An unmanned aerial vehicle highway inspection task intelligent planning management system

Abstract

The present application relates to the technical field of industrial unmanned aerial vehicle application and intelligent scheduling, and particularly discloses an unmanned aerial vehicle highway inspection task intelligent planning management system, which comprises a task input module, an intelligent planning engine, a task execution management module, a data hub and a feedback optimization module; the task input module receives and checks and analyzes inspection task requirements and highway basic information; the intelligent planning engine decomposes task priorities based on highway multi-dimensional attributes, dynamically allocates unmanned aerial vehicle resources and generates an optimal path meeting multiple constraints; the task execution management module issues adaptive instructions, monitors unmanned aerial vehicle states and processes inspection data; the data hub uniformly stores whole-process data; and the feedback optimization module optimizes a decision model based on historical data. The present application improves inspection planning accuracy and resource utilization, guarantees data quality, realizes self-evolution and ensures efficient and reliable highway inspection.

Description

Technical Field

[0001] This invention relates to the field of industrial drone application and intelligent scheduling technology, specifically to an intelligent planning and management system for drone highway inspection tasks. Background Technology

[0002] In the existing technologies for applying industrial drones to highway inspection, the commonly available drone-based highway inspection systems have many shortcomings and are difficult to meet the needs for efficient, accurate, and stable inspection.

[0003] On the one hand, existing systems lack intelligent and refined design in the inspection task planning stage. Most systems fail to decompose tasks based on multi-dimensional road segment attributes (such as technical grade and historical damage data) in the highway infrastructure information, and cannot prioritize inspections according to the importance and risk of road segments. This results in untimely inspections of key road segments and wasted resources on non-key road segments. On the other hand, the allocation of drone resources relies heavily on simple matching logic, failing to fully consider the real-time performance status of the drone fleet (such as battery level and sensor accuracy) and predictive reliability. It also fails to dynamically allocate resources with the goal of maximizing overall task efficiency, which easily leads to resource mismatch problems such as high-capacity drones being idle and low-capacity drones undertaking high-difficulty tasks, affecting inspection efficiency.

[0004] On the other hand, existing systems have shortcomings in task execution management and data management capabilities. During task execution, some systems cannot adapt to the command formats of different UAV models, which can easily lead to command issuance failures or execution deviations. Insufficient monitoring of the UAV's real-time status (such as position, flight attitude, and battery life) makes it difficult to detect flight anomalies and task progress issues in a timely manner, potentially leading to task interruptions or UAV safety risks. In terms of data management, existing systems often suffer from the problem of "data silos"—data is stored in a scattered manner. Task input data, execution dynamic data, and historical task data are not uniformly integrated and standardized, resulting in low data reuse rates and failing to provide effective data support for subsequent task planning and optimization.

[0005] Therefore, an intelligent planning and management system for UAV highway inspection tasks is proposed to address the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent planning and management system for unmanned aerial vehicle (UAV) highway inspection tasks, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A smart planning and management system for unmanned aerial vehicle (UAV) highway inspection missions includes:

[0009] The task input module is used to receive the inspection task requirements and basic highway information configured by the user.

[0010] The intelligent planning engine, connected to the task input module, is used to intelligently decompose and plan the received inspection task requirements, including:

[0011] The task decomposition unit, based on the multi-dimensional road segment attributes in the basic highway information, including technical grade and historical damage data, decomposes the overall inspection task into inspection sub-tasks with different priorities through a preset weight model.

[0012] The resource allocation unit, connected to the task decomposition unit, is used to dynamically allocate appropriate drone resources to each sub-task based on the priority, spatiotemporal attributes, and real-time performance status and predicted reliability of the available drone fleet, with the goal of maximizing global task efficiency.

[0013] The path planning unit, connected to the resource allocation unit, is used to generate the optimal inspection path for each allocated UAV under multiple constraints such as flight safety, data acquisition quality, and endurance.

[0014] The task execution management module connects to the intelligent planning engine and is used to send planning schemes to the corresponding drones and receive the drones' real-time status and inspection data.

[0015] The data hub communicates and interacts with the task input module, the intelligent planning engine, and the task execution management module, respectively, and is used to store and provide the basic data required for task planning, dynamic data generated during task execution, and historical task data.

[0016] The intelligent planning engine's decision-making process relies on data from the data hub; the system also includes a feedback optimization module, which continuously optimizes the intelligent planning engine's decision-making model based on historical execution data.

[0017] As a preferred embodiment, the task input module includes a configuration receiving unit, a data verification unit, and a data parsing unit connected in sequence.

[0018] The configuration receiving unit is used to receive raw data of inspection task requirements and highway basic information input by the user through the human-machine interaction interface;

[0019] The data verification unit is used to perform format verification and integrity checks on the original data and generate verified data.

[0020] The data parsing unit is used to parse the verified data, extract multi-dimensional road segment attributes from the inspection task parameters and highway basic information, including technical grade and historical damage data, and output the extracted parameters and attributes to the data hub for storage, so that the intelligent planning engine can call them.

[0021] As a preferred approach, the task decomposition unit, based on multi-dimensional road segment attributes in the highway infrastructure information, including technical grade and historical damage data, decomposes the overall inspection task into inspection sub-tasks with different priorities using a pre-defined weight model, including:

[0022] Obtain basic highway information related to the inspection task from the data hub. The basic highway information includes attribute data of multiple road sections, and the attribute data includes at least technical grade and historical damage data.

[0023] The attribute data of each road segment is standardized, the technical level is mapped to a preset numerical level, and historical disease data is converted into standardized indicators that reflect the severity of the disease.

[0024] Based on a pre-set weighting model, the numerical level and standardized indicators of each road segment are weighted and fused to generate a comprehensive priority score for each road segment.

[0025] Based on the comprehensive priority score of each road segment, the road segments are divided into multiple priority intervals, and each priority interval corresponds to a priority level;

[0026] Road segments within the same priority range are grouped into independent inspection sub-tasks, and each inspection sub-task is assigned a corresponding priority level, thereby completing the decomposition of the overall inspection task.

[0027] As a preferred embodiment, the resource allocation unit is used to dynamically allocate appropriate UAV resources to each subtask based on its priority, spatiotemporal attributes, and the real-time performance status and predicted reliability of the available UAV fleet, with the goal of maximizing overall task efficiency. This includes:

[0028] The priority identifiers and spatiotemporal attribute information of each inspection subtask are obtained from the data hub, and the real-time performance status data and predictive reliability indicators of the available drone fleet are obtained simultaneously.

[0029] Based on real-time performance status data and predicted reliability indicators, the effectiveness of each UAV is evaluated, and a comprehensive capability score for each UAV is generated.

[0030] Based on the priority identifier and spatiotemporal attribute information of each inspection sub-task, and combined with the comprehensive capability score, the initial matching degree between each UAV and each inspection sub-task is calculated.

[0031] With the goal of maximizing global task efficiency, considering spatiotemporal constraints and resource competition, the initial matching degree is reconciled and optimized to generate the optimal UAV resource allocation sequence.

[0032] Based on the drone resource allocation sequence, each inspection subtask is dynamically bound to the corresponding drone to complete the resource allocation.

[0033] As a preferred embodiment, the task execution management module includes a task instruction encapsulation unit, a communication link management unit, a status monitoring unit, and a data processing unit;

[0034] The task instruction encapsulation unit is connected to the path planning unit in the intelligent planning engine. It is used to receive the optimal inspection path generated for each UAV and encapsulate each path and its corresponding inspection sub-task information into a structured task instruction set that is specific to the UAV model and can be directly executed.

[0035] The communication link management unit is connected to the task instruction encapsulation unit. Before the task is issued, it verifies the communication link status with the target UAV and ensures that the link is stable and reliable. Then, it securely and accurately issues the structured task instruction set to the corresponding UAV.

[0036] The status monitoring unit is connected to the communication link management unit and is used to continuously receive real-time status data uploaded by the UAV through the communication link during the mission execution. The real-time status data includes at least location information, flight attitude, remaining battery power and sensor working status. The unit also analyzes and monitors the real-time status data to generate a real-time assessment of the UAV's health status and mission execution progress.

[0037] The data processing unit is connected to the status monitoring unit and is used to receive and parse the data collected by the UAV in real time during the inspection process. It performs preliminary quality verification and classification on the parsed data and sends the verified inspection data and the real-time evaluation results generated by the status monitoring unit to the data center for storage.

[0038] As a preferred solution, the task execution management module is used to distribute planning schemes to the corresponding drones and receive real-time status and inspection data from the drones, including:

[0039] Receive the optimal inspection path generated for each UAV by the path planning unit in the intelligent planning engine, and encapsulate each path and its corresponding inspection subtask information into a structured task instruction set that is specific to the UAV model and can be directly executed;

[0040] Before issuing the task, verify the status of the communication link with the target drone. After ensuring that the link is stable and reliable, the structured task instruction set is safely and accurately issued to the corresponding drone.

[0041] During mission execution, the system continuously receives real-time status data uploaded by the UAV through the communication link. The real-time status data includes at least location information, flight attitude, remaining battery power, and sensor working status. The system also analyzes and monitors the real-time status data to generate a real-time assessment of the UAV's health status and mission execution progress.

[0042] The system receives and parses the data collected in real time during the inspection process from the drone, performs preliminary quality verification and classification on the parsed data, and sends the verified inspection data and the real-time evaluation results to the data center for storage.

[0043] As a preferred solution, the data hub is used to store and provide the basic data required for task planning, dynamic data generated during task execution, and historical task data, including:

[0044] The data receiving unit is used to receive basic highway information from the task input module, real-time status and inspection data from the task execution management module, and intermediate data generated during the planning process from the intelligent planning engine.

[0045] The data preprocessing unit, connected to the data receiving unit, is used to perform consistency verification, format standardization, and redundancy removal on the data received by the data receiving unit, and generate standardized data.

[0046] The data classification and archiving unit, connected to the data preprocessing unit, is used to classify standardized data into basic data, dynamic data, or historical task data according to their attributes and sources, and add corresponding data tags.

[0047] The data storage management unit, connected to the data classification and archiving unit, is used to store classified and tagged data into different dedicated databases according to categories and to create data indexes;

[0048] The data service and provision unit, connected to the data storage management unit, is used to respond to data requests from the intelligent planning engine or feedback optimization module, and retrieve and output the required data from the corresponding database based on the data index.

[0049] As a preferred approach, the feedback optimization module is used to continuously optimize the decision model of the intelligent planning engine based on historical execution data, including:

[0050] Historical task execution data is obtained from the data hub. This data includes task decomposition results, resource allocation schemes, path planning schemes, and corresponding actual execution data.

[0051] Based on the acquired historical task execution data, calculate task execution performance indicators, which include at least task completion rate, resource utilization rate, and data collection quality.

[0052] Compare and analyze the calculated actual performance indicators with the expected performance indicators in the planning stage to identify performance deviations;

[0053] Based on the results of the performance deviation analysis, adjust the parameters of the decision model in the intelligent planning engine. The decision model should include at least the weight model of the task decomposition unit and the matching model of the resource allocation unit.

[0054] The adjusted decision-making model was used to simulate and plan historical tasks, and the consistency between the simulation planning results and the actual execution data was evaluated.

[0055] If the degree of agreement reaches the preset optimization threshold, the adjusted decision model will be updated to the intelligent planning engine.

[0056] As can be seen from the technical solution provided by the present invention above, the intelligent planning and management system for UAV highway inspection tasks provided by the present invention has the following beneficial effects:

[0057] To ensure the reliability of the data source, the standardized verification and parsing of the task input module provides accurate and standardized data for the intelligent planning engine, avoiding planning deviations.

[0058] To achieve precise planning of inspection tasks, the intelligent planning engine breaks down task priorities, optimizes drone resource allocation and paths, and improves overall task efficiency and resource utilization.

[0059] To ensure stable mission execution, the mission execution management module adapts to UAV commands, monitors status, and verifies data, reducing the risk of mission interruption and ensuring the quality of recovered data.

[0060] Unified data management, with a data hub integrating data from the entire process, eliminates data silos, enables data sharing and reuse, and improves data consistency and utilization efficiency;

[0061] Drive the system's self-evolution; the feedback optimization module adjusts the decision-making model based on historical data, reducing manual maintenance costs and enhancing the system's adaptability to various scenarios and long-term performance.

[0062] Improve the quality and efficiency of highway inspections, prioritize the investigation of defects in key road sections, shorten inspection time, and provide reliable support for highway maintenance and traffic safety. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the overall structure of an intelligent planning and management system for unmanned aerial vehicle (UAV) highway inspection tasks according to the present invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0065] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0066] like Figure 1 As shown, this embodiment of the invention provides an intelligent planning and management system for unmanned aerial vehicle (UAV) highway inspection tasks, including:

[0067] The task input module is used to receive the inspection task requirements and basic highway information configured by the user.

[0068] The intelligent planning engine, connected to the task input module, is used to intelligently decompose and plan the received inspection task requirements, including:

[0069] The task decomposition unit, based on the multi-dimensional road segment attributes in the basic highway information, including technical grade and historical damage data, decomposes the overall inspection task into inspection sub-tasks with different priorities through a preset weight model.

[0070] The resource allocation unit, connected to the task decomposition unit, is used to dynamically allocate appropriate drone resources to each sub-task based on the priority, spatiotemporal attributes, and real-time performance status and predicted reliability of the available drone fleet, with the goal of maximizing global task efficiency.

[0071] The path planning unit, connected to the resource allocation unit, is used to generate the optimal inspection path for each allocated UAV under multiple constraints such as flight safety, data acquisition quality, and endurance.

[0072] The task execution management module connects to the intelligent planning engine and is used to send planning schemes to the corresponding drones and receive the drones' real-time status and inspection data.

[0073] The data hub communicates and interacts with the task input module, the intelligent planning engine, and the task execution management module, respectively, and is used to store and provide the basic data required for task planning, dynamic data generated during task execution, and historical task data.

[0074] The intelligent planning engine's decision-making process relies on data from the data hub; the system also includes a feedback optimization module, which continuously optimizes the intelligent planning engine's decision-making model based on historical execution data.

[0075] In this embodiment, the task input module serves as the initial data receiving and preprocessing stage of the UAV highway inspection task intelligent planning and management system. It is the foundation for ensuring the accuracy and efficiency of subsequent intelligent planning, task execution, and other processes. Through standardized human-computer interaction, data verification, and parsing processes, it transforms the inspection requirements configured by the user and the basic highway information into standardized data that the system can directly access, providing reliable data support for the intelligent planning engine. The task input module includes a configuration receiving unit, a data verification unit, and a data parsing unit connected in sequence.

[0076] The configuration receiving unit is used to receive raw data of inspection task requirements and highway basic information input by the user through the human-machine interaction interface;

[0077] Furthermore, the configuration receiving unit serves as the direct interface between the module and the user. Its core function is to receive two types of raw data input by the user through a human-computer interaction interface: one type is the inspection task requirements, covering the inspection scope, inspection time, inspection focus (such as special inspections of specific road sections), data collection accuracy requirements, etc., set by the user; the other type is basic highway information, containing attribute data of multiple road sections of the highway to be inspected, at least covering the technical grade of the road section (such as expressway, first-class highway, second-class highway, etc.), historical damage data (such as the occurrence time, location, and severity records of past road surface cracks, potholes, settlement, and other damages), and other key information. Through a user-friendly human-computer interaction design, this unit supports users to submit data through visual input, file import, and other methods, ensuring that users can conveniently and accurately convey the relevant requirements and basic information of the inspection task.

[0078] The data verification unit is used to perform format verification and integrity checks on the original data and generate verified data.

[0079] Furthermore, the data verification unit receives the raw data output by the configuration receiving unit. Its core function is to perform format verification and integrity checks on the raw data, generate verified data, and filter invalid data to ensure data quality in subsequent processes. Specifically, the operation is as follows:

[0080] Format validation: The system checks the raw data entered by the user according to the preset standard data format (such as date format, road segment code format, and defect data record format). For example, the inspection time must conform to the format of "year-month-day hour:minute:second", and the road segment technical grade must be selected from the preset options of "expressway, first-class highway, second-class highway, third-class highway, and fourth-class highway". If the format entered by the user does not conform to the preset standard, it is judged as a format error.

[0081] Completeness check: Compare the original data with the necessary field list of the inspection task and the highway basic information to check for any missing fields. For example, the inspection task requires necessary fields such as "inspection scope (segment number)" and "inspection deadline," while the highway basic information requires necessary fields such as "segment number, technical grade, and historical damage records (if any)." If any necessary field is missing, the data is considered incomplete. For original data with incorrect format or incomplete data, the data verification unit will provide the user with specific error information (such as "Inspection time format is incorrect, please enter it in the format of 'year-month-day hour:minute:second'" or "Segment number is missing, please supplement it"). After the user corrects the data, it will be received and verified again until the data format is compliant and the information is complete, generating the verified data.

[0082] The data parsing unit is used to parse the verified data, extract multi-dimensional road segment attributes from the inspection task parameters and highway basic information, including technical grade and historical damage data, and output the extracted parameters and attributes to the data hub for storage, so that the intelligent planning engine can call them.

[0083] Furthermore, the data parsing unit connects to the verified data output by the data verification unit. Its core function is to perform structured parsing of the verified data, extract the key parameters and attributes required by the intelligent planning engine, and output them to the data central storage. Specifically, the operation is as follows:

[0084] Inspection task parameter extraction: From the verified inspection task requirement data, the road segment number corresponding to the inspection range, the start and end time of the inspection, the data collection accuracy threshold, and the special inspection requirement identifier (such as whether to focus on the inspection of special road segments such as bridges and tunnels) are parsed and extracted to form a structured inspection task parameter set.

[0085] Highway basic information attribute extraction: From the verified highway basic information data, multi-dimensional attributes of each road segment are extracted through parsing. Key attributes include the technical grade of the road segment (e.g., extracting "expressway" as the technical grade attribute value for the road segment), historical damage data (e.g., extracting specific historical damage records for a road segment such as "October 5, 2023, K100+500, crack length 3 meters, width 5 millimeters," and converting them into structured damage information). After parsing, the data parsing unit formats the extracted inspection task parameter set and the multi-dimensional attribute data of the highway segments according to the storage specifications of the data hub, and then outputs them to the data hub for storage, so that the intelligent planning engine can call them in subsequent stages such as task decomposition and resource allocation.

[0086] The configuration receiving unit employs a human-computer interaction interface technology based on graphical user interface (GUI) design specifications, supporting users to input data through two core methods: first, visual form input, where the system provides preset input forms, allowing users to fill in or select information in corresponding fields (such as selecting road section technical level from a dropdown menu or selecting inspection time from a date selector), reducing input difficulty; second, standardized file import, supporting users to import files conforming to the system's preset formats (such as Excel and CSV), containing structured data of inspection task requirements and basic highway information. The system reads the file content through an interface and converts it into raw data that can be processed internally. This technology simplifies the user operation process, reduces human input errors, and ensures the initial accuracy of the raw data.

[0087] The data verification unit employs a dual verification technology of "rule matching + field comparison." Format verification is based on preset regular expressions and enumeration rules: for formatted data such as dates and codes, regular expressions (e.g., matching the format "YYYY-MM-DD HH:mm:ss") are used to verify data format compliance; for enumeration data such as technical levels and inspection types, the input value is compared with a system-preset list of enumerated values ​​to determine if it is within the legal range; integrity checks are based on a preset list of necessary fields, using a field traversal comparison method to check whether the original data contains all fields in the list, triggering an error if any field is missing. This technology, through explicit verification rules, ensures that the data entering the parsing stage has a unified format and complete information.

[0088] The data parsing unit employs structured parsing technology, based on a pre-defined data dictionary and parsing templates. The system pre-establishes a data dictionary for inspection tasks and a data dictionary for basic highway information, clearly defining the meaning, data type, and parsing rules for various data fields (e.g., extracting "K100-K150" as the road segment range parameter from "Inspection Range: K100-K150 section," and extracting attributes such as "Location (K100+500), Damage Type (Crack), Damage Size (3 meters long, 5 millimeters wide)" from historical damage descriptions). Simultaneously, corresponding parsing templates are designed for different types of input data (e.g., form input data, file import data), defining the field mapping relationships and extraction logic within the templates. During parsing, the system calls the corresponding template, performing field splitting and attribute extraction on the validated data according to the rules of the data dictionary, converting unstructured or semi-structured data into structured data to ensure the intelligent planning engine can directly access it.

[0089] The interactive storage technology between the data parsing unit and the data hub is based on a standardized application programming interface (API). The data parsing unit formats and encapsulates the parsed structured data according to the API data transmission protocol (such as the RESTful API protocol) preset by the data hub, generating a data packet that meets the requirements. Then, it sends a data storage request to the data hub through the API interface. After receiving the request, the data hub verifies the data packet format. If the verification is successful, it stores the data in the corresponding data table (such as the inspection task parameter table or the highway section attribute table) and returns a successful storage response to the data parsing unit. This technology ensures that the parsed data can be safely and accurately transmitted to the data hub, realizing reliable data storage and subsequent retrieval.

[0090] In this embodiment, the intelligent planning engine, as the core decision-making component of the intelligent planning and management system for UAV highway inspection tasks, connects to the task input module on one end to receive inspection task requirements and basic highway information. On the other end, it relies on basic data, dynamic data, and historical data provided by the data center to support decision-making. Through intelligent decomposition of inspection tasks, dynamic allocation of UAV resources, and generation of optimal inspection paths, it achieves precise matching of "task-resource-path," providing a scientifically feasible planning scheme for subsequent task execution. The core function of the intelligent planning engine is to perform intelligent processing of the received inspection task requirements throughout the entire process, covering three core stages: firstly, through the task input module... The task decomposition unit breaks down the overall inspection task into sub-tasks of different priorities based on the multi-dimensional road segment attributes, achieving refined task breakdown. Secondly, the resource allocation unit dynamically matches suitable drone resources to sub-tasks based on their characteristics and drone performance status, aiming to maximize overall task efficiency. Finally, the path planning unit generates the optimal inspection path for each drone under multiple constraints, including flight safety, data acquisition quality, and drone endurance. The entire decision-making process is supported by data from the data hub, ultimately outputting a complete planning scheme to the task execution management module to ensure efficient and orderly inspection tasks. This includes:

[0091] The task decomposition unit, based on multi-dimensional road segment attributes in the highway infrastructure information, including technical grade and historical damage data, decomposes the overall inspection task into inspection sub-tasks with different priorities through a pre-set weight model, thereby achieving differentiated division of inspection priorities; its specific processing flow is as follows:

[0092] The data center obtains basic highway information related to the inspection task. The basic highway information includes attribute data of multiple road sections, covering at least the technical grade of the road section (such as expressway, first-class highway, second-class highway, etc.) and historical damage data (such as the occurrence and severity of past road surface defects such as cracks, potholes, and settlement).

[0093] Data Acquisition: Retrieve basic highway information related to the current inspection task from the data center. This information includes attribute data for multiple road sections, covering at least the technical grade of the road section (such as expressway, first-class highway, second-class highway, etc.) and historical damage data (such as records of past road surface cracks, potholes, settlement, and other damages and their severity).

[0094] Data standardization processing: The attribute data of each road segment is uniformly standardized and converted, and the technical level of the road segment is mapped to a numerical level according to the system's preset rules (for example, highways are mapped to level 5, first-class highways are mapped to level 4, and so on, to achieve the quantification of different technical levels); at the same time, historical disease data is converted into standardized indicators reflecting the severity of disease (for example, based on parameters such as the scope and depth of disease impact, the severity of disease is divided into a quantitative index of 1-10, with higher values ​​indicating more severe disease).

[0095] Comprehensive priority score calculation: Based on the preset weight model, the standardized technical grade value and the severity index of defects for each road segment are weighted and integrated. In the weight model, different attributes (technical grade, historical defect data) are assigned preset weights (for example, the weight of historical defect data is higher than that of technical grade to highlight the inspection priority of road segments with high defect incidence). The comprehensive priority score of each road segment is obtained by weighted summation.

[0096] Priority interval division: Based on the comprehensive priority score of all road segments, multiple priority intervals are set (for example, a score of 8-10 is high priority, 5-7 is medium priority, and 1-4 is low priority), and each interval corresponds to a specific priority level;

[0097] Subtask combination: road segments within the same priority range are integrated into independent inspection subtasks, and each subtask is labeled with a corresponding priority level, thus completing the fine decomposition of the overall inspection task and providing clear task units for subsequent resource allocation;

[0098] The resource allocation unit, connected to the task decomposition unit, has the core function of dynamically allocating appropriate drone resources to each sub-task based on its priority, spatiotemporal attributes (such as inspection time window and road segment geographical location), combined with the real-time performance status of the available drone fleet (such as remaining battery power, sensor accuracy, and flight speed) and predicted reliability (such as recent failure probability and task completion rate), with the goal of maximizing overall task efficiency. Its specific processing flow is as follows:

[0099] Data Synchronization Acquisition: Two types of key data are retrieved simultaneously from the data hub. One type is the priority identifier (high / medium / low priority) and spatiotemporal attribute information (inspection start and end time, geographical coordinate range of the road section covered by the sub-task) of each inspection sub-task. The other type is the real-time performance status data of the available drone fleet (remaining battery power percentage, current sensor working accuracy, maximum flight speed) and predicted reliability indicators (probability of failure of this task calculated based on historical data, completion rate of similar tasks).

[0100] Comprehensive assessment of UAV capabilities: Based on the real-time performance status data and predicted reliability indicators of the UAVs, a comprehensive capability assessment system is established. For example, the remaining battery power determines the flight endurance, the accuracy of the sensors affects the data acquisition quality, and the predicted reliability reflects the stability of mission completion. By quantifying and weighting the scores of these indicators, a comprehensive capability score is generated for each UAV. The higher the score, the stronger the UAV's ability to perform the mission.

[0101] Initial matching degree calculation: Combining the priority identifier, spatiotemporal attributes and UAV comprehensive capability score of each inspection sub-task, the initial matching degree between each UAV and each sub-task is calculated; for example, high-priority sub-tasks are matched with UAVs with high comprehensive capability scores first, while also considering the distance (spatiotemporal attribute) between the current location of the UAV and the sub-task road segment. The closer the distance, the higher the matching degree, so as to reduce the time spent by the UAV round trip.

[0102] Conflict resolution and optimization: With the goal of maximizing global task efficiency, the conflict problem in the initial matching degree is analyzed (such as multiple drones matching the same high-priority sub-task at the same time, or insufficient drone resources due to the concentration of sub-tasks in a certain area); by introducing spatiotemporal constraints (to avoid drones flying overlapping at the same time) and resource competition coordination rules (to prioritize the resource needs of high-priority sub-tasks), the initial matching degree is optimized and adjusted to generate the optimal drone resource allocation sequence.

[0103] Subtasks are linked to drones: Based on the optimized drone resource allocation sequence, each inspection subtask is dynamically linked to the corresponding drone, clarifying the subtasks that each drone needs to perform, and completing the precise allocation of resources;

[0104] The path planning unit, connected to the resource allocation unit, has the core function of generating the optimal inspection path for each assigned UAV while satisfying multiple constraints, ensuring that the UAV completes the inspection task efficiently and safely. Its specific processing logic is as follows:

[0105] Constraints are defined: three core constraints must be met for path planning: First, flight safety constraints, requiring avoidance of no-fly zones (such as airport airspace and areas with dense high-voltage power lines) and obstacles (such as bridges and billboards), while ensuring that flight altitude complies with airspace management regulations; second, data acquisition quality constraints, requiring the path to cover all road segments corresponding to sub-tasks, and flight speed and altitude to be adapted to sensor acquisition requirements (e.g., low-speed flight to ensure clear acquisition of road surface defects); and third, endurance constraints, requiring the total path length to match the remaining battery power of the drone to ensure that the drone can complete the inspection task and return safely.

[0106] Optimal Path Generation: With the goal of "maximizing inspection efficiency (shortest total path length, least flight time), optimizing data acquisition quality, and minimizing safety risks," a path optimization algorithm is used to plan the inspection path, taking into account the UAV's flight performance parameters (such as maximum flight speed and turning radius). For example, within the geographical coordinate range of the sub-task segment, paths that cover the segment with straight lines are prioritized, while avoiding safety constraint areas. The flight speed is adjusted to balance efficiency and acquisition quality, ultimately generating the optimal inspection path that satisfies all constraints.

[0107] In this embodiment, the task execution management module serves as the core link between the intelligent planning engine and the UAV. On one end, it receives the inspection planning scheme output by the intelligent planning engine, and on the other end, it directly interacts with the UAV for command and data communication. It undertakes the full-process execution management responsibility of "command issuance - status monitoring - data collection" and is a key link to ensure that the inspection task is implemented from planning to data closure. The task execution management module includes a task command encapsulation unit, a communication link management unit, a status monitoring unit, and a data processing unit.

[0108] The task instruction encapsulation unit connects to the path planning unit in the intelligent planning engine. It receives the optimal inspection path generated for each UAV and encapsulates each path and its corresponding inspection sub-task information into a structured task instruction set specific to the UAV model and capable of direct execution. The task instruction encapsulation unit acts as a bridge connecting the intelligent planning engine to subsequent communication links. Its core function is to integrate the optimal inspection path generated by the path planning unit with the corresponding inspection sub-task information and encapsulate it into a structured instruction set adapted to the UAV model. Its specific processing flow is as follows:

[0109] Data reception: Receives output data from the path planning unit of the intelligent planning engine, including the optimal inspection path for each UAV (including geographic coordinate point sequence, flight altitude / speed parameters), and inspection sub-task information associated with the path (such as sub-task priority, key road sections and areas to be collected, and data collection accuracy requirements).

[0110] Model compatibility analysis: Read the system's preset drone model parameter library, identify the drone model to be issued the command (such as different types such as multi-rotor, fixed-wing, or specific models from different manufacturers), and determine the command format and data interaction protocol supported by the drone model (such as command field definition and parameter value range).

[0111] Structured instruction encapsulation: The coordinate sequence of the inspection path, flight parameters, and sub-task requirements are integrated into a structured task instruction set according to the instruction format of the target UAV model. The instruction set must include "path execution instructions" (clearly defining the UAV flight trajectory and operation parameters), "task parameter instructions" (clearly defining data acquisition trigger conditions and storage format), and "emergency instructions" (such as low battery return-to-home trigger threshold and hovering strategy after loss of contact), ensuring that the instruction set can be directly parsed and executed by the UAV, avoiding task initiation problems due to format incompatibility.

[0112] The communication link management unit is connected to the task instruction encapsulation unit. Before task issuance, it verifies the communication link status with the target UAV, ensuring a stable and reliable link before securely and accurately issuing the structured task instruction set to the corresponding UAV. The communication link management unit is crucial for ensuring secure instruction issuance and stable data transmission. Its core function is to verify the UAV's communication link status and accurately issue the structured instruction set. Its specific processing flow is as follows:

[0113] Link status verification: Before the command is issued, a link test signal is sent to the target drone through a preset communication protocol (such as 4G / 5G, LoRa, satellite communication, etc.) to detect the link's signal strength, transmission delay, packet loss rate and other indicators to determine whether the link meets the "stable and reliable" standard (such as signal strength ≥ -85dBm, packet loss rate ≤ 1%).

[0114] Link anomaly handling: If link instability is detected (such as weak signal or excessive packet loss rate), the link switching mechanism is activated (such as switching from 4G to LoRa, or enabling the backup communication module), and the link test is repeated; if the requirements still cannot be met after multiple switching, a "link anomaly" alarm is reported to the system, and command issuance is suspended, waiting for maintenance personnel to intervene and investigate.

[0115] Secure command delivery: After confirming the link is stable, the structured command set is sent to the target drone using an encrypted transmission method (such as AES encryption algorithm). At the same time, the command transmission progress is monitored in real time. If the transmission is interrupted, the interruption is automatically triggered to resume transmission, ensuring that the command is delivered completely. After the command is received, the "command confirmation" signal is received from the drone to complete the command delivery loop.

[0116] The status monitoring unit is connected to the communication link management unit and is used to continuously receive real-time status data uploaded by the UAV during mission execution. This real-time status data includes at least location information, flight attitude, remaining battery power, and sensor operating status. The unit analyzes and monitors this data to generate real-time assessments of the UAV's health status and mission progress. The status monitoring unit is the core component for real-time monitoring of the UAV's operational status and mission progress. Its core function is to continuously receive and analyze the real-time status data uploaded by the UAV to generate health status and mission progress assessments. The specific processing flow is as follows:

[0117] Real-time data reception: Through a stable link established by the communication link management unit, the system continuously receives real-time status data periodically uploaded by the UAV. The data types include at least four categories: First, location information (latitude and longitude, altitude), used to determine whether the UAV is flying along the planned path; second, flight attitude (pitch angle, roll angle, yaw angle), used to determine the flight stability of the UAV; third, remaining battery power (percentage, voltage value), used to predict whether the endurance meets the mission requirements; and fourth, sensor operating status (such as whether the camera is framing normally, whether the radar is measuring distance normally), used to determine whether the data acquisition equipment is available.

[0118] Data analysis and monitoring: The received real-time status data is structured and analyzed to extract key parameters and compare them with preset thresholds (such as a low battery threshold of 20% and an abnormal yaw angle threshold of ±15°). If the parameters are within the normal range, the real-time operation status dashboard of the drone is updated. If the parameters exceed the thresholds (such as a battery level below 20% or a position deviation from the planned path exceeding 50 meters), an "abnormal alarm" (such as a "low battery alarm" or "path deviation alarm") is immediately generated and synchronized to the data center and system alarm module.

[0119] Progress and health assessment: Based on the parsed location information, calculate the proportion of the inspection path length completed by the UAV to the total path length, and generate the task execution progress (e.g., "60% completed"); combine the battery level and sensor status to comprehensively judge the health status of the UAV (e.g., "healthy", "sub-healthy", "abnormal"), and periodically send the progress and health assessment results to the data center for storage.

[0120] The data processing unit is connected to the status monitoring unit and is used to receive and parse the real-time data collected by the UAV during the inspection process. It performs preliminary quality verification and classification on the parsed data, and sends the verified inspection data, along with the real-time evaluation results generated by the status monitoring unit, to the data hub for storage. The data processing unit is crucial for ensuring the quality of inspection data and achieving standardized data storage. Its core function is to receive and process the inspection data collected by the UAV, complete quality verification and classification, and then synchronize it to the data hub. Its specific processing flow is as follows:

[0121] Data Acquisition and Reception: Receives real-time data collected by the drone during the inspection process. The data type depends on the inspection requirements (such as road surface image data, road surface smoothness radar data, crack depth detection data, etc.). At the same time, it receives real-time evaluation results generated by the status monitoring unit (such as the drone's position and sensor status during the data collection of this batch).

[0122] Data parsing and quality verification: The collected data is parsed (e.g., image data is converted from the original compressed format to a standard format), and two quality verifications are performed: first, integrity verification, checking for missing frames or breaks in the data (e.g., whether the image sequence is continuous); second, validity verification, combining the sensor status at the time of acquisition (e.g., whether the camera is focused correctly) and location information (e.g., whether the target road segment is covered) to determine whether the data meets the acquisition quality requirements (e.g., image clarity ≥ 1080P, no missing road segments).

[0123] Data Classification and Labeling: For data that passes verification, classify it according to the inspection sub-task attributes (such as high-priority sub-task, low-priority sub-task), the collected road segment number, and the data type (such as image, radar data), and add data tags (such as "Sub-task 1-K100-K120 road segment-road surface image-qualified"); For data that fails verification (such as blurry images, missing road segment data), mark it "unqualified" and record the reason (such as "sensor focusing abnormality"), and store it separately in the "pending processing" partition of the data center for subsequent manual review or supplementary collection;

[0124] Data synchronization and storage: The qualified data after classification and labeling are associated with the real-time evaluation results of the status monitoring unit, packaged and sent to the "inspection data" partition of the data center for storage, ensuring that each batch of inspection data can be bound to the corresponding task execution status, which is convenient for subsequent traceability and analysis.

[0125] In this embodiment, the data hub, as the core data support link of the intelligent planning and management system for UAV highway inspection tasks, communicates and interacts with the task input module, intelligent planning engine, task execution management module, and feedback optimization module. It undertakes the full-process data management responsibility of "data reception - preprocessing - classified storage - on-demand provision," and is a key foundation for ensuring smooth data flow among system modules, reliable decision data, and traceable historical data. The data hub stores and provides the basic data required for task planning, dynamic data generated during task execution, and historical task data, including:

[0126] The data receiving unit receives basic highway information from the task input module, real-time status and inspection data from the task execution management module, and intermediate data generated during the planning process from the intelligent planning engine. It serves as the data entry point between the data hub and other modules of the system. Its core function is to receive various types of raw data from the task input module, task execution management module, and intelligent planning engine, ensuring that all data enters the central processing flow without omission or delay. Its specific processing content is as follows:

[0127] Receive task input module data: Receive basic highway information output by the task input module after parsing, including the technical grade of each road section, historical damage data, etc. This data is the basis for the intelligent planning engine to decompose tasks.

[0128] Receive data from the task execution management module: Receive two types of data uploaded by the task execution management module. One type is real-time status data of the UAV (such as location information, flight attitude, remaining battery power, and sensor working status), and the other type is inspection and collection data after preliminary verification (such as road images and radar detection data). These data are dynamic data during the task execution process and need to be incorporated into the central management in real time.

[0129] Receive data from the intelligent planning engine: Receive intermediate data generated by the intelligent planning engine during task decomposition, resource allocation, and path planning, such as the comprehensive priority score of road segments, the initial matching degree between the drone and the sub-task, and the intermediate calculation results of path planning. This data is used to trace the planning decision-making process and also provides reference data for the planning stage of the feedback optimization module. The data receiving unit receives data through a preset communication interface (API interface adapted to each module), and adds a receiving timestamp to each type of data, temporarily storing it in a temporary buffer for subsequent preprocessing.

[0130] The data preprocessing unit, connected to the data receiving unit, performs consistency checks, format standardization, and redundancy cleanup on the data received by the data receiving unit, generating standardized data. The data preprocessing unit is a crucial step in ensuring data quality; its core function is to standardize the raw data temporarily stored in the data receiving unit, eliminating data format differences, errors, and redundancy, and generating standardized data that meets the requirements for subsequent storage and retrieval. Its specific processing flow is as follows:

[0131] Consistency check: Check whether the content of the received data conforms to the preset rules, such as whether the road segment number in the highway basic information conforms to the coding standard (such as "K + number" format), whether the battery level in the drone real-time status data is within a reasonable range of 0-100%, and whether the inspection and collection data has data format corruption (such as image files that cannot be opened); if data inconsistency is found, it is marked as "abnormal data", stored separately, and the verification result is fed back to the data source module. The data will be received again after correction.

[0132] Format standardization: Convert data from different sources and in different formats into a unified standard format for the system. For example, convert the location data of different types of UAVs uploaded by the task execution management module (some in latitude and longitude format, some in Gaussian coordinate format) into WGS84 latitude and longitude format; convert the image files in the inspection and collection data (some in JPG format, some in PNG format) into JPG format to ensure that there is no need for repeated format conversion when storing and retrieving data, thereby improving data utilization efficiency.

[0133] Redundancy cleanup: Delete duplicate or useless data. For example, if the task execution management module repeatedly uploads the same batch of inspection data due to network fluctuations, the data preprocessing unit identifies and deletes duplicate data by comparing data timestamps and data content hash values. At the same time, it cleans up useless fields in the data (such as debugging parameters that are not related to the system in the UAV status data) to reduce storage usage. After completing the above processing, the data preprocessing unit transmits the generated standardized data to the data classification and archiving unit.

[0134] The data classification and archiving unit, connected to the data preprocessing unit, is used to classify standardized data into basic data, dynamic data, or historical task data based on their attributes and sources, and add corresponding data tags. The core function of the data classification and archiving unit is to classify and divide data according to its attributes and sources, and add unique data tags, laying the foundation for accurate storage and rapid retrieval. Its specific processing flow is as follows:

[0135] Data classification: Based on the attributes and sources of the data, standardized data is divided into three main categories:

[0136] Basic data: This mainly includes basic highway information (road section technical grade, historical damage data) provided by the task input module, basic parameters of the UAV preset by the system (model, maximum endurance, sensor accuracy), airspace management regulations (coordinates of no-fly zones), etc. This type of data is updated infrequently and is the foundation for the long-term operation of the system.

[0137] Dynamic data: mainly includes real-time status data of UAVs and inspection data uploaded by the task execution management module, as well as real-time planning schemes (subtask allocation results and optimal inspection paths) output by the intelligent planning engine. This type of data is generated in real time as the task is executed, with a high update frequency, reflecting the dynamic process of task execution.

[0138] Historical task data: This mainly includes the full-process data of completed inspection tasks, such as the task decomposition results (sub-task priority list), resource allocation plan (UAV-sub-task binding relationship), path planning plan, task execution report (completion rate, data qualification rate), and historical dynamic data (UAV status and collected data of completed tasks). This type of data is the core basis for the optimization decision model of the feedback optimization module.

[0139] Adding data tags: Add multi-dimensional tags to each type of data. Tag content includes data source (e.g., "Task Input Module - Highway Basic Information", "Task Execution Management Module - UAV 1 Status Data"), data associated task number (e.g., "Inspection Task 20240501-001"), data generation time (accurate to the second), and data type (e.g., "Road Segment Attribute Data", "Image Acquisition Data"). For example, the tags for historical road damage data of a certain road segment are "Data Source: Task Input Module; Associated Task: 20240501-001; Generation Time: 2024-05-01 08:30:00; Data Type: Highway Basic Data - Historical Damage".

[0140] After the classification and tag addition are completed, the data classification and archiving unit will transfer the data to the data storage management unit;

[0141] The data storage management unit, connected to the data classification and archiving unit, is used to store classified and tagged data into different dedicated databases according to categories and to create data indexes. The data storage management unit is the "storage warehouse" of the data hub; its core function is to store classified and tagged data into dedicated databases according to categories and to create data indexes, achieving secure data storage and efficient management. Its specific processing flow is as follows:

[0142] Dedicated database allocation: Based on the data classification results, different types of data are stored in corresponding dedicated databases to avoid data mixing.

[0143] Basic data is stored in a "basic database" using a relational database (such as MySQL) to facilitate the querying and maintenance of structured data (such as querying the technical grade of a certain road section).

[0144] Dynamic data is stored in a "dynamic database" using a time-series database (such as InfluxDB). Because dynamic data has time-series characteristics (such as a drone uploading status data once per second), time-series databases can efficiently handle high-frequency writes and time-dimensional queries (such as querying the battery level change of a drone from 10:00 to 10:30).

[0145] Historical task data is stored in the "historical task database" using a hybrid storage model. Structured data (such as task decomposition results and execution reports) is stored in a relational database, while unstructured data (such as historical inspection images) is stored in an object storage service (such as MinIO), balancing the needs of structured data querying and large-capacity storage.

[0146] Data indexing: Create dedicated indexes for each type of database to improve data retrieval speed.

[0147] The basic database uses "road segment number" and "drone number" as index keywords to facilitate quick retrieval of basic information for specific road segments or drones;

[0148] The dynamic database uses a composite index of "task number", "drone number", and "time stamp" to facilitate the retrieval of dynamic data by task, drone, or time range (e.g., querying the status data of drone 1 after 10:00 in task 20240501-001).

[0149] The historical task database uses "task number" as the first-level index and "data type" as the second-level index, which facilitates quick location of various types of data for a specific historical task (such as querying the resource allocation plan and inspection images for task 20240501-001).

[0150] Data storage maintenance: Regularly maintain the database, including data backup (daily automatic backup of the basic database and historical task database to prevent data loss), storage capacity monitoring (automatically clean up old data that exceeds the preset retention period when the dynamic database capacity reaches the threshold, such as retaining dynamic data within 3 months), and data integrity check (weekly verification of database data and repair of data damaged due to storage anomalies).

[0151] The data service and provision unit, connected to the data storage management unit, responds to data requests from the intelligent planning engine or feedback optimization module, retrieving and outputting the required data from the corresponding database based on the data index. The data service and provision unit is the "data outlet" of the data hub; its core function is to respond to data requests from system modules such as the intelligent planning engine and feedback optimization module, retrieving and outputting data from the corresponding database based on the data index, ensuring that each module can obtain the required data in a timely manner. Its specific processing flow is as follows:

[0152] Data Request Reception: Receives structured data requests from other modules in the system. The request content must clearly state the data requirements (e.g., "Intelligent Planning Engine requests basic highway information for task 20240501-001", "Feedback Optimization Module requests execution data for all historical tasks in April 2024"). At the same time, it verifies the permissions of the requesting module (only authorized modules are allowed to obtain the corresponding data to ensure data security).

[0153] Data retrieval: Based on the request content, the data requirements are parsed, the target database and search conditions are determined, and the index of the corresponding database is called for fast retrieval; for example, after receiving the intelligent planning engine's request for "basic highway information for task 20240501-001", the system locates the basic database, uses "associated task number: 20240501-001" as the search condition, and retrieves basic information of all road segments involved in the task through the "road segment number" index;

[0154] Data output: The retrieved data is packaged in the format required by the requesting module (such as JSON or CSV) and transmitted to the requesting module through a preset communication interface; if the search result is empty (such as the requested historical task data does not exist), the system sends a "data does not exist" message to the requesting module and records the request log for subsequent troubleshooting.

[0155] The data service and provision unit simultaneously records all data request and output logs, including the request module, request time, request content, and output data volume, for data flow traceability and system operation monitoring.

[0156] In this embodiment, the feedback optimization module, acting as the "self-evolution core" of the system, uses historical task execution data stored in the data hub as a foundation. Through a closed-loop process of "data analysis - deviation identification - parameter adjustment - simulation verification," it continuously optimizes the decision-making model (task decomposition weight model, resource allocation matching model) of the intelligent planning engine, ensuring that the decision-making capability of the intelligent planning engine continuously improves with task execution experience, thereby driving the continuous optimization of the efficiency of the entire inspection system. The feedback optimization module is used to continuously optimize the decision-making model of the intelligent planning engine based on historical execution data, including:

[0157] Historical task execution data is obtained from the data hub. This data includes task decomposition results, resource allocation schemes, path planning schemes, and corresponding actual execution data.

[0158] Based on the acquired historical task execution data, task execution performance indicators are calculated. These indicators include at least task completion rate, resource utilization rate, and data collection quality. The specific processing flow is as follows:

[0159] Data Requirements Definition: Based on the optimization objectives (such as optimizing the task decomposition weight model and resource allocation matching model), the scope of historical data to be acquired should be clearly defined, including but not limited to: task decomposition results of historical tasks (comprehensive priority score of each road segment, sub-task priority division), resource allocation schemes (UAV-sub-task binding relationship, UAV comprehensive capability score), path planning schemes (planned path coordinates, constraint parameters), and actual execution data (actual task completion rate, actual UAV resource consumption, inspection data pass rate, and abnormal event records).

[0160] Data retrieval and filtering: Send structured data requests to the data hub. Based on filtering conditions such as "task time range" (e.g., tasks completed in the last 3 months), "task type" (e.g., highway inspection, mountain road inspection), and "data completeness" (must include both planned and actual execution data), quickly retrieve target historical task data through the data hub's index. Perform preliminary filtering on the retrieved data to remove task records with missing data (e.g., no actual execution data) or abnormal data (e.g., execution data is not related to planned data), ensuring that the data entering subsequent analysis is complete and usable.

[0161] Data association and integration: The filtered historical task data is associated and integrated according to the "single task" dimension to establish a one-to-one correspondence between "planning data" and "actual execution data" (such as associating the "subtask priority planning result" of a task with the "actual completion time / data quality of the subtask", and associating the "UAV resource allocation plan" with the "actual utilization rate / fault record of the UAV"), forming a structured historical task dataset to provide data support for subsequent performance analysis and model adjustment;

[0162] Indicator System Construction: Based on the decision-making objective of the intelligent planning engine (maximizing global task efficiency), a three-category core performance indicator system is established to ensure that the indicators comprehensively reflect the actual execution effect of the planning scheme.

[0163] Task completion rate indicators include "subtask on-time completion rate" (number of subtasks completed on time / total number of subtasks) and "full task completion rate" (number of tasks that completed all subtasks / total number of tasks), reflecting the feasibility of the planning scheme;

[0164] Resource utilization indicators include "drone time utilization" (actual drone mission execution time / total drone available time) and "drone power utilization" (actual drone power consumption / total power consumption before mission), reflecting the rationality of resource allocation;

[0165] Data collection quality indicators include "inspection data qualification rate" (quantity of qualified inspection data / total amount of collected data) and "key road segment data coverage rate" (length of key road segments (high priority sub-tasks) with actual collected data / total length of key road segments), reflecting the ability of route planning and resource matching to ensure data quality;

[0166] Indicator Quantification Calculation: For each historical task in the historical task dataset, three types of performance indicators are quantified according to preset calculation rules. For example, when calculating the "on-time completion rate of sub-tasks", the planned completion time of sub-tasks is compared with the actual completion time, the number of sub-tasks completed on time is counted and compared with the total number. When calculating the "inspection data qualification rate", the ratio of qualified data volume to total collected data volume is counted based on the "data verification results" recorded by the task execution management module. All indicator calculation results are associated with and stored in relation to the "expected performance indicators in the planning stage" of the corresponding historical task (completion rate, utilization rate, and qualification rate estimated by the intelligent planning engine during planning).

[0167] The calculated actual performance indicators are compared and analyzed with the expected performance indicators in the planning stage to identify performance deviations. The specific processing procedure is as follows:

[0168] Deviation Identification: For each historical task, the "actual performance indicators" and "expected performance indicators" are compared item by item to calculate the deviation value (deviation value = actual indicator value - expected indicator value). Based on a preset deviation threshold (e.g., a deviation value exceeding 10% is considered a significant deviation), indicators with significant deviations and their corresponding tasks are filtered out. For example, if the expected "subtask on-time completion rate" for a task is 95%, but the actual rate is 70%, the deviation value is -25%, exceeding the threshold, and is therefore determined to be a significant deviation. If the expected "drone time utilization rate" for a task is 80%, but the actual rate is 78%, the deviation value is -2%, which does not exceed the threshold, and is therefore determined to be a normal deviation.

[0169] Deviation Attribution: For indicators with significant deviations, by combining "planning data" and "actual execution data" from historical tasks, the root causes of the deviations are traced, focusing on the decision-making model problems of the intelligent planning engine.

[0170] If the "on-time completion rate of subtasks" is significantly biased, and the analysis reveals that high-priority subtasks are delayed due to insufficient drone battery life, it indicates that the resource allocation matching model has a bias in its assessment of drone battery life, and the weight parameter of "battery life" in the model needs to be adjusted.

[0171] If the "inspection data pass rate" deviates significantly, and the analysis finds that a certain road section is "low" according to the planning priority, but in reality, the road section has frequent defects, resulting in high data collection demand, it indicates that the weight of "historical defect data" in the task decomposition weight model is set too low, and the weight parameter of this attribute needs to be adjusted.

[0172] Deviation Summary: Summarize all indicators with significant deviations, corresponding task information, and root cause analysis results to form an "Effectiveness Deviation Analysis Report," which clarifies the decision-making model (task decomposition weight model / resource allocation matching model) that needs to be optimized and the optimization direction (such as adjusting the weight of a certain attribute or correcting a certain evaluation parameter).

[0173] Based on the performance deviation analysis results, the parameters of the decision model in the intelligent planning engine are adjusted. The decision model includes at least a weight model for task decomposition units and a matching model for resource allocation units. The specific processing flow is as follows:

[0174] Adjustment Target Positioning: Based on the "Performance Deviation Analysis Report," determine the decision-making model and specific parameters that need adjustment.

[0175] If the root cause of the deviation points to the task decomposition weight model, then adjust the weight values ​​of multi-dimensional road segment attributes such as "technical grade" and "historical disease data" in the model (e.g., increase the weight of historical disease data from 0.4 to 0.6).

[0176] If the root cause of the deviation points to the resource allocation matching model, then adjust the weight of the comprehensive capability score of indicators such as "real-time performance of UAV" (endurance, sensor accuracy) and "prediction reliability" in the model, or adjust the matching rules of "subtask priority - UAV capability" (e.g., high-priority subtasks need to be matched with UAVs with a comprehensive capability score ≥80 points, the original rule was ≥70 points).

[0177] Parameter adjustment rules: Based on the degree of bias and the amount of historical data, a "gradual adjustment" rule is adopted to avoid sudden parameter changes that could lead to model instability.

[0178] If a certain type of deviation (such as insufficient weight of historical disease data) occurs in multiple batches of historical tasks and the degree of deviation is consistent, the weight of the corresponding attribute will be increased by a preset step size (such as 0.1 each time).

[0179] If the deviation only occurs in a specific scenario (such as mountain road inspection), the decision model parameters are adjusted only for that scenario, while the parameter settings for other scenarios are retained to ensure that the model is adapted to different inspection scenarios.

[0180] Adjustment Log: Record detailed information for each parameter adjustment, including adjustment time, parameter value before adjustment, parameter value after adjustment, and adjustment basis (corresponding to deviation analysis results), forming a "Model Parameter Adjustment Log" for easy subsequent traceability and rollback;

[0181] The adjusted decision-making model is used to simulate and plan historical tasks, and the consistency between the simulation planning results and the actual execution data is evaluated. The specific processing flow is as follows:

[0182] Simulation planning settings: Select an "independent historical task dataset" that did not participate in the deviation analysis (such as completed tasks that have not been used for deviation analysis in the past month), and use the adjusted decision model (task decomposition weight model / resource allocation matching model) as the core algorithm for simulation planning, while keeping other planning conditions (such as drone resources and road segment basic information) consistent with the original historical tasks;

[0183] Simulated planning execution: For each task in the "Independent Historical Task Dataset", the task is decomposed and resources are allocated again using the adjusted model to generate "Simulated planning results" (simulated sub-task priorities, simulated UAV-sub-task binding relationships).

[0184] Alignment Assessment: The "simulated planning results" are compared with the "actual execution data" of the task to calculate the "alignment" index.

[0185] Task decomposition fit: The degree to which the priority of simulated subtasks matches the importance of subtasks in actual execution (such as the actual order of completion);

[0186] Resource allocation fit: The degree of matching between the simulated drones and the best-performing drones in actual execution (e.g., no delays, high data quality); if the fit reaches the preset optimization threshold (e.g., ≥85%), it means that the adjusted model can more accurately match the actual scenario; if it does not reach the threshold, return to the "Model Parameter Adjustment Unit" to re-analyze the deviation and adjust the parameters until the fit reaches the standard.

[0187] If the degree of agreement reaches the preset optimization threshold, the adjusted decision model will be updated to the intelligent planning engine. The specific processing flow is as follows:

[0188] Backup before update: Before updating the decision model of the intelligent planning engine, perform a full backup of the currently used model parameters and store them in the "model backup library" of the data center. If problems are found in the updated model later, the system can be quickly rolled back to the original model to ensure system stability.

[0189] Model replacement: Through the system's preset interface, the adjusted decision model (task decomposition weight model / resource allocation matching model) that meets the consistency standard replaces the corresponding old model in the intelligent planning engine, ensuring that the new model takes effect immediately and is used for the planning of subsequent new inspection tasks;

[0190] Update Log and Notification: Record detailed information about model updates (update time, model parameters before and after the update, verification consistency, and operators), and generate a "Model Update Report"; at the same time, send a "Model Update Complete" notification to the system management interface to inform the operations and maintenance personnel of the update results, which will facilitate the subsequent tracking of the actual application effect of the new model.

[0191] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent planning and management system for unmanned aerial vehicle (UAV) highway inspection tasks, characterized in that: include: The task input module is used to receive the inspection task requirements and highway basic information configured by the user. The task input module includes a configuration receiving unit, a data verification unit and a data parsing unit connected in sequence. The configuration receiving unit is used to receive raw data of inspection task requirements and highway basic information input by the user through a human-computer interaction interface; The data verification unit is used to perform format verification and integrity checks on the original data and generate verified data. The data parsing unit is used to parse the verified data, extract multi-dimensional road segment attributes from the inspection task parameters and highway basic information, including technical grade and historical damage data, and output the extracted parameters and attributes to the data center for storage, so that the intelligent planning engine can call them. The intelligent planning engine, connected to the task input module, is used to intelligently decompose and plan the received inspection task requirements, including: The task decomposition unit, based on multi-dimensional road segment attributes in the highway infrastructure information, including technical grade and historical damage data, decomposes the overall inspection task into inspection sub-tasks with different priorities through a pre-set weight model, including: Obtain basic highway information related to the inspection task from the data hub. The basic highway information includes attribute data of multiple road sections, and the attribute data includes at least technical grade and historical damage data. The attribute data of each road segment is standardized, the technical level is mapped to a preset numerical level, and historical disease data is converted into standardized indicators that reflect the severity of the disease. Based on a pre-set weighting model, the numerical level and standardized indicators of each road segment are weighted and fused to generate a comprehensive priority score for each road segment. Based on the comprehensive priority score of each road segment, the road segments are divided into multiple priority intervals, and each priority interval corresponds to a priority level; Road segments within the same priority range are grouped into independent inspection sub-tasks, and each inspection sub-task is assigned a corresponding priority level, thereby completing the decomposition of the overall inspection task. The resource allocation unit, connected to the task decomposition unit, is used to dynamically allocate appropriate drone resources to each sub-task based on the priority, spatiotemporal attributes, and real-time performance status and predicted reliability of the available drone fleet, with the goal of maximizing global task efficiency. The path planning unit, connected to the resource allocation unit, is used to generate the optimal inspection path for each allocated UAV under multiple constraints such as flight safety, data acquisition quality, and endurance. The task execution management module connects to the intelligent planning engine and is used to send planning schemes to the corresponding drones and receive the drones' real-time status and inspection data. The data hub communicates and interacts with the task input module, the intelligent planning engine, and the task execution management module, respectively, and is used to store and provide the basic data required for task planning, dynamic data generated during task execution, and historical task data. The decision-making process of the intelligent planning engine relies on data from the data hub; the system also includes a feedback optimization module for continuously optimizing the decision-making model of the intelligent planning engine based on historical execution data.

2. The intelligent planning and management system for UAV highway inspection tasks according to claim 1, characterized in that: The resource allocation unit is used to dynamically allocate appropriate drone resources to each sub-task based on its priority, spatiotemporal attributes, and the real-time performance status and predicted reliability of the available drone fleet, with the goal of maximizing overall task efficiency. This includes: The priority identifiers and spatiotemporal attribute information of each inspection subtask are obtained from the data hub, and the real-time performance status data and predictive reliability indicators of the available drone fleet are obtained simultaneously. Based on real-time performance status data and predicted reliability indicators, the effectiveness of each UAV is evaluated, and a comprehensive capability score for each UAV is generated. Based on the priority identifier and spatiotemporal attribute information of each inspection sub-task, and combined with the comprehensive capability score, the initial matching degree between each UAV and each inspection sub-task is calculated. With the goal of maximizing global task efficiency, considering spatiotemporal constraints and resource competition, the initial matching degree is reconciled and optimized to generate the optimal UAV resource allocation sequence. Based on the drone resource allocation sequence, each inspection subtask is dynamically bound to the corresponding drone to complete the resource allocation.

3. The intelligent planning and management system for UAV highway inspection tasks according to claim 1, characterized in that: The task execution management module includes a task instruction encapsulation unit, a communication link management unit, a status monitoring unit, and a data processing unit; The task instruction encapsulation unit is connected to the path planning unit in the intelligent planning engine. It is used to receive the optimal inspection path generated for each UAV and encapsulate each path and its corresponding inspection sub-task information into a structured task instruction set that is specific to the UAV model and can be directly executed. The communication link management unit is connected to the task instruction encapsulation unit and is used to verify the communication link status with the target UAV before the task is issued. After ensuring that the link is stable and reliable, the structured task instruction set is safely and accurately issued to the corresponding UAV. The status monitoring unit is connected to the communication link management unit and is used to continuously receive real-time status data uploaded by the UAV through the communication link during the mission execution. The real-time status data includes at least location information, flight attitude, remaining battery power and sensor working status. The unit also analyzes and monitors the real-time status data to generate a real-time assessment of the UAV's health status and mission execution progress. The data processing unit is connected to the status monitoring unit and is used to receive and parse the data collected by the UAV in real time during the inspection process. It performs preliminary quality verification and classification on the parsed data and sends the verified inspection data and the real-time evaluation results generated by the status monitoring unit to the data center for storage.

4. The intelligent planning and management system for UAV highway inspection tasks according to claim 1, characterized in that: The task execution management module is used to distribute planning schemes to the corresponding drones and receive real-time status and inspection data from the drones, including: Receive the optimal inspection path generated for each UAV by the path planning unit in the intelligent planning engine, and encapsulate each path and its corresponding inspection subtask information into a structured task instruction set that is specific to the UAV model and can be directly executed; Before issuing the task, verify the status of the communication link with the target drone. After ensuring that the link is stable and reliable, the structured task instruction set is safely and accurately issued to the corresponding drone. During mission execution, the system continuously receives real-time status data uploaded by the UAV through the communication link. The real-time status data includes at least location information, flight attitude, remaining battery power, and sensor working status. The system also analyzes and monitors the real-time status data to generate a real-time assessment of the UAV's health status and mission execution progress. The system receives and parses the data collected in real time during the inspection process from the drone, performs preliminary quality verification and classification on the parsed data, and sends the verified inspection data and the real-time evaluation results to the data center for storage.

5. The intelligent planning and management system for UAV highway inspection tasks according to claim 1, characterized in that: The data hub is used to store and provide basic data required for task planning, dynamic data generated during task execution, and historical task data, including: The data receiving unit is used to receive basic highway information from the task input module, real-time status and inspection data from the task execution management module, and intermediate data generated during the planning process from the intelligent planning engine. The data preprocessing unit, connected to the data receiving unit, is used to perform consistency verification, format standardization, and redundancy removal on the data received by the data receiving unit, and generate standardized data. The data classification and archiving unit, connected to the data preprocessing unit, is used to classify standardized data into basic data, dynamic data, or historical task data according to their attributes and sources, and add corresponding data tags. The data storage management unit, connected to the data classification and archiving unit, is used to store classified and tagged data into different dedicated databases according to categories and to create data indexes; The data service and provision unit, connected to the data storage management unit, is used to respond to data requests from the intelligent planning engine or feedback optimization module, and retrieve and output the required data from the corresponding database based on the data index.

6. The intelligent planning and management system for UAV highway inspection tasks according to claim 1, characterized in that: The feedback optimization module is used to continuously optimize the decision model of the intelligent planning engine based on historical execution data, including: Historical task execution data is obtained from the data hub. This data includes task decomposition results, resource allocation schemes, path planning schemes, and corresponding actual execution data. Based on the acquired historical task execution data, calculate task execution performance indicators, which include at least task completion rate, resource utilization rate, and data collection quality. Compare and analyze the calculated actual performance indicators with the expected performance indicators in the planning stage to identify performance deviations; Based on the results of the performance deviation analysis, adjust the parameters of the decision model in the intelligent planning engine. The decision model should include at least the weight model of the task decomposition unit and the matching model of the resource allocation unit. The adjusted decision-making model was used to simulate and plan historical tasks, and the consistency between the simulation planning results and the actual execution data was evaluated. If the degree of agreement reaches the preset optimization threshold, the adjusted decision model will be updated to the intelligent planning engine.

Images