fraction, trajectory dwell ratio, or sampling coverage. Indicates source Prior credibility; Represents the natural exponential function; Indicates the current system time; Indicates source Most recent action on the node timestamp; Indicates source The duration of failure; Indicates the smoothing term. When When empty, the processor directly instructs... The value is zero; when the failure duration of a data source is incorrectly configured to be zero, a minimum validity period can be used instead. In this embodiment, the minimum validity period can be taken as zero. Its value is set according to the numerical stability requirements; when all weights are lower than the preset lower limit, the node's credibility drops to a low credibility state and enters conservative processing.

[0063] In one alternative implementation, the prior confidence level of the building GIS data can be taken as... Construction reports can be submitted. Visual detection of moving obstacles is feasible Low-altitude meteorological interpolation fields can be taken Communication quality test data is available. The navigation quality estimate is acceptable. The above values ​​can be obtained based on historical error statistics and offline calibration of long-term sensor source deviations.

[0064] S602, the processor executes node evolution rules based on the intensity of node environment changes and node trustworthiness. When the intensity of node environment changes is not lower than a preset splitting threshold and the current level is lower than a preset maximum level, the node is split into eight child nodes; when the intensity of node environment changes is not higher than a preset merging threshold, the node trustworthiness is not lower than a preset trustworthiness threshold, the variance of the occupancy risk of sibling nodes is not greater than a preset variance threshold, and the number of instantiated sibling nodes reaches a preset lower limit, the sibling nodes are merged into the parent node; when the node trustworthiness is lower than a preset trustworthiness threshold, the node is not deleted, but the node structure is maintained and conservative occupancy updates are implemented; in other cases, the current level is maintained, and only node attributes are updated. In this embodiment, the splitting threshold can be taken as... The merging threshold can be taken as follows: The confidence threshold can be taken as follows: The threshold for the variance of risk occupied by sibling nodes can be taken as follows: These thresholds can be obtained through joint optimization of boundary representation error, invalid subdivision node ratio, expired node misuse rate, and single-cycle update time in offline digital twin scenarios. To avoid triggering a merge before nodes have formed a valid comparison basis, the condition that the number of instantiated sibling nodes reaches a preset lower limit is used to ensure that the merge decision is based on a sufficient foundation of the same group of nodes.

[0065] S603, the processor generates a node occupancy risk value based on the five-dimensional environment vector and the low-confidence penalty term. Specifically, the node occupancy risk value is obtained by summing the corresponding fusion weights applied to the node's geometric occupancy factor, dynamic obstacle factor, weather disturbance factor, communication quality risk factor, navigation quality risk factor, and confidence inverse vector. In this embodiment, the fusion weights for geometric occupancy, dynamic obstacle, weather disturbance, communication risk, navigation risk, and low-confidence penalty can be taken as follows: , , , , and Its value is obtained through offline search and calibration using indicators such as collision proxy risk, false alarm rate, and risk field continuity. If the result exceeds the range of zero to one, the processor truncates it within the legal range.

[0066] S604. The processor compares the node occupancy risk value with a preset occupancy threshold to generate a node occupancy status. When the occupancy risk value is not lower than the threshold, the node is marked as occupied or high-risk and untraversable; when the occupancy risk value is lower than the threshold, the node is marked as traversable or low-risk and awaiting passage. In this embodiment, the occupancy threshold can be set to... This value can be selected offline based on the principle of balancing the false alarm rate and the missed detection rate. After this processing, the node retains both the continuous risk value and the binary occupancy result, making it convenient for different downstream systems to call upon it as needed.

[0067] Step S7: Generation of node timeliness trust records and handling of abnormal degradation.

[0068] S701, the processor writes the most recent update time, effective deadline, credibility, and change source vector to each node. The effective deadline is not simply the minimum of all data source deadlines, but is determined through a cumulative process of sorting and assigning spatial contribution weights to the deadlines of each effective data source. This suppresses the excessive dominance of a single low-weight, short-lived noise source on the overall validity period of the node. First, a candidate deadline is generated for each effective data source within the node, its value equal to the sum of the source's most recent action time and its failure duration. Then, each candidate deadline and its corresponding spatial contribution weight form a time-weighted sequence, which is sorted from smallest to largest deadline. Next, the cumulative weight ratio of the sorted sequence is calculated. When the cumulative weight ratio reaches a preset ratio, the corresponding smallest candidate deadline is taken as the node's effective deadline. In this embodiment, the preset ratio can be taken as... Its value is obtained through offline calibration of node freshness stability and short-term noise suppression effect. When a node does not have a valid data source, the processor sets the valid deadline to the most recent update time and writes a timeliness exception flag.

[0069] S702. The processor executes an anomaly degradation mechanism. When a node meets any of the following conditions: no valid data source, trust level below a preset trust threshold, or timeliness anomaly marked as true, the node is switched to conservative occupancy mode. Conservative occupancy mode is not simply full occupancy; instead, it generates a degradation value based on the parent node's occupancy risk, the average risk of the same-layer neighborhood, and historical occupancy status. The corresponding relationships are as follows: , in, Represents a node Degraded occupancy risk value under conservative occupancy mode; , and These represent the degradation fusion weights for parent node risk, average neighborhood risk, and historical occupancy risk, respectively. Represents a node The risk value occupied by the parent node; Represents a node The average occupancy risk value of the same-layer neighborhood; Represents a node The occupancy risk value of the previous period; This indicates the time of the previous update cycle. In this embodiment, the three degradation fusion weights can be respectively taken as... , and Its value is obtained by offline calibration through weighted optimization of the risk jump magnitude and the proportion of high-risk missed packets in the anomaly perception packet loss scenario.

[0070] The same-layer neighborhood is explicitly defined as the set of instantiated nodes that are at the same level as the current node and are adjacent in three-dimensional space, and can be determined according to the three-dimensional twenty-six-neighborhood. When a node has no parent node, i.e. the root node is abnormal, the corridor's global average occupancy risk can be used instead of the parent node's risk; when the neighborhood is empty, the occupancy risk of the node in the previous cycle is used instead of the neighborhood's average risk; when historical values ​​are missing, the parent node's value is used instead. In this way, even if individual data sources temporarily fail, the node state can still maintain continuity under the joint constraints of spatial and temporal contexts.

[0071] S703. The processor performs partial rollback or reconstruction under abnormal conditions. When the proportion of abnormal nodes exceeds a preset upper limit, the average model confidence level falls below a preset lower limit, or abnormal states occur for multiple consecutive periods, the processor locates the local subtree containing the abnormal nodes from historical stable version model snapshots and performs model rollback or reconstruction on that local subtree. Rollback is suitable for situations where the historical stable structure can still be directly inherited, while reconstruction is suitable for situations where the local environment has changed significantly during the abnormal period and multi-source data needs to be remapped. By processing only the local subtree containing the abnormal nodes without clearing or fully reconstructing the global model, both model stability and online update efficiency can be balanced.

[0072] Step S8: Model organization and output release.

[0073] S801: The processor organizes all nodes into a queryable model library. Each node outputs at least the node code, spatial boundary, hierarchy, occupancy status, occupancy risk value, environment vector, source of change, last update time, effective deadline, credibility, and parent-child relationship index. This output structure preserves both the octree hierarchy and the environment and time-sensitive credibility attributes that can be directly used for decision-making.

[0074] S802, the processor publishes model snapshots to downstream path planning, conflict detection, capacity assessment, or digital twin systems via a communication interface. The output format can be a binary tree structure file, a key-value database with hierarchical indexes, or a memory-resident query interface. If the average model confidence level falls below a preset lower limit or the proportion of abnormal nodes exceeds a preset upper limit for several consecutive periods, the system automatically rolls back to the previous stable version model snapshot and reconstructs the data for the abnormal periods separately. In this embodiment, consecutive... The average model reliability within a given period is lower than Or the proportion of abnormal nodes is higher than A rollback can be triggered at any time; stable version model snapshots can be defined as continuous. The average model reliability within each period is no less than The proportion of abnormal nodes is not higher than Furthermore, the update time should be the most recent model version that meets real-time requirements. These thresholds can be configured offline based on historical operational stability requirements.

[0075] In this embodiment, offline parameter calibration and online updating adopt an engineering process of online updating after parameter calibration, without relying on neural network training. In the offline phase, the processor scans change weights, risk fusion weights, degradation fusion weights, and various thresholds in the digital twin environment; in the online phase, the processor... Steps S1 to S8 are executed repeatedly at a fixed period. Exemplary deployment hardware may employ a clock speed of at least [missing information]. 16-core general-purpose CPU and no less than The memory, without dedicated GPU acceleration, can be stably maintained at no more than [a certain value]. There are 1 active node; when the regular parallel computing library is enabled, the single-cycle update time can meet the update cycle requirements.

[0076] Example 2: Online Multidimensional Octree Modeling for Urban Logistics Route Corridors

[0077] This embodiment focuses on a low-altitude airway corridor for the routine operation of urban logistics drones. The corridor is [length missing]. Width is High coverage to The area contains high-rise residential buildings, communication towers, roadside trees, temporary tower crane construction areas, and intermittent vehicle hoisting operations, and is also affected by valley winds and localized communication shadows. Traditional fixed-resolution 3D voxel modeling, if using global subdivision to cover the details of the construction area, would lead to a surge in the number of nodes; if a coarser resolution is used, the edges of the construction area, the tops of the towers, and the tree canopy disturbance areas would be averaged, resulting in distortion of local spatial representation.

[0078] In this embodiment, the processor collects data from four types of interfaces: one GIS building model, one database of poles and trees, one airborne sensing feedback trajectory stream of moving obstacles, and one stream each of low-altitude meteorological and communication grid data. GIS static data is updated daily, and construction reports are submitted daily. Update once, moving obstacle perception every Update once, the weather grid every Update once, communication quality grid per Update once; the system's unified modeling update cycle remains the same. The processor projects all data onto a local East-North-Sky coordinate system and performs normalization and time alignment according to the rules described in Example 1. In this scenario, the processor establishes the root node bounding box as... The maximum subdivision level is . No. The edge lengths of the layer nodes are respectively , and It can provide a detailed representation of building edges, tower cladding, and construction boom areas.

[0079] The following describes a representative node update process within a specific processing cycle. Assume the current system time is... Select the first A node in a layer has a spatial boundary of... , , The environment vector for this node in the previous cycle was... The first to fifth dimensions correspond to the geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor, and navigation quality risk factor, respectively; the reliability of this node in the previous period is... The risk value occupied in the previous cycle was After the current cycle mapping, the processor obtains the volume fraction of the intersection between the building's cantilevered edges and nodes as follows: Temporary hoisting operations expand their influence and form dynamic factors. The normalized meteorological disturbance in the gust area is Communication quality risk is Navigation quality risk is Therefore, the current period node environment vector is updated as follows: .

[0080] The most recent real update time of this node is The effective deadline is Based on the time-related disturbance relationship described in Example 1, it can be seen that the node has a non-zero time-related disturbance, reflecting that although it is still within its validity period, it has begun to accumulate unrefreshed uncertainty. Furthermore, the geometric occupancy change is... The dynamic obstacle change is The change in meteorological disturbance is The change in communication quality risk is The change in navigation quality risk is After calculation based on the offline-calibrated combination of change weights, the environmental change intensity of this node is: .

[0081] This node simultaneously maps three main data sources: building GIS source, construction reporting source, and visual perception source of mobile obstacles, with their spatial contribution weights being respectively... , and The prior confidence levels are respectively , and The most recent action times are respectively , and The corresponding failure durations are as follows: , and After calculation according to the credibility evolution rule, the current credibility of this node is: Then, according to the occupancy risk fusion rule, the five-dimensional environment vector and the low-confidence penalty term are jointly projected into the node occupancy risk value, from which the node occupancy risk value can be obtained. Since the intensity of the change is below the splitting threshold, and the sibling node group to which the current node belongs does not simultaneously meet the merging conditions, the node remains at its current level in this cycle, neither splitting nor merging. Furthermore, since its occupancy risk value is below the occupancy threshold, the output state is "travelable but dynamic disturbances need to be noted." This result, along with the update time, effective deadline, confidence level, and source of change, is written back to the node record and enters the next cycle as a historical state.

[0082] In this embodiment, an urban low-altitude digital twin environment is used for exemplary simulation. The simulation object is... The simulation duration for the logistics route corridor is... Under normal operating conditions, this includes stable building clusters, trees, and regular communication coverage; under impact conditions... to Introducing a temporary construction area, in to Introducing mobile hoisting operations, in to Superimposed gust bands. The comparison scheme one uses a fixed-resolution voxel model, with voxel sizes uniformly set. Scheme 2 is a static octree model containing only geometric occupancy. Evaluation metrics include boundary representation error of high-variability regions, effective node attribute completeness rate, total number of nodes, and single-cycle update time. Three gradient scenarios (low perturbation, medium perturbation, and high perturbation) are set up, with each scenario repeated. Wheel, together Wheel. Statistical results show that under low-disturbance operating conditions, the boundary expression error of the proposed solution is: Under medium disturbance conditions Under high disturbance conditions .all After round-summarizing, the completeness rate of effective node attributes in the present invention is [percentage missing]. The total number of nodes is The time consumed for a single cycle update is Compared to the fixed-resolution voxel model, the boundary representation error under high-perturbation conditions is reduced by... Down to The effective node attribute completeness rate is determined by... Upgraded to This indicates that the newly added nodes are mainly concentrated in areas of real change, rather than spreading indiscriminately throughout the corridor. Subsequently, based on the above calculations and processing results, the processor continues to execute the subsequent steps as described in Example 1, completing operations such as node state maintenance, model organization, and version release. The specific execution logic is consistent with Example 1 and will not be repeated here.

[0083] Example 3: Timeliness Reliability Modeling and Anomaly Degradation for Navigation Inspection Corridors

[0084] This embodiment focuses on a navigation inspection corridor on the outskirts of a city. The corridor's length is also... However, the environment mainly consists of power poles, forest belts, and open factory areas, and the complexity of static obstacles is lower than in Example 2. The main problem lies in the instability of some sensing sources: there is packet loss in the mobile obstacle sensing link, low-altitude meteorological stations are sparse, and the communication coverage grid update delay is significant. Even if the traditional static octree can save existing obstacle records, it cannot determine whether these records are continuously valid real states or gradually lose credibility due to long-term lack of updates.

[0085] In this embodiment, the system hardware and parameter architecture are consistent with those in Embodiment 1, but the source failure duration is slightly adjusted for the specific scenario: the failure duration of the forest belt update data is set to... Perception of moving obstacles is still The weather grid is Both communication quality and navigation quality are The processor still operates as follows. Execute steps S1 to S8 periodically.

[0086] Assume the current time is Select the first Layer nodes, whose spatial boundaries are , , The risk value of this node in the previous cycle was... Credibility is In the current cycle, the obstacle perception link is interrupted, the communication grid is not refreshed, and the weather grid retains only old values. The processor reads the current environment vector as follows: The most recent actual update time for this node is... The effective deadline is After calculation according to the timeliness disturbance rules, the timeliness disturbance of this node has reached the upper limit, indicating that the high-timeliness data has exceeded the timeliness limit as a whole.

[0087] The only valid sources for node credibility calculation in the current period are the forest belt database and the old meteorological grid, with spatial weights of respectively. and The prior confidence levels are respectively and The most recent action times are respectively and The corresponding failure durations are as follows: and After calculation according to the credibility evolution rule, the current credibility of this node has decreased to [value missing]. Since this value is below the preset trust threshold, the processor no longer directly uses the normal occupancy risk output, but instead triggers a conservative occupancy degradation mechanism. Let the occupancy risk of the parent node of this node be... The average risk of the same-level neighborhood is Historical occupancy risk is The conservative degradation occupancy result after combining degradation fusion weights is: Since the result, while below the occupancy threshold, is close to it, the system marks the node as a low-confidence area requiring verification, rather than directly classifying it as fully passable. If obstacle perception recovers in the next cycle, the processor recalculates the actual occupancy risk and removes the degradation mode.

[0088] In this embodiment, an exemplary simulation is performed using an edge navigation inspection digital twin environment, with a simulation duration of [duration missing]. Under normal operating conditions, the moving obstacle sensing link is stable; under impact conditions... to Random introduction The perceived packet loss, in to Introducing a weather update delay, in to Overlay communication grid refresh is slowed down. The control scheme is a static octree model without a time-sensitive reliability mechanism. Evaluation metrics include expired node misuse rate, high-risk missed label ratio, abnormal cycle occupancy risk jump magnitude, and single-cycle update time. Three gradient operating conditions—low time-sensitive distortion, medium time-sensitive distortion, and high time-sensitive distortion—are set up, with each group repeated. Wheel, together Wheel. Statistical results show that under low time-distortion conditions, the misuse rate of expired nodes in the proposed solution is [missing information]. Under medium-age distortion conditions Under high-time-distortion conditions .all After rounds of summarization, the high-risk omission rate of the proposed solution is: The abnormal cycle occupancy risk jump range is The time consumed for a single cycle update is Compared to the static octree model without a time-sensitive reliability mechanism, the misuse rate of expired nodes is reduced from... Down to The proportion of high-risk missed bids was... Down to The magnitude of the abnormal cycle occupancy risk jump is from Down to This demonstrates that the node timeliness reliability mechanism, the conservative degradation mechanism, and the local rollback or reconstruction mechanism under abnormal conditions can maintain the continuity of the risk field under different levels of data anomalies and reduce the probability of expired data being mistakenly included in modeling. Subsequently, based on the above calculation and processing results, the processor continues to execute the subsequent steps as described in Example 1, completing operations such as node layer maintenance, abnormal version detection, and model release. The specific execution logic is consistent with Example 1 and will not be repeated here.

[0089] Example 4: Electronic Equipment

[0090] This embodiment provides an electronic device, including a memory, a processor, a communication interface, and a data acquisition interface. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the octree multidimensional modeling method for low-altitude airway corridors described in any one of embodiments one through three.

[0091] The processor can be a central processing unit, a digital signal processor, a field-programmable gate array, an application-specific integrated circuit, or other programmable logic devices; the memory can be a read-only memory, a random access memory, flash memory, a solid-state drive, or other non-transient storage media; the communication interface is used to interact with downstream path planning, conflict detection, traffic capacity assessment, or digital twin systems; the data acquisition interface is used to receive multi-source environmental data such as buildings, poles, trees, temporary facilities, moving obstacles, low-altitude meteorology, communication quality, and navigation quality.

[0092] When the processor runs the computer program, it sequentially executes the following steps: multi-source environmental data acquisition and parameter dictionary establishment, unified preprocessing of spatial and temporal attributes, octree spatial model construction and node field initialization, multi-source environmental mapping and node environmental vector generation, node environmental change intensity calculation, node credibility calculation, node splitting and merging and state update, node timeliness credibility record generation and abnormal degradation processing, model organization and output release, thereby outputting an online rolling update octree multidimensional environmental model of the low-altitude airway corridor.

[0093] It should be understood that relational terms such as "first" and "second" are used only to distinguish different objects or steps, and do not indicate an actual order or hierarchy. The terms "including," "comprising," and their variations are open-ended qualifiers and do not exclude other elements or steps not explicitly listed.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the principles and spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-dimensional octree modeling method for low-altitude airway corridors, applied to multi-source environment modeling and dynamic updating scenarios of airway corridors in the low-altitude traffic field, characterized in that... include: Acquire multi-source environmental data within the low-altitude airway corridor and write source type identifier, observation time, source confidence initial value, and failure duration for each data record; The spatial reference, time reference and attribute format of the multi-source environmental data are unified. The spatial data is converted into a unified three-dimensional coordinate frame. Records within the same update cycle are time-aligned according to the latest priority rule. Continuous numerical attributes are normalized into comparable risk quantities. Construct an octree spatial model of the target area, map the multi-source environmental data to the corresponding octree nodes, form a five-dimensional environmental vector in each node that includes at least geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor and navigation quality risk factor, and write the node’s most recent update time and effective deadline. Based on the difference between the five-dimensional environmental vectors of the current period and the previous period, as well as the node timeliness disturbance, the intensity of node environmental change is calculated. The node credibility is calculated based on the prior credibility of the data source mapped to the node, the most recent action time, the failure duration, and the spatial contribution weight. The splitting, merging, and state updating of octree nodes are controlled according to the intensity of node environmental change and the node credibility. When a node has no valid data source, the node's credibility is lower than the preset credibility threshold, or the node has timeliness anomalies, a conservative degraded occupancy result is generated based on the parent node's occupancy risk, the average risk of the same-layer neighborhood, and the historical occupancy status. A low-altitude multidimensional environment model containing node status information and hierarchical relationships is then output. The node status information includes at least the occupancy status, occupancy risk value, source of change, most recent update time, effective expiration time, and reliability.

2. The octree multidimensional modeling method for low-altitude airway corridors according to claim 1, characterized in that, The process of mapping multi-source environmental data to corresponding octree nodes, forming a five-dimensional environmental vector within each node, includes: Perform spatial intersection calculations between the entity bounding volume and the node boundary box for buildings, towers, trees, and temporary facilities to obtain the geometric occupancy contribution; The trajectory point sequence of the moving obstacle is extracted according to the time window, and the dynamic obstacle contribution is obtained based on the dwell ratio of the trajectory point relative to the node and the speed influence range. Low-air meteorological data, communication quality data, and navigation quality data are aggregated according to the sampling results within each node to obtain the meteorological disturbance contribution, communication quality risk contribution, and navigation quality risk contribution, respectively. Contributions of the same type are weighted and fused according to the current credibility weight of the corresponding data source to generate the node's geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor, and navigation quality risk factor.

3. The octree multidimensional modeling method for low-altitude airway corridors according to claim 2, characterized in that, The aggregation of low-altitude meteorological data, communication quality data, and navigation quality data based on intra-node sampling results includes: When the node side length is greater than a preset multiple of the original sampling step length of the corresponding continuous field, the average value of the sampling points inside the node is used as the aggregation result of the continuous field on the node, so as to suppress noise on coarse-scale nodes. When the node side length is not greater than the preset multiple of the original sampling step length of the corresponding continuous field, the maximum risk value of the sampling points inside the node is used as the aggregation result of the continuous field on the node, so as to maintain a local high risk peak on the fine-scale node. The continuous field includes at least one of the following: low-altitude meteorological field, communication quality field, and navigation quality field.

4. The octree multidimensional modeling method for low-altitude airway corridors according to claim 1, characterized in that, The intensity of environmental changes at the computing nodes includes: Read the five-dimensional environment vector of the node in the current period and the five-dimensional environment vector of the previous period, and calculate the changes in geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor and navigation quality risk factor respectively. Based on the current system time, the most recent update time of the node, and the effective deadline of the node, calculate the time-dependent disturbance term that characterizes the degree of uncertainty growth of the node not being refreshed; The intensity of nodal environmental change is obtained by summing the changes of each environmental factor and the time-dependent disturbance term after applying corresponding weights and then performing amplitude limiting processing. Based on the proportion of the weighted change of each environmental factor in the total change, a change source vector for each node is generated.

5. The octree multidimensional modeling method for low-altitude airway corridors according to claim 4, characterized in that, The method of controlling the splitting, merging, and state updating of octree nodes based on the intensity of changes in the node environment and the node's credibility includes: When the intensity of the change in the node's environment is not lower than the preset splitting threshold and the node's level is lower than the preset maximum level, the current node will be split into child nodes. When the intensity of the change in the node environment is not higher than the preset merging threshold, the node credibility is not lower than the preset credibility threshold, the variance of the occupation risk of the sibling node is not greater than the preset variance threshold, and the number of instantiated sibling nodes reaches the preset lower limit, the sibling node is merged into the parent node. When the node's credibility is lower than the preset credibility threshold, the node structure is maintained and a conservative degenerate occupation update is initiated. In all other cases, maintain the current level and update the node environment vector, source of change, occupancy risk value, and occupancy status.

6. The octree multidimensional modeling method for low-altitude airway corridors according to claim 1, characterized in that, The reliability of the computing nodes includes: Determine the set of data sources mapped to the node; weight each data source according to its spatial contribution weight within the node, and exponentially decay the data based on the initial source credibility value of each data source and the time difference between the current system time and the most recent action time of the data source, according to the failure duration. The credibility contributions of each data source after attenuation are normalized and fused to obtain the node credibility. When a node is not mapped to a valid data source or the sum of the spatial contribution weights of each data source is lower than a preset lower limit, the node's credibility is set to a low credibility state.

7. The octree multidimensional modeling method for low-altitude airway corridors according to claim 6, characterized in that, The effective deadline for writing nodes includes: For each valid data source within a node, a candidate cutoff time is determined based on its most recent active time and failure duration. Each candidate deadline and its corresponding spatial contribution weight are combined to form a time-weighted sequence, and then sorted by candidate deadline from smallest to largest. Calculate the cumulative weight ratio of the sorted sequence, and take the minimum candidate deadline time corresponding to the cumulative weight ratio reaching the preset ratio as the effective deadline time of the node. When a node does not have a valid data source, set the node's valid expiration time to the node's most recent update time and write an expiration exception flag.

8. The octree multidimensional modeling method for low-altitude airway corridors according to claim 1, characterized in that, The generation of conservative degenerate occupancy results includes: Obtain the occupancy risk of the parent node of the abnormal node, the occupancy risk of neighboring nodes at the same level as the abnormal node, and the historical occupancy risk of the abnormal node; The parent node occupancy risk, the average risk of the same-layer neighborhood, and the historical occupancy risk are fused according to a preset degradation weight to obtain a conservative degradation occupancy result; The same-layer neighborhood is a set of instantiated same-layer nodes that are adjacent to the abnormal node in three-dimensional space. When the proportion of abnormal nodes or the model credibility meets the preset abnormal conditions, the model is rolled back or rebuilt based on the historical stable version of the local subtree where the abnormal node is located.

9. The octree multidimensional modeling method for low-altitude airway corridors according to claim 1, characterized in that, The generation of the occupancy risk value includes: The node occupancy risk value is obtained by summing the corresponding fusion weights applied to the node's geometric occupancy factor, dynamic obstacle factor, meteorological disturbance factor, communication quality risk factor, navigation quality risk factor, and low reliability penalty term. The node occupancy risk value is compared with a preset occupancy threshold. When the node occupancy risk value is not lower than the preset occupancy threshold, the node is marked as occupied. When the node occupancy risk value is lower than the preset occupancy threshold, the node is marked as traversable. The low-confidence penalty term is determined by the inverse vector of node confidence.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the low-altitude airway corridor octree multidimensional modeling method as described in any one of claims 1 to 9.

Images