Multi-unmanned aerial vehicle low-altitude service scheduling result generation method and device, equipment and medium

Abstract

The invention discloses a multi-unmanned aerial vehicle low-altitude service scheduling result generation method and device, equipment and a medium, and relates to the technical field of unmanned aerial vehicle scheduling, and the method comprises the steps: constructing a low-altitude service semantic grid, converting the service demand characteristics of each region of a city into a region semantic vector, and mapping an unmanned aerial vehicle task to the grid to generate a task semantic vector; a semantic enhancement mechanism is utilized to calculate the semantic association strength of a task and a region, a multi-dimensional fairness index system is constructed, and a dynamic fairness preference weight is generated through self-adaptive correction. And finally, task allocation and path planning are performed based on the dynamic weight and the semantic association strength, so that the dynamic balance service efficiency and fairness of multi-unmanned aerial vehicle scheduling are improved, regional service deviation is dynamically corrected, the service accessibility of vulnerable regions and crowds is ensured, and the method is adaptive to complex urban environments and abnormal scenes.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) scheduling technology, and in particular to a method, apparatus, equipment, and medium for generating scheduling results for low-altitude services of multiple UAVs. Background Technology

[0002] Current drone mission scheduling technology mainly follows the principle of optimal efficiency, with scheduling strategies designed around indicators such as mission completion time, execution cost, and energy consumption. It is generally based on fixed constraints, deterministic requirements, and static scenario assumptions for scheduling planning. To alleviate the problem of uneven regional services, some existing studies have attempted to introduce fairness-related strategies into scheduling, such as setting response time thresholds and increasing the weight of tasks in remote areas. However, these strategies mostly rely on human experience or static rules, which are significantly insufficient in terms of service fairness and are difficult to adapt to the complex and ever-changing service demands in real urban environments. On the one hand, efficiency-optimized scheduling strategies lead to significant discrepancies between central urban areas and peripheral areas in terms of average response time, task arrival rate, and service accessibility. Vulnerable groups face passive service disadvantages and cannot obtain balanced service guarantees. On the other hand, fairness strategies that rely on human experience or static rules lack effective utilization of multi-source semantic information, making it difficult to play a stable, automatic, and efficient role in complex urban environments in the long term, and unable to dynamically correct regional service biases. In addition, existing methods have not built a sound mechanism for handling abnormal scenarios. In the event of emergencies, drone malfunctions, or surges in demand, it is difficult to achieve reasonable resource allocation and post-event service compensation, further exacerbating the problem of service unfairness.

[0003] Therefore, there is an urgent need for a method to generate low-altitude service scheduling results for multiple UAVs in order to improve the efficiency and fairness of dynamic balancing services for multiple UAV scheduling and to adapt to abnormal scenarios. Summary of the Invention

[0004] The main purpose of this application is to provide a method, apparatus, device and medium for generating multi-UAV low-altitude service scheduling results, aiming to solve the technical problem of how to improve the efficiency and fairness of dynamic balance service in multi-UAV scheduling and adapt to abnormal scenarios.

[0005] To achieve the above objectives, this application proposes a method for generating multi-UAV low-altitude service scheduling results, including: A low-altitude service semantic grid is constructed based on a three-dimensional spatial structure and multi-source urban data, and regional semantic vectors are obtained. The three-dimensional spatial structure is based on geographic coordinates and flight altitude ranges. The multi-source urban data includes population census data, POI data, road network data and terrain model data. The regional semantic vectors represent the service demand characteristics of each area of ​​the city. The unmanned aerial vehicle (UAV) mission to be executed is mapped to the low-altitude service semantic grid to generate a mission semantic vector, wherein the mission semantic vector and the region semantic vector have the same dimensions. The semantic association strength is obtained by processing the semantic vectors of the region and the semantic vector region of the task through a semantic enhancement mechanism. Based on the regional semantic vector and the task semantic vector, a multidimensional fairness index system is constructed, and an initial fairness preference weight is set. The multidimensional fairness index system includes a regional coverage deviation index, a regional service balance index, and a weak regional compensation index. The initial fairness preference weights are adaptively adjusted to generate dynamic fairness preference weights; Based on the dynamic fairness preference weights and semantic association strength, the unmanned aerial vehicle (UAV) tasks to be executed are allocated and paths are planned as a whole to obtain the target path planning scheme.

[0006] In one embodiment, the step of constructing a low-altitude service semantic grid based on three-dimensional spatial structure and multi-source urban data, and obtaining regional semantic vectors, includes: The urban low-altitude space is divided into multiple three-dimensional cells according to geographical coordinates and flight altitude ranges, wherein the three-dimensional cells do not overlap. Regional features are extracted from each of the three-dimensional cells from multi-source urban data, wherein the regional features include population density, proportion of vulnerable groups, coverage of public facilities, regional economic activity, and regional accessibility; The extracted regional features are normalized to obtain normalized regional features; Initial weights are configured for the normalized regional features based on the service scenario type, and semantic weights for each three-dimensional cell are generated through weighted fusion, wherein the service scenario type includes at least public service scenarios and commercial service scenarios; Based on the spatial coordinate information of the three-dimensional cells, the regional features, and the semantic weights, a low-altitude service semantic grid is constructed, and a regional semantic vector corresponding to each region is obtained.

[0007] In one embodiment, the step of mapping the unmanned aerial vehicle (UAV) mission to be executed to the low-altitude service semantic grid and generating a mission semantic vector includes: The drone mission to be executed is analyzed to obtain attribute information, which includes mission type, mission priority, service object category, mission urgency and expected workload. The attribute information is structured and organized to form a set of original features for the task; The original feature set of the task is normalized to obtain the normalized task features; The normalized task features are matched with the semantic dimensions of the low-altitude service semantic grid to obtain the dimension matching result. A task semantic vector is constructed based on the dimension matching results.

[0008] In one embodiment, the step of processing the region semantic vector and the task semantic vector region through a semantic enhancement mechanism to obtain the semantic association strength includes: The cosine similarity algorithm is used to calculate the vector similarity between the semantic vector of the region and the semantic vector of the task; The basic matching degree is obtained based on the vector similarity. Based on the semantic weights in the regional semantic vector and historical service coverage data, the regional service demand score corresponding to the regional semantic vector is calculated. The regional service demand score is processed by the Sigmoid function to generate a demand enhancement coefficient, wherein the demand enhancement coefficient increases as the regional service demand increases. The initial semantic association value is obtained by multiplying the basic matching degree with the demand enhancement coefficient by weight. The initial semantic association value is normalized to obtain the semantic association strength.

[0009] In one embodiment, the step of constructing a multidimensional fairness index system based on the region semantic vector and the task semantic vector, and setting initial fairness preference weights, includes: The theoretical service demand value of each region is calculated based on the semantic weights in the region semantic vector. Based on the theoretical service demand value and historical task allocation data, the actual service coverage value of each region is obtained; Calculate the regional coverage deviation between the theoretical service demand value and the actual service coverage value, and construct a regional coverage deviation index based on the regional coverage deviation value; Calculate the variance of the regional coverage deviation for all regions, and construct a regional service balance index based on the variance; Based on the proportion of disadvantaged groups in the semantic vector of the region and the coverage deviation value of the region, a compensation index for disadvantaged regions is constructed. By integrating the regional coverage deviation index, the regional service balance index, and the compensation index for disadvantaged areas, a multi-dimensional fairness index system is constructed. Based on the service scenario requirements, initial fairness preference weights are set for each indicator in the multidimensional fairness indicator system, wherein the sum of each initial fairness preference weight is a preset value, and the service scenario requirements include at least public service scenario requirements and commercial service scenario requirements.

[0010] In one embodiment, the step of adaptively adjusting the initial fairness preference weights to generate dynamic fairness preference weights includes: Acquire historical service performance data, including service deviation trends, task completion rates, and service satisfaction in each region. Real-time collection of current task execution status data, including task execution progress, UAV resource usage, and real-time changes in regional demand; Based on the historical service performance data and the current task execution status data, the service deviation trend value of each region is calculated using the first-order difference method. The direction of weight adjustment is determined based on the service deviation trend value; The initial fairness preference weights are initially corrected based on the weight adjustment direction to obtain the first weights; The first weight is subjected to boundary constraint processing to obtain the second weight; The second weight is normalized to generate dynamic fairness preference weights.

[0011] In one embodiment, the step of performing overall allocation and path planning for the unmanned aerial vehicle (UAV) task to be executed based on the dynamic fairness preference weights and semantic association strength to obtain a target path planning scheme includes: Obtain the equipment capability parameters of each UAV, including endurance, payload limit, flight speed, and communication range; By combining the dynamic fairness preference weight, the semantic association strength, and the device capability parameters, a task allocation utility function is constructed, wherein the task allocation utility function includes semantic fit, fairness constraints, and task urgency. With the goal of maximizing the task allocation utility function, a task allocation model is constructed by combining UAV payload constraints, flight range constraints, and unique task execution constraints. The task allocation model is solved using a reinforcement learning algorithm to obtain the allocation result. The reward function of the reinforcement learning algorithm includes a semantic matching reward, a fairness reward, and a cost penalty term. The allocation result includes tasks, drones, and regions. Based on the allocation results and combined with low-altitude airspace traffic rules, the optimal flight path and mission execution sequence are planned for each UAV, generating a path planning scheme for a common scenario. Real-time collection of drone operating status, regional environmental status, and mission execution status; A comprehensive anomaly metric is calculated based on the UAV's operating status, regional environmental status, and mission execution status, wherein the comprehensive anomaly metric is the weighted sum of each status parameter; When the comprehensive anomaly metric exceeds the preset anomaly trigger threshold, an anomaly event is determined to have occurred, and the system switches to emergency dispatch mode. In the emergency scheduling mode, the task allocation utility function is reconstructed with semantic adaptability and task urgency as the core weights, and the upper limit of the fairness constraint is adjusted according to the preset emergency degradation factor to obtain the adjusted scheduling utility function. Based on the adjusted scheduling utility function, drone resources are allocated to high-risk areas first, generating emergency scenario allocation results; When not in the emergency dispatch mode, calculate the regional service gap of each non-high-risk area during the abnormal event, where the regional service gap is the difference between the expected service demand and the actual service demand; Based on the regional service gap, the dynamic fairness preference weight for the next scheduling cycle is compensated and adjusted. Compensatory scheduling is implemented for regions with regional service gaps to obtain the target path planning scheme.

[0012] Furthermore, to achieve the above objectives, this application also proposes a multi-UAV low-altitude service scheduling result generation device, which includes: The semantic grid construction module is used to construct a low-altitude service semantic grid based on a three-dimensional spatial structure and multi-source urban data, and obtain regional semantic vectors. The three-dimensional spatial structure is based on geographic coordinates and flight altitude ranges. The multi-source urban data includes population census data, POI data, road network data and terrain model data. The regional semantic vectors represent the service demand characteristics of each area of ​​the city. The task mapping module is used to map the UAV task to be executed to the low-altitude service semantic grid and generate a task semantic vector, wherein the task semantic vector and the region semantic vector have the same dimensions. The semantic enhancement module is used to process the semantic vector of the region and the semantic vector region of the task through a semantic enhancement mechanism to obtain the semantic association strength; The fairness index construction module is used to construct a multi-dimensional fairness index system based on the regional semantic vector and the task semantic vector, and to set initial fairness preference weights. The multi-dimensional fairness index system includes regional coverage deviation index, regional service balance index, and weak region compensation index. The weight optimization module is used to adaptively correct the initial fairness preference weights to generate dynamic fairness preference weights. The results module is used to perform overall allocation and path planning for the unmanned aerial vehicle (UAV) tasks to be executed based on the dynamic fairness preference weights and semantic association strengths, so as to obtain the target path planning scheme.

[0013] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the multi-UAV low-altitude service scheduling result generation method described above.

[0014] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the multi-UAV low-altitude service scheduling result generation method described above.

[0015] This application constructs a low-altitude service semantic grid, transforming the service demand characteristics of various urban areas into regional semantic vectors, and mapping UAV tasks to this grid to generate task semantic vectors. A semantic enhancement mechanism is used to calculate the semantic association strength between tasks and regions, constructing a multi-dimensional fairness index system, and generating dynamic fairness preference weights through adaptive correction. Finally, task allocation and path planning are performed based on dynamic weights and semantic association strength, improving the dynamic balancing service efficiency and fairness of multi-UAV scheduling, dynamically correcting regional service deviations, ensuring service accessibility for disadvantaged areas and populations, and adapting to complex urban environments and abnormal scenarios. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the first embodiment of the method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) according to this application. Figure 2 This is a flowchart illustrating the second embodiment of the method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) according to this application. Figure 3 This is a flowchart illustrating the third embodiment of the method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) according to this application. Figure 4 This is a schematic diagram of the module structure of the multi-UAV low-altitude service scheduling result generation device in the first embodiment of the multi-UAV low-altitude service scheduling result generation method of this application. Figure 5 This is a schematic diagram of the hardware operating environment involved in the method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) in this application embodiment.

[0018] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0020] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0021] Current drone mission scheduling technology mainly follows the principle of optimal efficiency, with scheduling strategies designed around indicators such as mission completion time, execution cost, and energy consumption. It is generally based on fixed constraints, deterministic requirements, and static scenario assumptions for scheduling planning. To alleviate the problem of uneven regional services, some existing studies have attempted to introduce fairness-related strategies into scheduling, such as setting response time thresholds and increasing the weight of tasks in remote areas. However, these strategies mostly rely on human experience or static rules, which are significantly insufficient in terms of service fairness and are difficult to adapt to the complex and ever-changing service demands in real urban environments. On the one hand, efficiency-optimized scheduling strategies lead to significant discrepancies between central urban areas and peripheral areas in terms of average response time, task arrival rate, and service accessibility. Vulnerable groups often face passive service disadvantages and cannot obtain balanced service guarantees. On the other hand, fairness strategies that rely on human experience or static rules lack effective utilization of multi-source semantic information, making it difficult to play a stable, automatic, and efficient role in complex urban environments in the long term, and unable to dynamically correct regional service biases. In addition, existing methods have not built a sound mechanism for handling abnormal scenarios. In the event of emergencies, drone malfunctions, or surges in demand, it is difficult to achieve reasonable resource allocation and post-event service compensation, further exacerbating the problem of service unfairness.

[0022] Therefore, this application proposes a method for generating low-altitude service scheduling results for multiple UAVs to improve the efficiency and fairness of dynamic balancing services in multi-UAV scheduling and adapt to abnormal scenarios. The main solution of this application's embodiments is as follows: A low-altitude service semantic grid is constructed based on a three-dimensional spatial structure and multi-source urban data, and regional semantic vectors are obtained; the UAV tasks to be executed are mapped to the low-altitude service semantic grid to generate task semantic vectors; the regional semantic vectors and task semantic vectors are processed through a semantic enhancement mechanism to obtain semantic association strength; a multi-dimensional fairness index system is constructed based on the regional semantic vectors and the task semantic vectors, and initial fairness preference weights are set; the initial fairness preference weights are adaptively corrected to generate dynamic fairness preference weights; based on the dynamic fairness preference weights and semantic association strengths, the UAV tasks to be executed are allocated and path planned as a whole to obtain a target path planning scheme.

[0023] Based on the above, this application also provides a method for generating multi-UAV low-altitude service scheduling results, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) according to this application.

[0024] In this embodiment, the method for generating multi-UAV low-altitude service scheduling results includes steps S10 to S60: Step S10: Construct a low-altitude service semantic grid based on the three-dimensional spatial structure and multi-source urban data, and obtain regional semantic vectors.

[0025] It should be noted that the three-dimensional spatial structure is a spatial framework based on geographic coordinates and flightable altitude ranges, used to delineate the low-altitude range of a city, providing a spatial foundation for the construction of the low-altitude service semantic grid. Multi-source urban data refers to various datasets used to characterize urban regional features, including census data, POI (Point of Interest) data, road network data, and terrain model data, providing data support for extracting regional service demand characteristics. The low-altitude service semantic grid is a semantic grid system constructed based on the three-dimensional spatial structure and multi-source urban data. Through the discretization of urban low-altitude space and its association with semantic features, it achieves a structured expression of regional service demand. Regional semantic vectors are vectors extracted from the low-altitude service semantic grid to represent the service demand characteristics of various urban areas. Their dimensions correspond to the regional semantic features, and they can quantitatively reflect the intensity and differences in regional service demand.

[0026] Understandably, step S10 also includes: firstly, dividing the urban low-altitude airspace into multiple three-dimensional cells based on geographic coordinates and flight altitude ranges. Specifically, let the entire urban low-altitude airspace range be: ,in Geographic coordinates Indicates the range of flightable altitudes. Indicates the urban area in geographic coordinates The minimum and maximum range of latitude (usually corresponding to planar longitude). Indicates the urban area in geographic coordinates The minimum and maximum range of dimensions (usually corresponding to latitude in a plane). Indicates the urban area at flight altitude For subsequent semantic modeling, the minimum and maximum ranges of the dimensions are defined, and the aforementioned urban low-altitude spatial range is divided into... There are three-dimensional cells, and the three-dimensional cells do not overlap. The spatial coordinate information of the three-dimensional cells is denoted as... .

[0027] Next, regional features for each 3D cell are extracted from the multi-source city data, including population density. The proportion of vulnerable groups Public facilities coverage Regional economic activity and regional accessibility The above features are combined to form regional features. ,in This indicates the number of cells. It should be noted that for each three-dimensional cell, the population density is calculated by dividing the number of permanent residents within the cell's spatial area by the cell's area, derived from census data. Next, the number of facilities providing services to vulnerable groups such as the elderly and healthcare within the cell is selected from POI data, and this number is divided by the total number of facilities to obtain the proportion of vulnerable groups. Then, the coverage area percentage of public facilities (such as parks and community centers) within the cell is calculated from POI data to obtain the public facility coverage. Following this, indicators such as the frequency of commercial activities and consumption data corresponding to the cell are extracted from regional economic statistics to quantify the regional economic activity. Finally, the road network density and average travel time within the cell are calculated from road network data to obtain the regional accessibility.

[0028] Next, the extracted regional features are normalized to obtain normalized regional features. Specifically, since different semantic features have different scales, normalization is required to ensure that the features participate in subsequent weight calculations under a unified dimension. In this embodiment, Min-Max normalization is used, and the specific formula is as follows: in Indicates the first The first cell Each feature is normalized to construct the semantic feature vector of that cell: .

[0029] Then, initial weights are configured for the normalized regional features based on the service scenario type, and semantic weights for each 3D cell are generated through weighted fusion. It should be noted that the service scenario type refers to the different application areas targeted by the low-altitude service, including at least public service scenarios and commercial service scenarios. Initial weights are weight values ​​pre-set for the normalized regional features based on the service scenario type; the sum of all weights is a preset fixed value (e.g., 1), used to reflect the importance of different features in their respective scenarios. Weighted fusion is the process of multiplying the normalized regional features by their corresponding initial weights and then summing the products; it is the core operation for generating semantic weights. Semantic weights are numerical values ​​obtained through weighted fusion that quantify the intensity of service demand and scenario adaptability of the 3D cell. Specifically, to reflect the importance of each semantic feature in different application scenarios, a vector of initial weights is set. : The semantic vector is weighted and fused with the initial weight vector to obtain the semantic weight of the cell's comprehensive service requirements. : Among them, the initial weights The settings can be flexibly adjusted according to the scenario. For example, the weight of vulnerable groups can be increased in public service scenarios, and the weight of economic activity can be increased in commercial scenarios. Finally, based on the spatial coordinate information, regional features, and semantic weights of the three-dimensional cells, a low-altitude service semantic grid is constructed, and a regional semantic vector corresponding to each region is obtained. According to the above steps, each three-dimensional spatial cell corresponds to a semantic structure triplet, which can be represented as: Then the low-altitude semantic grid of the entire city can be represented as: Then, the regional features and semantic weights of each cell are concatenated by dimension to obtain the regional semantic vector of the corresponding region.

[0030] Step S20: Map the drone mission to be executed to the low-altitude service semantic grid to generate a mission semantic vector.

[0031] It should be noted that drone missions refer to various low-altitude service-related tasks that drones need to perform, including delivery, inspection, and emergency support, and possess attributes such as mission type, priority, and service recipients. The mission semantic vector is a vector generated by mapping the drone mission to be performed onto a low-altitude service semantic grid. The mission semantic vector and the region semantic vector have the same dimensions and are used to achieve structured expression and semantic matching of mission attributes.

[0032] It is understandable that step S20 also includes: parsing the drone mission to be executed to obtain attribute information, wherein the attribute information includes the mission type. Task Priority Service recipient categories urgency of the task and expected workload The attribute information is then structured and organized to form a set of original task features. The original task feature set is normalized to obtain normalized task features. These normalized features are then matched with the semantic dimensions of the low-altitude service semantic grid to obtain dimension matching results. The normalized task features are matched one-to-one with the semantic dimensions of the low-altitude service semantic grid (corresponding to the five dimensions of the region semantic vector), for example, task type matches the regional economic activity dimension, and service object category matches the proportion of vulnerable groups dimension, resulting in dimension matching results. Finally, based on the dimension matching results, the normalized task features are concatenated into a vector form according to the matching dimension order to construct the task semantic vector, specifically represented as follows: .

[0033] Step S30: The semantic vector of the region and the semantic vector of the task are processed through a semantic enhancement mechanism to obtain the semantic association strength.

[0034] It should be noted that semantic enhancement mechanism refers to a processing mechanism used to enhance the correlation between regional semantic vectors and task semantic vectors. It optimizes vector matching results by combining the intensity of regional service demand, emphasizing fairness. Semantic correlation strength refers to the quantified value of the closeness of the correlation between regional semantic vectors and task semantic vectors obtained after processing by the semantic enhancement mechanism, providing a semantic adaptation basis for task allocation.

[0035] Understandably, to determine which regions are suitable for task execution, it is necessary to calculate the correlation between the task semantic vector and the region semantic vector. Let the first region be... The first task and the first The semantic similarity between the regions is: in, Indicates the first The task semantic vector of each task Indicates the first The semantic vectors of each region. To reflect the need for regional fairness in scheduling, this embodiment scores the service demand of each region. In conjunction with semantic similarity, a fairness enhancement function is introduced. The sigmoid function, which exhibits high stability with real-world data, is selected as the function influencing service demand. Specifically, it is expressed as: Where parameters Used to control the intensity of the impact of regional service demand on the outcome. When a region is chronically underserved, its Higher, making Increase; conversely, if a region is already over-serviced, then the corresponding... This will tend to stabilize, preventing further widening of the differences. Taking into account both task semantics and regional requirements, the task-region semantic association strength is constructed, specifically represented as follows: in Indicates the first The first task and the first The semantic association strength between regions is considered. In this process, similarity reflects whether a task is suitable for a particular region, while the reinforcement function reflects whether that region should be prioritized for service. Combining these two aspects ensures that task mapping not only matches capabilities but also reflects service fairness. A higher value indicates a stronger task... and region It has stronger semantic relevance and a higher degree of matching with fairness requirements.

[0036] Step S40: Construct a multi-dimensional fairness index system based on the region semantic vector and the task semantic vector, and set the initial fairness preference weights.

[0037] It should be noted that the multidimensional fairness index system refers to a set of multidimensional indicators used to assess the fairness of low-altitude services. This system includes regional coverage deviation indicators, regional service balance indicators, and disadvantaged area compensation indicators, providing quantitative standards for fair scheduling. The regional coverage deviation indicator is constructed based on the difference between the theoretical service demand and the actual service coverage of a region, used to characterize the degree of difference between regional service demand and actual service received. The regional service balance indicator is constructed by calculating the variance of the regional coverage deviation values ​​of all regions, used to assess the degree of balance in service allocation among different regions. The disadvantaged area compensation indicator is constructed based on the proportion of disadvantaged groups in the regional semantic vector and the regional coverage deviation value, used to quantify the intensity of service compensation demand in disadvantaged areas. Initial fairness preference weights refer to the initial weights set for each indicator in the multidimensional fairness index system according to service scenario requirements. The sum of all weights is a preset fixed value (e.g., 1), used for configuring fairness constraints in initial scheduling decisions.

[0038] Understandably, in order to quantify the service preference of a task for different regions, the strength of the semantic association between the task and the region is used. For the region Calculate its task coverage, because By combining task semantic matching degree and regional service requirement enhancement, the regional service coverage is constructed by directly summing the task semantic matching degree and regional service requirement enhancement, specifically expressed as follows: in Indicates the area Regional service coverage. Coverage measures the theoretical potential service strength that the system can provide to a region in the semantic space, providing a basic parameter for subsequent fairness assessment. Next, to characterize the difference between the actual semantic demand of a region and its obtained service coverage, a regional service demand score is used. and regional service coverage Obtain regional coverage deviation Specifically, it is expressed as: when This indicates that the demand for services in a region exceeds the coverage of those services, indicating an underservice situation. This indicates that the regional service coverage is higher than the regional service demand, indicating an over-service phenomenon; when This indicates that regional services are becoming more balanced.

[0039] To assess the evenness of service deviation distribution across the entire region, this method is based on the deviation of all regions. To calculate the service balance index, assuming the total number of regions is N, first calculate the mean deviation, specifically expressed as: in The mean of the deviation is used as the basis for calculating the variance of the service deviation, thus characterizing the degree of dispersion of fairness among regions. Specifically, this is expressed as follows: in, A larger value indicates a more dispersed and uneven distribution of regional deviations. A smaller value indicates more concentrated regional bias and better overall fairness. (Comprehensive regional service coverage) Service deviation and variance Construct a multi-dimensional fairness indicator system: To ensure that different indicators reflect different levels of importance in subsequent scheduling, initial fairness preference weights are set, with the specific weight vector as follows: in, This reflects the system's level of attention to regional coverage deviations. This reflects the system's emphasis on compensating disadvantaged areas. This reflects the system's emphasis on the balance of regional services.

[0040] Step S50: Adaptively adjust the initial fairness preference weights to generate dynamic fairness preference weights.

[0041] It should be noted that the dynamic fairness preference weight refers to the weight obtained after adaptively correcting the initial fairness preference weight. It can be dynamically adjusted according to historical service performance and real-time task execution status to achieve automatic correction of regional service deviations.

[0042] Further, step S50 also includes: acquiring historical service performance data; collecting current task execution status data in real time; calculating the service deviation trend value for each region based on the historical service performance data and the current task execution status data using the first-order difference method; determining the weight adjustment direction based on the service deviation trend value; initially correcting the initial fairness preference weights based on the weight adjustment direction to obtain the first weight; applying boundary constraints to the first weight to obtain the second weight; and normalizing the second weight to generate dynamic fairness preference weights. It should be noted that historical service performance data refers to a data set recording service-related situations within past scheduling cycles, including the service deviation change trend, task completion rate, and service satisfaction for each region. Current task execution status data refers to the dynamic data of tasks and resources collected in real time within the current scheduling cycle, including task execution progress, UAV resource occupancy, and real-time changes in regional demand, providing a real-time basis for weight correction. The first-order difference method is a calculation method that analyzes the speed and direction of service deviation changes by calculating the difference between service deviation values ​​in two adjacent cycles, used to obtain the service deviation trend value. The service deviation trend value is a numerical value calculated using the first-order difference method, reflecting the direction and magnitude of service deviation changes in each region, and is used to determine the direction of weight adjustment. The weight adjustment direction refers to the adjustment direction of the initial fairness preference weights determined based on the service deviation trend value, i.e., increasing or decreasing the weight of the corresponding indicator. The first weight is the weight obtained after preliminary correction of the initial fairness preference weights according to the weight adjustment direction, before boundary constraint processing. Boundary constraint processing refers to the method of limiting the first weight within a preset range (e.g., 0.1-0.5) to avoid the weight being too high or too low and affecting the scheduling effect. The second weight is the weight obtained after boundary constraint processing; its value is within a preset reasonable range and requires further normalization. The dynamic fairness preference weight is the weight generated after normalizing the second weight, which can be dynamically adjusted based on historical and real-time data to achieve automatic correction of regional service deviations.

[0043] Understandably, the first step is to extract historical service data from the scheduling system database for a preset period (e.g., 30 days), filter out the service deviation trends, task completion rates, and service satisfaction rates for each region, and compile historical service performance data. This data is then used to define the regions. In the Service deviation at the end of each scheduling cycle Specifically, it is expressed as: in, This indicates the score for regional service demand. For the first The service coverage of the region in each scheduling cycle is then used to obtain the historical deviation sequence, which is specifically represented as follows: This sequence is used to determine whether a region has been in an underservice or overservice state over a period of time, providing a dynamic information basis for weight updates.

[0044] Simultaneously, real-time data on the current task execution progress, drone resource utilization, and real-time changes in regional demand are collected through the drone communication module and regional sensors. This is done to obtain historical and real-time dual-dimensional data support required for weight correction. Secondly, based on service deviation data from historical service performance data and real-time service deviation from current task execution status data, a first-order difference method is used for calculation, i.e., through the formula: The service deviation trend value for each region is obtained to quantify the direction and magnitude of service deviation changes, providing a precise basis for weight adjustment. Then, the direction of weight adjustment is determined based on the service deviation trend value: if the trend value is positive, it indicates that the regional service deviation is expanding, and the weight of the corresponding fairness indicator needs to be increased; if the trend value is negative, it indicates that the deviation is shrinking, and the corresponding weight can be decreased. Based on this direction, the initial fairness preference weights are adjusted by increasing or decreasing to obtain the first weight. This is done to ensure that the weight adjustment aligns with the dynamic changes in regional service deviation. Next, boundary constraints are applied to the first weight, limiting each weight value to a preset range (e.g., 0.1-0.5). If the value exceeds the upper limit, the upper limit value is used; if it falls below the lower limit, the lower limit value is used, resulting in the second weight. This is to avoid scheduling imbalance caused by excessively high or low weights. Specifically, the initial fairness preference weight vector is designed... The adaptive update rules enable the system to automatically optimize towards "reducing service discrepancies" during execution. The specific update rule formula is as follows: in, This indicates the trend of the region where the deviation has worsened most severely within the current cycle. Represents the learning rate, and When the deviation in certain areas continues to increase (underservice worsens), The system automatically increases When regional bias is improved, , Gradually decrease, while ensuring the sum of the three weights is maintained. .

[0045] Let the feasible intervals for the weights of the three types of fairness preferences be as follows: ,in Then the boundary constraint rules are: in, This represents the updated weights after the above adaptive update. This represents the minimum allowable value for the weight. This represents the maximum allowable weight. This represents the second weight. Finally, the second weight is normalized to ensure that the sum of all weights is 1, generating the dynamic fairness preference weight, specifically represented as follows: This measure is to ensure that the weights meet the constraints of the scheduling model and that the fairness constraints are effectively implemented.

[0046] Step S60: Based on the dynamic fairness preference weight and semantic association strength, perform overall allocation and path planning for the drone tasks to be executed to obtain the target path planning scheme.

[0047] It should be noted that the target path planning scheme refers to the final scheme obtained after overall allocation and path planning of UAV tasks based on dynamic fairness preference weights and semantic association strength, which can take into account both service efficiency and fairness.

[0048] Further, step S60 includes: obtaining the equipment capability parameters of each UAV. It should be noted that the equipment capability parameters include endurance, payload limit, flight speed and communication range. By combining dynamic fairness preference weights, semantic association strength, and device capability parameters, a task allocation utility function is constructed. It should be noted that the task allocation utility function includes semantic fit, fairness constraints, and task urgency. Semantic association strength is used as the quantitative indicator of semantic fit, the correlation between dynamic fairness preference weights and regional service deviation is used as the fairness constraint, and the quantified task urgency is used as the task urgency. At the same time, the fit coefficient corresponding to the device capability parameters is incorporated to construct the task allocation utility function.

[0049] With the objective of maximizing the task allocation utility function, a task allocation model is constructed by combining UAV payload constraints, flight range constraints, and task uniqueness execution constraints. Specifically, let the set of UAVs be . Among them, drones Define scheduling decision variables: The comprehensive scheduling objective function of the task allocation model is then obtained as follows: in Indicates task Assigned to region Task allocation utility, representing The total number of tasks. This represents the path cost of the drone. This is the cost weighting coefficient. The UAV payload constraint is... Flight range constraint is ,in For drones The available flight range, and the unique execution constraint for the mission are: .

[0050] A reinforcement learning algorithm is used to solve the task allocation model, yielding the allocation results. The reward function of the reinforcement learning algorithm includes a semantic matching reward, a fairness reward, and a cost penalty term. The allocation results include tasks, drones, and regions. Specifically, the aforementioned reward function... Specifically, it is expressed as follows: in, Indicates selection of drone Execute the task And dispatched to the region , Indicates drone Execute action Path cost, Indicates the area Service deviation, This represents the corresponding penalty parameter. Specifically, a reward function is constructed with semantic matching reward, fairness reward, and cost penalty term as its core. The task allocation model is used as the decision environment for the reinforcement learning algorithm. By iteratively selecting the matching combination of drone and task, the reward function value corresponding to each decision is calculated and the policy is optimized. The corresponding policy gradient method is used to update the parameters as follows: Finally, the task allocation model is solved by a convergent optimal strategy to obtain the allocation result.

[0051] Based on the allocation results and low-altitude airspace traffic rules, an optimal flight path and mission execution sequence are planned for each UAV, generating a path planning scheme for a typical scenario. It should be noted that low-altitude airspace traffic rules refer to the normative requirements established for low-altitude flight activities, covering flight altitude restrictions, route priorities, airspace avoidance criteria, etc. The optimal flight path refers to the shortest distance or lowest energy consumption flight route connecting the UAV's starting point, mission execution point, and destination, while satisfying low-altitude airspace traffic rules and avoiding obstacles. The mission execution sequence refers to the order in which multiple tasks to be performed by a single UAV are determined, taking into account factors such as mission urgency and mission location distribution, to improve mission execution efficiency. The typical scenario path planning scheme refers to a complete planning scheme, including flight path and mission execution sequence, developed for each UAV under normal operational scenarios without unexpected abnormal situations. Specifically, based on the task allocation results, a task list for each UAV is defined, low-altitude airspace traffic rules are retrieved as constraints, and the optimal flight path for each UAV is calculated through path planning algorithms by combining the UAV equipment capability parameters and task location coordinates. At the same time, the task execution order is optimized according to the task urgency and location distribution, and the integrated results generate a conventional scenario path planning scheme.

[0052] Real-time data collection of UAV operating status, regional environmental status, and mission execution status; calculation of comprehensive anomaly metric based on UAV operating status, regional environmental status, and mission execution status, wherein the comprehensive anomaly metric is the weighted sum of each status parameter.

[0053] When the comprehensive anomaly metric exceeds the preset anomaly trigger threshold, an anomaly is determined to have occurred, and the system switches to emergency scheduling mode. In emergency scheduling mode, the task allocation utility function is reconstructed using semantic adaptability and task urgency as core weights, and the upper limit of fairness constraints is adjusted according to a preset emergency degradation factor, resulting in the adjusted scheduling utility function. It should be noted that the preset anomaly trigger threshold is a pre-set critical value for the comprehensive anomaly metric used to determine whether an anomaly has occurred; an anomaly is determined to have occurred when the real-time calculated comprehensive anomaly metric exceeds this value. The upper limit of fairness constraints refers to the upper limit of fairness constraint items set in the normal scheduling mode to ensure regional service fairness, used to limit the influence of fairness constraints in scheduling decisions. The preset emergency degradation factor is a pre-set coefficient used to reduce the upper limit of fairness constraints; the new upper limit in emergency mode is obtained by calculating this coefficient with the original upper limit of fairness constraints. Specifically, the comprehensive anomaly metric calculated in real time is compared with the preset anomaly trigger threshold. When the comprehensive anomaly metric exceeds the threshold, an anomaly event is determined to have occurred, and the system switches to emergency scheduling mode. In emergency scheduling mode, semantic adaptability and task urgency are set as core weights, and the upper limit of fairness constraints is adjusted according to the preset emergency degradation factor. The task allocation utility function is reconstructed based on the adjusted weights and constraint upper limits to obtain the adjusted scheduling utility function.

[0054] Based on the adjusted scheduling utility function, drone resources are prioritized for allocation to high-risk areas, generating emergency scenario allocation results. When not in emergency scheduling mode, the regional service gap for each non-high-risk area during an abnormal event is calculated. The dynamic fairness preference weights for the next scheduling cycle are then adjusted compensatorily based on the regional service gap, and compensatory scheduling is implemented for areas with regional service gaps to obtain the target path planning scheme. It should be noted that high-risk areas refer to task execution areas that are significantly affected by abnormal events, where task interruption may lead to serious consequences, or where there are significant safety hazards; these are the priority allocation targets for resources in emergency scheduling. Emergency scenario allocation results refer to the task allocation scheme obtained after prioritizing drone resource allocation to high-risk areas based on the adjusted scheduling utility function. Non-high-risk areas refer to other task execution areas not identified as high-risk areas; they are relatively less affected by abnormal events, and their task execution urgency is lower than that of high-risk areas. The regional service gap is the difference between the expected service demand and the actual service demand in non-high-risk areas during an abnormal event. Specifically, based on the adjusted scheduling utility function, the matching utility value between each high-risk area and the drone is calculated, and drone resources are allocated to the high-risk area with the highest utility value first, generating emergency scenario allocation results. After exiting the emergency scheduling mode, the expected service demand and actual service demand of each non-high-risk area during the abnormal event are statistically analyzed, and the difference between the two is calculated to obtain the regional service gap. According to the size of the regional service gap, the dynamic fairness preference weight of the corresponding area in the next scheduling cycle is increased to complete the compensatory correction. Based on the corrected weight, compensatory scheduling is implemented for the areas with gaps. Combining the emergency scenario allocation results, the path planning and execution order are planned to obtain the target path planning scheme.

[0055] In emergency scenarios, prioritizing resource allocation to high-risk areas can minimize the negative impact of abnormal events. After the anomaly ends, calculating regional service gaps and implementing compensatory corrections and scheduling can make up for service deficiencies in non-high-risk areas, ensuring overall service fairness. The final target path planning scheme takes into account both emergency protection and fairness compensation, improving the emergency response capability and service balance of the scheduling system.

[0056] This embodiment constructs a low-altitude service semantic grid, transforming the service demand characteristics of various urban areas into regional semantic vectors, and mapping UAV tasks to this grid to generate task semantic vectors. A semantic enhancement mechanism is used to calculate the semantic association strength between tasks and regions, constructing a multi-dimensional fairness index system, and generating dynamic fairness preference weights through adaptive correction. Finally, task allocation and path planning are performed based on dynamic weights and semantic association strength, improving the dynamic balancing service efficiency and fairness of multi-UAV scheduling, dynamically correcting regional service deviations, ensuring service accessibility for disadvantaged areas and populations, and adapting to complex urban environments and abnormal scenarios.

[0057] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 The method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) further includes steps S201 to S206 in step S30: Step S201: The cosine similarity algorithm is used to calculate the region semantic vector and the task semantic vector to obtain the vector similarity.

[0058] It should be noted that the cosine similarity algorithm is a calculation method used to measure the degree of directional similarity between two vectors in space. It represents the similarity by calculating the cosine value of the angle between the two vectors. The result ranges from -1 to 1. The closer the value is to 1, the more consistent the directions of the two vectors are and the higher the degree of similarity.

[0059] Using the cosine similarity algorithm to calculate the similarity between the region semantic vector and the task semantic vector can effectively avoid the interference of the magnitude of the vector dimension and accurately focus on the semantic feature matching degree represented by the vector direction.

[0060] Step S202: Obtain the basic matching degree based on vector similarity.

[0061] It should be noted that the vector similarity calculated using the cosine similarity algorithm is directly used as the basic matching degree between the task and the corresponding region, without the need to introduce any additional correction factors or weight coefficients. The basic matching degree obtained based on vector similarity can directly and objectively reflect the adaptation relationship between the task and the region at the semantic feature level, providing a unified and standardized initial benchmark for subsequent optimization of the matching degree by incorporating other factors.

[0062] Step S203: Calculate the regional service demand score corresponding to the regional semantic vector based on the semantic weight in the regional semantic vector and historical service coverage data.

[0063] It should be noted that semantic weight refers to the numerical value obtained by weighted fusion and normalization of regional features, which can quantitatively reflect the intensity of service demand and scenario adaptability of the three-dimensional cell. Historical service coverage data refers to historical data recording the coverage frequency, coverage duration, and service completion quality of drone services in each region within past scheduling cycles, which can reflect the past service demand of the region being met. Specifically, the semantic weight is extracted from the regional semantic vector, the historical service coverage data of the corresponding region is retrieved, and the semantic weight and the historical service coverage data are weighted and calculated based on a preset calculation logic to obtain the regional service demand score corresponding to the regional semantic vector.

[0064] Step S204: Process the regional service demand score using the Sigmoid function to generate a demand enhancement coefficient.

[0065] It should be noted that the demand enhancement coefficient is the coefficient value obtained after processing the regional service demand score using the Sigmoid function. This coefficient value is positively correlated with the regional service demand score, increasing as regional service demand increases. It is used to enhance the weight of high-demand regions in task allocation. The calculated regional service demand score is substituted into the Sigmoid function for a non-linear transformation, and a value between 0 and 1 is obtained through function mapping; this value is the demand enhancement coefficient.

[0066] Step S205: Multiply the basic matching degree and the demand enhancement coefficient by weight to obtain the initial semantic association value.

[0067] Step S206: Normalize the initial semantic association value to obtain the semantic association strength.

[0068] This embodiment calculates the similarity between the region semantic vector and the task semantic vector using a cosine similarity algorithm to obtain a basic matching degree. Combining the weights of the region semantic vector and historical service data, a region service demand score is calculated, and a demand enhancement coefficient is generated using a sigmoid function to strengthen the matching degree between the task and the region. The basic matching degree and the demand enhancement coefficient are weighted and multiplied to obtain an initial semantic association value, which is then normalized to obtain the semantic association strength. This improves the accuracy and fairness of UAV task allocation, optimizes the allocation of urban low-altitude service resources, and enhances service efficiency and coverage, particularly significantly improving service accessibility in remote and disadvantaged areas.

[0069] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 The method for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs) further includes steps S301 to S307 in step S40: Step S301: Calculate the theoretical service demand value of each region based on the semantic weights in the region semantic vector.

[0070] Specifically, the first step is to identify key semantic features that influence service demand, such as population density, infrastructure completeness, and transportation accessibility. Each semantic feature is assigned a weight value based on its degree of influence on service demand. Then, the semantic feature value for each region is multiplied by its corresponding semantic weight, and these products are summed to obtain a semantic vector for each region. Finally, using the information in the semantic vectors, the theoretical service demand value for each region is estimated through specific calculation methods (such as weighted summation).

[0071] Step S302: Based on the theoretical service demand value and historical task allocation data, obtain the actual service coverage value of each region.

[0072] It should be noted that historical task allocation data refers to historical records of the frequency, duration, and coverage of UAV task allocations in various regions during past scheduling cycles, reflecting the actual service resource input received by the region in the past. Specifically, firstly, the theoretical service demand values ​​pre-calculated for each region are retrieved, and specific records such as task allocation frequency, service duration, and coverage corresponding to each region are extracted from the historical task allocation data. This is done to obtain the basic data support needed to calculate the actual service coverage value. Then, according to preset association calculation rules, the theoretical service demand value and the corresponding historical task allocation data are weighted and fused for calculation. This is done to ensure that the calculation result reflects both the inherent demand attributes of the region and the actual service situation in the past. Finally, the calculated data is organized to obtain the actual service coverage value corresponding to each region. This is done to output a quantitative indicator that can intuitively reflect the degree to which the region's past service demand is met.

[0073] Step S303: Calculate the regional coverage deviation value between the theoretical service demand value and the actual service coverage value, and construct a regional coverage deviation index based on the regional coverage deviation value.

[0074] Specifically, first, the theoretical service demand and actual service coverage for each region are determined, and the difference between the two is calculated to obtain the regional coverage deviation value for each region. This is done to quantify the gap between the inherent service demand and the actual service supply of a region. Next, the regional coverage deviation values ​​for all regions are collected and weighted by combining the region's service priority weight and area weight. This is done to eliminate the impact of differences in the basic attributes of different regions on the deviation assessment. Finally, based on the weighted integration result, a regional coverage deviation index is constructed that can uniformly measure the service supply and demand matching situation of all regions. This is done to form a standardized evaluation basis for subsequent scheduling strategy optimization.

[0075] Step S304: Calculate the variance of the regional coverage deviation for all regions, and construct a regional service balance index based on the variance.

[0076] It's important to note that, firstly, all calculated regional coverage deviation values ​​for each area are collected to ensure data completeness and unique identifiers for each area. This is done to obtain the basic sample data needed to calculate the variance. Then, the variance calculation formula is used to calculate the variance of all deviation values, yielding the variance of the regional coverage deviation. This is done because variance effectively reflects the degree of data dispersion, aligning with the need to measure the evenness of coverage deviation distribution. Finally, the calculated variance is used as the core parameter and adjusted in conjunction with preset standardization coefficients to construct a regional service evenness index. This is done to transform the variance results into an intuitive and unified standard for evaluating evenness.

[0077] Step S305: Based on the proportion of disadvantaged groups and the regional coverage deviation value in the regional semantic vector, construct a compensation index for disadvantaged areas.

[0078] It should be noted that, firstly, the proportion of vulnerable groups recorded in the semantic vectors of each region is extracted, and the corresponding regional coverage deviation value is retrieved. This is done to obtain the core correlation elements for constructing the indicator and anchor the relationship between vulnerability attributes and service gaps. Then, weights are assigned to the proportion of vulnerable groups and the regional coverage deviation value, with a higher weight for the proportion of vulnerable groups to highlight fairness. The two are then weighted and summed to balance the degree of vulnerability with service gaps and strengthen attention to areas with high vulnerability and low coverage. Finally, the weighted result is normalized to obtain the vulnerable area compensation index for each region. This is done to form a unified and comparable quantitative standard for compensation priority.

[0079] Step S306: Integrate regional coverage deviation indicators, regional service balance indicators, and disadvantaged area compensation indicators to construct a multi-dimensional fairness indicator system.

[0080] It should be noted that, firstly, all quantitative data for the regional coverage deviation index, regional service balance index, and disadvantaged area compensation index are aggregated to ensure that each index matches the corresponding regional identifier. This is done to integrate the core data foundation for multi-dimensional fairness assessment. Secondly, differentiated weights are assigned to each of the three indicators, with the disadvantaged area compensation index having a higher weight than the other two, thus highlighting the focus on people's livelihood. Then, a weighted fusion algorithm is used to integrate the three indicators. This is done to balance the multiple fairness requirements of supply and demand gaps, overall balance, and compensation for the disadvantaged. Finally, the weighted fusion result is standardized and mapped to a unified assessment interval, completing the multi-dimensional fairness indicator system. This is done to form a comprehensive and comparable quantitative evaluation standard for fairness.

[0081] Step S307: Set initial fairness preference weights for each indicator in the multi-dimensional fairness indicator system according to the service scenario requirements.

[0082] It should be noted that the sum of the initial fairness preference weights is a preset value (e.g., 1). The initial fairness preference weights refer to the weight values ​​assigned to each fairness indicator in the initial stage of the multi-dimensional fairness indicator system. These weight values ​​reflect the relative importance of each indicator in assessing service fairness under the service scenario requirements. The initial fairness preference weights need to be set according to the specific service scenario requirements to ensure that the weight allocation meets the specific requirements for service fairness in different scenarios. Service scenario requirements include at least public service scenario requirements and commercial service scenario requirements. Public service scenario requirements typically refer to the need to provide basic public services to all members of society, such as healthcare and education. These scenarios have higher requirements for service fairness. Commercial service scenario requirements, on the other hand, involve more economic benefits and market competitiveness, and will adjust the allocation of services according to market demand and customer affordability.

[0083] This embodiment constructs a multi-dimensional fairness index system, comprehensively considering factors such as regional service demand, historical service coverage, service balance, and compensation for disadvantaged areas, to provide decision support for fairness optimization in UAV mission scheduling. It can dynamically adjust the allocation of service resources according to the needs of different service scenarios, improve the fairness and efficiency of services, and ensure the rationality and impartiality of resource allocation, especially in different scenarios such as public services and commercial services.

[0084] Based on the first embodiment of this application, this application also provides a multi-UAV low-altitude service scheduling result generation device, please refer to... Figure 4 The device includes: Semantic grid construction module 10 is used to construct a low-altitude service semantic grid based on three-dimensional spatial structure and multi-source urban data, and obtain regional semantic vectors. The three-dimensional spatial structure is based on geographic coordinates and flight altitude range. The multi-source urban data includes population census, POI data, road network data and terrain model data. The regional semantic vectors represent the service demand characteristics of each area of ​​the city. The task mapping module 20 is used to map the UAV task to be executed to the low-altitude service semantic grid and generate a task semantic vector, wherein the task semantic vector and the region semantic vector have the same dimensions. The semantic enhancement module 30 is used to process the region semantic vector and the task semantic vector region through a semantic enhancement mechanism to obtain the semantic association strength; The fairness index construction module 40 is used to construct a multi-dimensional fairness index system based on the regional semantic vector and the task semantic vector, and to set the initial fairness preference weights. The multi-dimensional fairness index system includes regional coverage deviation index, regional service balance index and disadvantaged area compensation index. The weight optimization module 50 is used to adaptively correct the initial fairness preference weights and generate dynamic fairness preference weights. The results module 60 is used to perform overall allocation and path planning for the unmanned aerial vehicle (UAV) tasks to be executed based on dynamic fairness preference weights and semantic association strength, so as to obtain the target path planning scheme.

[0085] The multi-UAV low-altitude service scheduling result generation device provided in this application, employing the multi-UAV low-altitude service scheduling result generation method described in the above embodiments, can solve the technical problem of how to improve the dynamic balancing service efficiency and fairness of multi-UAV scheduling and adapt to abnormal scenarios. Compared with the prior art, the beneficial effects of the multi-UAV low-altitude service scheduling result generation device provided in this application are the same as those of the multi-UAV low-altitude service scheduling result generation method described in the above embodiments, and other technical features in the multi-UAV low-altitude service scheduling result generation device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0086] In one embodiment, the semantic grid construction module 10 is further configured to: divide the urban low-altitude space into multiple three-dimensional cells according to geographic coordinates and flight altitude ranges, wherein the three-dimensional cells do not overlap; extract regional features of each three-dimensional cell from multi-source urban data, wherein the regional features include population density, proportion of vulnerable groups, coverage of public facilities, regional economic activity, and regional accessibility; normalize the extracted regional features to obtain normalized regional features; configure initial weights for the normalized regional features according to service scenario types, and generate semantic weights for each three-dimensional cell through weighted fusion, wherein the service scenario types include at least public service scenarios and commercial service scenarios; construct a low-altitude service semantic grid based on the spatial coordinate information of the three-dimensional cells, the regional features, and the semantic weights, and obtain a regional semantic vector corresponding to each region.

[0087] In one embodiment, the task mapping module 20 is further configured to parse the UAV task to be executed to obtain attribute information, wherein the attribute information includes task type, task priority, service object category, task urgency, and expected workload; to structure and organize the attribute information to form a set of original task features; to normalize the set of original task features to obtain normalized task features; to match the normalized task features with the semantic dimensions of the low-altitude service semantic grid to obtain dimension matching results; and to construct a task semantic vector based on the dimension matching results.

[0088] In one embodiment, the semantic enhancement module 30 is further configured to: calculate the region semantic vector and the task semantic vector using a cosine similarity algorithm to obtain a vector similarity; obtain a basic matching degree based on the vector similarity; calculate a region service demand score corresponding to the region semantic vector based on the semantic weights in the region semantic vector and historical service coverage data; process the region service demand score using a sigmoid function to generate a demand enhancement coefficient, wherein the demand enhancement coefficient increases as the region service demand increases; multiply the basic matching degree and the demand enhancement coefficient by weight to obtain an initial semantic association value; and normalize the initial semantic association value to obtain the semantic association strength.

[0089] In one embodiment, the fairness index construction module 40 is further configured to: calculate the theoretical service demand value of each region based on the semantic weights in the regional semantic vector; obtain the actual service coverage value of each region based on the theoretical service demand value and historical task allocation data; calculate the regional coverage deviation value between the theoretical service demand value and the actual service coverage value, and construct a regional coverage deviation index based on the regional coverage deviation value; calculate the variance of the regional coverage deviation of all regions, and construct a regional service balance index based on the variance; construct a disadvantaged region compensation index based on the proportion of disadvantaged groups in the regional semantic vector and the regional coverage deviation value; integrate the regional coverage deviation index, the regional service balance index, and the disadvantaged region compensation index to construct a multi-dimensional fairness index system; and set initial fairness preference weights for each index in the multi-dimensional fairness index system according to service scenario requirements, wherein the sum of each initial fairness preference weight is a preset value, and the service scenario requirements include at least public service scenario requirements and commercial service scenario requirements.

[0090] In one embodiment, the weight optimization module 50 is further configured to acquire historical service performance data, wherein the historical service performance data includes the service deviation change trend, task completion rate and service satisfaction of each region; collect current task execution status data in real time, wherein the current task execution status data includes task execution progress, UAV resource occupancy and real-time change data of regional demand; calculate the service deviation trend value of each region based on the historical service performance data and the current task execution status data using the first-order difference method; determine the weight adjustment direction according to the service deviation trend value; preliminarily correct the initial fairness preference weight based on the weight adjustment direction to obtain a first weight; perform boundary constraint processing on the first weight to obtain a second weight; and perform normalization processing on the second weight to generate dynamic fairness preference weight.

[0091] In one embodiment, the result module 60 is further configured to acquire the equipment capability parameters of each UAV, wherein the equipment capability parameters include endurance, payload limit, flight speed, and communication range; combine the dynamic fairness preference weight, the semantic association strength, and the equipment capability parameters to construct a task allocation utility function, wherein the task allocation utility function includes semantic fit, fairness constraints, and task urgency; with the goal of maximizing the task allocation utility function, combine the UAV payload constraints, flight range constraints, and unique task execution constraints to construct a task allocation model; use a reinforcement learning algorithm to solve the task allocation model to obtain the allocation result, wherein the reward function of the reinforcement learning algorithm includes semantic matching reward, fairness reward, and cost penalty, and the allocation result includes tasks, UAVs, and regions; based on the allocation result and combined with low-altitude airspace traffic rules, plan the optimal flight path and task execution order for each UAV to generate a conventional scenario path planning scheme; collect the UAV operating status, regional environmental status, and task execution status in real time; and according to the... The system calculates a comprehensive anomaly metric based on the drone's operational status, regional environmental status, and task execution status. This comprehensive anomaly metric is a weighted sum of all status parameters. When the comprehensive anomaly metric exceeds a preset anomaly trigger threshold, an anomaly event is identified, and the system switches to emergency scheduling mode. In emergency scheduling mode, the task allocation utility function is reconstructed using semantic adaptability and task urgency as core weights. The upper limit of fairness constraints is adjusted according to a preset emergency downgrade factor, resulting in an adjusted scheduling utility function. Based on the adjusted scheduling utility function, drone resources are prioritized for allocation to high-risk areas, generating emergency scenario allocation results. When the system is no longer in emergency scheduling mode, the regional service gap for each non-high-risk area during the anomaly event is calculated, where the regional service gap is the difference between expected and actual service demand. The dynamic fairness preference weights for the next scheduling cycle are compensated based on the regional service gap, and compensatory scheduling is implemented for areas with regional service gaps to obtain a target path planning scheme.

[0092] This application provides a multi-UAV low-altitude service scheduling result generation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the multi-UAV low-altitude service scheduling result generation method in the above embodiment 1.

[0093] The following is for reference. Figure 5This document illustrates a structural schematic diagram of a multi-UAV low-altitude service scheduling result generation device suitable for implementing embodiments of this application. The multi-UAV low-altitude service scheduling result generation device in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The multi-UAV low-altitude service scheduling result generation device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0094] like Figure 5 As shown, the multi-UAV low-altitude service scheduling result generation device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the multi-UAV low-altitude service scheduling result generation device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following can be connected to I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the multi-UAV low-altitude service scheduling result generation device to communicate wirelessly or wiredly with other devices to exchange data. Although various multi-UAV low-altitude service scheduling result generation devices are shown in the figure, it should be understood that implementation or possession of all of them is not required. More or fewer may be implemented alternatively.

[0095] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0096] The multi-UAV low-altitude service scheduling result generation device provided in this application, employing the multi-UAV low-altitude service scheduling result generation method described in the above embodiments, can solve the technical problem of how to improve the dynamic balancing service efficiency and fairness of multi-UAV scheduling and adapt to abnormal scenarios. Compared with the prior art, the beneficial effects of the multi-UAV low-altitude service scheduling result generation device provided in this application are the same as those of the multi-UAV low-altitude service scheduling result generation method described in the above embodiments, and other technical features in this multi-UAV low-altitude service scheduling result generation device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0097] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0098] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0099] This application provides a computer-readable medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the multi-UAV low-altitude service scheduling result generation method in the above embodiments.

[0100] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable medium may be any tangible medium containing or storing a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0101] The aforementioned computer-readable medium may be included in the multi-UAV low-altitude service scheduling result generation device; or it may exist independently and not be assembled into the multi-UAV low-altitude service scheduling result generation device.

[0102] The aforementioned computer-readable medium carries one or more programs that, when executed by a multi-UAV low-altitude service scheduling result generation device, enable the device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0103] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, all blocks in the flowcharts or block diagrams may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that all blocks in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using dedicated hardware-based implementations that perform the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0104] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0105] The readable medium provided in this application is a computer-readable medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for generating multi-UAV low-altitude service scheduling results. This method addresses the technical problem of improving the efficiency and fairness of dynamic balancing services in multi-UAV scheduling and adapting to abnormal scenarios. Compared with the prior art, the beneficial effects of the computer-readable medium provided in this application are the same as those of the multi-UAV low-altitude service scheduling result generation method provided in the above embodiments, and will not be elaborated upon here.

[0106] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the multi-UAV low-altitude service scheduling result generation method described above.

[0107] The computer program product provided in this application can solve the technical problem of how to improve the efficiency and fairness of dynamic balancing services for multi-UAV scheduling and adapt to abnormal scenarios. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the multi-UAV low-altitude service scheduling result generation method provided in the above embodiments, and will not be repeated here.

[0108] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for generating scheduling results for low-altitude services using multiple unmanned aerial vehicles (UAVs), characterized in that, The method includes: A low-altitude service semantic grid is constructed based on a three-dimensional spatial structure and multi-source urban data, and regional semantic vectors are obtained. The three-dimensional spatial structure is based on geographic coordinates and flight altitude ranges. The multi-source urban data includes population census data, POI data, road network data and terrain model data. The regional semantic vectors represent the service demand characteristics of each area of ​​the city. The unmanned aerial vehicle (UAV) mission to be executed is mapped to the low-altitude service semantic grid to generate a mission semantic vector, wherein the mission semantic vector and the region semantic vector have the same dimensions. The semantic association strength is obtained by processing the semantic vectors of the region and the semantic vector region of the task through a semantic enhancement mechanism. Based on the regional semantic vector and the task semantic vector, a multidimensional fairness index system is constructed, and an initial fairness preference weight is set. The multidimensional fairness index system includes a regional coverage deviation index, a regional service balance index, and a weak regional compensation index. The initial fairness preference weights are adaptively adjusted to generate dynamic fairness preference weights; Based on the dynamic fairness preference weights and semantic association strength, the unmanned aerial vehicle (UAV) tasks to be executed are allocated and paths are planned as a whole to obtain the target path planning scheme.

2. The method as described in claim 1, characterized in that, The steps of constructing a low-altitude service semantic grid based on three-dimensional spatial structure and multi-source urban data, and obtaining regional semantic vectors, include: The urban low-altitude space is divided into multiple three-dimensional cells according to geographical coordinates and flight altitude ranges, wherein the three-dimensional cells do not overlap. Regional features are extracted from each of the three-dimensional cells from multi-source urban data, wherein the regional features include population density, proportion of vulnerable groups, coverage of public facilities, regional economic activity, and regional accessibility; The extracted regional features are normalized to obtain normalized regional features; Initial weights are configured for the normalized regional features based on the service scenario type, and semantic weights for each three-dimensional cell are generated through weighted fusion, wherein the service scenario type includes at least public service scenarios and commercial service scenarios; Based on the spatial coordinate information of the three-dimensional cells, the regional features, and the semantic weights, a low-altitude service semantic grid is constructed, and a regional semantic vector corresponding to each region is obtained.

3. The method as described in claim 1, characterized in that, The step of mapping the unmanned aerial vehicle (UAV) mission to be executed to the low-altitude service semantic grid and generating a mission semantic vector includes: The drone mission to be executed is analyzed to obtain attribute information, which includes mission type, mission priority, service object category, mission urgency and expected workload. The attribute information is structured and organized to form a set of original features for the task; The original feature set of the task is normalized to obtain the normalized task features; The normalized task features are matched with the semantic dimensions of the low-altitude service semantic grid to obtain the dimension matching result. A task semantic vector is constructed based on the dimension matching results.

4. The method as described in claim 1, characterized in that, The step of processing the semantic vector of the region and the semantic vector region of the task through a semantic enhancement mechanism to obtain the semantic association strength includes: The cosine similarity algorithm is used to calculate the vector similarity between the semantic vector of the region and the semantic vector of the task; The basic matching degree is obtained based on the vector similarity. Based on the semantic weights in the regional semantic vector and historical service coverage data, the regional service demand score corresponding to the regional semantic vector is calculated. The regional service demand score is processed by the Sigmoid function to generate a demand enhancement coefficient, wherein the demand enhancement coefficient increases as the regional service demand increases. The initial semantic association value is obtained by multiplying the basic matching degree with the demand enhancement coefficient by weight. The initial semantic association value is normalized to obtain the semantic association strength.

5. The method as described in claim 1, characterized in that, The step of constructing a multi-dimensional fairness index system based on the region semantic vector and the task semantic vector, and setting initial fairness preference weights, includes: The theoretical service demand value of each region is calculated based on the semantic weights in the region semantic vector. Based on the theoretical service demand value and historical task allocation data, the actual service coverage value of each region is obtained; Calculate the regional coverage deviation between the theoretical service demand value and the actual service coverage value, and construct a regional coverage deviation index based on the regional coverage deviation value; Calculate the variance of the regional coverage deviation for all regions, and construct a regional service balance index based on the variance; Based on the proportion of disadvantaged groups in the semantic vector of the region and the coverage deviation value of the region, a compensation index for disadvantaged regions is constructed. By integrating the regional coverage deviation index, the regional service balance index, and the compensation index for disadvantaged areas, a multi-dimensional fairness index system is constructed. Based on the service scenario requirements, initial fairness preference weights are set for each indicator in the multidimensional fairness indicator system, wherein the sum of the initial fairness preference weights is a preset value, and the service scenario requirements include at least public service scenario requirements and commercial service scenario requirements.

6. The method as described in claim 1, characterized in that, The step of adaptively adjusting the initial fairness preference weights to generate dynamic fairness preference weights includes: Acquire historical service performance data, including service deviation trends, task completion rates, and service satisfaction in each region. Real-time collection of current task execution status data, including task execution progress, UAV resource usage, and real-time changes in regional demand; Based on the historical service performance data and the current task execution status data, the service deviation trend value of each region is calculated using the first-order difference method. The direction of weight adjustment is determined based on the service deviation trend value; The initial fairness preference weights are initially corrected based on the weight adjustment direction to obtain the first weights; The first weight is subjected to boundary constraint processing to obtain the second weight; The second weight is normalized to generate dynamic fairness preference weights.

7. The method as described in claim 1, characterized in that, The step of performing overall allocation and path planning for the unmanned aerial vehicle (UAV) tasks to be executed based on the dynamic fairness preference weights and semantic association strength to obtain a target path planning scheme includes: Obtain the equipment capability parameters of each UAV, including endurance, payload limit, flight speed, and communication range; By combining the dynamic fairness preference weight, the semantic association strength, and the device capability parameters, a task allocation utility function is constructed, wherein the task allocation utility function includes semantic fit, fairness constraints, and task urgency. With the goal of maximizing the task allocation utility function, a task allocation model is constructed by combining UAV payload constraints, flight range constraints, and unique task execution constraints. The task allocation model is solved using a reinforcement learning algorithm to obtain the allocation result. The reward function of the reinforcement learning algorithm includes a semantic matching reward, a fairness reward, and a cost penalty term. The allocation result includes tasks, drones, and regions. Based on the allocation results and combined with low-altitude airspace traffic rules, the optimal flight path and mission execution sequence are planned for each UAV, generating a path planning scheme for a common scenario. Real-time collection of drone operating status, regional environmental status, and mission execution status; A comprehensive anomaly metric is calculated based on the UAV's operating status, regional environmental status, and mission execution status, wherein the comprehensive anomaly metric is the weighted sum of each status parameter; When the comprehensive anomaly metric exceeds the preset anomaly trigger threshold, an anomaly event is determined to have occurred, and the system switches to emergency dispatch mode. In the emergency scheduling mode, the task allocation utility function is reconstructed with semantic adaptability and task urgency as the core weights, and the upper limit of the fairness constraint is adjusted according to the preset emergency degradation factor to obtain the adjusted scheduling utility function. Based on the adjusted scheduling utility function, drone resources are allocated to high-risk areas first, generating emergency scenario allocation results; When not in the emergency dispatch mode, calculate the regional service gap of each non-high-risk area during the abnormal event, where the regional service gap is the difference between the expected service demand and the actual service demand; Based on the regional service gap, the dynamic fairness preference weight for the next scheduling cycle is compensated and adjusted. Compensatory scheduling is implemented for regions with regional service gaps to obtain the target path planning scheme.

8. A device for generating scheduling results for low-altitude services of multiple unmanned aerial vehicles (UAVs), characterized in that, The device includes: The semantic grid construction module is used to construct a low-altitude service semantic grid based on a three-dimensional spatial structure and multi-source urban data, and obtain regional semantic vectors. The three-dimensional spatial structure is based on geographic coordinates and flight altitude ranges. The multi-source urban data includes population census data, POI data, road network data and terrain model data. The regional semantic vectors represent the service demand characteristics of each area of ​​the city. The task mapping module is used to map the UAV task to be executed to the low-altitude service semantic grid and generate a task semantic vector, wherein the task semantic vector and the region semantic vector have the same dimensions. The semantic enhancement module is used to process the semantic vector of the region and the semantic vector region of the task through a semantic enhancement mechanism to obtain the semantic association strength; The fairness index construction module is used to construct a multi-dimensional fairness index system based on the regional semantic vector and the task semantic vector, and to set initial fairness preference weights. The multi-dimensional fairness index system includes regional coverage deviation index, regional service balance index, and weak region compensation index. The weight optimization module is used to adaptively correct the initial fairness preference weights to generate dynamic fairness preference weights. The results module is used to perform overall allocation and path planning for the unmanned aerial vehicle (UAV) tasks to be executed based on the dynamic fairness preference weights and semantic association strengths, so as to obtain the target path planning scheme.

9. A device for generating low-altitude service scheduling results for multiple unmanned aerial vehicles (UAVs), characterized in that, The device includes: a memory, a processor, and a multi-UAV low-altitude service scheduling result generation program stored in the memory and running on the processor, the multi-UAV low-altitude service scheduling result generation program being configured to implement the steps of the multi-UAV low-altitude service scheduling result generation method as described in any one of claims 1-7.

10. A storage medium, characterized in that, The storage medium stores a multi-UAV low-altitude service scheduling result generation program, which, when executed by a processor, implements the steps of the multi-UAV low-altitude service scheduling result generation method as described in any one of claims 1-7.

Images