Unmanned aerial vehicle deck platform laying path planning system and method based on knowledge graph

Abstract

The invention discloses an unmanned aerial vehicle deck platform laying path planning system and method based on a knowledge graph, and relates to the technical field of unmanned aerial vehicle laying path planning, and the method comprises the following steps: obtaining ship navigation data and storm state data of a target object; and carrying out stability calculation on the ship navigation data according to a sliding window, judging whether a stability calculation result is greater than a preset judgment threshold value or not, and if so, carrying out constraint curve fitting on the ship navigation data in the corresponding sliding window according to the wind wave state data to generate a route prediction sequence. According to the scheme, stability calculation is carried out on ship navigation data according to a sliding window, and constraint curve fitting is carried out in combination with wind wave state data, so that unmanned aerial vehicle laying can dynamically adapt to navigation state changes, a local wind wave trend field is generated based on the wind wave data, and a take-off and landing time window set is adaptively extracted; and the laying time and position are strictly constrained by the environment.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) deployment path planning technology, specifically to a UAV deck platform deployment path planning system and method based on knowledge graphs. Background Technology

[0002] In modern maritime operations, drones, as flexible and efficient aerial work platforms, are widely used in tasks such as ship monitoring, cargo placement, and rescue delivery. The deployment efficiency and operational safety of drones largely depend on the deployment path planning of the deck platform. The core issue is how to combine the ship's motion state and environmental factors to formulate a reasonable take-off and landing sequence and route. Ships are greatly affected by wind and waves during navigation, especially under high sea state conditions, where the ship's speed and course are constantly changing, which makes the take-off and landing and path planning of drones face a high degree of uncertainty.

[0003] Existing technologies typically use statistical sea state risk assessment methods to assist in UAV take-off and landing decisions. For example, in existing solutions, future course and speed are predicted by fitting ship navigation data to a time series, thereby determining the UAV's operating window. However, ship trajectories may exhibit alternating stable and unstable segments in a short period. When ships exhibit unstable behaviors such as frequent turns, using time series fitting will lead to a rapid accumulation of prediction errors, resulting in a deviation between the predicted UAV take-off and landing points and the determined time. Furthermore, when using fixed thresholds to filter take-off and landing periods based on wind and wave conditions, the temporal characteristics of wind and waves are ignored. More importantly, the interaction between the uncertainty of trajectory prediction and the temporal characteristics of wind and waves will amplify the defects of either situation. The resulting double error will make it difficult for the path planning scheme to guarantee deployment safety. Summary of the Invention

[0004] The purpose of this invention is to provide a knowledge graph-based deployment path planning system and method for unmanned aerial vehicle (UAV) deck platforms to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention discloses a method for planning the deployment path of an unmanned aerial vehicle (UAV) deck platform based on a knowledge graph, comprising the following steps:

[0007] Acquire the target object's ship navigation data and wind and wave status data;

[0008] The stability calculation of the ship navigation data is performed using a sliding window. If the stability calculation result is greater than a preset threshold, then the ship navigation data in the corresponding sliding window is fitted with a constraint curve based on the wind and wave state data to generate a route prediction sequence.

[0009] Otherwise, statistical modeling is performed on the ship navigation data within the corresponding sliding window to generate a heading and turning probability matrix;

[0010] The wind and wave state data is decomposed into a time series using a sliding window to generate a local wind and wave trend field. The set of takeoff and landing time windows is then extracted from each sliding window based on an adaptive threshold. The adaptive threshold is obtained by performing quantile statistics on the wind and wave state data within the corresponding sliding window.

[0011] A temporary candidate set is extracted from the route prediction sequence based on the set of takeoff and landing time windows, and a deployment priority value is obtained by weighting the temporary candidate set according to the heading turning probability matrix;

[0012] Based on the deployment priority value, the temporary candidate set is locally de-conflicted to generate a path planning scheme.

[0013] Secondly, this invention discloses a knowledge graph-based deployment path planning system for unmanned aerial vehicle (UAV) deck platforms, comprising:

[0014] The data acquisition module is used to acquire the ship's navigation data and wind and wave status data of the target object;

[0015] The navigation data analysis module is used to perform stability calculations on the ship navigation data in a sliding window, determine whether the stability calculation result is greater than a preset discrimination threshold, and if so, perform constraint curve fitting on the ship navigation data in the corresponding sliding window based on the wind and wave state data to generate a route prediction sequence.

[0016] Otherwise, statistical modeling is performed on the ship navigation data within the corresponding sliding window to generate a heading and turning probability matrix;

[0017] The take-off and landing time extraction module is used to perform time-series decomposition on the wind and wave state data according to a sliding window, generate a local wind and wave trend field, and extract the take-off and landing time window set from each sliding window according to an adaptive threshold.

[0018] The priority calculation module is used to extract a temporary candidate set from the route prediction sequence based on the set of take-off and landing time windows, and to calculate the deployment priority value of the temporary candidate set by weighting the result based on the heading turning probability matrix.

[0019] The path planning module is used to perform local conflict resolution on the temporary candidate set based on the deployment priority value and generate a path planning scheme.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] 1. This solution performs stability calculations on ship navigation data using a sliding window and combines wind and wave state data with constraint curve fitting, enabling UAV deployment to dynamically adapt to changes in navigation conditions. Based on the local wind and wave trend field generated from wind and wave data and the adaptively extracted set of take-off and landing time windows, the deployment time and location are strictly constrained by the environment, ensuring the safe intervals and reasonable spatial distribution of UAV take-off and landing. This significantly improves the safety, feasibility, and overall operational efficiency of UAV deck platform deployment under dynamic and complex sea conditions.

[0022] 2. This scheme can quantify the specific impact of wind and wave conditions on UAV flight energy consumption by accurately registering the route prediction sequence with the local wind and wave trend during the candidate path planning stage. By accumulating the cumulative energy consumption of each candidate orientation by distance integration, it can accurately reflect the energy consumption trend from the starting point to the candidate node, thereby significantly improving the accuracy, reliability and overall energy efficiency of UAV deck platform deployment path planning. Attached Figure Description

[0023] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0024] Figure 1 This is a flowchart illustrating the steps of the knowledge graph-based unmanned aerial vehicle (UAV) deck platform deployment path planning method of the present invention.

[0025] Figure 2 A schematic diagram of the process for generating a temporary candidate set provided by the present invention;

[0026] Figure 3 This is a schematic diagram of the process for generating a timing conflict matrix provided by the present invention;

[0027] Figure 4 A schematic diagram of the process for generating the deployment sequence provided by the present invention;

[0028] Figure 5 This is a schematic diagram of the module functions of the knowledge graph-based unmanned aerial vehicle (UAV) deck platform deployment path planning system provided by the present invention. Detailed Implementation

[0029] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0030] Application Overview:

[0031] During ship navigation, the ship's speed and course are dynamically changed due to the influence of wind and waves. The navigation trajectory is identified as having alternating stable and unstable segments. When the ship performs unstable behaviors such as frequent turns, time series fitting methods are applied, leading to the accumulation of prediction errors. At the same time, the assessment of wind and wave conditions uses a fixed threshold to screen take-off and landing periods, ignoring the temporal characteristics of wind and waves. The interaction between the two is amplified, causing deviations in the determination of UAV take-off and landing points and times based on this prediction, which affects the reliability of the path planning scheme and the safety of deployment.

[0032] For example, when ships are carrying out cargo deployment missions at sea, they frequently change course to maintain navigation stability due to wind and waves. Existing methods perform time series fitting on ship navigation data, which increases the deviation between the predicted results and the actual trajectory. Wind and wave status data are used to determine take-off and landing periods with fixed thresholds. Short-term fluctuations in wind and waves can lead to misjudgments of the operational window, increasing the risk of inappropriate UAV take-off and landing timing and raising the possibility of mismatch between deployment paths and the actual movement of the ship.

[0033] If the above problems are not addressed, the route planning scheme will be unable to adapt to the dynamic changes in the ship's navigation status and the temporal characteristics of the wind and wave environment. Prediction errors will continue to accumulate, safety risks will increase, the possibility of equipment operation failure or mission interruption will be enhanced, and the efficiency and reliability of offshore operations will be weakened.

[0034] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0035] Example 1:

[0036] Please see Figure 1 A knowledge graph-based method for planning the deployment path of unmanned aerial vehicle (UAV) deck platforms includes the following steps:

[0037] Acquire the target object's ship navigation data and wind and wave status data;

[0038] Stability calculations are performed on ship navigation data using a sliding window. If the stability calculation result is greater than a preset threshold, constraint curve fitting is performed on the ship navigation data within the corresponding sliding window based on wind and wave state data to generate a route prediction sequence.

[0039] Otherwise, statistical modeling is performed on the ship navigation data within the corresponding sliding window to generate a heading and turning probability matrix;

[0040] The wind and wave state data is decomposed into a time series using a sliding window to generate a local wind and wave trend field. The set of take-off and landing time windows is then extracted from each sliding window based on an adaptive threshold. The adaptive threshold is obtained by performing quantile statistics on the wind and wave state data within the corresponding sliding window.

[0041] A temporary candidate set is extracted from the route prediction sequence based on the set of takeoff and landing time windows, and the deployment priority value is obtained by weighting the temporary candidate set according to the heading turning probability matrix.

[0042] Based on the deployment priority value, the temporary candidate set is locally de-conflicted to generate a path planning scheme.

[0043] Among them, ship navigation data refers to a set of structured time-series data used to describe the motion state of a target object at sea;

[0044] Wave and wind condition data refers to a set of structured time series data used to describe the wave and wind conditions of the water body surrounding the target object;

[0045] Stability calculation refers to the data processing procedure used to quantify the stability and predictability of a ship's motion state within a certain time window;

[0046] The preset discrimination threshold refers to the numerical limit used to quantify whether the stability index of navigation data meets the constraint curve fitting. It is obtained by performing stability calculations on historical ship navigation data within a specified sliding window and performing distribution statistics on the stability calculation results, based on the quantiles (such as the 90th percentile) of the distribution statistics.

[0047] Constrained curve fitting refers to the data processing process of mathematically modeling ship navigation data over a specific time period using wind and wave state data;

[0048] Route prediction sequence refers to the spatiotemporal sequence of future navigation trajectories obtained through calculation and modeling based on ship navigation data and wind and wave state data;

[0049] Statistical modeling refers to the data processing process of performing probability distribution analysis and parametric modeling on ship navigation data;

[0050] The heading and turning probability matrix is ​​a structured description of the heading change behavior of a target object within a specific time window;

[0051] Temporal decomposition refers to the data processing process of extracting trend and periodic features from continuous wind and wave state data according to time sequence;

[0052] Local wind and wave trend field refers to the structured data representation that models and quantifies the dynamic changes of factors such as wind speed, wave height, and wave direction in the ship's operating environment within a specific time window;

[0053] Adaptive threshold refers to the numerical limit used to determine whether the wind and wave conditions within a certain time period meet the requirements for safe take-off and landing of drones.

[0054] The take-off and landing time window set refers to the set of time intervals suitable for UAVs to take off and land on deck platforms, extracted from the local wind and wave trend field.

[0055] The temporary candidate set refers to the set of candidate nodes that can be used as UAV deployment target nodes, selected from the flight path prediction sequence based on the set of take-off and landing time windows.

[0056] Deployment priority value refers to a numerical indicator used to measure the priority and selectivity of each candidate UAV deployment node in path planning;

[0057] Local conflict resolution refers to the process of selecting a set of non-conflicting candidate nodes to form a preliminary deployment plan by analyzing time, space and risk constraints during the initial sorting or generation of candidate nodes.

[0058] Path planning refers to the specific spatiotemporal arrangement and operational sequence of multiple UAVs performing deployment tasks on a deck platform.

[0059] This scheme effectively identifies stable and unstable navigation intervals through stability calculation and threshold judgment using a sliding window, ensuring that the route prediction sequence is generated only within reliable navigation intervals, thus improving prediction accuracy. In unstable intervals, it models the heading and turning probability matrix to provide statistical risk references for deployment decisions, thereby reducing deployment conflicts and unexpected risks caused by sudden heading changes. By decomposing wind and wave data over time to generate a local wind and wave trend field, and combining this with adaptive thresholds to extract takeoff and landing time windows, it enables dynamic adaptation of UAV takeoff and landing arrangements to environmental changes, ensuring operational safety and feasibility. By combining the route prediction sequence and the heading and turning probability matrix to perform weighted calculations on the temporary candidate set, it ensures that high-confidence, low-risk candidate nodes are selected first. A path planning scheme is generated through a local conflict resolution strategy, which not only optimizes the coordination of parallel deployment of multiple UAVs but also ensures that the selection of each node is traceable through structured data processing. This significantly improves the success rate of UAV deployment, operational safety, and overall mission efficiency, while providing an operational data foundation for subsequent real-time adjustments and dynamic optimization.

[0060] The above describes a complete scheme for the deployment path planning method of UAV deck platforms based on knowledge graphs. The following section describes how to obtain the target object's ship navigation data and wind and wave status data, specifically including:

[0061] Ship navigation data of the target object is acquired through ship sensors and navigation equipment; ship navigation data includes, but is not limited to, position coordinates, heading angle, speed, acceleration and timestamp, etc.

[0062] The wind and wave status data of the target object are obtained through meteorological sensors; the wind and wave status data includes, but is not limited to, wind speed, wind direction, wave height, and wave direction;

[0063] Stability calculations for ship navigation data are performed using a sliding window method, and the specific calculation formula is as follows:

[0064] ;

[0065] In the formula, Represents a sliding window Stability calculation results of domestic ship navigation data, Represents a sliding window The total number of data points for domestic ship navigation data. Represents a sliding window Inner Individual ship navigation data, Represents a sliding window The average of the internal ship navigation data; all of the above data have been normalized during the calculation.

[0066] If the stability calculation result is greater than the preset discrimination threshold, the wind and wave state data and ship navigation data in the corresponding sliding window are aligned according to the time series, and the ship navigation data in the sliding window are fitted based on the constraint curve fitting method (such as weighted least squares method) to generate a continuous route prediction sequence.

[0067] Otherwise, the ship navigation data within the corresponding sliding window is statistically organized according to time series to establish a set of heading and turning events. Based on statistical methods (such as frequency statistics, histogram distribution, or Markov chain modeling), the turning probability of each heading state to the adjacent possible heading state is calculated. At the same time, the turning probability is weighted and corrected by combining the ship's speed and position information at the corresponding time points to generate a heading and turning probability matrix. The specific calculation formula is as follows:

[0068] ;

[0069] In the formula, Represents the heading state in the heading change probability matrix. Turn to heading status The probability of heading change, This represents the heading status obtained from statistics within the sliding window. Turn to heading status Number of events This represents the heading status obtained from statistics within the sliding window. Turn to heading status The number of events, Indicates the time when the turning event occurs. Instantaneous speed, This indicates a reference speed value (e.g., the average speed of a sliding window). Indicates the time when the turning event occurs. The spatial distance between the location of occurrence and the reference location, Indicates a reference distance (e.g., the maximum distance offset within a sliding window). and These represent the corresponding weighted correction factors. All the above data have been normalized during the calculation.

[0070] The heading and turning probability matrix has rows representing the current heading state, columns representing turning states, and matrix elements representing the corresponding heading and turning probabilities.

[0071] The wind and wave status data is read in chronological order and sliced ​​using a sliding window with a fixed length and a fixed step size. Then, within each sliding window, the multi-dimensional wind and wave status data such as wind speed and wave height are subjected to time-series decomposition processing (including but not limited to local trend term extraction, fast disturbance term separation, and window stationarity reconstruction) to generate the corresponding local wind and wave trend field.

[0072] Within the same sliding window, quantile statistics (e.g., 75th percentile) are performed on all wind and wave status data to obtain the adaptive threshold of the sliding window. This adaptive threshold is then compared point by point with the local wind and wave trend field on the same time axis. Continuous time periods where the local wind and wave trend field is lower than the adaptive threshold are selected as candidate take-off and landing intervals. Interval merging is performed on these candidate take-off and landing intervals (adjacent candidate take-off and landing intervals with a time interval of less than 2 seconds are merged) to generate take-off and landing time windows. The take-off and landing time windows of each sliding window are integrated to generate a set of take-off and landing time windows covering all sliding windows.

[0073] The above describes how to obtain the ship's navigation data and wind and wave status data for the target object. The following section describes how, after extracting the takeoff and landing time window set, the process also includes generating an energy consumption matrix, specifically:

[0074] Spatiotemporal registration is performed on the route prediction sequence and the local wind and wave trend field, and the spatiotemporal registration results are discretized to generate candidate azimuths.

[0075] For each candidate bearing, the cumulative energy consumption value of the corresponding candidate bearing is generated by accumulating the distance integral based on the wind-navigation coupling factor, and the cumulative energy consumption values ​​of each candidate bearing are combined to generate an energy consumption matrix.

[0076] The wind-navigation coupling factor is obtained by mapping the flight path prediction sequence of the corresponding candidate orientation.

[0077] Spatiotemporal registration refers to the data processing procedure of aligning spatial-temporal data from different sources or at different times to the same reference coordinate system and unified time base.

[0078] Azimuth discretization refers to the data processing process that converts the continuous heading information obtained by spatiotemporal registration of the route prediction sequence and the local wind and wave trend field into a finite set of candidate headings according to predefined heading segmentation or discretization rules.

[0079] Candidate orientation refers to the flight direction or heading angle that each candidate node can choose within the take-off and landing time window during the deployment of UAVs;

[0080] The wind-heading coupling factor refers to the degree of influence of the interaction between wind speed, wind direction and the current heading of the UAV on flight energy consumption at a given time and spatial location.

[0081] Distance integral accumulation refers to the data processing process used to quantify and accumulate the energy consumption of each candidate orientation along its corresponding flight path.

[0082] Cumulative energy consumption value refers to a numerical indicator that quantifies the energy consumption of each candidate orientation on the planned flight path.

[0083] An energy consumption matrix is ​​a structured data used to describe the energy consumption of each candidate location under different time and space conditions.

[0084] The above content will be described in detail below:

[0085] Read the flight path prediction sequence and the local wind and wave trend field, and align and match the flight path prediction sequence and the local wind and wave trend field with timestamps to achieve temporal registration. At the same time, perform nearest neighbor or interpolation registration on the flight path prediction sequence and the local wind and wave trend field based on spatial coordinates to complete spatial alignment, thereby generating spatiotemporal registration results.

[0086] For each route prediction point in the spatiotemporal registration result corresponding to the route prediction sequence, the continuous azimuth is divided into discrete angle intervals according to the heading angle and the route segment direction (e.g., discrete once every 5° or 10°), and each discrete angle interval is used as a candidate azimuth.

[0087] For each candidate bearing, the cumulative energy consumption value for that candidate bearing is generated by accumulating the distance integral based on the wind-navigation coupling factor. An energy consumption matrix is ​​generated by combining the cumulative energy consumption values ​​of each candidate orientation. The specific calculation formula is as follows:

[0088] ;

[0089] In the formula, Indicates candidate orientation The total number of route prediction points, Indicates candidate orientation At the route prediction point The wind-navigation coupling factor Indicates the route prediction point The increase in sailing distance; all the above data have been normalized during the calculation.

[0090] The wind-navigation coupling factor is obtained by mapping the flight path prediction sequence of the corresponding candidate azimuth, and its specific calculation formula is as follows:

[0091] ;

[0092] In the formula, This represents the function representing the impact of wind environment on energy consumption. Indicates the route prediction point wind speed, Indicates the route prediction point The angle between the wind direction and the heading. Indicates the route prediction point The waves are high. This represents a function that describes the influence of ship parameters on energy consumption. Indicates the ship's speed. The above data, which indicates the ship's heading, has been normalized during calculation.

[0093] This solution achieves precise quantification and traceable management of the energy consumption of UAVs on the deck platform deployment path by performing spatiotemporal registration of the flight path prediction sequence with the local wind and wave trend field and generating discrete candidate azimuths. It also calculates the cumulative energy consumption of each candidate azimuth by combining the wind-navigation coupling factor and constructing an energy consumption matrix. This enables the early assessment of the energy consumption cost of different candidate azimuths and time points during the path planning stage, providing a reliable quantitative basis for the parallel deployment of multiple UAVs, thereby significantly improving the efficiency and feasibility of deployment path optimization and reducing operational risks.

[0094] The above describes the process of extracting the takeoff and landing time window set, which also includes generating an energy consumption matrix. The following describes how to extract a temporary candidate set from the flight path prediction sequence based on the takeoff and landing time window set. Please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of the process for generating a temporary candidate set provided in an embodiment of this application. Generating a temporary candidate set specifically includes:

[0095] The predicted flight times within the flight prediction sequence are compared with the set of take-off and landing time windows at each time point. The take-off and landing time windows containing the predicted flight times are extracted, and the difference between the take-off and landing time windows and the corresponding predicted flight times is calculated.

[0096] If the element of the corresponding route prediction time in the energy consumption matrix is ​​greater than the preset energy consumption threshold of the corresponding take-off and landing time window, then if the difference calculation result of the corresponding route prediction time is greater than the preset take-off and landing preparation threshold of the corresponding take-off and landing time window, then integrate the corresponding route prediction times to generate a temporary candidate set.

[0097] The preset energy consumption threshold is obtained by performing quantile statistics on the elements of the energy consumption matrix contained in the corresponding take-off and landing time window.

[0098] Among them, the preset energy consumption threshold refers to the numerical limit used to screen whether the predicted time of the flight route can be used as a temporary candidate point.

[0099] The difference calculation result refers to the time interval obtained by quantitatively comparing the predicted flight route time with the end time of the corresponding take-off and landing time window.

[0100] The preset takeoff and landing preparation threshold is a numerical limit used to determine whether a predicted time for a certain route can be included in the temporary candidate set. It is calculated by statistically analyzing the time difference distribution of actual takeoff and landing operations in the historical data within each takeoff and landing time window, and calculating the minimum safe preparation time for each takeoff and landing time window according to the statistical quantile (e.g., 95th percentile), and using this quantile value as the preset takeoff and landing preparation threshold.

[0101] The above content will be described in detail below:

[0102] The predicted flight times within the flight prediction sequence are compared with the set of take-off and landing time windows at each time point. Take-off and landing time windows that include the flight prediction times or are reachable from the flight prediction times are extracted. The difference between the end time of the take-off and landing time window and the corresponding flight prediction time is calculated to obtain the remaining length of the take-off and landing time window. The remaining length of the window is used as the result of the difference calculation.

[0103] If the element of the corresponding route prediction time in the energy consumption matrix is ​​greater than the preset energy consumption threshold of the corresponding take-off and landing time window, then it is further determined whether the difference calculation result of the corresponding route prediction time is greater than the preset take-off and landing preparation threshold of the corresponding take-off and landing time window. If so, the corresponding route prediction times are integrated to generate a temporary candidate set.

[0104] The preset energy consumption threshold is obtained by performing quantile statistics (e.g., the 90th or 95th percentile) on the elements of the energy consumption matrix contained in the corresponding take-off and landing time window.

[0105] This solution accurately filters route nodes that match the actual available time windows by comparing the predicted route times with the take-off and landing time windows point by point in time. This ensures that each candidate route time is physically and operationally feasible, avoiding take-off and landing failures or conflicts caused by timing mismatches. By judging the energy consumption matrix elements corresponding to the predicted route times with the preset energy consumption thresholds within the window, high-energy-consuming infeasible routes are effectively eliminated, ensuring that the deployment plan is optimized in terms of energy consumption. At the same time, the threshold is adaptively set using quantile statistics to take into account the energy differences of different windows, achieving dynamic adjustment and high adaptability. By comparing the difference between the predicted route times and the take-off and landing windows with the preset take-off and landing preparation thresholds, sufficient operational buffer time is ensured for each candidate route, improving the safety and reliability of UAVs under conditions of wind and wave changes or complex deck operations. This makes the overall path planning both energy-efficient and safe in the parallel deployment of multiple UAVs, significantly improving the mission success rate, executability, and overall scheduling efficiency.

[0106] The above describes the extraction of a temporary candidate set from the flight path prediction sequence based on the set of takeoff and landing time windows. The following describes the weighted calculation of the deployment priority value from the temporary candidate set using the heading change probability matrix, specifically including:

[0107] The elements of the energy consumption matrix, the difference calculation results, and the heading and turning probability matrix contained in the take-off and landing time window are weighted and calculated to obtain the deployment priority value of each route prediction time in the temporary candidate set.

[0108] The weights in the weighted calculation process are obtained by evaluating the confidence levels of the route prediction sequence, energy consumption matrix, and takeoff and landing time window set, respectively.

[0109] Among them, confidence assessment refers to the data processing process used to quantify the reliability and credibility of each type of input data in the calculation of deployment priority value.

[0110] The above content will be described in detail below:

[0111] We perform a weighted calculation on the elements of the energy consumption matrix, the difference calculation results, and the heading and turning probability matrix included in the takeoff and landing time window to obtain the deployment priority value of each route's predicted time within the temporary candidate set. The specific calculation formula is as follows:

[0112] ;

[0113] In the formula, Indicates the predicted time of the flight path The corresponding elements of the energy consumption matrix, This represents the maximum value among the elements of the energy consumption matrix corresponding to the predicted time of each flight route within the temporary candidate set. Indicates the predicted time of the flight path The corresponding difference calculation results, This represents the maximum value of the difference calculation results corresponding to the predicted times of each route within the temporary candidate set. Indicates the predicted time of the flight path The elements of the corresponding heading and turning probability matrix, , and These represent the corresponding weighting factors. All the data above have been normalized during the calculation.

[0114] The weights in the weighted calculation process are obtained by evaluating the confidence levels of the route prediction sequence, energy consumption matrix, and takeoff and landing time window set, respectively. The specific calculation formula is as follows:

[0115] ;

[0116] ;

[0117] ;

[0118] In the formula, This represents the confidence assessment result corresponding to the energy consumption matrix. This indicates the confidence level assessment result corresponding to the difference calculation result. This represents the confidence assessment result corresponding to the heading and turning probability matrix. All the above data have been normalized during the calculation.

[0119] Taking the calculation of the confidence assessment result corresponding to the energy consumption matrix as an example, the specific calculation formula is as follows:

[0120] ;

[0121] In the formula, This represents the standard deviation of the elements of the energy consumption matrix corresponding to the predicted time of each flight route within the temporary candidate set. This represents the mean of the elements of the energy consumption matrix corresponding to the predicted time for each flight route within the temporary candidate set. The above data has been normalized during the calculations, indicating positive numbers.

[0122] The same method was used to calculate the confidence assessment results corresponding to the difference calculation results and the confidence assessment results corresponding to the heading and turning probability matrix.

[0123] This solution uses a confidence-weighted approach to automatically adapt deployment priorities to the uncertainties and accuracy differences of various input data, reducing deployment conflicts or route infeasibility risks caused by data bias. By combining the heading and turning probability matrix with energy consumption differences, it can prioritize the selection of spatiotemporal points with low heading risk while maintaining energy consumption optimization, thus improving overall mission safety. The feasibility of the time window is evaluated using the confidence of the take-off and landing time window set, enabling the deployment strategy to dynamically respond to constraints such as sea waves, deck space, and UAV take-off and landing capabilities. This improves the executability and real-time adaptability of multi-UAV deployment schemes, significantly enhancing the safety, energy efficiency, and execution reliability of multi-target deployment planning on UAV deck platforms, while ensuring data transparency and traceability in the decision-making process.

[0124] The above describes how to calculate the deployment priority value from the temporary candidate set using a weighted average of the heading and turning probability matrix. The following describes how to perform local conflict resolution on the temporary candidate set based on the deployment priority value to generate a path planning scheme, specifically including:

[0125] Conflict identification is performed on the temporary candidate set based on deployment priority and take-off and landing time window to generate a temporal conflict matrix;

[0126] Local rearrangement optimization is performed on the temporal conflict matrix and deployment priority values ​​to generate a deployment sequence;

[0127] A path score matrix is ​​generated by constructing entity nodes and weighted relation edges and calculating path cumulative indicators for the route prediction sequence, deployment sequence and take-off and landing time window set;

[0128] Perform path consistency verification on the path scoring matrix and the temporal conflict matrix to generate a path planning scheme.

[0129] Among them, conflict identification refers to the data processing process of assessing the temporal and spatial feasibility of any two nodes in the temporary candidate set;

[0130] The temporal conflict matrix is ​​a data structure used to characterize the potential temporal and spatial conflict relationships between candidate nodes.

[0131] Local rearrangement optimization refers to the process of adjusting the order of candidates or replacing them, under the premise of satisfying time-space constraints and conflict restrictions, to reduce the overall conflict cost, optimize energy consumption distribution, and improve path feasibility and confidence, based on the initially generated deployment candidate sequence (i.e., temporary candidate set).

[0132] A deployment sequence refers to an ordered set of candidate nodes generated according to priority and conflict constraints to enable simultaneous take-off and landing of multiple UAVs and task allocation.

[0133] The path accumulation index is a numerical indicator used to quantitatively evaluate the deployment path of drones.

[0134] The path scoring matrix refers to structured data used to quantify the comprehensive evaluation of feasible paths from the deck to each candidate node;

[0135] Path consistency verification refers to the data processing process of systematically checking the temporal and spatial states of all nodes in each candidate path scheme.

[0136] The above content will be described in detail below:

[0137] The initial deployment sequence is formed by selecting the top Z flight path prediction times from the temporary candidate set based on the deployment priority value; where Z represents a positive integer between 10 and 15.

[0138] For each predicted flight path in the initial deployment sequence, based on the take-off and landing time window in which it is located, the predicted flight path times included in that take-off and landing time window are processed by the difference between each pair of adjacent predicted flight path times, and the minimum value among them is taken as the safe take-off and landing interval for that take-off and landing time window.

[0139] Calculate the difference between the predicted times of any two routes in the initial deployment sequence, and determine whether the difference is greater than the corresponding safe take-off and landing interval. If so, it is determined that there is a conflict between the predicted times of the two routes and recorded as 1.

[0140] Otherwise, it is determined that there is no conflict between the predicted times of the two routes and they are recorded as 0, and then integrated to generate a time-series conflict matrix;

[0141] In determining whether the difference is greater than the corresponding safe take-off and landing interval, when the take-off and landing time windows of the predicted times of the two routes are different, that is, the predicted times of the two routes each correspond to a safe take-off and landing interval, the larger of the two safe take-off and landing intervals is taken as the corresponding safe take-off and landing interval in the judgment process.

[0142] The confidence assessment results of the temporal conflict matrix, safe takeoff and landing interval, and energy consumption matrix are weighted to obtain the cumulative conflict cost of the corresponding route prediction time within the initial deployment sequence. The specific calculation formula is as follows:

[0143] ;

[0144] In the formula, This represents the set of predicted flight path times within the initial deployment sequence. Indicates the predicted time of the flight path The corresponding safe takeoff and landing interval, Represents the flight path prediction time in the time-series conflict matrix and flight path prediction time elements, , and These represent the corresponding weighting coefficients. All the data above have been normalized during the calculation.

[0145] For each flight path prediction time in the initial deployment sequence, perform pairwise enumeration and swapping:

[0146] For any two predicted flight paths in the initial deployment sequence, the positions are swapped. This generates an enumerated initial deployment sequence after the swaps. The cumulative conflict cost of the corresponding two predicted flight paths in the enumerated initial deployment sequence after the swaps is recalculated. The difference in cumulative conflict cost is then calculated. The cumulative energy consumption of the corresponding two predicted flight paths in the enumerated initial deployment sequence after the swaps is also recalculated. The cumulative energy consumption change is then calculated. The energy consumption change benefit of the current enumerated initial deployment sequence after the swaps is obtained by subtracting a certain multiple (e.g., 0.5 times) of the cumulative conflict cost difference from the cumulative energy consumption change.

[0147] The initial deployment sequence after enumeration and swapping that maximizes the benefit of energy consumption change is selected as the deployment sequence.

[0148] Each flight path prediction time is written as a knowledge graph node into the entity table. At the same time, all attribute fields in the flight path prediction sequence, deployment sequence, and corresponding take-off and landing time window set are fully mapped to the knowledge graph node. An edge set is constructed on the knowledge graph node set. The edge generation is based on the difference analysis between node attributes, including but not limited to time interval, azimuth deviation, energy consumption difference, and window matching difference. The edge weight is calculated based on these differences. The edge weight is calculated by normalizing the difference analysis results corresponding to the knowledge graph node and then weighting the combination, while also adjusting it based on the confidence evaluation results.

[0149] Starting from each candidate location, a path search is performed on the knowledge graph in conjunction with the flight route prediction sequence to generate a path. During the search process, edge weight constraints and node attribute conditions are strictly followed, including but not limited to time adjacency, location continuity and energy consumption advantages and disadvantages.

[0150] Temporal adjacency refers to the requirement that the time of consecutive knowledge graph nodes on the path must satisfy the order and operability. That is, the route prediction time of the subsequent knowledge graph node must not be earlier than the previous knowledge graph node, and the time interval between adjacent knowledge graph nodes must be greater than or equal to the safe take-off and landing interval of their respective take-off and landing time windows to prevent take-off and landing conflicts.

[0151] Orientation continuity refers to the fact that the difference in orientation change of continuous knowledge graph nodes on the path should be less than a preset constraint threshold, so as to ensure that the UAV does not have excessive or uncontrollable angle changes when turning, thereby ensuring flight feasibility and stability.

[0152] Energy efficiency constraint refers to the process of comparing the cumulative energy consumption values ​​of adjacent knowledge graph nodes during path generation, and tending to select knowledge graph nodes with lower cumulative energy consumption values ​​in order to optimize overall navigation energy efficiency.

[0153] The path is comprehensively evaluated based on the energy consumption matrix, the local wind and wave trend field, and the heading change probability matrix to generate a path score matrix. The specific calculation formula is as follows:

[0154] ;

[0155] ;

[0156] In the formula, Represents the path in the path rating matrix The comprehensive evaluation results Representing a path Knowledge Graph Nodes The corresponding element in the energy consumption matrix, Representing a path The maximum value of the corresponding element in the energy consumption matrix for the knowledge graph node. Representing a path Knowledge Graph Nodes The risk value, Representing a path The knowledge graph node corresponds to the maximum value among the risk values. and These represent the corresponding weighting factors. Knowledge graph nodes representing local wind and wave trend fields The confidence value, Represents knowledge graph nodes in the heading and turning probability matrix and knowledge graph nodes Corresponding elements between them and These represent the corresponding weighting coefficients. All the data above have been normalized during the calculation.

[0157] The search nodes are constructed based on the predicted time of each flight path in the deployment sequence and the path in the corresponding path score matrix, thus forming a complete search space;

[0158] When performing a constrained search within the search space, each expansion of a search node is feasibility determined by the take-off and landing time window of the predicted flight path. Specifically, the start and end times of the take-off and landing time window corresponding to the predicted flight path are read, and it is verified whether the time index of the node to be expanded falls within the allowed start and end times. If so, the elements related to the node to be expanded are read from the temporal conflict matrix and averaged. If the averaged result is less than the preset conflict threshold, it is determined to be an expanded node. The expanded node and the search node are then integrated to generate constrained search results.

[0159] Using the constrained search results as keys, the corresponding values ​​of the route prediction sequence and the local wind and wave trend field are obtained, and spatiotemporal consistency calculations are performed on each constrained search result: including but not limited to comparing whether the deviation between the predicted heading of each node in the constrained search results and the wind direction in the wind and wave trend field exceeds the preset value, whether the predicted speed change matches the local wave height gradient, whether the time interval between nodes is consistent with the time step of the route prediction sequence, and whether the node displacement is continuous with the predicted displacement field. If any indicator is not satisfied, the node is marked as inconsistent and removed, thereby generating the path after node filtering, and using it as the constrained search result after spatiotemporal consistency verification.

[0160] The constrained search results, after being verified for spatiotemporal consistency, are written into a structured combination template in chronological order along with the corresponding route prediction times in the deployment sequence to generate a complete route planning scheme data package.

[0161] This scheme identifies conflicts in a temporary candidate set based on deployment priority values ​​and takeoff / landing time windows, generating a temporal conflict matrix. This allows for early identification of potential temporal and spatial conflicts between candidate nodes, ensuring subsequent path optimization is performed within a controllable range, thus reducing conflict rates and mission failure risks. By locally rearranging the temporal conflict matrix and deployment priority values ​​to generate a deployment sequence, the scheme effectively adjusts the task order and deployment node timing while ensuring high-priority tasks are executed, optimizing overall deployment efficiency. It also considers the coordination of parallel UAV operations. This is achieved by constructing a set of flight path prediction sequences, deployment sequences, and takeoff / landing time windows. The system generates weighted relation edges by constructing body nodes and calculates path cumulative indicators to generate a path scoring matrix. This enables a comprehensive quantification of energy consumption, risk, time adaptability, and confidence for each candidate path, providing a precise basis for path selection. By verifying the path consistency of the path scoring matrix and the time-series conflict matrix, the system verifies the time window, conflict weight, flight path safety, and environmental risks node by node, ensuring that the final path planning scheme is feasible under all operational constraints. This significantly improves the safety, reliability, and traceability of the deployment scheme, while reducing the risks of UAV collisions, energy waste, and time delays during mission execution, achieving efficient closed-loop control for multi-UAV collaborative deployment missions.

[0162] The above describes how to locally resolve conflicts in a temporary candidate set based on deployment priority to generate a path planning scheme. The following describes how to identify conflicts in the temporary candidate set based on deployment priority and takeoff / landing time windows to generate a temporal conflict matrix. Please refer to [link / reference]. Figure 3 , Figure 3 This is a flowchart illustrating the process of generating a timing conflict matrix according to an embodiment of this application. Generating the timing conflict matrix specifically includes:

[0163] The initial deployment sequence is formed by selecting the top Z flight path prediction times from the temporary candidate set based on the deployment priority value.

[0164] For each predicted flight path in the initial deployment sequence, based on the take-off and landing time window in which it is located, the predicted flight path times included in that take-off and landing time window are processed by the difference between each pair of adjacent predicted flight path times, and the minimum value among them is taken as the safe take-off and landing interval for that take-off and landing time window.

[0165] Calculate the difference between the predicted times of any two routes in the initial deployment sequence, determine whether the difference is greater than the corresponding safe take-off and landing interval, if so, determine that there is a conflict between the predicted times of the two routes, and then generate a time-series conflict matrix.

[0166] The initial deployment sequence refers to the ordered list of the first Z flight path prediction times selected from the temporary candidate set in descending order according to the deployment priority value of each candidate.

[0167] Safe takeoff and landing interval refers to the minimum time interval within the same takeoff and landing time window that ensures the safe takeoff and landing of drones and avoids time overlap or close proximity that could lead to operational risks.

[0168] This part has already been described in detail above, so I will not repeat it here.

[0169] This scheme forms an initial deployment sequence by selecting predicted flight path times according to deployment priority values, giving high-priority flight paths priority consideration in the candidate sequence. This improves the execution guarantee of critical flight paths in the overall mission planning. By statistically analyzing the pairwise differences between predicted flight path times within each take-off and landing time window and taking the minimum value to generate a safe take-off and landing interval, the system achieves quantitative constraints on flight path density within the same time window, ensuring that the UAV take-off and landing time interval is not lower than the safety threshold and avoiding operational conflicts or collision risks caused by too close a time interval. In the process of calculating the difference between predicted flight path times of any two flight paths in the initial sequence and comparing it with the safe take-off and landing interval, the system can automatically mark potential conflicts and form a complete temporal conflict matrix. This allows subsequent deployment optimization algorithms to directly use the matrix for local conflict resolution or overall path optimization, improving planning efficiency and safety, and effectively controlling time conflicts and operational risks in the deployment of multiple UAVs.

[0170] The above describes conflict identification of a temporary candidate set based on deployment priority and takeoff / landing time windows, generating a temporal conflict matrix. The following describes a local rearrangement optimization of the temporal conflict matrix and deployment priority to generate a deployment sequence. Please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram of the process for generating a deployment sequence provided in an embodiment of this application. Generating a deployment sequence specifically includes:

[0171] The confidence assessment results of the timing conflict matrix, safe take-off and landing interval, and energy consumption matrix are weighted and calculated to obtain the cumulative conflict cost of the corresponding route prediction time within the initial deployment sequence.

[0172] The initial deployment sequence is enumerated and swapped based on the cumulative conflict cost, and the energy consumption change benefit is calculated. The initial deployment sequence after the enumeration and swapping with the largest energy consumption change benefit is selected as the deployment sequence.

[0173] Among them, the cumulative conflict cost refers to the overall risk used to quantify the temporal and spatial conflicts between candidate nodes in the initial deployment sequence.

[0174] Enumeration swap refers to the data processing process of swapping the positions or order of two or more nodes in the initial placement sequence.

[0175] Energy consumption change benefit refers to a quantitative indicator used to evaluate the energy consumption optimization effect of exchanging or replacing candidate nodes during local rearrangement on the overall deployment scheme.

[0176] This part has already been described in detail above, so I will not repeat it here.

[0177] This scheme quantifies the temporal and spatial conflict relationships between different deployment nodes by weighting the cumulative conflict costs of each candidate node in the initial deployment sequence at the time of flight path prediction. This allows the system to accurately identify high-risk conflict nodes during local optimization, providing clear data for subsequent adjustments. By using an enumeration exchange strategy to calculate the energy consumption change benefits, the energy consumption information and conflict costs of candidate nodes are considered together. This ensures that each node exchange not only reduces conflict but also optimizes overall energy consumption, achieving a dual improvement in both safety and energy efficiency of the deployment scheme. By selecting the exchange sequence with the greatest energy consumption change benefit as the final deployment scheme, the system can maximize energy savings and conflict mitigation effects while retaining deployment priority constraints, thereby improving the overall execution efficiency and mission completion reliability of parallel deployment of multiple UAVs.

[0178] The above describes the local rearrangement and optimization of the temporal conflict matrix and deployment priority values ​​to generate a deployment sequence. The following describes how to generate a path score matrix by constructing entity nodes and weighted relational edges and calculating path cumulative indicators from the flight path prediction sequence, deployment sequence, and takeoff and landing time window set. Specifically, this includes:

[0179] The predicted flight path times are used as knowledge graph nodes to construct a knowledge graph. The predicted flight path sequence, deployment sequence, and take-off and landing time window sets are written into the attributes of the knowledge graph nodes. The edge weights of the knowledge graph are obtained by performing a difference analysis on the attributes of the corresponding two knowledge graph nodes.

[0180] Using knowledge graphs as constraints, paths are generated for each candidate orientation by combining route prediction sequences. The paths are then comprehensively evaluated based on energy consumption matrix, local wind and wave trend field, and heading change probability matrix to generate a path score matrix.

[0181] Among them, a knowledge graph refers to a directed weighted graph composed of entity nodes and relation edges;

[0182] A path refers to an ordered spatiotemporal node sequence that starts from the drone deck platform, passes through one or more nodes in the candidate sequence, and reaches a specific candidate endpoint node.

[0183] The path scoring matrix refers to structured data that quantifies and evaluates the deployment path of unmanned aerial vehicle (UAV) deck platforms.

[0184] This part has already been described in detail above, so I will not repeat it here.

[0185] This scheme uses a path scoring matrix to uniformly quantify the cumulative energy consumption, risk level, and time window suitability of each path from the deck to the candidate node. This makes path selection not only rely on a single indicator but also comprehensively consider multi-dimensional constraints, greatly improving the rationality and safety of the deployment scheme. By performing difference analysis on the node attributes and edge weights of the knowledge graph and combining the heading and turning probability matrix with the local wind and wave trend field to score the path, the path evaluation fully considers the dynamic environmental factors and the uncertainty of the route prediction, thus improving the accuracy and feasibility of the path planning.

[0186] The above describes how to generate a path score matrix by constructing entity nodes and weighted relational edges and calculating path cumulative indicators from the route prediction sequence, deployment sequence, and takeoff and landing time window set. The following describes how to perform path consistency verification on the path score matrix and the temporal conflict matrix to generate a path planning scheme, specifically including:

[0187] The search space is constructed using the path scoring matrix, deployment sequence, and temporal conflict matrix as execution objects. A constrained search is performed on the path within the search space, and feasibility is determined through take-off and landing time windows during the constrained search process.

[0188] Based on the route prediction sequence and the local wind and wave trend field, the constrained search results are subjected to spatiotemporal consistency verification. The constrained search results and deployment sequence after spatiotemporal consistency verification are then combined in a structured manner to generate a route planning scheme.

[0189] The search space refers to the set of directed nodes established using the candidate nodes in the placement sequence as the basic units, according to the optimal path information from the starting point to the candidate node in the path scoring matrix.

[0190] Constrained search refers to the process of automatically generating path combinations that satisfy multiple constraints within a constructed search space, based on the placement of candidate nodes and a scoring matrix.

[0191] Feasibility determination refers to the data processing procedure used to determine whether a candidate node can be legally added to a path planning scheme.

[0192] Spatiotemporal consistency verification refers to the data processing process of systematically verifying the candidate deployment paths obtained by constrained search in both time and space dimensions.

[0193] Structured assembly refers to a data processing operation that systematically and systematically integrates data.

[0194] This part has already been described in detail above, so I will not repeat it here.

[0195] This solution constructs a search space with a path scoring matrix, deployment sequence, and time-series conflict matrix as execution objects. It automatically considers the time window constraints and potential conflicts of each candidate node during the path generation phase, enabling feasibility assessment of parallel deployment of multiple UAVs. This reduces the risk of human intervention and operational errors. During the constrained search process, the time, space, energy consumption, and risk indicators of each candidate node are calculated and judged in real time using the scoring matrix and conflict matrix. This balances the generated path scheme among multiple objectives such as optimal energy consumption, minimum conflict, and shortest task completion time, improving the overall reliability and execution efficiency of the planning scheme. Combined with the flight path prediction sequence and local wind and wave trend field for spatiotemporal consistency verification, it can verify in real time whether each path meets safety and operational constraints in the actual environment, significantly reducing the probability of mission failure due to wind and wave disturbances or heading deviations. By structurally combining the verified paths with the deployment sequence, the path planning scheme generated by the system can not only be directly used for UAV control execution, but each path also retains a complete source index and data link.

[0196] Example 2:

[0197] Please see Figure 5 A knowledge graph-based deployment path planning system for unmanned aerial vehicle (UAV) deck platforms includes:

[0198] The data acquisition module is used to acquire the ship's navigation data and wind and wave status data of the target object;

[0199] The navigation data analysis module is used to perform stability calculations on ship navigation data in a sliding window, determine whether the stability calculation result is greater than a preset threshold, and if so, perform constraint curve fitting on the ship navigation data in the corresponding sliding window based on wind and wave state data to generate a route prediction sequence.

[0200] Otherwise, statistical modeling is performed on the ship navigation data within the corresponding sliding window to generate a heading and turning probability matrix;

[0201] The takeoff and landing time extraction module is used to perform time-series decomposition on wind and wave state data according to a sliding window, generate a local wind and wave trend field, and extract the set of takeoff and landing time windows from each sliding window according to an adaptive threshold.

[0202] The priority calculation module is used to extract a temporary candidate set from the route prediction sequence based on the set of take-off and landing time windows, and to calculate the deployment priority value by weighting the temporary candidate set according to the heading turning probability matrix.

[0203] The path planning module is used to perform local conflict resolution on the temporary candidate set based on the deployment priority value and generate a path planning scheme.

[0204] This embodiment has the same technical effects as Embodiment 1.

[0205] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The data mentioned in this application have undergone normalization and other preprocessing to unify dimensions during formula calculations.

[0206] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A knowledge graph-based unmanned aerial vehicle deck platform laying path planning method, characterized in that, The method comprises the following steps: Obtaining ship navigation data and wind and wave state data of a target object; Performing stability calculation on the ship navigation data in a sliding window, and determining whether the stability calculation result is greater than a preset discrimination threshold; if yes, performing constraint curve fitting on the ship navigation data in the corresponding sliding window according to the wind and wave state data, and generating a route prediction sequence; Otherwise, performing statistical modeling on the ship navigation data in the corresponding sliding window, and generating a heading turning probability matrix; Performing time series decomposition on the wind and wave state data in a sliding window, generating a local wind and wave trend field, and extracting a take-off and landing time window set from each sliding window according to an adaptive threshold; wherein the adaptive threshold is obtained by quantile statistics on the wind and wave state data in the corresponding sliding window; Extracting a critical time set from the route prediction sequence according to the take-off and landing time window set, and obtaining a deployment priority value by weighted calculation according to the heading turning probability matrix; Performing local conflict elimination on the critical time set according to the deployment priority value, and generating a path planning scheme. 2.The knowledge graph based unmanned aerial vehicle deck platform path planning method of claim 1, wherein: After extracting the take-off and landing time window set, an energy consumption matrix is generated, specifically including: Performing space-time registration on the route prediction sequence and the local wind and wave trend field, and performing azimuth discretization on the space-time registration result to generate each candidate azimuth; For each candidate azimuth, the cumulative energy consumption value of the corresponding candidate azimuth is generated by distance integration accumulation according to a wind-ship coupling factor, and the energy consumption matrix is generated by integrating the cumulative energy consumption values of all candidate azimuths; Wherein, the wind-ship coupling factor is obtained by mapping the route prediction sequence of the corresponding candidate azimuth. 3.The knowledge graph based UAV deck platform path planning method of claim 2, wherein: Extracting a critical time set from the route prediction sequence according to the take-off and landing time window set specifically includes: Comparing the route prediction time in the route prediction sequence with the take-off and landing time window set point by point, extracting the take-off and landing time window containing the route prediction time, and performing difference calculation on the take-off and landing time window and the corresponding route prediction time; Determining whether the corresponding route prediction time is greater than the preset energy consumption threshold of the corresponding take-off and landing time window in the energy consumption matrix; if yes, determining whether the difference calculation result of the corresponding route prediction time is greater than the preset take-off and landing preparation threshold of the corresponding take-off and landing time window; if yes, integrating the corresponding route prediction time to generate a critical time set; Wherein, the preset energy consumption threshold is obtained by quantile statistics on the elements of the energy consumption matrix contained in the corresponding take-off and landing time window. 4.The knowledge graph based UAV deck platform path planning method of claim 3, wherein: Obtaining a deployment priority value by weighted calculation according to the heading turning probability matrix specifically includes: Performing weighted calculation on the elements of the energy consumption matrix contained in the take-off and landing time window, the difference calculation result and the heading turning probability matrix to obtain the deployment priority value of each route prediction time in the critical time set; Wherein, the weights in the weighted calculation process are obtained by confidence evaluation on the route prediction sequence, the energy consumption matrix and the take-off and landing time window set, respectively. 5.The knowledge graph based UAV deck platform path planning method of claim 4, wherein: Performing local conflict elimination on the critical time set according to the deployment priority value, and generating a path planning scheme specifically includes: According to the deployment priority and the take-off and landing time window, conflict identification is performed on the temporary time set to generate a time sequence conflict matrix; Local rearrangement optimization is performed on the time sequence conflict matrix and the deployment priority to generate a deployment sequence; A path score matrix is generated by constructing entity nodes and weighted relationship edges and calculating path cumulative indicators based on the route prediction sequence, the deployment sequence, and the take-off and landing time window set; Path consistency verification is performed on the path score matrix and the time sequence conflict matrix to generate a path planning scheme. 6.The knowledge graph based UAV deck platform path planning method of claim 5, wherein: According to the deployment priority and the take-off and landing time window, conflict identification is performed on the temporary time set to generate a time sequence conflict matrix, specifically including: An initial deployment sequence is formed by taking the first Z route prediction times from the temporary time set according to the deployment priority; For each route prediction time in the initial deployment sequence, a difference value is calculated for each two adjacent route prediction times included in the take-off and landing time window in which the route prediction time is located, and the minimum value is taken as the safe take-off and landing interval of the take-off and landing time window; The difference value between any two route prediction times in the initial deployment sequence is calculated to determine whether the difference value is greater than the corresponding safe take-off and landing interval. If yes, it is determined that a conflict exists between the two route prediction times, and a time sequence conflict matrix is generated. 7.The knowledge graph based UAV deck platform path planning method of claim 6, wherein: Local rearrangement optimization is performed on the time sequence conflict matrix and the deployment priority to generate a deployment sequence, specifically including: The initial deployment sequence is enumerated and exchanged according to the cumulative conflict cost to calculate the energy consumption change benefit, and the initial deployment sequence after enumeration and exchange with the maximum energy consumption change benefit is selected as the deployment sequence. A path score matrix is generated by constructing entity nodes and weighted relationship edges and calculating path cumulative indicators based on the route prediction sequence, the deployment sequence, and the take-off and landing time window set, specifically including: 8.The knowledge graph based UAV deck platform path planning method of claim 5, wherein: The route prediction time is taken as a knowledge graph node, and a knowledge graph is constructed. The route prediction sequence, the deployment sequence, and the take-off and landing time window set are written into the knowledge graph node attributes. The edge weight of the knowledge graph is obtained by difference analysis of the corresponding two knowledge graph node attributes; Based on the knowledge graph as a constraint, a path of each candidate direction is generated combined with the route prediction sequence, and the path is comprehensively evaluated based on the energy consumption matrix, the local wind and wave trend field, and the heading turning probability matrix to generate a path score matrix. Path consistency verification is performed on the path score matrix and the time sequence conflict matrix to generate a path planning scheme, specifically including: 9.The knowledge graph based UAV deck platform path planning method of claim 8, wherein: A search space is constructed based on the path score matrix, the deployment sequence, and the time sequence conflict matrix as execution objects, and a constrained search is performed on the path in the search space. During the constrained search process, feasibility is determined by the take-off and landing time window. ​ According to the spatio-temporal consistency verification performed on the constrained search result and the local wind wave trend field according to the route prediction sequence, the path planning scheme is generated by structurally combining the spatio-temporal consistency verified constrained search result and the deployment sequence.

10. A knowledge graph based unmanned aerial vehicle deck platform emplacement path planning system, characterized in that, Comprise: A data acquisition module for acquiring ship navigation data and wind wave state data of a target object; A navigation data analysis module for performing stability calculation on the ship navigation data according to a sliding window, and determining whether the stability calculation result is greater than a preset discrimination threshold; if yes, performing constraint curve fitting on the ship navigation data in the corresponding sliding window according to the wind wave state data to generate a route prediction sequence; Otherwise, performing statistical modeling on the ship navigation data in the corresponding sliding window to generate a heading turning probability matrix; A take-off and landing time extraction module for performing time sequence decomposition on the wind wave state data according to a sliding window to generate a local wind wave trend field, and extracting a take-off and landing time window set from each sliding window according to an adaptive threshold from the local wind wave trend field; A priority calculation module for extracting a time selection set from the route prediction sequence according to the take-off and landing time window set, and obtaining a deployment priority value by weighted calculation on the time selection set according to the heading turning probability matrix; A path planning module for performing local conflict removal on the time selection set according to the deployment priority value to generate a path planning scheme.

Images