A shared unmanned aerial vehicle cabinet data analysis management method and system

Abstract

The application provides a shared unmanned aerial vehicle cabinet data analysis management method and system, which comprises the following steps: obtaining the behavior record data set of historical unmanned aerial vehicles, calculating the cabinet health score through the behavior record data set, and establishing the behavior dependence weight between the unmanned aerial vehicles and the cabinets; obtaining the spatial coordinates of the cabinets and the unmanned aerial vehicles and the historical abnormal rate of the cabinets, and generating a dynamic space-time graph containing multiple time slices; predicting the number of empty spaces and the health state change trend of the cabinets in the future period by using the dynamic space-time graph; generating the cabinet access recommendation priority according to the number of empty spaces and the health state change trend of the cabinets in the future period, combining the current position of the unmanned aerial vehicle, and delivering the cabinet access recommendation priority to the unmanned aerial vehicle for execution; and feeding back the behavior health graph based on the access result, and dynamically adjusting the health score and the behavior dependence weight.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of shared unmanned aerial vehicle cabinets, and particularly relates to a shared unmanned aerial vehicle cabinet data analysis management method and system. BACKGROUND

[0002] With the rise of emerging application scenarios such as urban intelligent logistics, security inspection and emergency distribution, shared unmanned aerial vehicle systems are being widely deployed and applied as a flexible and efficient means of transportation and service. As an important part of shared unmanned aerial vehicle systems, distributed unmanned aerial vehicle cabinets bear the key functions of charging, docking, data interaction and scheduling of unmanned aerial vehicles, and have gradually become an important node of urban Internet of Things and unmanned aerial vehicle networks. However, the existing unmanned aerial vehicle cabinet systems still have many limitations in management and data analysis, and are difficult to meet the needs of the increasingly complex distributed shared unmanned aerial vehicle system. Specifically, the existing technology generally takes a single cabinet as a management unit, lacks collaborative analysis and management capabilities for multiple cabinets and cross-site, resulting in low cabinet resource utilization, unbalanced unmanned aerial vehicle scheduling, and problems such as overload, insufficient space, and deteriorating health status of some cabinets. At the same time, the existing unmanned aerial vehicle access strategy usually adopts simple rules based on current space or nearby access, without considering dynamic factors such as historical usage behavior, health wear and tear of the cabinet, urgency of the task, and remaining power of the unmanned aerial vehicle, lacking precise prediction of future resource load trends and risks, and thus in actual application, problems such as queuing, drifting, access failure and even safety hazards of unmanned aerial vehicles during access peak often occur. In addition, traditional systems usually adopt a passive management mode, only reporting simple data and updating status after access or task execution, without realizing active collaboration and global optimization among cabinets, unmanned aerial vehicles and users. Especially in complex environments with multiple cabinets, independent operation of each cabinet and closed data make it difficult to achieve dynamic balance and intelligent scheduling of regional resources. The existing technology lacks an integrated solution for health status prediction, resource collaborative scheduling and dynamic intelligent analysis in the context of shared unmanned aerial vehicles, multiple cabinets and distributed scenarios, which seriously hinders the operational efficiency and user experience of the system in high-load and dynamic environments. In view of the above problems, it is urgent to design an innovative system that can comprehensively analyze the complex interaction between unmanned aerial vehicles and cabinets, and realize intelligent prediction and dynamic management based on multiple cabinets and a global perspective, in order to realize the intelligentization and sustainable operation and maintenance of shared unmanned aerial vehicle systems. SUMMARY

[0003] The purpose of the present application is to provide a shared unmanned aerial vehicle cabinet data analysis management method and system, which can realize precise prediction of cabinet resource pressure, health status and future load trend by establishing a unified analysis framework of multi-dimensional information such as unmanned aerial vehicle behavior, cabinet usage, resource load and environmental characteristics, and provide intelligent decision-making basis for unmanned aerial vehicle access management and resource coordination between cabinets.

[0004] To achieve the above object, in the first aspect of the present application provides a shared unmanned aerial vehicle cabinet data analysis management method, the method comprises the following steps:

[0005] Obtain the behavior record data set of historical unmanned aerial vehicles, including access frequency, charging duration and task intensity, and the resource use record of the cabinet, and construct a behavior health atlas, calculate the cabinet health score through the behavior record data set, and establish the behavior dependence weight between the unmanned aerial vehicle and the cabinet;

[0006] Obtain the spatial coordinates of the cabinet and the unmanned aerial vehicle and the historical abnormal rate of the cabinet, generate a dynamic space-time graph containing multiple time slices in combination with the behavior health atlas, wherein each time slice corresponds to the calculation of the edge weight of the dynamic space-time graph once, and the edge weight is updated in real time to reflect the cabinet resource pressure and health risk; the dynamic space-time graph is used to support multi-period prediction;

[0007] Predict the number of empty slots and the change trend of the health status of the cabinet in the future period by using the dynamic space-time graph;

[0008] According to the number of empty slots and the change trend of the health status of the cabinet in the future period, generate a cabinet access recommendation priority in combination with the current position of the unmanned aerial vehicle, and deliver it to the unmanned aerial vehicle for execution;

[0009] Optimize the behavior health atlas based on the access result feedback, dynamically adjust the health score and the behavior dependence weight; wherein the access result feedback includes the access success rate, the actual deviation of the cabinet health and the access behavior deviation.

[0010] Further, the resource use record includes the number of times of slot activation, total charging energy consumption, number of times of start and stop of the cooling system and fault abnormal record.

[0011] Further, the cabinet health score is calculated by fusing the number of times of slot activation, charging energy consumption and number of times of start and stop of the cooling system of the cabinet, and the behavior dependence weight is determined based on the access frequency, task intensity and charging duration of the unmanned aerial vehicle.

[0012] Further, the edge weight calculation of the dynamic space-time graph includes: combining the cabinet health score, the distance of the spatial coordinates of the cabinet and the unmanned aerial vehicle, and the historical abnormal rate, and introducing an abnormal rate control factor to balance resource utilization and health risk; the abnormal rate control factor is set according to the system safety strategy.

[0013] Further, the use of the dynamic space-time graph to predict the number of empty slots and the change trend of the health status of the cabinet in the future period specifically includes:

[0014] Obtain the dynamic space-time graph, construct a prediction input;

[0015] By jointly modeling the temporal and structural features of the nodes in the graph, joint optimization is performed, and the optimization objective is expressed as follows:

[0016]

[0017] Where Θ represents the model parameters, M represents the total number of cabinets, and i represents the current cabinet unit; Cabinet C represents the actual and predicted future time period T. i The number of vacant positions; ΔH i (t+T) and β represents the change in actual and predicted health scores, and β∈[0,1] is the balancing factor for the loss in health trend prediction.

[0018] By combining graph structure embedding and temporal recursion, the prediction matrix P = {V} for the next time period T is output. i (t+T),ΔH i (t+T)};where V i (t+T),ΔH i (t+T) represent the number of empty rack slots and the trend of health status changes in the future time period, respectively.

[0019] Furthermore, the predicted input specifically includes:

[0020] Obtain the cabinet health scores of the current target cabinet and adjacent cabinets, and combine them with the edge weights of the dynamic spatiotemporal graph to obtain the prediction input by summing all the current target cabinets and adjacent cabinets.

[0021] Furthermore, the recommendation priority is generated through a scoring mechanism; the scoring mechanism is obtained by integrating the predicted number of vacant spaces, the real-time distance between the drone and the cabinet, and the health change trend, and dynamically adjusts the weights to optimize resource allocation.

[0022] Furthermore, the scoring mechanism lowers the recommendation priority for racks with rapidly deteriorating health and raises the priority for racks with ample space and close proximity.

[0023] Furthermore, the optimization of the behavioral health graph based on access result feedback, and the dynamic adjustment of health scores and behavioral dependency weights, specifically includes:

[0024] Get feedback on the access results;

[0025] The behavioral health graph is adaptively updated based on the feedback of the access results to obtain the updated rack health score and behavioral dependency weight.

[0026] A second aspect of the present invention provides a shared drone cabinet data analysis and management system, the system comprising:

[0027] A health atlas construction module is configured to acquire a behavior record data set of historical unmanned aerial vehicles, including access frequency, charging duration and task intensity, and resource usage records of the cabinets, and to construct a behavior health atlas, calculate a cabinet health score based on the behavior record data set, and establish a behavior dependency weight between the unmanned aerial vehicles and the cabinets;

[0028] A dynamic topology generation module is configured to acquire spatial coordinates of the cabinets and the unmanned aerial vehicles and historical abnormal rates of the cabinets, generate a dynamic space-time graph containing multiple time slices in combination with the behavior health atlas, wherein each time slice corresponds to calculation of an edge weight of the dynamic space-time graph once, and the edge weight is updated in real time to reflect cabinet resource pressure and health risks, and the dynamic space-time graph is used to support multi-period prediction;

[0029] A resource prediction module is configured to predict the number of empty spaces and health state change trends of the cabinets in a future period based on the dynamic space-time graph;

[0030] An intelligent recommendation module is configured to generate a cabinet access recommendation priority in combination with a current position of the unmanned aerial vehicles based on the number of empty spaces and health state change trends of the cabinets in the future period, and to deliver the recommendation priority to the unmanned aerial vehicles for execution;

[0031] A feedback optimization module is configured to optimize the behavior health atlas based on access result feedback, and to dynamically adjust the health score and the behavior dependency weight, wherein the access result feedback includes an access success rate, a cabinet health actual deviation and an access behavior deviation.

[0032] The present application has at least the following beneficial technical effects:

[0033] The present application addresses the problems of lack of intelligent resource prediction capability, cooperative scheduling mechanism and health state management in existing shared unmanned aerial vehicle cabinet systems, and through establishment of a unified analysis framework of multi-dimensional information such as unmanned aerial vehicle behavior, cabinet usage, resource load and environmental characteristics, the present application can accurately predict cabinet resource pressure, health condition and future load trends, and provide intelligent decision basis for unmanned aerial vehicle access management and resource coordination between cabinets. At the same time, based on dynamic monitoring and analysis of the cabinet operating state, the system can intelligently generate a cooperative scheduling strategy for multiple cabinets, guide the unmanned aerial vehicles to reasonably select target cabinets in the access process, relieve resource pressure and wear and tear risks of some cabinets, and improve the unmanned aerial vehicle access success rate and system stability. Unlike existing scheduling modes that only rely on instant empty spaces or fixed rules, the present application realizes intelligent closed-loop management of prediction-scheduling-feedback by integrating historical behavior, health prediction and space-time characteristics, and breaks through the limitations of passive response of traditional systems to dynamic changes in the environment.

[0034] Through the above mechanism, the application can significantly optimize the cabinet network resource distribution, improve the intelligent scheduling capability and health management level of the shared unmanned aerial vehicle system, and ultimately realize efficient collaboration and system-level optimization among the unmanned aerial vehicle, the cabinet and user demand. BRIEF DESCRIPTION OF DRAWINGS

[0035] The application is further described by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the application. For ordinary skilled in the art, other drawings can be obtained without creative labor according to the following drawings.

[0036] Figure 1 A flow chart of a shared unmanned aerial vehicle cabinet data analysis management method of the application. DETAILED DESCRIPTION

[0037] The embodiments of the application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the application, and cannot be understood as a limitation on the application.

[0038] As shown in Figure 1 , the shared unmanned aerial vehicle cabinet data analysis management method provided by the embodiment of the application comprises:

[0039] S1, acquiring a behavior record data set of a historical unmanned aerial vehicle, including access frequency, charging duration and task intensity, and resource usage record of the cabinet, and constructing a behavior health graph, calculating a cabinet health score through the behavior record data set, and establishing a behavior dependence weight between the unmanned aerial vehicle and the cabinet.

[0040] Specifically, this step aims to establish a behavior-health relationship graph between the unmanned aerial vehicle and the cabinet, and provide global data support for subsequent resource prediction and access scheduling links. For the problem that in the shared unmanned aerial vehicle cabinet system, the behaviors of high-frequency access and heavy task execution of the unmanned aerial vehicle cause different degrees of long-term impact on the hardware resources (such as power supply, slot, cooling system) of the cabinet, the application proposes a behavior-health graph construction mechanism (Behavior-Health Graph, abbreviated as BHG-Graph) of unmanned aerial vehicle-cabinet. This mechanism systematically mines the correlation between the behavior data of the unmanned aerial vehicle and the resource usage data of the cabinet, establishes a graph data structure for dynamic health perception, and makes up for the shortcomings of the existing system that only focuses on the real-time resource state of the cabinet and ignores the historical loss accumulation and behavior dependence.

[0041] Specifically, the system first collects a behavior record data set D U of a historical unmanned aerial vehicle, including access frequency f u , charging duration tu and task intensity m u And the server rack resource usage records D C Includes the number of times the slot has been used (n) c Total charging energy consumption e c Number of times the cooling system starts and stops (s) c With fault and abnormal record a c Subsequently, the system will... j With rack C i As nodes in the graph, node attributes embed their respective historical behaviors and resource characteristics, and the computer cabinet's health score H is based on a behavior-resource coupling degree scoring mechanism. i The specific definition is as follows:

[0042]

[0043] Among them, H i Indicates rack C i The health score ranges from [0,1], with smaller values ​​indicating worse health; N represents the cabinet C. i Total number of drones served in the past; For the j-th drone to cabinet C i Access frequency; For the j-th drone in C i The cumulative charging time (in hours); The task intensity of the j-th UAV (e.g., m=1 for normal inspection tasks, m=2 for emergency tasks); n c For rack C i Total number of slot activations; e c The cumulative charging energy consumption of this cabinet (in kWh); s c α represents the total number of start-stop cycles of the server rack's cooling system (add 1 to avoid a denominator of 0); α, β, and γ are the task factors for fitting historical data, satisfying α + β + γ = 1.

[0044] In addition, to quantify the load dependence of a single drone on a specific rack, the system represents drone node U in the diagram. j With rack node C i Establish directed edges between nodes, and define the behavior dependency weight r of each edge. ij The formula is as follows:

[0045]

[0046] Where, r ij For U drones j With rack C i The larger the edge weight, the denser and more intense the drone's access to the cabinet. Indicates access frequency. represents the task intensity, represents the charging duration.

[0047] Through the above mechanism, the system finally outputs the behavior health graph G1={N, E, H i ,r ij}, where N is the node set in the graph, E is the directed edge set in the graph, H i is the cabinet health score, and r ij is the behavior-dependent edge weight. The graph will be directly used as an input feature of the dynamic spatio-temporal topology in the subsequent steps, providing dynamic and reliable behavior-health correlation information for the prediction model and intelligent scheduling module.

[0048] S2, obtain the spatial coordinates of the cabinet and the unmanned aerial vehicle and the historical abnormal rate of the cabinet, combine the behavior health graph, and generate a dynamic spatio-temporal graph containing multiple time slices, wherein each time slice corresponds to a calculation of the edge weight of the dynamic spatio-temporal graph, and the edge weight is updated in real time to reflect the cabinet resource pressure and health risk; the dynamic spatio-temporal graph is used to support multi-period prediction.

[0049] Specifically, this step focuses on the actual needs of multi-cabinet distribution and dynamic changes of unmanned aerial vehicle access in a shared unmanned aerial vehicle cabinet system, and aims to build a dynamic spatio-temporal topology structure for resource prediction and health management, providing accurate and scenario-adaptive structured input for subsequent prediction tasks. In view of the particularity of the shared unmanned aerial vehicle cabinet system, we propose a health-behavior-driven spatio-temporal graph construction mechanism (HBD-STG), the core idea of which is: using the health graph G1={N, E, H i ,r ij} generated in step 1, combining the current spatial positions of the cabinet and the unmanned aerial vehicle, realizing the comprehensive optimization modeling of the graph node attributes and the edge weight, and thus generating a dynamic spatio-temporal graph G2(t) of multiple time slices. In traditional spatio-temporal modeling methods, only the spatial distance or time sequence interaction between nodes is considered, while in this scheme, the calculation of the edge weight considers three-dimensional factors: behavior frequency, health status, and spatial proximity, and innovative control items are introduced for the patent scene to improve the applicability of the graph structure in resource scheduling prediction.

[0050] The specific construction mechanism is as follows: first, the system obtains the health score H i and the behavior-dependent weight r ij output by step 1, and synchronously collects the spatial coordinates (x i , y i ) and (x j , y j ) of the cabinet and the unmanned aerial vehicle. On this basis, the system defines the edge weight calculation formula as follows:

[0051]

[0052] wherein w ij represents the UAV U j and the cabinet C i The edge weight in the graph G2; r ij is the behavior-dependent weight, derived from step 1; d ij is the spatial distance between the UAV and the cabinet, H i is the health score of the cabinet; a i represents the historical abnormality rate (such as the frequency of failure or alarm records) of the cabinet C i , is the average abnormality rate of all cabinets; λ is an abnormality rate control factor, λ ∈ [0, 1], which is set according to the system safety policy.

[0053] The abnormality rate control term is introduced in the formula embodies the forward-looking management capability of the present application for the running risk of the cabinet in the shared cabinet scenario. Assuming that the historical abnormality frequency of a cabinet is high ij , the edge weight w ij is amplified, reminding the system that the cabinet is in a potential high-risk state, and the weight will be more inclined to low-risk cabinets in subsequent prediction, ensuring the safety of UAV access and the stability of resources. In addition, this mechanism can dynamically adjust λ, balancing the operation goals of maximizing resource utilization and minimizing health risk.

[0054] In the time dimension, to cope with the continuous changes of UAV access mode and cabinet health status, the system constructs a dynamic graph G2(t) for multiple time slices t, each time slice corresponds to a w ij calculation, thus forming a set of spatio-temporal graph sequences {G2(t1), G2(t2),..., G2(t n )}. This spatio-temporal sequence is input into the subsequent prediction model to realize global perception of future resource load and health status trends.

[0055] Through this mechanism, the present application realizes efficient conversion from the behavior health atlas G1 to the dynamic prediction graph G2(t), providing a strongly associated input structure for system prediction and scheduling. This mechanism, particularly for the problems of lack of health perception and lagging scheduling in shared UAV cabinet systems, innovatively introduces abnormality control terms and multi-factor edge weight modeling, significantly improving the sensitivity of the system to cabinet state changes and the control ability of scheduling safety, avoiding the one-sidedness of relying only on the current spatial distance or a single load indicator in the prior art.

[0056] S3, using the dynamic spatio-temporal graph to predict the number of empty spaces and the trend of health status changes of the cabinet in the future period.

[0057] In particular, this step uses the dynamic spatio-temporal graph G2(t) generated in step 2 to predict the evolution trend of the resource load state (such as the number of empty spaces) and the health state of the distributed cabinets within the future time window. The purpose is to provide prediction support for subsequent unmanned aerial vehicle access recommendations. In a shared unmanned aerial vehicle cabinet system, affected by factors such as unmanned aerial vehicle access peaks, task types, and uneven spatial distribution, the cabinet resources and health state are highly dynamic in time and space. Traditional short-term static prediction cannot effectively cope with this. Therefore, this step innovatively designs a health-driven graph time series prediction mechanism (referred to as HDGTP), which combines dynamic spatio-temporal graph structures to achieve collaborative prediction of resources and health under multiple cabinets. We take the dynamic graph sequence G2(t) = {N, E, w ij (t), H i (t)} output by step 2 as input, where w ij (t) is the edge weight of the spatio-temporal graph (combining space, health, and behavior dependency information), and H i (t) is the health score of cabinet C i at the current time. By jointly modeling the temporal and structural features of the nodes in the graph, the system can capture both the local spatial influence (neighborhood unmanned aerial vehicle-cabinet relationship) and the effect of health factors on resource load changes.

[0058] In the design of the prediction model, we propose a health-coupled graph time series prediction model (H-CGTP) with the following core optimization objectives:

[0059]

[0060] where Θ is the model parameter, M is the total number of cabinets, V i (t+T) and V are the real and predicted number of empty spaces for cabinet C i over the next T time periods, ΔH i (t+T) and ΔH are the real and predicted health score changes, and β ∈ [0, 1] is the balance factor for health trend prediction loss. To strengthen the coupling modeling of spatio-temporal dependency and health score on resource load, we particularly introduce a health-spatio-temporal correlation term as an input feature to construct a graph-based prediction input:

[0061]

[0062] where x i (t) is the comprehensive input feature of node C i at time t, N(i) is the set of adjacent cabinets for cabinet C i , w ij (t) is the behavior-space edge weight between the unmanned aerial vehicle and the cabinet at time t, H i (t) and Hj (t) is the cabinet C i and its adjacent cabinet C j health score.

[0063] Through the design of x i (t), the system realizes the input mechanism of adjacent health perception + space dependence + behavior coupling, which is beneficial to capture the potential influence of the surrounding cabinet health change trend on the resource load of the current cabinet. For example, when the health of the adjacent cabinet deteriorates, the system can predict that the current cabinet will face higher access pressure, so as to predict the trend of resource shortage in advance.

[0064] Finally, the HDGTP mechanism outputs the prediction matrix P = {V i (t+T), ΔH i (t+T)} based on the G2(t) spatiotemporal graph input, through the combination of graph structure embedding and temporal recursion, and provides it for the subsequent access recommendation system.

[0065] S4, according to the number of empty positions and the health state change trend of the cabinet in the future period, combined with the current position of the unmanned aerial vehicle, generate the cabinet access recommendation priority, and issue it to the unmanned aerial vehicle for execution.

[0066] Specifically, this step focuses on receiving the prediction matrix P = {V i (t+T), ΔH i (t+T)} output by step 3, and designing an unmanned aerial vehicle access recommendation mechanism for a distributed shared unmanned aerial vehicle cabinet system. The goal of this step is to provide optimal cabinet access suggestions for each unmanned aerial vehicle to be accessed based on the system's predicted empty position change trend V i (t+T) and health state evolution ΔH i (t+T) within the future T period, to improve the resource utilization and health risk avoidance ability of the system. The present application proposes a health-load collaborative access recommendation mechanism (HLRI-Mechanism), which comprehensively considers the empty position change trend, health degradation trend of the cabinet within the future time window, and the spatial distance between the current unmanned aerial vehicle position and the cabinet, and establishes an access priority ranking for unmanned aerial vehicles.

[0067] The specific recommendation mechanism is as follows: the system calculates the access recommendation score R i for each cabinet C i based on the prediction results, and the score mechanism is based on three indicators of resource availability, health safety and space scheduling cost, and the formula is as follows:

[0068]

[0069] Wherein, R i is the cabinet C iAccess recommendation rating; V i (t+T) represents the predicted number of future empty rack spaces; d ij For the drone U to be connected j With rack C i The current spatial distance between them, in meters, to avoid a zero denominator plus a 1 balance factor; ΔH i (t+T) represents the change in the health score of the cabinet within the future time period T (the absolute value reflects the rate of health degradation); α and β are the scoring weights, satisfying α+β=1, which can be dynamically adjusted according to the system strategy.

[0070] The innovation is reflected in:

[0071] Vacancy-Distance Coupling Term It dynamically integrates resource abundance and spatial proximity, preventing the one-way deviation in traditional scheduling schemes where high-resource racks are located far away or close-range racks are short of resources, thus reflecting a dynamic balance mechanism between space and resources.

[0072] Health Change Penalty Item | ΔH i (t+T)| can automatically reduce the recommended priority of cabinets whose health status is rapidly deteriorating, and has health perception capabilities to prevent drones from densely accessing cabinets that are about to decline in health.

[0073] Subsequently, the system based on R i Sort the drones from highest to lowest priority to generate the access priority matrix S = {r1, r2, ..., r...} n} where r1 is the rack number with the highest recommendation priority, and n is the number of candidate racks. The recommendation matrix S will be issued to the drone as the final access command, guiding it to select racks with abundant resources, proximity, and better health trends, thus achieving proactive scheduling for the future.

[0074] Through the HLRI-Mechanism, the system establishes a prediction-driven + multi-factor fusion access recommendation scheme in the access recommendation stage. It fully absorbs the results of previous predictions and forms a dynamic access recommendation capability with triple perception of available space, space, and health. This solves industry pain points such as strong access blindness, high health risk, and serious resource imbalance in distributed UAV systems, and supports the complete realization of the system's intelligent access management function.

[0075] S5. Optimize the behavioral health graph based on access result feedback, and dynamically adjust the health score and behavior dependency weights; wherein the access result feedback includes access success rate, actual rack health offset and access behavior deviation.

[0076] Specifically, this step designs an adaptive optimization mechanism for the behavior health graph G1, completes the self-learning closed loop of the system, improves the ability of the health graph to describe the relationship between the behavior of the unmanned aerial vehicle and the health of the cabinet under dynamic environment, and ensures that the graph structure continues to evolve with the changes in the environment and behavior.

[0077] Specifically, this step proposes a graph self-optimization mechanism based on access feedback (BFGO-Mechanism). The system receives the access recommendation execution data of step 4, including the following three core feedback indicators:

[0078] Access success rate δ j : Whether the unmanned aerial vehicle U j successfully accesses according to the recommendation, δ j = 1 for success and δ j = 0 for failure;

[0079] Actual cabinet health deviation η i : The difference between the actual change and the predicted health change of the cabinet C i health score, i.e.

[0080] Access behavior deviation The difference between the access behavior of the unmanned aerial vehicle U j in the non-recommended cabinet and the expected access behavior.

[0081] The system takes the above feedback as input to optimize the health score H i of the cabinet node in G1 and the unmanned aerial vehicle-cabinet behavior edge weight r ij , and the specific adaptive update is as follows:

[0082] H i ′ = H i - λ1·η i (7)

[0083]

[0084] where H i ′ and r ij ′ are the updated cabinet health score and behavior edge weight, respectively; λ1, λ2 are adaptive learning rate factors; reflects the change trend of the actual access behavior of the unmanned aerial vehicle U j to the cabinet C i , if the unmanned aerial vehicle does not access according to the recommendation (δ j = 0), the system automatically increases the adjustment of r ij , reflecting that the current behavior relationship has deviated from the historical dependence of the graph.

[0085] This is the first time that the access deviation degree The index is specifically used for sensing the deviation of recommended-actual access, solves the problem that the traditional health atlas is difficult to capture dynamic behavior mode changes, and the health deviation η proposed by the application i The health score H i in the atlas is dynamically adjusted to be more consistent with the real resource usage state, which significantly improves the timeliness and health sensing accuracy of the atlas.

[0086] As Figure 1 shown, the application embodiment further provides a shared unmanned aerial vehicle cabinet data analysis management system, which comprises:

[0087] A health atlas construction module is configured to acquire a historical unmanned aerial vehicle behavior record data set, including access frequency, charging duration and task intensity, and cabinet resource usage records, and to construct a behavior health atlas, calculate a cabinet health score through the behavior record data set, and establish a behavior dependence weight between the unmanned aerial vehicle and the cabinet.

[0088] A dynamic topology generation module is configured to acquire spatial coordinates of the cabinet and the unmanned aerial vehicle and historical abnormality rate of the cabinet, generate a dynamic space-time graph containing multiple time slices in combination with the behavior health atlas, wherein each time slice corresponds to calculation of an edge weight of the dynamic space-time graph once, and the edge weight is updated in real time to reflect cabinet resource pressure and health risk; and the dynamic space-time graph is used to support multi-period prediction.

[0089] A resource prediction module is configured to predict the number of cabinet empty spaces and health state change trend in a future period based on the dynamic space-time graph.

[0090] An intelligent recommendation module is configured to generate a cabinet access recommendation priority in combination with a current position of the unmanned aerial vehicle according to the number of cabinet empty spaces and the health state change trend in the future period, and to issue the recommendation priority to the unmanned aerial vehicle for execution.

[0091] A feedback optimization module is configured to optimize the behavior health atlas based on access result feedback, and to dynamically adjust the health score and the behavior dependence weight; wherein the access result feedback includes access success rate, cabinet health actual deviation and access behavior deviation.

[0092] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0093] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiments is merely a logical function division, and there can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0094] The functions described above can be implemented in software, firmware, hardware, or any combination thereof. Moreover, the functions will be implemented in software as functions of an application program running on a computer, in one embodiment. Furthermore, if implemented in software, the functions can be stored in or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media include both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, or twisted pair, then the coaxial cable, fiber optic cable, or twisted pair are included in the definition of medium. Disk and disc, as used herein, include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), and Blu-Ray® disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0095] In light of the above, it should be understood that numerous variations and modifications can be made to the embodiments of the present application without departing from the principles and spirit of the application. Accordingly, the scope of the application is defined only by the following claims and equivalents thereto.

Claims

1. A method for sharing drone cabinet data analysis management, characterized in that, The method comprises the following steps: Obtain the behavior record data set of historical unmanned aerial vehicles, including access frequency, charging duration and task intensity, and the resource usage record of the cabinet, and construct a behavior health atlas, calculate the cabinet health score through the behavior record data set and the resource usage record of the cabinet, and establish the behavior dependence weight between the unmanned aerial vehicle and the cabinet; the resource usage record includes the number of times of slot activation, total charging energy consumption, number of times of starting and stopping of the heat dissipation system, and fault anomaly record; Obtain the spatial coordinates of the cabinet and the unmanned aerial vehicle and the historical abnormal rate of the cabinet, generate a dynamic space-time graph containing multiple time slices in combination with the behavior health atlas, wherein each time slice corresponds to the calculation of the edge weight of the dynamic space-time graph once, and the edge weight is updated in real time to reflect the cabinet resource pressure and health risk; the dynamic space-time graph is used to support multi-period prediction; Predict the number of empty slots and the change trend of the health status of the cabinet in the future period by using the dynamic space-time graph; Generate a cabinet access recommendation priority according to the number of empty slots and the change trend of the health status of the cabinet in the future period, in combination with the current position of the unmanned aerial vehicle, and deliver it to the unmanned aerial vehicle for execution; Optimize the behavior health atlas based on the access result feedback, and dynamically adjust the health score and the behavior dependence weight; wherein the access result feedback includes the access success rate, the actual deviation of the cabinet health and the access behavior deviation; The edge weight calculation of the dynamic space-time graph is as follows: ； wherein, with is the spatial coordinate of the cabinet and the UAV, represents the UAV and the cabinet in the dynamic space-time graph ; is the behavior-dependent weight; is the spatial distance between the UAV and the cabinet, ; is the health score of the cabinet; represents the historical abnormality rate of the cabinet , is the average abnormality rate of all cabinets; is the abnormality rate control factor, according to the system safety policy; is the abnormality rate control term, used to balance resource utilization maximization and health risk.

2. The method of claim 1, wherein, The cabinet health score is calculated by fusing the number of times of slot activation, charging energy consumption and the number of times of starting and stopping of the heat dissipation system of the cabinet, and the behavior dependence weight is determined based on the access frequency, task intensity and charging duration of the unmanned aerial vehicle.

3. The method of claim 1, wherein, The prediction of the number of empty slots and the change trend of the health status of the cabinet in the future period by using the dynamic space-time graph comprises: Obtain the dynamic space-time graph and construct a prediction input; Jointly model the time sequence features and structure features of the nodes in the graph, perform joint optimization, and the optimization target is represented as follows: ； wherein, is a model parameter, is the total number of cabinets; i is the current cabinet; and are the real and predicted future number of vacant slots in the cabinet during the time period; and are the real and predicted change in health score, is a balancing factor for the health trend prediction loss. By combining graph structure embedding with temporal recurrence, the output prediction matrix of future time period ; wherein, respectively represent the number of empty positions and the health status change trend of the cabinet in the future time period.

4. The method of claim 3, wherein, The prediction input specifically comprises: Obtain the cabinet health score of the current target cabinet and the adjacent cabinet, combine the edge weight of the dynamic space-time graph, and obtain the prediction input by summing all the current target cabinet and the adjacent cabinet.

5. The method of claim 1, wherein, The recommendation priority is generated by a scoring mechanism; the scoring mechanism is obtained by fusing the predicted number of empty slots, the real-time distance between the unmanned aerial vehicle and the cabinet, and the health change trend, and the weight is dynamically adjusted to optimize resource allocation.

6. The method of claim 5, wherein, The scoring mechanism reduces the recommendation priority of the cabinet with rapid health deterioration, and increases the priority of the cabinet with rich empty slots and close distance.

7. The method of claim 1, wherein, The optimization of the behavior health atlas based on the access result feedback, and the dynamic adjustment of the health score and the behavior dependence weight, specifically comprises: Obtain the access result feedback; Adaptively update the behavior health atlas according to the access result feedback to obtain the updated cabinet health score and behavior dependence weight.

8. A shared drone cabinet data analytics management system, characterized by, The system comprises: The health atlas construction module is configured to obtain a behavior record data set of historical unmanned aerial vehicles, including access frequency, charging duration, and task intensity, and resource usage records of the cabinets, and to construct a behavior health atlas. The health score of the cabinet is calculated based on the behavior record data set and the resource usage records of the cabinet, and the behavior dependence weight between the unmanned aerial vehicle and the cabinet is established. The resource usage records include the number of times of slot activation, total charging energy consumption, number of times of start and stop of the heat dissipation system, and fault anomaly records. The dynamic topology generation module is configured to obtain spatial coordinates of the cabinet and the unmanned aerial vehicle and historical anomaly rates of the cabinet, and to generate a dynamic space-time graph containing multiple time slices in combination with the behavior health atlas. Each time slice corresponds to the calculation of the edge weight of the dynamic space-time graph once, and the edge weight is updated in real time to reflect the cabinet resource pressure and health risk. The dynamic space-time graph is used to support multi-period prediction. The resource prediction module is configured to predict the number of empty slots and the change trend of the health status of the cabinet in a future period based on the dynamic space-time graph. The intelligent recommendation module is configured to generate a cabinet access recommendation priority based on the number of empty slots and the change trend of the health status of the cabinet in the future period in combination with the current position of the unmanned aerial vehicle, and to deliver the recommendation priority to the unmanned aerial vehicle for execution. The feedback optimization module is configured to optimize the behavior health atlas based on access result feedback, and to dynamically adjust the health score and the behavior dependence weight. The access result feedback includes access success rate, actual deviation of cabinet health, and access behavior deviation. The edge weight calculation of the dynamic space-time graph is as follows: ； wherein, with is the spatial coordinate of the cabinet and the UAV, represents the UAV and the cabinet in the dynamic space-time graph ; is the behavior-dependent weight; is the spatial distance between the UAV and the cabinet, ; is the health score of the cabinet; represents the historical abnormality rate of the cabinet , is the average abnormality rate of all cabinets; is the abnormality rate control factor, according to the system safety policy; is the abnormality rate control term, used to balance the maximum resource utilization and health risk.

Images