A low-altitude logistics information recommendation method based on big data analysis

Abstract

The application discloses a low-altitude logistics information recommendation method based on big data analysis, and relates to the technical field of big data intelligent analysis, and comprises the following steps: step one, collecting low-altitude logistics multi-source operation data and generating low-altitude logistics event fingerprint data; step two, constructing a multi-constraint flight path reachable shell layer and generating a candidate flight path set; step three, generating flight path segment fingerprint data and constructing airspace pressure field data; step four, constructing a dynamic hypergraph; step five, inputting the dynamic hypergraph into an improved TDHNN model, performing pressure gradient propagation, double-flow propagation, risk breakpoint gating and hyperedge blocking processing, and generating a flight path risk breakpoint sequence; step six, performing disturbance analysis, generating flight path continuity data and counterfactual stability data; and step seven, performing candidate flight path sorting and generating a low-altitude logistics information recommendation result. The application realizes intelligent recommendation of a low-altitude logistics flight path.

Description

Technical Field

[0001] This invention relates to the fields of big data intelligent analysis and low-altitude logistics technology, and in particular to a method for recommending low-altitude logistics information based on big data analysis. Background Technology

[0002] With the rapid development of the low-altitude economy and the large-scale application of drone logistics delivery, how to achieve safe and efficient route planning and logistics information recommendation in complex low-altitude airspace environments has become an important research direction in the field of low-altitude logistics. For route selection and delivery decisions in low-altitude logistics tasks, existing technologies mainly employ methods based on shortest path search, rule constraint matching, or static risk assessment for route recommendation, and combine drone operating status, airspace openness, and meteorological information to complete logistics task scheduling. However, the following problems commonly exist in actual low-altitude logistics scenarios: Traditional route recommendation methods typically construct route decision models based on a single dimension or a limited number of constraints, making it difficult to comprehensively characterize the complex relationships between logistics orders, UAV operating status, airspace environment, and historical flight paths. This results in insufficient adaptability of the recommendation results to dynamic environmental changes. Existing methods mostly use ordinary graph structures or rule network structures to describe route relationships, failing to effectively represent the high-order correlation features between logistics events, route segments, and airspace status, thus limiting the ability to model risk propagation in complex low-altitude operation scenarios. Furthermore, existing risk assessment methods primarily focus on the current risk status, lacking mechanisms to identify the risk evolution process and the location of risk abrupt changes, making it difficult to accurately identify key risk nodes affecting route safety. In addition, existing recommendation models typically rank routes based solely on static risk indicators, failing to comprehensively analyze the stability and continuity characteristics of routes under disturbance conditions. This results in insufficient robustness of the recommendation results under airspace, weather, and mission changes, making it difficult to meet the comprehensive requirements of route safety, continuity, and stability in complex low-altitude logistics scenarios.

[0003] Therefore, how to provide a method for recommending low-altitude logistics information based on big data analysis is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose a low-altitude logistics information recommendation method based on big data analysis. This invention constructs a dynamic hypergraph structure that integrates low-altitude logistics events, route segments, and airspace pressure fields. It combines pressure gradient propagation, dual-flow propagation, risk breakpoint gating, and hyperedge blocking mechanisms in the improved TDHNN model to achieve modeling of the low-altitude logistics risk propagation process and identification of risk breakpoints. Furthermore, it introduces a counterfactual analysis mechanism based on single-factor perturbation to evaluate the continuity and stability of candidate routes, ultimately realizing intelligent recommendation and dynamic risk control of low-altitude logistics routes in complex low-altitude environments.

[0005] A method for recommending low-altitude logistics information based on big data analysis according to an embodiment of the present invention includes the following steps: Step 1: Collect multi-source operation data of low-altitude logistics, perform time alignment on the multi-source operation data of low-altitude logistics, and generate low-altitude logistics event fingerprint data; Step 2: Construct a multi-constraint route reachability shell based on the low-altitude logistics event fingerprint data to generate a candidate route set; Step 3: Divide the candidate route set into route segments, generate route segment fingerprint data, and construct airspace pressure field data; Step 4: Construct a dynamic hypergraph based on the route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data; Step 5: Input the dynamic hypergraph into the improved TDHNN model, construct the pressure gradient propagation path based on the airspace pressure field data, update the node state and dynamic hyperedge connection relationship in the dynamic hypergraph in a time series, and perform dual-stream propagation, risk breakpoint gating and hyperedge blocking to generate a route risk breakpoint sequence. Step 6: Perform disturbance analysis based on the route risk breakpoint sequence to generate route continuity data and counterfactual stability data; Step 7: Based on the route continuity data and counterfactual stability data, sort the candidate routes and generate low-altitude logistics information recommendation results.

[0006] Optionally, step one specifically includes: Collect logistics order records, drone operation records, airspace operation records, meteorological monitoring records, historical flight track records, and logistics execution records, and establish data mapping relationships based on data source identifiers; Extract order number, starting location, destination location, cargo weight, cargo volume, and delivery time information from logistics order records to construct an order attribute dataset; Extract spatial location, flight speed, flight altitude, remaining battery power, payload status, and mission status information from the drone's operation records to construct a drone attribute dataset; Extract route node information, route connection information, airspace open status information, and airspace capacity information from airspace operation records; extract wind speed information, wind direction information, precipitation information, and visibility information from meteorological monitoring records; and construct an environmental attribute dataset. The order attribute dataset, drone attribute dataset, and environmental attribute dataset are sorted by time based on timestamp information. Data with time intervals less than the current data arrival frequency are divided into the same time period unit, and a time sequence relationship is established based on the consecutive time period units. By associating order number, drone number, starting position, target position, time period, and spatial location with order attribute dataset, drone attribute dataset, environmental attribute dataset, historical flight track records, and logistics execution records, a low-altitude logistics event unit is constructed. Numerical normalization, category coding, and time-series coding are performed on the order attribute dataset, UAV attribute dataset, environmental attribute dataset, historical flight track record, and logistics execution record in the low-altitude logistics event unit, respectively, and then they are concatenated according to a unified field order to generate low-altitude logistics event fingerprint data.

[0007] Optionally, step two specifically involves: The order attribute data, drone attribute data, environmental attribute data, and historical flight track data in the low-altitude logistics event fingerprint data are analyzed to extract the starting position, target position, remaining battery power, payload status, delivery time limit, airspace status, and weather status. Using the route node corresponding to the starting position as the central node, the achievable range of power is calculated based on the remaining power and historical energy consumption data per unit distance, and a power constraint shell is constructed. The load passage range is determined based on the load status and UAV load parameters, and a load constraint shell is constructed. Based on the airway openness, airway connectivity, and airspace capacity information in the airspace status, airway nodes that meet the passage conditions are mapped to the airspace topology network to construct an airspace constraint shell; Based on wind speed, wind direction, precipitation, and visibility information in the meteorological conditions, the navigable range is determined and a meteorological constraint shell is constructed. Based on the delivery time limit and the flight duration information in historical flight track data, the time limit reachable range is determined and a time limit constraint shell is constructed. Spatial overlay calculations are performed on the energy constraint shell, load constraint shell, airspace constraint shell, meteorological constraint shell, and time constraint shell. Route nodes and route connections that are simultaneously located within the five types of constraint shells are retained to generate a set of reachable regions. Based on the route nodes and route connection relationships in the reachable area set, a path combination is performed to construct a candidate route set.

[0008] Optionally, step three specifically includes: Obtain each candidate route from the candidate route set, and segment each candidate route according to the connection relationship between adjacent route nodes and the location of route attribute changes to construct multiple route segments; Extract spatial coordinate information, route length information, flight altitude information, and heading information corresponding to each route segment to construct spatial location feature data; Spatial matching is performed based on the spatial coordinates of flight route segments and meteorological monitoring records, airspace operation records, and historical flight track records to extract meteorological status data, airspace status data, historical traffic data, and congestion status data; Numerical normalization, category coding, and time-series coding are performed on spatial location feature data, meteorological status data, airspace status data, historical traffic data, and congestion status data, and they are combined according to the route segment identifier to construct route segment fingerprint data. Based on historical flight track records, statistical data were compiled on the changes in the number of drones, take-off and landing frequency, and airway passage frequency in each spatial area; Based on airspace operation records, statistical analysis was conducted on airspace capacity occupancy data for each space area, and based on meteorological monitoring records, statistical analysis was conducted on wind speed variation data for each space area. Spatial mapping is performed on data such as changes in the number of drones, take-off and landing frequency, airway passage frequency, airspace capacity occupancy, and wind speed changes. Pressure accumulation calculation is then performed according to spatial grid units to generate the pressure value distribution corresponding to each spatial grid unit, thus constructing airspace pressure field data.

[0009] Optionally, step four specifically involves: Read route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data, and establish a unified time-series index based on timestamp information; The fingerprint data of each airway segment is mapped to airway segment nodes, the fingerprint data of low-altitude logistics events is mapped to logistics event nodes, and the spatial grid cells in the airspace pressure field data are mapped to airspace pressure nodes. Based on the starting position, target position, and route segment connection relationship in the candidate route set corresponding to the logistics event node, establish the association relationship between the logistics event node and the route segment node; The association between route segment nodes and airspace pressure nodes is established based on the spatial location of the route segment nodes and the spatial grid location of the airspace pressure nodes. Based on a unified time-series index, logistics event nodes, air route segment nodes, and airspace pressure nodes are associated in a time sequence to construct a node association set; Using logistics event nodes, route segment nodes, and airspace pressure nodes within the same time period unit as hyperedge connection objects, dynamic hyperedges are constructed based on the node association set, and the time and space identifiers corresponding to each dynamic hyperedge are recorded. A dynamic hypergraph is constructed based on dynamic hyperedges and node relationships.

[0010] Optionally, step five specifically includes: The dynamic hypergraph is input into the improved TDHNN model; The improved TDHNN model incorporates a pressure gradient propagation module, a dual-stream propagation module, a risk breakpoint gating module, and a super-edge blocking module in the dynamic node update structure and dynamic super-edge update structure of the TDHNN model. The pressure gradient propagation module performs pressure gradient propagation processing based on spatial pressure nodes and dynamic hyperedges, generating dynamic hyperedge propagation paths. The dual-stream propagation module constructs forward propagation links and reverse propagation links based on the dynamic hyperedge propagation path, and performs hypergraph message propagation along the forward and reverse propagation links to logistics event nodes, route segment nodes, and airspace pressure nodes, generating forward node states and reverse node states. Perform state fusion processing on the forward node state and the reverse node state to generate node time sequence state; The risk breakpoint gating module performs risk breakpoint identification processing based on the node timing state to generate risk breakpoint nodes; The hyperedge blocking module retrieves dynamic hyperedges connected to the risk breakpoint nodes based on the risk breakpoint nodes, performs connection state analysis and connection relationship update processing on the dynamic hyperedges, and generates an updated dynamic hypergraph. Based on the updated dynamic hypergraph, node state updates and dynamic hyperedge updates are performed to generate risk state values ​​corresponding to each route segment node. Arrange the risk status values ​​according to the connection order of the route segments in the candidate route to generate a route risk breakpoint sequence.

[0011] Optionally, the pressure gradient propagation process specifically includes: Read the pressure value corresponding to each airspace pressure node, and retrieve the adjacent airspace pressure nodes connected to each airspace pressure node. Calculate the pressure difference between each airspace pressure node and adjacent airspace pressure nodes, and establish a pressure gradient propagation path from high-pressure nodes to low-pressure nodes based on the pressure difference. Map the pressure gradient propagation path to dynamic hyperedges, determine the propagation direction corresponding to each dynamic hyperedge, and count the total pressure difference corresponding to each dynamic hyperedge within the same propagation link. Calculate the proportion of the pressure difference corresponding to each dynamic hyperedge to the total pressure difference, generate the propagation weight, and construct the dynamic hyperedge propagation path based on the propagation direction and the propagation weight.

[0012] Optionally, the risk breakpoint identification process specifically includes: Calculate the node timing state difference between adjacent time period units for each route segment node; The node timing state differences corresponding to each segment node within the same candidate route are statistically analyzed, and the mean and standard deviation of the node timing state differences are calculated. The adaptive gating limit is determined based on the mean of the node timing state difference and the standard deviation of the node timing state difference; Compare the node timing state difference corresponding to each route segment node with the adaptive gating limit; Nodes in a route segment whose timing state difference is greater than the adaptive gating limit are marked as risk breakpoint nodes. The distribution of risk breakpoint nodes in the candidate routes is statistically analyzed, and risk breakpoint identification information is generated based on the risk breakpoint node distribution results; the risk breakpoint identification information is arranged according to the connection order of route segments in the candidate routes to generate a risk breakpoint sequence.

[0013] Optionally, step six specifically includes: Based on the sequence of risk breakpoints in the route, the locations of risk breakpoints in the candidate routes are identified, and the route segments before the breakpoints and the route segments after the breakpoints are divided. Wind speed disturbance, airspace restriction disturbance, take-off and landing point congestion disturbance, power disturbance and delivery time limit disturbance are applied sequentially to the route segments in the candidate routes to construct a set of disturbed route scenarios; The data corresponding to each set of disturbance route scenarios are re-input into the improved TDHNN model to generate a sequence of risk state values ​​corresponding to each disturbance route scenario. Calculate the state difference between the risk state value sequence corresponding to each disturbance route scenario and the original risk state value sequence to generate a risk offset sequence; Calculate the mean and standard deviation of risk offset for each risk offset sequence, and determine the offset limit based on the mean and standard deviation of risk offset; The number of route segments exceeding the offset limit is counted, and the proportion of the number of route segments to the total number of candidate route segments is calculated to generate the disturbance impact ratio. Calculate the route stability index based on the proportion of disturbance impact corresponding to each disturbance route scenario; Calculate the difference sequence between the risk status values ​​corresponding to the set of route segments before the breakpoint and the set of route segments after the breakpoint, and calculate the breakpoint continuity index based on the difference sequence; Route continuity data and counterfactual stability data are generated based on route stability indicators and discontinuity continuity indicators.

[0014] Optionally, step seven specifically includes: Read the route continuity data and counterfactual stability data corresponding to the candidate routes; Calculate the continuity score, stability score, and risk score for each candidate route; The continuity score, stability score, and risk score are normalized to generate continuity evaluation value, stability evaluation value, and risk evaluation value. A candidate route evaluation sequence is constructed based on continuity evaluation values, stability evaluation values, and risk evaluation values; The candidate route evaluation sequence is sorted to generate the candidate route ranking result; Extract the risk breakpoint sequence corresponding to the top N candidate routes from the candidate route ranking results; Statistically analyze the number of risk breakpoints and their distribution locations for each risk breakpoint sequence. The first N candidate routes are screened based on the number and location of risk breakpoints to determine the target route. Read the route segment information, risk breakpoint sequence and logistics task information corresponding to the target route, and generate low-altitude logistics information recommendation results.

[0015] The beneficial effects of this invention are: This invention constructs a dynamic hypergraph structure that integrates low-altitude logistics event fingerprint data, route segment fingerprint data, and airspace pressure field data. Combined with a multi-constraint route reachability shell construction mechanism and a collaborative design of an improved TDHNN model, it addresses the problems of insufficient correlation capabilities of multi-source heterogeneous data in low-altitude logistics scenarios, difficulty in characterizing route risk propagation processes, and insufficient stability of recommendation results in complex airspace environments. It proposes a multi-constraint route selection strategy based on power constraints, load constraints, airspace constraints, meteorological constraints, and time constraints to achieve refined matching between low-altitude logistics tasks and route resources. In the graph modeling stage, a dynamic hypergraph structure composed of logistics event nodes, route segment nodes, and airspace pressure nodes is introduced. Dynamic hyperedges enable a unified expression of high-order relationships between logistics tasks, route status, and airspace operation status. In the model inference stage, a pressure ladder is introduced. The system comprises a degree propagation module, a dual-flow propagation module, a risk breakpoint gating module, and a super-edge blocking module. It utilizes the airspace pressure field to construct a pressure gradient propagation path and combines forward and reverse propagation links to model the risk evolution process. Simultaneously, it identifies risk breakpoint nodes and dynamically adjusts abnormal propagation links, effectively enhancing the model's ability to perceive the characteristics of route risk propagation and the location of risk abrupt changes. In the stability analysis phase, a counterfactual analysis mechanism based on single-factor disturbances is constructed. Disturbance route scenarios are generated through wind speed disturbances, airspace flight restriction disturbances, take-off and landing point congestion disturbances, power disturbances, and delivery time limit disturbances. Robustness assessment of candidate routes is achieved by combining route continuity data and counterfactual stability data. Finally, through candidate route ranking and risk breakpoint screening mechanisms, low-altitude logistics information recommendation results are output, realizing dynamic risk perception, stability assessment, and intelligent recommendation of low-altitude logistics routes. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall process of a low-altitude logistics information recommendation method based on big data analysis proposed in this invention; Figure 2 This is an overall structural diagram of the improved TDHNN model in the low-altitude logistics information recommendation method based on big data analysis proposed in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-2 A method for recommending low-altitude logistics information based on big data analysis includes the following steps: Step 1: Collect multi-source operation data of low-altitude logistics, perform time alignment on the multi-source operation data of low-altitude logistics, and generate low-altitude logistics event fingerprint data; Step 2: Construct a multi-constraint route reachability shell based on low-altitude logistics event fingerprint data to generate a candidate route set; Step 3: Divide the candidate route set into route segments, generate route segment fingerprint data, and construct airspace pressure field data; Step 4: Construct a dynamic hypergraph based on route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data; Step 5: Input the dynamic hypergraph into the improved TDHNN model, construct the pressure gradient propagation path based on the airspace pressure field data, update the node state and dynamic hyperedge connection relationship in the dynamic hypergraph in a time series, and perform dual-stream propagation, risk breakpoint gating and hyperedge blocking to generate a route risk breakpoint sequence. Step 6: Perform disturbance analysis based on the route risk breakpoint sequence to generate route continuity data and counterfactual stability data; Step 7: Based on route continuity data and counterfactual stability data, rank the candidate routes and generate low-altitude logistics information recommendation results.

[0019] In this embodiment, step one specifically includes: Collect logistics order records, drone operation records, airspace operation records, meteorological monitoring records, historical flight track records, and logistics execution records, and establish data mapping relationships based on data source identifiers; Extract order number, starting location, destination location, cargo weight, cargo volume, and delivery time information from logistics order records to construct an order attribute dataset; Extract spatial location, flight speed, flight altitude, remaining battery power, payload status, and mission status information from the drone's operation records to construct a drone attribute dataset; Extract route node information, route connection information, airspace open status information, and airspace capacity information from airspace operation records; extract wind speed information, wind direction information, precipitation information, and visibility information from meteorological monitoring records; and construct an environmental attribute dataset. Based on timestamp information, the order attribute dataset, drone attribute dataset, and environmental attribute dataset are sorted by time. Data with time intervals less than the current data arrival frequency are divided into the same time period unit, and a time sequence relationship is established based on the consecutive time period units. By associating order number, drone number, starting position, target position, time period, and spatial location with order attribute dataset, drone attribute dataset, environmental attribute dataset, historical flight track records, and logistics execution records, a low-altitude logistics event unit is constructed. Numerical normalization, category coding, and time-series coding are performed on the order attribute dataset, UAV attribute dataset, environmental attribute dataset, historical flight track record, and logistics execution record in the low-altitude logistics event unit, respectively, and then they are concatenated according to a unified field order to generate low-altitude logistics event fingerprint data.

[0020] In this implementation, data mapping relationships are used to establish a unified indexing system among data from different sources. This unified indexing system includes order identifiers, drone identifiers, spatial area identifiers, and time identifiers. Data arrival frequency is obtained based on the statistical data received within a continuous time period, and the time scale is adaptively adjusted according to changes in data arrival frequency to maintain temporal consistency among data from different sources. Temporal correlation relationships are established based on the entity continuity between adjacent time period units, including order continuity, drone continuity, and spatial location continuity. Category coding uses discrete coding rules corresponding to field categories to generate category code values, and temporal coding generates temporal code values ​​based on the time sequence corresponding to timestamps. The unified field order includes order attribute fields, drone attribute fields, environmental attribute fields, historical flight track fields, and execution record fields. These fields are concatenated in a fixed order to maintain a unified dimensional structure for low-altitude logistics event fingerprint data, providing a consistent data input foundation for subsequent route reachability shell construction and dynamic hypergraph construction.

[0021] In this embodiment, step two specifically involves: Analyze order attribute data, drone attribute data, environmental attribute data, and historical flight track data from low-altitude logistics event fingerprint data to extract starting position, target position, remaining battery power, payload status, delivery time limit, airspace status, and weather status; Using the route node corresponding to the starting position as the central node, the range of achievable power is calculated based on the remaining power and historical energy consumption data per unit distance, and a power constraint shell is constructed. The load passage range is determined based on the load status and UAV load parameters, and a load constraint shell is constructed. Based on the airway openness, airway connectivity, and airspace capacity information in the airspace status, airway nodes that meet the passage conditions are mapped to the airspace topology network to construct an airspace constraint shell; Based on wind speed, wind direction, precipitation, and visibility information in the meteorological conditions, the navigable range is determined and a meteorological constraint shell is constructed. Based on the delivery time limit and the flight duration information in historical flight track data, the time-limited reachable range is determined and a time-limited constraint shell is constructed. Spatial overlay calculations are performed on the energy constraint shell, load constraint shell, airspace constraint shell, meteorological constraint shell, and time constraint shell. Route nodes and route connections that are simultaneously located within the five types of constraint shells are retained to generate a set of reachable regions. Based on the route nodes and route connections in the reachable area set, perform path combinations to construct a candidate route set.

[0022] In this implementation, route nodes are derived from the set of nodes in the low-altitude airspace topology network, and route connections are derived from the directed connections between route nodes. The reachable range of power is determined based on the remaining power and historical energy consumption statistics per unit distance. The load passage range is determined based on the current load as a percentage of the rated load and the route operating conditions. Airspace capacity information is represented by the number of UAVs allowed to pass per unit time, and the suitable airspace is determined based on the passage conditions corresponding to wind speed, precipitation, and visibility levels. Spatial overlay calculation uses a unified spatial coordinate system to perform gridded mapping on each constraint shell and performs overlap determination on each grid cell. The route nodes in the reachable area set are used to construct a path search space according to their connectivity. The connecting paths between the route nodes corresponding to the starting position and the route nodes corresponding to the target position are filtered, and path combinations that meet the connectivity conditions are retained to ensure that the candidate route set is consistent with the operational constraints of the current low-altitude logistics task.

[0023] In this embodiment, step three specifically includes: Obtain each candidate route from the candidate route set, and segment each candidate route according to the connection relationship between adjacent route nodes and the location of route attribute changes to construct multiple route segments; Extract spatial coordinate information, route length information, flight altitude information, and heading information corresponding to each route segment to construct spatial location feature data; Spatial matching is performed based on the spatial coordinates of flight route segments and meteorological monitoring records, airspace operation records, and historical flight track records to extract meteorological status data, airspace status data, historical traffic data, and congestion status data; Numerical normalization, category coding, and time-series coding are performed on spatial location feature data, meteorological status data, airspace status data, historical traffic data, and congestion status data, and then combined according to route segment identifiers to construct route segment fingerprint data. Based on historical flight track records, statistical data were collected on the changes in the number of drones, take-off and landing frequency, and airway passage frequency in each spatial area. Based on airspace operation records, the airspace capacity occupancy data of each space area is statistically analyzed, and based on meteorological monitoring records, the wind speed change data of each space area is statistically analyzed. Spatial mapping is performed on data such as changes in the number of drones, take-off and landing frequency, airway passage frequency, airspace capacity occupancy, and wind speed changes. Pressure accumulation calculation is then performed according to spatial grid units to generate the pressure value distribution corresponding to each spatial grid unit, thus constructing airspace pressure field data.

[0024] In this implementation, the locations of changes in route attributes include changes in flight altitude, heading, airspace status, and meteorological status. Spatial matching adopts a unified spatial coordinate system, mapping the spatial range corresponding to route segments to meteorological monitoring areas, airspace management areas, and historical flight track coverage areas. Congestion status data is determined based on the ratio between the number of UAVs passing through the route segment per unit time and the route capacity. Pressure accumulation calculation uses spatial grid cells as the basic calculation unit, normalizing data on changes in the number of UAVs, takeoff and landing frequencies, route passage frequencies, airspace capacity occupancy, and wind speed changes, and then accumulating them according to their corresponding weights. Grid connections are established between adjacent spatial grid cells, giving the airspace pressure field data continuous spatial distribution characteristics, providing a spatial constraint basis for subsequent pressure gradient propagation path construction and risk propagation calculation in the dynamic hypergraph.

[0025] In this embodiment, step four specifically includes: Read route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data, and establish a unified time-series index based on timestamp information; The fingerprint data of each airway segment is mapped to airway segment nodes, the fingerprint data of low-altitude logistics events is mapped to logistics event nodes, and the spatial grid cells in the airspace pressure field data are mapped to airspace pressure nodes. Based on the starting position, target position, and route segment connection relationship in the candidate route set corresponding to the logistics event node, establish the association relationship between the logistics event node and the route segment node; The association between route segment nodes and airspace pressure nodes is established based on the spatial location of the route segment nodes and the spatial grid location of the airspace pressure nodes. Based on a unified time-series index, logistics event nodes, air route segment nodes, and airspace pressure nodes are associated in a time sequence to construct a node association set; Using logistics event nodes, route segment nodes, and airspace pressure nodes within the same time period unit as hyperedge connection objects, dynamic hyperedges are constructed based on the node association set, and the time and space identifiers corresponding to each dynamic hyperedge are recorded. A dynamic hypergraph is constructed based on dynamic hyperedges and node relationships.

[0026] In this implementation, the unified temporal index is constructed using an ascending timestamp arrangement, with time identifiers corresponding to time period unit numbers. Route segment nodes correspond to individual route segments, logistics event nodes to individual low-altitude logistics event units, and airspace pressure nodes to individual spatial grid units. Node association sets are organized using node identifier pairs, time identifiers, and spatial identifiers, where the spatial identifiers correspond to the positional relationship between the spatial region where the route segment is located and the spatial region where the spatial grid unit is located. Dynamic hyperedges employ a multi-node joint connection structure, with the same dynamic hyperedge simultaneously connecting at least one logistics event node, one route segment node, and one airspace pressure node. Time identifiers characterize the time period unit to which the dynamic hyperedge belongs, and spatial identifiers characterize the spatial region corresponding to the dynamic hyperedge, enabling the dynamic hypergraph to possess both temporal and spatial association characteristics, providing a graph structure foundation for subsequent pressure gradient propagation, dual-flow propagation, and risk breakpoint gating.

[0027] In this embodiment, step five specifically includes: Input dynamic hypergraphs into the improved TDHNN model; The TDHNN model is improved by adding a pressure gradient propagation module, a two-stream propagation module, a risk breakpoint gating module, and a super-edge blocking module to the dynamic node update structure and dynamic super-edge update structure of the TDHNN model. The pressure gradient propagation module performs pressure gradient propagation processing based on spatial pressure nodes and dynamic hyperedges, generating dynamic hyperedge propagation paths. The dual-stream propagation module is based on the dynamic hyperedge propagation path. It constructs forward propagation links and reverse propagation links respectively, and performs hypergraph message propagation on logistics event nodes, route segment nodes and airspace pressure nodes along the forward propagation links and reverse propagation links to generate forward node status and reverse node status. Perform state fusion processing on the forward and reverse node states to generate node temporal states; The risk breakpoint gating module performs risk breakpoint identification processing based on the node timing status and generates risk breakpoint nodes. The hyperedge blocking module is based on risk breakpoint nodes, retrieves dynamic hyperedges connected to risk breakpoint nodes, performs connection state analysis and connection relationship update processing on dynamic hyperedges, and generates updated dynamic hypergraphs. Based on the updated dynamic hypergraph, node state updates and dynamic hyperedge updates are performed to generate risk state values ​​corresponding to each route segment node. Arrange the risk status values ​​according to the connection order of the route segments in the candidate route to generate a route risk breakpoint sequence.

[0028] In this embodiment, the improved TDHNN model inherits the dynamic hypergraph learning framework of the TDHNN model. Based on the dynamic node update structure and the dynamic hyperedge update structure, it models the dynamic hypergraph composed of logistics event nodes, route segment nodes and airspace pressure nodes, and uses dynamic hyperedges to describe the multi-dimensional association relationships between nodes, so as to realize the node state update and hyperedge relationship update in the dynamic hypergraph. The improved TDHNN model incorporates a pressure gradient propagation module, a dual-stream propagation module, a risk breakpoint gating module, and a super-edge blocking module based on the dynamic node update structure and the dynamic super-edge update structure. The pressure gradient propagation module constructs a dynamic super-edge propagation path based on the spatial pressure field; the dual-stream propagation module establishes forward and reverse propagation links respectively; the risk breakpoint gating module identifies risk breakpoint nodes based on the temporal state of nodes; and the super-edge blocking module updates the dynamic super-edge connection relationship based on the risk breakpoint nodes. The pressure gradient propagation module introduces the spatial pressure distribution in the airspace pressure field into the dynamic hypergraph propagation process, giving the dynamic hyperedge directional propagation characteristics; the dual-flow propagation module simultaneously describes the risk diffusion process and the backtracking propagation process; the risk breakpoint gating module realizes the identification of the location of sudden changes in route risk; the hyperedge blocking module realizes the dynamic pruning of abnormal propagation links; the synergistic effect of these modules enables the dynamic hypergraph to characterize the dynamic coupling relationship between low-altitude logistics tasks, route status, and airspace pressure, improving the modeling capability of the route risk evolution process and the accuracy of risk breakpoint identification.

[0029] In this embodiment, the pressure gradient propagation process is specifically as follows: Read the pressure value corresponding to each airspace pressure node, and retrieve the adjacent airspace pressure nodes connected to each airspace pressure node. Calculate the pressure difference between each airspace pressure node and adjacent airspace pressure nodes, and establish a pressure gradient propagation path from high-pressure nodes to low-pressure nodes based on the pressure difference. The pressure gradient propagation path is mapped to the dynamic hyperedge, the propagation direction corresponding to each dynamic hyperedge is determined, and the total pressure difference corresponding to each dynamic hyperedge within the same propagation link is counted. Calculate the proportion of the pressure difference corresponding to each dynamic hyperedge to the total pressure difference, generate the propagation weight, and construct the propagation path of the dynamic hyperedge based on the propagation direction and the propagation weight.

[0030] In this embodiment, the risk breakpoint identification process is specifically as follows: Calculate the node timing state difference between adjacent time period units for each route segment node; The node timing state differences corresponding to each segment node within the same candidate route are statistically analyzed, and the mean and standard deviation of the node timing state differences are calculated. The adaptive gating limit is determined based on the mean of the node timing state difference and the standard deviation of the node timing state difference; Compare the node timing state difference corresponding to each route segment node with the adaptive gating limit; Nodes in a route segment whose temporal state difference is greater than the adaptive gating limit are marked as risk breakpoint nodes. The distribution of risk breakpoint nodes in the candidate routes is statistically analyzed, and risk breakpoint identification information is generated based on the risk breakpoint node distribution results; the risk breakpoint identification information is arranged according to the connection order of route segments in the candidate routes to generate a risk breakpoint sequence.

[0031] In this embodiment, step six specifically includes: Based on the sequence of risk breakpoints in the route, the locations of risk breakpoints in the candidate routes are identified, and the route segments before the breakpoints and the route segments after the breakpoints are divided. Wind speed disturbance, airspace restriction disturbance, take-off and landing point congestion disturbance, power disturbance and delivery time limit disturbance are applied sequentially to the route segments in the candidate routes to construct a set of disturbed route scenarios; The data corresponding to each set of disturbance route scenarios are re-input into the improved TDHNN model to generate a sequence of risk state values ​​corresponding to each disturbance route scenario. Calculate the state difference between the risk state value sequence corresponding to each disturbance route scenario and the original risk state value sequence to generate a risk offset sequence; Calculate the mean and standard deviation of risk offset for each risk offset sequence, and determine the offset limit based on the mean and standard deviation of risk offset; The number of route segments exceeding the offset limit is counted, and the proportion of the number of route segments to the total number of candidate route segments is calculated to generate the disturbance impact ratio. Calculate the route stability index based on the proportion of disturbance impact corresponding to each disturbance route scenario; Calculate the difference sequence between the risk status values ​​corresponding to the set of route segments before the breakpoint and the set of route segments after the breakpoint, and calculate the breakpoint continuity index based on the difference sequence; Route continuity data and counterfactual stability data are generated based on route stability indicators and discontinuity continuity indicators.

[0032] In this implementation, the disturbance route scenarios are constructed using a single-factor disturbance approach. Only one disturbance factor's data is adjusted at a time, while the remaining data remain unchanged to ensure comparability between different disturbance scenarios. Wind speed disturbances correspond to changes in meteorological monitoring data; airspace restriction disturbances correspond to changes in airspace openness; take-off and landing point congestion disturbances correspond to changes in take-off and landing area traffic flow; battery disturbances correspond to changes in the remaining battery power of drones; and delivery time limit disturbances correspond to changes in the delivery time window. The risk offset mean is calculated based on all offset values ​​in the risk offset sequence, and the risk offset standard deviation is calculated based on the deviation between each offset value and the risk offset mean. The route stability index is used to characterize the sensitivity of candidate routes to different disturbance scenarios. The breakpoint continuity index is used to characterize the continuity of risk transmission between route segments before and after a risk breakpoint. Route continuity data and counterfactual stability data correspond to the route structural stability characteristics and disturbance response characteristics, respectively, providing a quantitative evaluation basis for ranking candidate routes.

[0033] In this embodiment, step seven specifically includes: Read the route continuity data and counterfactual stability data corresponding to the candidate routes; Calculate the continuity score, stability score, and risk score for each candidate route; The continuity score, stability score, and risk score are normalized to generate continuity evaluation value, stability evaluation value, and risk evaluation value. A candidate route evaluation sequence is constructed based on continuity evaluation values, stability evaluation values, and risk evaluation values; The candidate route evaluation sequence is sorted to generate the candidate route ranking result; Extract the risk breakpoint sequence corresponding to the top N candidate routes from the candidate route ranking results; Statistically analyze the number of risk breakpoints and their distribution locations for each risk breakpoint sequence. The first N candidate routes are screened based on the number and location of risk breakpoints to determine the target route. Read the route segment information, risk breakpoint sequence and logistics task information corresponding to the target route, and generate low-altitude logistics information recommendation results.

[0034] In this implementation, the continuity score is used to characterize the smoothness of the change in risk status along the route direction in the candidate route; the stability score is used to characterize the ability of the candidate route to maintain its state under different disturbance route scenarios; and the risk score is used to characterize the overall risk level corresponding to the candidate route. N is a positive integer greater than 1, determined based on the number of candidate routes. The normalization process adopts a unified numerical interval mapping method to ensure that different types of evaluation results have consistent dimensions. The candidate route evaluation sequence establishes a mapping relationship between the candidate route identifier and the corresponding evaluation result. The number of risk breakpoints is counted by the number of times the breakpoint identifier appears in the risk breakpoint sequence, and the distribution location of risk breakpoints is characterized by the relative position of the risk breakpoints in the candidate routes. During the screening process, candidate routes with fewer risk breakpoints and whose risk breakpoint distribution locations are far from the starting route segment are prioritized for retention. The low-altitude logistics information recommendation results include the target route identifier, the route segment set, the risk breakpoint identifier, and the corresponding logistics task identifier.

[0035] Example 1: To verify the feasibility of this invention in practice, it was applied to a low-altitude logistics delivery platform in a city. The platform is deployed between the city's core business district, logistics distribution centers, and community delivery stations, utilizing multiple logistics drones to perform on-demand delivery tasks. The platform receives approximately 12,000 logistics orders daily, manages 480 drones, covers 2,150 low-altitude airway nodes and 3,560 airspace monitoring grid units, and receives real-time meteorological monitoring data, airspace operation data, and drone operation data.

[0036] In actual operation, multiple accessible routes often exist for the same delivery task. Due to the continuous changes in airspace openness, local wind field variations, drone operating density, and congestion levels at take-off and landing points, some routes, although meeting the shortest distance requirements, are prone to local congestion, risk accumulation, and delivery delays during actual flight. Traditional route recommendation methods typically select routes based on distance or static risk indicators, making it difficult to identify the risk propagation process and the location of risk abrupt changes, resulting in insufficient stability of the recommended results in complex low-altitude environments.

[0037] When applying this invention, firstly, logistics order records, UAV operation records, airspace operation records, meteorological monitoring records, historical flight track records, and logistics execution records are collected, and low-altitude logistics event fingerprint data is constructed. Then, a candidate route set is constructed using power constraint shells, load constraint shells, airspace constraint shells, meteorological constraint shells, and time constraint shells. The candidate route set is then segmented into route segments, and route segment fingerprint data and airspace pressure field data are constructed. Furthermore, a dynamic hypergraph is constructed using logistics event nodes, route segment nodes, and airspace pressure nodes, and input into an improved TDHNN model.

[0038] The improved TDHNN model establishes a pressure gradient propagation path based on the airspace pressure field, enabling airspace pressure changes to propagate along the dynamic superedge. The dual-flow propagation module simultaneously establishes forward and reverse propagation links to jointly model the risk diffusion and risk backtracking processes. The risk breakpoint gating module identifies the locations of risk mutations. The superedge blocking module dynamically prunes abnormal propagation links to obtain a sequence of route risk breakpoints.

[0039] Subsequently, multiple disturbance route scenarios were constructed using wind speed disturbances, airspace restriction disturbances, take-off and landing point congestion disturbances, battery power disturbances, and delivery time limit disturbances. These disturbance scenarios were then re-input into an improved TDHNN model for inference analysis to obtain risk changes under different disturbance conditions. Based on the risk status changes before and after the disturbances, route continuity data and counterfactual stability data were calculated to rank candidate routes, and finally, low-altitude logistics information recommendation results were output.

[0040] This invention differs from existing low-altitude logistics recommendation methods that rely solely on distance, time, or static risk indicators for route selection. It maps low-altitude logistics events, route segments, and airspace pressure fields onto a unified dynamic hypergraph structure, establishing high-order correlations between logistics tasks, route status, and airspace operational status through dynamic hyperedges. Furthermore, it introduces a pressure gradient propagation module based on the TDHNN model, transforming the spatial pressure distribution in the airspace pressure field into dynamic hyperedge propagation directions, giving the risk propagation process spatial evolution characteristics. It further introduces a risk breakpoint gating module and a hyperedge blocking module to identify abrupt changes in route risk and dynamically adjust abnormal propagation links, thereby achieving refined modeling of the risk propagation process. In addition, this invention constructs a counterfactual analysis mechanism based on single-factor perturbations. By independently analyzing wind speed, airspace openness, takeoff and landing point congestion, battery status, and delivery time limits, it evaluates the stability and continuity characteristics of candidate routes under different operating conditions. This ensures that the recommendation results not only reflect the current risk status but also the adaptability under future environmental changes, thereby improving the reliability and robustness of low-altitude logistics information recommendation results.

[0041] To verify the actual effect of the present invention, it was compared with the traditional shortest path recommendation method and the static risk scoring recommendation method. The test was conducted continuously for 30 days, with a total of 10,000 delivery tasks. The experimental results are shown in Table 1.

[0042] Table 1. Comparison of the effectiveness of recommended low-altitude logistics routes

[0043] As shown in Table 1, the method of this invention outperforms the comparative methods in terms of average delivery time, on-time delivery rate, number of risk events, accuracy of risk breakpoint identification, and stability under disturbance scenarios. Specifically, the on-time delivery rate reaches 97.8%, an improvement of 9.2 percentage points compared to the traditional shortest path method; the number of risk events decreases from 418 to 79, a reduction of 81.1%; the accuracy of risk breakpoint identification reaches 95.6%, indicating that the improved TDHNN model can effectively identify risk abrupt changes in the route; and the stability score under disturbance scenarios reaches 0.92, demonstrating that the counterfactual stability analysis mechanism constructed in this invention can accurately assess the adaptability of the route to complex environmental changes.

[0044] Test results show that this invention achieves collaborative modeling between low-altitude logistics operation status, airspace pressure status, and route risk status through dynamic hypergraph modeling, pressure gradient propagation mechanism, risk breakpoint gating mechanism, and counterfactual stability analysis mechanism. It can realize route risk perception, stability assessment, and intelligent recommendation in complex low-altitude logistics scenarios, improve delivery efficiency, and reduce operational risks, and has good engineering application value.

[0045] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for recommending low-altitude logistics information based on big data analysis, characterized in that, Includes the following steps: Step 1: Collect multi-source operation data of low-altitude logistics, perform time alignment on the multi-source operation data of low-altitude logistics, and generate low-altitude logistics event fingerprint data; Step 2: Construct a multi-constraint route reachability shell based on the low-altitude logistics event fingerprint data to generate a candidate route set; Step 3: Divide the candidate route set into route segments, generate route segment fingerprint data, and construct airspace pressure field data; Step 4: Construct a dynamic hypergraph based on the route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data; Step 5: Input the dynamic hypergraph into the improved TDHNN model, construct the pressure gradient propagation path based on the airspace pressure field data, update the node state and dynamic hyperedge connection relationship in the dynamic hypergraph in a time series, and perform dual-stream propagation, risk breakpoint gating and hyperedge blocking to generate a route risk breakpoint sequence. Step 6: Perform disturbance analysis based on the route risk breakpoint sequence to generate route continuity data and counterfactual stability data; Step 7: Based on the route continuity data and counterfactual stability data, sort the candidate routes and generate low-altitude logistics information recommendation results.

2. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step one specifically involves: Collect logistics order records, drone operation records, airspace operation records, meteorological monitoring records, historical flight track records, and logistics execution records, and establish data mapping relationships based on data source identifiers; Extract order number, starting location, destination location, cargo weight, cargo volume, and delivery time information from logistics order records to construct an order attribute dataset; Extract spatial location, flight speed, flight altitude, remaining battery power, payload status, and mission status information from the drone's operation records to construct a drone attribute dataset; Extract route node information, route connection information, airspace open status information, and airspace capacity information from airspace operation records; extract wind speed information, wind direction information, precipitation information, and visibility information from meteorological monitoring records; and construct an environmental attribute dataset. The order attribute dataset, drone attribute dataset, and environmental attribute dataset are sorted by time based on timestamp information. Data with time intervals less than the current data arrival frequency are divided into the same time period unit, and a time sequence relationship is established based on the consecutive time period units. By associating order number, drone number, starting position, target position, time period, and spatial location with order attribute dataset, drone attribute dataset, environmental attribute dataset, historical flight track records, and logistics execution records, a low-altitude logistics event unit is constructed. Numerical normalization, category coding, and time-series coding are performed on the order attribute dataset, UAV attribute dataset, environmental attribute dataset, historical flight track record, and logistics execution record in the low-altitude logistics event unit, respectively, and then they are concatenated according to a unified field order to generate low-altitude logistics event fingerprint data.

3. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step two specifically involves: The order attribute data, drone attribute data, environmental attribute data, and historical flight track data in the low-altitude logistics event fingerprint data are analyzed to extract the starting position, target position, remaining battery power, payload status, delivery time limit, airspace status, and weather status. Using the route node corresponding to the starting position as the central node, the achievable range of power is calculated based on the remaining power and historical energy consumption data per unit distance, and a power constraint shell is constructed. The load passage range is determined based on the load status and UAV load parameters, and a load constraint shell is constructed. Based on the airway openness, airway connectivity, and airspace capacity information in the airspace status, airway nodes that meet the passage conditions are mapped to the airspace topology network to construct an airspace constraint shell; Based on wind speed, wind direction, precipitation, and visibility information in the meteorological conditions, the navigable range is determined and a meteorological constraint shell is constructed. Based on the delivery time limit and the flight duration information in historical flight track data, the time limit reachable range is determined and a time limit constraint shell is constructed. Spatial overlay calculations are performed on the energy constraint shell, load constraint shell, airspace constraint shell, meteorological constraint shell, and time constraint shell. Route nodes and route connections that are simultaneously located within the five types of constraint shells are retained to generate a set of reachable regions. Based on the route nodes and route connection relationships in the reachable area set, a path combination is performed to construct a candidate route set.

4. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step three specifically involves: Obtain each candidate route from the candidate route set, and segment each candidate route according to the connection relationship between adjacent route nodes and the location of route attribute changes to construct multiple route segments; Extract spatial coordinate information, route length information, flight altitude information, and heading information corresponding to each route segment to construct spatial location feature data; Spatial matching is performed based on the spatial coordinates of flight route segments and meteorological monitoring records, airspace operation records, and historical flight track records to extract meteorological status data, airspace status data, historical traffic data, and congestion status data; Numerical normalization, category coding, and time-series coding are performed on spatial location feature data, meteorological status data, airspace status data, historical traffic data, and congestion status data, and then combined according to route segment identifiers to construct route segment fingerprint data. Based on historical flight track records, statistical data were compiled on the changes in the number of drones, take-off and landing frequency, and airway passage frequency in each spatial area; Based on airspace operation records, statistical analysis was conducted on airspace capacity occupancy data for each space area, and based on meteorological monitoring records, statistical analysis was conducted on wind speed variation data for each space area. Spatial mapping is performed on data such as changes in the number of drones, take-off and landing frequency, airway passage frequency, airspace capacity occupancy, and wind speed changes. Pressure accumulation calculation is then performed according to spatial grid units to generate the pressure value distribution corresponding to each spatial grid unit, thus constructing airspace pressure field data.

5. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step four specifically involves: Read route segment fingerprint data, low-altitude logistics event fingerprint data, and airspace pressure field data, and establish a unified time-series index based on timestamp information; The fingerprint data of each airway segment is mapped to airway segment nodes, the fingerprint data of low-altitude logistics events is mapped to logistics event nodes, and the spatial grid cells in the airspace pressure field data are mapped to airspace pressure nodes. Based on the starting position, target position, and route segment connection relationship in the candidate route set corresponding to the logistics event node, establish the association relationship between the logistics event node and the route segment node; The association between route segment nodes and airspace pressure nodes is established based on the spatial location of the route segment nodes and the spatial grid location of the airspace pressure nodes. Based on a unified time-series index, logistics event nodes, air route segment nodes, and airspace pressure nodes are associated in a time sequence to construct a node association set; Using logistics event nodes, route segment nodes, and airspace pressure nodes within the same time period unit as hyperedge connection objects, dynamic hyperedges are constructed based on the node association set, and the time and space identifiers corresponding to each dynamic hyperedge are recorded. A dynamic hypergraph is constructed based on dynamic hyperedges and node relationships.

6. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step five specifically involves: The dynamic hypergraph is input into the improved TDHNN model; The improved TDHNN model incorporates a pressure gradient propagation module, a dual-stream propagation module, a risk breakpoint gating module, and a super-edge blocking module in the dynamic node update structure and dynamic super-edge update structure of the TDHNN model. The pressure gradient propagation module performs pressure gradient propagation processing based on spatial pressure nodes and dynamic hyperedges, generating dynamic hyperedge propagation paths. The dual-stream propagation module constructs forward propagation links and reverse propagation links based on the dynamic hyperedge propagation path, and performs hypergraph message propagation along the forward and reverse propagation links to logistics event nodes, route segment nodes, and airspace pressure nodes, generating forward node states and reverse node states. Perform state fusion processing on the forward node state and the reverse node state to generate node time sequence state; The risk breakpoint gating module performs risk breakpoint identification processing based on the node timing state to generate risk breakpoint nodes; The hyperedge blocking module retrieves dynamic hyperedges connected to the risk breakpoint nodes based on the risk breakpoint nodes, performs connection state analysis and connection relationship update processing on the dynamic hyperedges, and generates an updated dynamic hypergraph. Based on the updated dynamic hypergraph, node state updates and dynamic hyperedge updates are performed to generate risk state values ​​corresponding to each route segment node. Arrange the risk status values ​​according to the connection order of the route segments in the candidate route to generate a route risk breakpoint sequence.

7. The method for recommending low-altitude logistics information based on big data analysis according to claim 6, characterized in that, The pressure gradient propagation process specifically involves: Read the pressure value corresponding to each airspace pressure node, and retrieve the adjacent airspace pressure nodes connected to each airspace pressure node. Calculate the pressure difference between each airspace pressure node and adjacent airspace pressure nodes, and establish a pressure gradient propagation path from high-pressure nodes to low-pressure nodes based on the pressure difference. Map the pressure gradient propagation path to dynamic hyperedges, determine the propagation direction corresponding to each dynamic hyperedge, and count the total pressure difference corresponding to each dynamic hyperedge within the same propagation link. Calculate the proportion of the pressure difference corresponding to each dynamic hyperedge to the total pressure difference, generate the propagation weight, and construct the dynamic hyperedge propagation path based on the propagation direction and the propagation weight.

8. The method for recommending low-altitude logistics information based on big data analysis according to claim 6, characterized in that, The risk breakpoint identification and processing specifically involves: Calculate the node timing state difference between adjacent time period units for each route segment node; The node timing state differences corresponding to each segment node within the same candidate route are statistically analyzed, and the mean and standard deviation of the node timing state differences are calculated. The adaptive gating limit is determined based on the mean of the node timing state difference and the standard deviation of the node timing state difference; Compare the node timing state difference corresponding to each route segment node with the adaptive gating limit; Nodes in a route segment whose timing state difference is greater than the adaptive gating limit are marked as risk breakpoint nodes. The distribution of risk breakpoint nodes in the candidate routes is statistically analyzed, and risk breakpoint identification information is generated based on the risk breakpoint node distribution results; the risk breakpoint identification information is arranged according to the connection order of route segments in the candidate routes to generate a risk breakpoint sequence.

9. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step six specifically involves: Based on the sequence of risk breakpoints in the route, the locations of risk breakpoints in the candidate routes are identified, and the route segments before the breakpoints and the route segments after the breakpoints are divided. Wind speed disturbance, airspace restriction disturbance, take-off and landing point congestion disturbance, power disturbance and delivery time limit disturbance are applied sequentially to the route segments in the candidate routes to construct a set of disturbed route scenarios; The data corresponding to each set of disturbance route scenarios are re-input into the improved TDHNN model to generate a sequence of risk state values ​​corresponding to each disturbance route scenario. Calculate the state difference between the risk state value sequence corresponding to each disturbance route scenario and the original risk state value sequence to generate a risk offset sequence; Calculate the mean and standard deviation of risk offset for each risk offset sequence, and determine the offset limit based on the mean and standard deviation of risk offset; The number of route segments exceeding the offset limit is counted, and the proportion of the number of route segments to the total number of candidate route segments is calculated to generate the disturbance impact ratio. Calculate the route stability index based on the proportion of disturbance impact corresponding to each disturbance route scenario; Calculate the difference sequence between the risk status values ​​corresponding to the set of route segments before the breakpoint and the set of route segments after the breakpoint, and calculate the breakpoint continuity index based on the difference sequence; Route continuity data and counterfactual stability data are generated based on route stability indicators and discontinuity continuity indicators.

10. The method for recommending low-altitude logistics information based on big data analysis according to claim 1, characterized in that, Step seven specifically involves: Read the route continuity data and counterfactual stability data corresponding to the candidate routes; Calculate the continuity score, stability score, and risk score for each candidate route; The continuity score, stability score, and risk score are normalized to generate continuity evaluation value, stability evaluation value, and risk evaluation value. A candidate route evaluation sequence is constructed based on continuity evaluation values, stability evaluation values, and risk evaluation values; The candidate route evaluation sequence is sorted to generate the candidate route ranking result; Extract the risk breakpoint sequence corresponding to the top N candidate routes from the candidate route ranking results; Statistically analyze the number of risk breakpoints and their distribution locations for each risk breakpoint sequence. The first N candidate routes are screened based on the number and location of risk breakpoints to determine the target route. Read the route segment information, risk breakpoint sequence and logistics task information corresponding to the target route, and generate low-altitude logistics information recommendation results.

Images