A method and system for remote patrol monitoring of a drone

Abstract

The application discloses an unmanned aerial vehicle remote patrol monitoring method and system, relates to the field of nest type unmanned aerial vehicle patrol and dispatch, and comprises the following steps: acquiring historical operation data and equipment basic data, analyzing the data to obtain sample data, extracting and sorting the sample data to construct an unmanned aerial vehicle patrol knowledge graph, acquiring a current patrol task, analyzing the patrol task based on the unmanned aerial vehicle patrol knowledge graph to obtain task requirements and screening to obtain an effective candidate unmanned aerial vehicle group, generating an unmanned aerial vehicle dispatch strategy based on the effective candidate unmanned aerial vehicle group by using an improved traditional genetic algorithm-particle swarm hybrid algorithm, and performing unmanned aerial vehicle remote patrol monitoring based on the unmanned aerial vehicle dispatch strategy to realize abnormal detection and early warning. The application realizes multi-source data fusion and constraint verification by relying on the knowledge graph, the dispatch scheme is optimal and has strong adaptability, a multi-target weighted optimization is combined with the improved GA-PSO algorithm, and the solving efficiency and dynamic adjustment capability are improved.

Description

Technical Field

[0001] This invention relates to the field of nested unmanned aerial vehicle (UAV) inspection and scheduling, and in particular to a method and system for remote inspection and monitoring of UAVs. Background Technology

[0002] With the rapid development of drone technology and autonomous control technology, nested drones, with their core advantages of "autonomous take-off and landing, automatic charging, and remote control," have been widely penetrated into many key areas such as power line inspection, security patrol, environmental monitoring, traffic control, and emergency rescue, becoming core equipment for remote inspection operations. However, in practical applications, current nested drone remote inspection systems still face many technical bottlenecks, severely restricting inspection efficiency, operational safety, and management controllability. These bottlenecks are manifested in the following ways: First, there is insufficient integration of multi-source heterogeneous data, resulting in a lack of knowledge support. Patrol scenarios involve various types of data, including drone parameters (endurance, payload, sensor accuracy, etc.), drone nest resources (location, charging power, capacity, etc.), dynamic environmental data (wind speed, visibility, obstacle distribution, etc.), and task requirements (priority, time limit, inspection accuracy, etc.). Existing systems mostly use distributed storage, resulting in prominent data silos and a lack of structured knowledge modeling and correlation analysis methods. This leads to the inability to fully utilize various types of data during scheduling decisions, easily causing problems such as mismatch between drones and task requirements, and path planning ignoring dynamic obstacles.

[0003] Secondly, multi-nest, multi-drone collaborative scheduling suffers from low efficiency and poor optimization. Traditional scheduling relies heavily on manual experience or simple rule matching, making it difficult to handle complex scenarios involving multiple parallel tasks and multi-device collaboration. On the one hand, it fails to achieve precise matching between tasks and drone capabilities, easily leading to unreasonable situations such as high-priority tasks being assigned to low-performance drones or drone endurance not covering task duration. On the other hand, the scheduling objective is singular (often only considering task completion time), lacking comprehensive optimization of multiple objectives such as energy consumption, equipment utilization, and priority matching, resulting in high operating costs, high equipment idle rates, and high-priority tasks easily being occupied by low-priority tasks. Furthermore, traditional optimization algorithms often use random initialization for solving, easily generating invalid solutions (such as paths entering no-fly zones or exceeding drone endurance), making it difficult to guarantee solution efficiency and feasibility.

[0004] Third, the system suffers from poor adaptability to dynamic scenarios and a lack of flexibility in scheduling schemes. Patrols face numerous dynamic interference factors, such as sudden drone malfunctions, abrupt changes in environmental parameters (excessive wind speed, decreased visibility), and the addition of urgent tasks. Existing systems lack rapid response and dynamic adjustment mechanisms: when a malfunction occurs, it is difficult to quickly select alternative drones and replan routes; when the environment deteriorates, flight parameters cannot be adjusted in a timely manner; and when new tasks are added, they are prone to time conflicts with existing tasks, leading to operational interruptions and task delays, severely impacting the continuity and reliability of patrols.

[0005] Fourth, the system suffers from poor scenario adaptability and insufficient versatility. The task requirements and environmental characteristics of inspection scenarios in different industries vary significantly (e.g., emergency rescue requires prioritizing task completion speed, while routine inspections require prioritizing energy consumption control). The scheduling rules and optimization objectives of the existing system are mostly fixed settings and cannot be flexibly adjusted according to the scenario, resulting in insufficient rationality of the scheduling scheme in specific scenarios and difficulty in meeting diverse operational needs.

[0006] To address these issues, there is an urgent need for a method and system for remote inspection and monitoring using drones. Summary of the Invention

[0007] To address the aforementioned issues, this application proposes a remote inspection and monitoring method and system for unmanned aerial vehicles (UAVs), aiming to improve the adaptability of UAVs while enhancing deployment efficiency and dynamic adjustment capabilities.

[0008] On the one hand, this application mentions a method for remote inspection and monitoring using unmanned aerial vehicles, including the following steps: S1. Obtain historical operation data and equipment basic data, and analyze the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and abnormal data; S2. Extract and organize the sample data to construct a knowledge graph of drone patrols; S3. Obtain the current inspection task, analyze the inspection task based on the drone inspection knowledge graph to obtain the task requirements, and filter out the effective candidate drone groups. S4. Based on the effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies. S5. Based on the drone scheduling strategy, remote drone patrol and monitoring is implemented to achieve anomaly detection and early warning.

[0009] Preferably, the specific content of extracting and organizing sample data to construct a drone patrol knowledge graph in S2 includes: Using ontology modeling, we define the core entities, core relationships, and entity attributes in the inspection scenario, and construct the schema layer of the knowledge graph. For different types of data, a differentiated knowledge extraction strategy is adopted to extract knowledge from the data and obtain the initial knowledge triple (subject-relation-object). The entities in the initial knowledge triples are aligned using an algorithm based on attribute similarity and semantic similarity to merge duplicate entities; Attribute fusion is performed by selecting or weighting conflicting attributes based on data reliability. By integrating relationships and eliminating redundancy and contradictions, a knowledge graph of drone patrols is obtained.

[0010] Preferably, the current inspection task is obtained, and the specific content of the task requirements is obtained by parsing the inspection task based on the drone inspection knowledge graph: Select patrol task entities from the entities in the knowledge graph, extract core attributes from the patrol task entities, and generate a task basic information table. The task basic information table includes the task identifier T. i (i is the task number, i=1,2,...,n), Task type Type(T) i Task priority P(T) i ), Patrol Area (T) i Task deadline (T) i The data requirement is DataReq(T) i ); Obtain the current inspection task and match it with the task information table to obtain the task requirements; From the entities in the knowledge graph, select entities in the patrol area and environmental elements, extract geographical and environmental features, and quantify them into scheduling constraint parameters: The geographical and environmental features include regional topographic complexity: Terrain (A) i Obstacle distribution: Obstacle(A) i Real-time environmental parameters, Env(T) i ); Based on geographical and environmental characteristics, combined with regional area and inspection accuracy, the estimated task time Test(T) is calculated. i ).

[0011] By combining knowledge graph reasoning, hard and soft constraints corresponding to the task are generated, and scheduling sub-goals are broken down.

[0012] Preferably, the specific content of the selected effective candidate drone groups includes: From the drone entities in the knowledge graph, select all drones whose status is marked as available and form an initial drone pool; The knowledge graph extracts the core attribute information of each drone in the initial drone pool. The core attribute information includes the drone's endurance, maximum flight altitude, type and accuracy of the sensors it carries, its home nest information, and the straight-line distance between its home nest and the center of the current patrol area. Based on the hard constraints derived from the task analysis, each drone in the initial drone pool is verified one by one, and only drones that fully satisfy all hard constraints are retained to form an intermediate candidate pool. For drones in the intermediate candidate pool, a weighted score is applied based on the soft constraint dimension in combination with the mission requirements to quantify the suitability of each drone for the mission. Based on the soft-constraint weighted scoring results, the drones in the intermediate candidate pool are sorted from high to low, and finally the top n% of the drones are selected as valid candidate drone groups.

[0013] Preferably, the specific content of generating a drone scheduling strategy based on an improved genetic algorithm-particle swarm optimization algorithm using effective candidate drone groups is as follows: Define the drone allocation variable, mission execution order variable, flight path variable, and charging arrangement variable as decision variables; Four initial objective functions are constructed based on task completion efficiency, resource consumption, equipment utilization, and task priority matching degree; Configure multiple constraints for the objective function, including task allocation constraints, UAV load constraints, time sequence constraints, path constraints, and variable constraints; Based on the scenario requirements, a multi-objective optimization model is obtained by weighting the four initial objective functions; An improved genetic algorithm-particle swarm optimization hybrid algorithm is used to solve the multi-objective optimization model to obtain the objective solution; The objective solution is transformed into an executable scheduling scheme, and the feasibility is verified by a knowledge graph reasoning engine to obtain a feasible UAV scheduling strategy.

[0014] Preferably, the objective solution obtained by solving the multi-objective optimization model using the improved genetic algorithm-particle swarm optimization hybrid algorithm includes: Initialization: Generate an initial solution based on the candidate set of unmanned aerial vehicles. The initial solution contains decision variables. Fitness calculation: The fitness function is a weighted single objective function, and the fitness value of each initial solution is calculated. Iterative optimization is performed by iteratively solving the problem using improved selection, crossover, and mutation operators and particle update rules. Knowledge graph constraint verification is incorporated into the iteration process to ensure that the solution in each generation meets the constraints. The number of iterations is set to 50-100 generations, and the convergence condition is that the fitness value fluctuation is ≤1% for 5 consecutive generations. After the iteration is completed, the solution with the lowest fitness value is selected as the target solution, which corresponds to a complete set of scheduling parameters.

[0015] Preferably, the specific content of feasibility verification using a knowledge graph reasoning engine includes: Drone allocation: Clearly define the patrol mission corresponding to each candidate drone, mark the core requirements of the mission, and ensure that the drone's capabilities are accurately matched with the mission requirements; Flight path planning: Generate detailed flight trajectories based on path variables, mark the coordinates of the route, key obstacle avoidance points, and key inspection areas, and combine them with GIS map visualization to ensure safe path and complete coverage of the inspection area; Nest scheduling: Clearly define the nests where each drone will dock, the timing of its charging, and the duration of its charging to avoid conflicts over nest charging resources and ensure sufficient drone battery life; Task execution order: Clearly define the execution order of multiple tasks, the start and estimated completion times of each task, and prioritize the time nodes of high-priority tasks to facilitate staff tracking of task progress; On the other hand, this application mentions a drone remote patrol and monitoring system, comprising: Data acquisition unit: Acquires historical operational data and basic equipment data, and analyzes the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and anomaly data; Knowledge graph construction unit: Extracts and organizes sample data to construct a knowledge graph for drone patrols; Initial strategy generation unit: Obtain the current inspection task, parse the inspection task based on the UAV inspection knowledge graph to obtain task requirements, and filter out effective candidate UAV groups; Target strategy generation unit: Based on effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies. Based on the UAV scheduling strategies, remote UAV patrol and monitoring is carried out to achieve anomaly detection and early warning.

[0016] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program in the memory to implement the content of a remote patrol and monitoring method for unmanned aerial vehicles.

[0017] A storage medium storing computer-executable instructions, which, when loaded and executed by a processor, implement the content of a remote patrol and monitoring method for unmanned aerial vehicles (UAVs).

[0018] In summary, the unmanned aerial vehicle (UAV) remote patrol and monitoring method and system of the present invention have the following advantages compared with traditional technologies: 1. This application adopts ontology modeling, multi-source knowledge extraction and fusion technology to construct a structured knowledge network for patrol scenarios, realize the association and integration of multiple types of data such as UAVs, UAV nests, tasks, and environment, completely break down data silos, provide knowledge support for scheduling decisions such as dynamic constraint verification, historical experience reuse, and abnormal pattern matching, significantly improve the accuracy of task and equipment adaptation, avoid ineffective scheduling decisions, and ensure the feasibility of dynamic adjustment of the plan. 2. This application uses knowledge graph constraints as a priori, optimizes the initial solution generation and iteration process, eliminates invalid solutions, and shortens the convergence time, improving the solution efficiency by more than 40% compared with traditional algorithms; through a multi-objective weighted fusion model, it balances task time consumption, energy consumption, equipment utilization and priority matching degree, and adapts to different scenario requirements.

[0019] 3. This application adapts to multiple industry scenarios such as power inspection and emergency rescue by customizing the modeling of entities and relationships in knowledge graphs and dynamically adjusting the target weights of scheduling algorithms.

[0020] The technical method of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of a remote patrol and monitoring method using unmanned aerial vehicles (UAVs) according to the present invention. Figure 2 This is a unit diagram of a drone remote patrol and monitoring system according to the present invention. Detailed Implementation

[0022] The technical method of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0023] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the scope of this application and its application or use.

[0024] Techniques, systems, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the instruction manual.

[0025] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0026] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0027] Example 1 This embodiment mentions a method for remote patrol and monitoring using unmanned aerial vehicles (UAVs), such as... Figure 1 The above includes the following steps: S1. Obtain historical operation data and equipment basic data, and analyze the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and anomaly data.

[0028] Drone data includes model, flight time, payload, maximum flight altitude, cruising speed, sensor type and accuracy, etc.

[0029] The data on the drone nest includes its location, the number of drones it can hold, its charging speed, its maintenance status, and its communication range.

[0030] Equipment operating status data includes drone battery level, attitude, sensor data, and the status of the drone's charging equipment.

[0031] Inspection data includes task type (routine inspection, anomaly investigation, emergency rescue, etc.), inspection area range (latitude and longitude boundaries, key areas of concern), task priority (high / medium / low), task time limit, and task requirements (such as inspection accuracy, shooting angle, data upload frequency, etc.).

[0032] Historical patrol mission records (mission completion status, time spent, drone utilization rate, abnormal event records, etc.).

[0033] Abnormal data includes historical equipment failure data and user feedback data (evaluation of inspection effectiveness, suggestions for scheduling optimization, etc.).

[0034] Environmental data includes real-time meteorological data (wind speed, wind direction, rainfall, visibility, temperature, etc.), geographic environment data (topography, obstacle distribution, no-fly zones, etc.), and real-time traffic data (road congestion, vehicle traffic status, etc., for traffic patrol scenarios).

[0035] S2. Extract and organize the sample data to construct a knowledge graph of drone patrols; Furthermore, the specific content of extracting and organizing sample data to construct a drone patrol knowledge graph in S2 includes: Using ontology modeling, we define the core entities, core relationships, and entity attributes in the inspection scenario, and construct the schema layer of the knowledge graph.

[0036] Core entities include: drones, drone nests, patrol missions, patrol areas, environmental elements, abnormal events, and users; core relationships include: drone-nest (ownership, docking), drone-patrol mission (execution, adaptation), nest-patrol area (coverage, distance), patrol mission-patrol area (association, focus), and environmental element-patrol mission (impact, limitations); entity attributes include various parameters and status information from the data acquisition layer mentioned above. For example, the attributes of the "drone" entity include model, battery life, power level, and sensor accuracy, while the attributes of the "patrol mission" entity include priority, time limit, and patrol range.

[0037] For different types of data, a differentiated knowledge extraction strategy is adopted to extract knowledge from the data to obtain initial knowledge triples (subject-relation-object), such as (drone A-owner-nest B) and (patrol task C-execution-drone A).

[0038] For structured data (such as UAV parameter tables and mission plan tables), SQL queries and ETL tools are used to directly map it to entities and attributes in the knowledge graph; for semi-structured data (such as historical mission records and equipment failure reports), a combination of rule matching and regular expressions is used to extract knowledge; for unstructured data (such as user feedback text and abnormal event descriptions), natural language processing (NLP) techniques (such as entity recognition and relation extraction models) are used to extract key knowledge.

[0039] The entities in the initial knowledge triples are aligned, and duplicate entities are merged using an algorithm based on attribute similarity (such as drone model and nest location) and semantic similarity.

[0040] Attribute fusion is performed by selecting or weighting conflicting attributes based on data reliability.

[0041] By integrating relationships and eliminating redundancy and contradictions, a knowledge graph of drone patrols is obtained.

[0042] S3. Obtain the current inspection task, analyze the inspection task based on the drone inspection knowledge graph to obtain the task requirements, and filter out the effective candidate drone groups.

[0043] Based on the task entity attributes in the knowledge graph, the patrol task is analyzed to clarify the core requirements of the task, including the geographical features of the patrol area (such as terrain complexity and obstacle distribution), task priority, time limit requirements, and data collection accuracy requirements. At the same time, combined with the environmental element entities in the knowledge graph, the impact of the current environment on the task is analyzed (such as whether the windy weather is suitable for flight and whether the visibility meets the shooting requirements), providing constraints for subsequent scheduling.

[0044] Further, obtain the current inspection task, and parse the inspection task based on the UAV inspection knowledge graph. The specific content of the task requirements is as follows: Select the inspection task entity from the entities of the knowledge graph, extract the core attributes in the inspection task entity, and generate a task basic information table.

[0045] The task basic information table includes the task identifier T i (where i is the task serial number, i = 1, 2,..., n), task type Type(T i ), task priority P(T i ), inspection area Area(T i ), task time limit Deadline(T i ), data requirement DataReq(T i ).

[0046] Obtain the current inspection task, and match the task requirements based on the task basic information table.

[0047] Select the inspection area entity and environmental factor entity from the entities of the knowledge graph, extract the geographical and environmental features, and quantify them into scheduling constraint parameters: The geographical and environmental features include regional terrain complexity: Terrain(A i ), obstacle distribution: Obstacle(A i ), real-time environmental parameters, Env(T i ).

[0048] Based on the geographical and environmental features, combined with the regional area and inspection accuracy, calculate the estimated task time consumption Test(T i ).

[0049] .

[0050] Among them, is the inspection area (unit: km 2 ); is the terrain complexity coefficient (default 0.3, the more complex the terrain, the larger the value); is the standard cruising speed of the UAV (unit: km / h, extracted from the "UAV" entity in the knowledge graph as the default value); is the environmental impact coefficient (β = 1 when the wind speed Vw ≤ 3m / s, β = 0.7 when 3 < Vw ≤ 6m / s, β = 0 when V_w > 6m / s); is the inspection accuracy (unit: m, the higher the accuracy requirement the smaller).

[0051] Combined with the knowledge graph reasoning, generate the hard constraints and soft constraints corresponding to the task, and disassemble the scheduling sub-goals.

[0052] Hard constraints include visibility and the maximum flight altitude of drones.

[0053] Soft constraints include prioritizing the drone that is closest to the nest.

[0054] Sub-objectives such as "emergency rescue mission" have the sub-objective of "shortest mission completion time" and "routine inspection" have the sub-objective of "lowest energy consumption".

[0055] The inference rules are based on entity relationships and attribute constraints. For example, "high patrol task priority means selecting drones with a flight time greater than or equal to 1.2 times the estimated task time," "the presence of high-altitude obstacles in the patrol area means selecting drones with a maximum flight altitude greater than or equal to the obstacle height plus 50 meters," and "requiring infrared imaging means selecting drones equipped with infrared sensors." Simultaneously, considering the location attributes of the drone nest entity, drones in the nest closest to the patrol area are prioritized for selection, reducing round-trip energy consumption and time.

[0056] Furthermore, the specific details for selecting valid candidate drone groups include: From the drone entities in the knowledge graph, select all drones whose status is marked as available to form an initial drone pool.

[0057] The knowledge graph extracts the core attribute information of each drone in the initial drone pool. The core attribute information includes the drone's endurance, maximum flight altitude, type and accuracy of the sensors it carries, its home nest information, and the straight-line distance between its home nest and the center of the current patrol area.

[0058] Based on the hard constraints derived from task analysis, each drone in the initial drone pool is verified one by one, and only drones that fully satisfy all hard constraints are retained to form an intermediate candidate pool.

[0059] For drones in the intermediate candidate pool, a weighted score is applied based on the soft constraint dimension in conjunction with the mission requirements to quantify the suitability of each drone for the mission.

[0060] Based on the soft-constraint weighted scoring results, the drones in the intermediate candidate pool are sorted from high to low, and finally the top n% of the drones are selected as valid candidate drone groups.

[0061] S4. Based on the effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies.

[0062] Furthermore, the specific details of generating a drone scheduling strategy based on an improved genetic algorithm-particle swarm optimization algorithm using effective candidate drone groups are as follows: Define the drone allocation variable, mission execution order variable, flight path variable, and charging arrangement variable as decision variables.

[0063] Drone assignment variable: x ij (0-1 variable, x) ij =1 indicates that the drone U j Execute task T i x ij =0 indicates no execution), used to clarify the correspondence between tasks and drones, and the variable value needs to be determined in combination with the adaptability of the candidate drone set.

[0064] Task execution order variable: y il (0-1 variable, y) il =1 indicates task T i In Task T l Previously executed, y il =0 indicates after), in multi-task parallel scenarios, the execution order of high-priority tasks is guaranteed first to avoid task conflicts.

[0065] Flight path variable: Path(U j ,T i )={(Lon j1 ,Lat j1 ),...,(Lon jt ,Lat jt )} (UAV) j Execute task T i The route (a sequence of coordinate points) needs to be planned in conjunction with information from the knowledge graph, such as the terrain of the patrol area, the distribution of obstacles, and no-fly zones, to ensure a safe and unobstructed path.

[0066] Charging schedule variable: z jn (0-1 variable, z) jn =1 indicates that the drone U j In nest N n Charging, z jn =0 indicates no charging), which is used to clarify the charging plan between drone missions. It is determined by combining data such as drone nest charging resources and the drone's remaining power to ensure the drone's continuous availability.

[0067] Four initial objective functions are constructed based on task completion efficiency, resource consumption, equipment utilization, and task priority matching degree.

[0068] Shortest total task completion time: .

[0069] in, The key to reducing the time for drones to travel between their nests and the patrol area is to optimize drone allocation and routes, thereby reducing mission execution and round-trip time, and especially prioritizing the timely completion of high-priority tasks.

[0070] Drones have the lowest total energy consumption: .

[0071] in, For U drones j The energy consumption per unit time (unit: kWh / h, extracted from the knowledge graph) aims to reduce overall energy consumption and improve scheduling economy by rationally allocating drones and planning paths.

[0072] Highest equipment utilization rate: .

[0073] in, For U drones j The available time (unit: h) is primarily aimed at preventing drones from being idle, ensuring that the battery life of each available drone is fully utilized, and improving the overall equipment utilization efficiency.

[0074] The task with the highest priority match: .

[0075] in, The task start time. To ensure high-priority tasks are executed on time and that low-priority tasks do not consume core resources, the system sets task completion times to improve the efficiency of task execution.

[0076] Configure multiple constraints for the objective function, including task allocation constraints, UAV load constraints, time sequence constraints, path constraints, and variable constraints; Based on the scenario requirements, a multi-objective optimization model is obtained by weighting the four initial objective functions.

[0077] An improved genetic algorithm-particle swarm optimization (GA-PSO) hybrid algorithm is used to solve the multi-objective optimization model to obtain the objective solution.

[0078] The objective solution is transformed into an executable scheduling scheme, and the feasibility is verified by a knowledge graph reasoning engine to obtain a feasible UAV scheduling strategy.

[0079] An improved genetic algorithm-particle swarm optimization (GA-PSO) hybrid algorithm is used to solve the optimization model. Constraints from the knowledge graph are introduced as prior knowledge to optimize the solution process and improve efficiency and optimality. The specific solution process is as follows: Initialization: Generate an initial solution based on the candidate drone set. The initial solution includes decision variables to avoid randomly generating invalid solutions (such as allocation schemes that exceed the drone's endurance). The population size is set to 50-100 to ensure the diversity of solutions.

[0080] Fitness calculation uses a weighted single objective function as the fitness function to calculate the fitness value of each initial solution (i.e., scheduling scheme candidate). The lower the fitness value, the better the scheduling scheme.

[0081] Iterative optimization is performed by iteratively solving the problem using improved selection, crossover, mutation operators, and particle update rules.

[0082] Knowledge graph constraint verification is incorporated into the iteration process to ensure that each generation of solutions meets the constraints. The number of iterations is set to 50-100 generations, and the convergence condition is that the fitness value fluctuation is ≤1% for 5 consecutive generations, so as to avoid excessive iteration leading to increased time consumption.

[0083] After the iteration is completed, the solution with the lowest fitness value is selected as the target solution. This target solution corresponds to a complete set of scheduling parameters, which provides the basis for the output of subsequent solutions.

[0084] The optimal solution obtained is transformed into an executable scheduling scheme. At the same time, the final feasibility is verified by a knowledge graph reasoning engine to ensure that the scheme is free from constraints and conflicts and adapts to the actual scenario. The final output scheme includes four core components: UAV allocation: clearly define the patrol task corresponding to each candidate UAV, mark the core requirements of the task, and ensure that the UAV capabilities are accurately matched with the task requirements.

[0085] Flight path planning: Generates detailed flight trajectories based on path variables, marking coordinate points along the route, key obstacle avoidance points, and key inspection areas. Combined with GIS map visualization, it ensures safe path and complete coverage of the inspection area.

[0086] Nest scheduling: Clearly define the nesting location, charging time, and charging duration for each drone to avoid conflicts in nesting charging resources and ensure sufficient drone battery life.

[0087] Task execution order: Clearly define the execution order of multiple tasks, the start and expected completion times of each task, and prioritize the time nodes of high-priority tasks to facilitate staff tracking of task progress.

[0088] S5. Based on the drone scheduling strategy, remote drone patrol and monitoring is implemented to achieve anomaly detection and early warning.

[0089] When dynamic changes occur during patrols (such as sudden drone malfunctions, environmental degradation, or new urgent tasks), the scheduling plan is dynamically adjusted based on real-time updated data from the knowledge graph. For example, when a drone malfunctions, alternative drones are quickly selected through knowledge reasoning, and routes and task allocations are replanned; when an urgent task is added, the scheduling plan for low-priority tasks is adjusted first to ensure that high-priority tasks are executed first. During the adjustment process, the feasibility of the plan is verified in real time through the knowledge graph (such as whether alternative drones meet the task requirements and whether there are new obstacles on the path), ensuring that the adjusted plan is optimal and feasible.

[0090] An anomaly detection model is constructed based on anomaly event knowledge (such as UAV failure modes, environmental anomaly thresholds, and mission anomaly characteristics) from a knowledge graph and real-time data streams. Through knowledge reasoning and threshold judgment, the model detects the following anomalies: UAV malfunctions (sudden battery drop, abnormal attitude, sensor malfunctions, etc.), nesting device anomalies (charging failures, communication interruptions, etc.), environmental anomalies (excessive wind speed, extremely low visibility, etc.), and mission execution anomalies (mission delays, unqualified data acquisition, etc.). When an anomaly is detected, the system immediately triggers an alert (audio alert, pop-up alert, SMS notification) and recommends emergency response plans based on the knowledge graph (such as replacement plans for faulty UAVs and mission adjustment suggestions).

[0091] Example 2 This embodiment mentions a remote patrol and monitoring system for unmanned aerial vehicles (UAVs), such as... Figure 2 As shown, it includes: Data acquisition unit: Acquires historical operational data and basic equipment data, and analyzes the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and anomaly data.

[0092] Knowledge graph construction unit: Extract and organize sample data to construct a knowledge graph for drone patrol.

[0093] Initial strategy generation unit: Obtain the current inspection task, analyze the inspection task based on the UAV inspection knowledge graph to obtain task requirements, and filter out effective candidate UAV groups.

[0094] Target strategy generation unit: Based on effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies. Based on the UAV scheduling strategies, remote UAV patrol and monitoring is carried out to achieve anomaly detection and early warning.

[0095] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program in the memory to implement the content of a remote patrol and monitoring method for unmanned aerial vehicles.

[0096] A storage medium storing computer-executable instructions, which, when loaded and executed by a processor, implement the content of a remote patrol and monitoring method for unmanned aerial vehicles (UAVs).

[0097] In summary, the knowledge graph-based remote patrol optimization scheduling and monitoring method for nested UAVs proposed in this application effectively solves the problems of low scheduling efficiency, weak monitoring capabilities, and poor scene adaptability of current nested UAV remote patrols by constructing a knowledge graph that integrates multi-source data and combining multi-objective optimization algorithms with real-time monitoring technology. This provides a brand-new technical path for the intelligent management of remote patrols.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical methods of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical methods of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical methods to deviate from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for remote inspection and monitoring using unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1. Obtain historical operation data and equipment basic data, and analyze the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and abnormal data; S2. Extract and organize the sample data to construct a knowledge graph of drone patrols; S3. Obtain the current inspection task, analyze the inspection task based on the drone inspection knowledge graph to obtain the task requirements, and filter out the effective candidate drone groups. S4. Based on the effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies. S5. Based on the drone scheduling strategy, remote drone patrol and monitoring is implemented to achieve anomaly detection and early warning.

2. The method for remote patrol and monitoring of unmanned aerial vehicles according to claim 1, characterized in that, The specific content of extracting and organizing sample data to construct a drone patrol knowledge graph in S2 includes: Using ontology modeling, we define the core entities, core relationships, and entity attributes in the inspection scenario, and construct the schema layer of the knowledge graph. For different types of data, a differentiated knowledge extraction strategy is adopted to extract knowledge from the data and obtain the initial knowledge triple (subject-relation-object). The entities in the initial knowledge triples are aligned using an algorithm based on attribute similarity and semantic similarity to merge duplicate entities; Attribute fusion is performed by selecting or weighting conflicting attributes based on data reliability. By integrating relationships and eliminating redundancy and contradictions, a knowledge graph of drone patrols is obtained.

3. The method for remote inspection and monitoring of unmanned aerial vehicles according to claim 1, characterized in that, To obtain the current inspection task, the specific requirements of the task are analyzed based on the drone inspection knowledge graph: Select patrol task entities from the entities in the knowledge graph, extract core attributes from the patrol task entities, and generate a task basic information table. The task basic information table includes the task identifier T. i (i is the task number, i=1,2,...,n), Task type Type(T) i Task priority P(T) i ), Patrol Area (T) i Task deadline (T) i The data requirement is DataReq(T) i ); Obtain the current inspection task and match it with the task information table to obtain the task requirements; From the entities in the knowledge graph, select entities in the patrol area and environmental elements, extract geographical and environmental features, and quantify them into scheduling constraint parameters: The geographical and environmental features include regional topographic complexity: Terrain (A) i Obstacle distribution: Obstacle(A) i Real-time environmental parameters, Env(T) i ); Based on geographical and environmental characteristics, combined with regional area and inspection accuracy, the estimated task time Test(T) is calculated. i ); By combining knowledge graph reasoning, hard and soft constraints corresponding to the task are generated, and scheduling sub-goals are broken down.

4. The method for remote inspection and monitoring of unmanned aerial vehicles according to claim 3, characterized in that, The specific details of the selected valid candidate drone groups include: From the drone entities in the knowledge graph, select all drones whose status is marked as available and form an initial drone pool; The knowledge graph extracts the core attribute information of each drone in the initial drone pool. The core attribute information includes the drone's endurance, maximum flight altitude, type and accuracy of the sensors it carries, its home nest information, and the straight-line distance between its home nest and the center of the current patrol area. Based on the hard constraints derived from the task analysis, each drone in the initial drone pool is verified one by one, and only drones that fully satisfy all hard constraints are retained to form an intermediate candidate pool. For drones in the intermediate candidate pool, a weighted score is applied based on the soft constraint dimension in combination with the mission requirements to quantify the suitability of each drone for the mission. Based on the soft-constraint weighted scoring results, the drones in the intermediate candidate pool are sorted from high to low, and the top n% of drones are finally selected as the effective candidate drone group.

5. The method for remote inspection and monitoring of unmanned aerial vehicles according to claim 3, characterized in that, The specific content of generating a drone scheduling strategy based on an improved genetic algorithm-particle swarm optimization hybrid algorithm based on effective candidate drone groups is as follows: Define the drone allocation variable, mission execution order variable, flight path variable, and charging arrangement variable as decision variables; Four initial objective functions are constructed based on task completion efficiency, resource consumption, equipment utilization, and task priority matching degree; Configure multiple constraints for the objective function, including task allocation constraints, UAV load constraints, time sequence constraints, path constraints, and variable constraints; Based on the scenario requirements, a multi-objective optimization model is obtained by weighting the four initial objective functions; An improved genetic algorithm-particle swarm optimization hybrid algorithm is used to solve the multi-objective optimization model to obtain the objective solution; The objective solution is transformed into an executable scheduling scheme, and the feasibility is verified by a knowledge graph reasoning engine to obtain a feasible UAV scheduling strategy.

6. The method for remote inspection and monitoring of unmanned aerial vehicles according to claim 3, characterized in that, The objective solution obtained by solving the multi-objective optimization model using an improved genetic algorithm-particle swarm optimization hybrid algorithm includes: Initialization: Generate an initial solution based on the candidate set of unmanned aerial vehicles. The initial solution contains decision variables. Fitness calculation: The fitness function is a weighted single objective function, and the fitness value of each initial solution is calculated. Iterative optimization is performed by iteratively solving the problem using improved selection, crossover, and mutation operators and particle update rules. Knowledge graph constraint verification is incorporated into the iteration process to ensure that the solution in each generation meets the constraints. The number of iterations is set to 50-100 generations, and the convergence condition is that the fitness value fluctuation is ≤1% for 5 consecutive generations. After the iteration is completed, the solution with the lowest fitness value is selected as the target solution, which corresponds to a complete set of scheduling parameters.

7. The method for remote inspection and monitoring of unmanned aerial vehicles according to claim 3, characterized in that, The specific content of feasibility verification using a knowledge graph reasoning engine includes: Drone allocation: Clearly define the patrol mission corresponding to each candidate drone, mark the core requirements of the mission, and ensure that the drone's capabilities are accurately matched with the mission requirements; Flight path planning: Generate detailed flight trajectories based on path variables, mark the coordinates of the route, key obstacle avoidance points, and key inspection areas, and combine them with GIS map visualization to ensure safe path and complete coverage of the inspection area; Nest scheduling: Clearly define the nests where each drone will dock, the timing of its charging, and the duration of its charging to avoid conflicts over nest charging resources and ensure sufficient drone battery life; Task execution order: Clearly define the execution order of multiple tasks, the start and expected completion times of each task, and prioritize the time nodes of high-priority tasks to facilitate staff tracking of task progress.

8. A remote patrol and monitoring system for unmanned aerial vehicles (UAVs), characterized in that, include: Data acquisition unit: Acquires historical operational data and basic equipment data, and analyzes the above data to obtain sample data including drone data, nest data, inspection data, environmental data, and anomaly data; Knowledge graph construction unit: Extracts and organizes sample data to construct a knowledge graph for drone patrols; Initial strategy generation unit: Obtain the current inspection task, parse the inspection task based on the UAV inspection knowledge graph to obtain task requirements, and filter out effective candidate UAV groups; Target strategy generation unit: Based on effective candidate UAV groups, an improved traditional genetic algorithm-particle swarm optimization hybrid algorithm is used to generate UAV scheduling strategies. Based on the UAV scheduling strategies, remote UAV patrol and monitoring is carried out to achieve anomaly detection and early warning.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the content of the unmanned aerial vehicle remote patrol and monitoring method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the content of the unmanned aerial vehicle (UAV) remote patrol and monitoring method as described in any one of claims 1 to 7.

Images