Unmanned aerial vehicle cluster task allocation and cooperative control method for low-altitude economy

Abstract

The invention discloses an unmanned aerial vehicle cluster task allocation and cooperative control method for low-altitude economy, and belongs to the technical field of unmanned aerial vehicle clusters. According to the method, the problems of inaccurate task allocation, unintelligent cooperative control and failure prediction and maintenance in time in the existing method are solved, and by establishing the ontology knowledge base and the multi-dimensional semantic entity of the low-altitude economic tasks, the sequence constraint and the overlapping region between the tasks are identified, and the space-time buffer region is dynamically divided, so that the efficiency of the low-altitude economic tasks is improved. Conflicts and interferences among the unmanned aerial vehicles are effectively avoided, and the safety and efficiency of task execution are improved; the task priority, the unmanned aerial vehicle state and the task execution efficiency are comprehensively analyzed through a multi-objective optimization algorithm, and it is ensured that high-priority tasks are executed preferentially; the task allocation scheme is dynamically adjusted according to real-time feedback, the unmanned aerial vehicle in a good healthy state is preferentially allocated to execute the task, a standby task scheme is reserved for the unmanned aerial vehicle which may have a fault, and the reliability and continuity of task execution are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm technology, specifically to a method for task allocation and collaborative control of UAV swarms for the low-altitude economy. Background Technology

[0002] With the rapid development of the low-altitude economy, drones have been widely used in logistics delivery, agricultural plant protection, environmental monitoring, and power line inspection. Drone swarm operations can improve mission efficiency and reduce costs, but current methods for drone swarm task allocation and collaborative control have the following problems:

[0003] The task allocation is not precise enough and cannot be dynamically adjusted according to task requirements and drone status;

[0004] The collaborative control is not intelligent enough, lacks effective communication and navigation methods, and is prone to task conflicts and communication interruptions; moreover, it cannot predict and maintain faults in a timely manner, affecting the smooth execution of tasks.

[0005] Therefore, to meet current needs, a method for task allocation and collaborative control of UAV swarms, geared towards the low-altitude economy, is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide a method for task allocation and collaborative control of UAV swarms for low-altitude economic purposes. By establishing an ontology knowledge base and multi-dimensional semantic entities for low-altitude economic tasks, the method identifies the sequence constraints, overlapping areas, and dynamically divides spatiotemporal buffers between tasks, effectively avoiding conflicts and interference between UAVs and improving the safety and efficiency of task execution. A multi-objective optimization algorithm is used to comprehensively analyze task priority, UAV status, and task execution efficiency, ensuring that high-priority tasks are executed first. The task allocation scheme is dynamically adjusted based on real-time feedback, prioritizing tasks assigned to UAVs in good health and reserving backup task schemes for UAVs that may malfunction, thus improving the reliability and continuity of task execution and solving the problems mentioned in the background art.

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

[0008] A method for task allocation and collaborative control of UAV swarms for the low-altitude economy includes the following steps:

[0009] S1. Based on the requirements of low-altitude economic missions, establish a mission model, including mission type, mission area, mission time, mission priority, and mission objectives; perform visual inspection on the mission area to obtain environmental information and target object information within the mission area;

[0010] S2. Obtain the status information of the drone cluster, including but not limited to the drone's battery level, payload, flight status, flight capability, communication status, and health status. Based on the task model and the drone cluster's status information, use a multi-objective optimization algorithm to allocate tasks and assign tasks to drones suitable for performing the task.

[0011] S3. Establish a ground calibration system to calibrate the UAV's navigation system; use satellite antennas to receive satellite signals to provide positioning information for the UAV; establish a communication network, including communication links between UAVs and communication links between UAVs and the ground control center, to realize information sharing and transmission of collaborative control commands between UAVs;

[0012] S4. Based on the task allocation results and the real-time status information of the UAV, formulate a collaborative control strategy. The collaborative control strategy includes, but is not limited to, flight path planning, task execution sequence, communication coordination, and conflict avoidance.

[0013] Furthermore, it also includes the following steps:

[0014] S5. Multiple sensors are mounted on the drone to collect real-time operating data, including but not limited to motor speed, battery voltage, flight attitude, temperature and vibration data.

[0015] A fault prediction model is established using a convolutional neural network to extract and learn features from operational data, and to predict the types and probabilities of possible faults in the UAV.

[0016] Establish a health management mechanism to maintain and service drones in a timely manner based on fault prediction results; when the fault prediction model predicts that a drone has a fault risk, take corresponding maintenance measures according to the fault type and risk level.

[0017] Establish health records for drones, recording their operational data and maintenance history.

[0018] Furthermore, in S1, a mission model is established based on the requirements of low-altitude economic missions, including the following steps:

[0019] Based on the task model, an ontology knowledge base for low-altitude economic tasks is established, the task content is converted into multi-dimensional semantic entities, and detailed attributes and relationships are defined for each semantic entity.

[0020] Identify the sequential constraints between tasks from multi-dimensional semantic entities;

[0021] By analyzing the geographical coordinate range of entities in the task area, overlapping areas are identified; and spatiotemporal buffer zones are dynamically divided according to task priority and task type to avoid conflicts and interference between drones.

[0022] Identify task groups that require similar equipment or drones with the same skills, as well as collaborative task packages that can be combined for execution;

[0023] Based on the recognition results, multiple task templates are preset for multi-dimensional semantic entities.

[0024] Furthermore, in S2, based on the task model and the state information of the UAV swarm, a multi-objective optimization algorithm is used to allocate tasks, including the following steps:

[0025] When a task request is received to identify and encode a specific object, the complete digital profile of the specific object is obtained, and the physical characteristics, operation interface and historical service records of the specific object are parsed.

[0026] Based on the analysis results, a preset task template is matched, and personalized task parameters are generated;

[0027] A multi-objective optimization algorithm is used to allocate tasks. Based on task priority, high-priority tasks are assigned to drones with higher task execution efficiency, thereby reducing task completion time.

[0028] Based on the drone's payload capacity, equipment type, and skill level, medium-priority and low-priority tasks are assigned to appropriate drones.

[0029] The system monitors task execution and drone status in real time, regularly sends task progress reports and its own status information to the system, and dynamically adjusts the task allocation plan based on real-time information. If an anomaly occurs during task execution, the system will automatically activate the anomaly handling mechanism to reassign tasks to drones that are currently not on a task or adjust the task path.

[0030] The task priority is dynamically adjusted based on real-time feedback during task execution.

[0031] Furthermore, in S2, based on the task model and the state information of the UAV swarm, a multi-objective optimization algorithm is used to allocate tasks, which also includes the following steps:

[0032] Based on the real-time feedback data, identify problems and potential optimization points that arise during task execution;

[0033] Based on the identified problems and optimization points, the parameters and processes of the task template are automatically adjusted;

[0034] By continuously collecting and analyzing task execution data, we dynamically optimize task templates to ensure that they can adapt to different task scenarios and environmental changes.

[0035] During the task allocation process, priority is given to assigning tasks to drones in good health, and backup task plans are reserved for drones that may malfunction.

[0036] Furthermore, in S2, based on the task model and the state information of the UAV swarm, a multi-objective optimization algorithm is used to allocate tasks, which also includes the following steps:

[0037] The health status of drones is predicted in real time based on a fault prediction model. If a fault is predicted in a drone, an early warning will be issued.

[0038] During the task allocation process, priority is given to assigning tasks to drones in good health; for drones that may malfunction, backup task plans will be generated in advance.

[0039] The health status of the drone is updated in real time to the task allocation system, ensuring that task allocation decisions are always based on the latest health status information.

[0040] Furthermore, in S4, a collaborative control strategy is formulated based on the task allocation results and the real-time status information of the UAV, including the following steps:

[0041] A multi-agent cooperative communication mechanism is introduced, defining a communication agent for each UAV, enabling real-time information sharing and collaborative decision-making among UAVs through a communication network;

[0042] When any drone detects a new obstacle, it transmits the real-time obstacle information to neighboring drones. The neighboring drones then dynamically adjust their flight paths and mission execution strategies based on this information to avoid collisions with obstacles, thus achieving intelligent collaborative control of the swarm.

[0043] When multiple drones are conducting patrol missions on the same route or in the same area, the flight paths and mission execution order of the drones are dynamically adjusted, and the drones share the patrol results in real time through the communication network to ensure the efficient execution of the mission.

[0044] Based on the mission priority and the UAV's communication status, the parameters of the communication link are dynamically adjusted to ensure the communication needs of high-priority missions; when a communication link interruption or weak signal is detected, the system will automatically switch to a backup communication mode or re-establish the communication link.

[0045] Furthermore, S1 also includes the following steps:

[0046] The target object is identified and encoded by converting its feature information into a unique identifier and storing it in the database.

[0047] Based on the task type and task area characteristics, the task is decomposed into multiple subtasks according to preset rules; each subtask has an independent task model, including subtask type, subtask area, subtask time, subtask priority and subtask objective.

[0048] Furthermore, in step S2, the allocation of tasks using a multi-objective optimization algorithm based on the task model and the state information of the UAV cluster specifically includes the following steps:

[0049] The real-time flight data of the UAV during the execution of the mission is acquired, and the real-time flight data includes at least: the current remaining battery percentage, the current communication signal-to-noise ratio, and the current environmental perception confidence level.

[0050] Based on the real-time flight data, the task execution matching index of the UAV for the task is calculated using a preset calculation formula; the calculation formula is:

[0051] ;

[0052] In the formula, The matching degree index is used to determine the task execution. The remaining battery percentage is 0 to 1. The signal-to-noise ratio of the current communication signal is expressed in decibels. This is a preset communication quality benchmark threshold, in decibels. The current environment perception confidence level, with a value ranging from 0 to 1, is used to characterize the accuracy probability of the UAV visual sensor in extracting environmental features. The estimated remaining time required to perform the task described above, in seconds; This represents the maximum remaining flight time of the drone, in seconds. is the base of the natural logarithm; is the signal sensitivity adjustment coefficient, which is a constant greater than 0; Energy weighting coefficient, This is the risk penalty weighting coefficient, and and All are dimensionless constants;

[0053] The task allocation scheme is dynamically weighted and adjusted based on the calculated task execution matching index. When the task execution matching index is lower than a preset threshold, the currently allocated task is automatically removed and a reallocation mechanism is triggered.

[0054] Furthermore, in step S4, the step of formulating a collaborative control strategy based on the task allocation results and the real-time status information of the UAV also includes the following steps:

[0055] A distributed semantic fragment reconstruction mechanism is established, which is triggered when the bandwidth of the communication network between the drones is lower than a preset security value.

[0056] The UAV extracts features from the collected environmental image data and transforms the extracted key features into lightweight semantic fragment data packets. The lightweight semantic fragment data packets contain only obstacle category identification codes, obstacle three-dimensional coordinate center points, and obstacle boundary vector information.

[0057] The drone broadcasts the lightweight semantic fragment data packet through the communication network;

[0058] The adjacent UAV receives the lightweight semantic fragment data packet and, in conjunction with the locally pre-stored ontology knowledge base, reconstructs a three-dimensional holographic mapping model of the surrounding environment in the local virtual space.

[0059] Based on the reconstructed 3D holographic mapping model, the adjacent UAVs use the artificial potential field method to calculate the obstacle avoidance resultant force, achieving collaborative obstacle avoidance and dynamic path correction without transmitting the original image data.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] 1. In this invention, by establishing an ontology knowledge base and multi-dimensional semantic entities for low-altitude economic tasks, the sequential constraints, overlapping areas, and dynamic division of spatiotemporal buffer zones between tasks are identified, effectively avoiding conflicts and interference between UAVs and improving the safety and efficiency of task execution; and by identifying task groups and collaborative task packages of UAVs with the same type of equipment or the same skills, the task allocation strategy is further optimized, improving the collaborative operation capability of UAV swarms.

[0062] 2. In this invention, by using a multi-objective optimization algorithm to comprehensively analyze task priority, UAV status, and task execution efficiency during the task allocation process, high-priority tasks are ensured to be executed first. The task allocation scheme is dynamically adjusted based on real-time feedback, enhancing the system's adaptability and reliability. Furthermore, UAVs in good health are prioritized for task allocation, while backup task schemes are reserved for UAVs that may malfunction, further improving the reliability and continuity of task execution. Simultaneously, by continuously collecting and analyzing task execution data, task templates are dynamically optimized to ensure that task templates can adapt to different task scenarios and environmental changes, thereby improving the overall performance of UAV swarms in low-altitude economic tasks. Attached Figure Description

[0063] Figure 1 This is a flowchart of the UAV swarm task allocation and collaborative control method for low-altitude economy according to the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] To address the technical issues of existing UAV swarm task allocation and collaborative control methods, such as inaccurate task allocation (unable to dynamically adjust based on task requirements and UAV status), insufficient intelligent collaborative control (lacking effective communication and navigation methods, leading to task conflicts and communication interruptions), and inability to predict and maintain faults in a timely manner, thus affecting successful task execution, please refer to [link to relevant documentation]. Figure 1 This embodiment provides the following technical solution:

[0066] A method for task allocation and collaborative control of UAV swarms for the low-altitude economy includes the following steps:

[0067] S1. Based on the requirements of low-altitude economic missions, establish a mission model, including mission type (e.g., logistics distribution, agricultural plant protection, environmental monitoring, power line inspection, etc.), mission area, mission time, mission priority (divided according to the urgency and importance of the mission, such as high, medium, and low levels), and mission objectives (e.g., quantity of goods delivered, plant protection area, monitoring indicators, etc.). Perform visual inspection on the mission area to obtain environmental information and target object information within the mission area, such as topography, building distribution, and the location and characteristics of target objects. For example, for logistics distribution missions, the target object may be a delivery point; for agricultural plant protection missions, the target object may be a crop area; for environmental monitoring missions, the target object may be a monitoring point or a pollution source, etc. This includes the following steps:

[0068] Based on the task model, an ontology knowledge base for low-altitude economic tasks is established, converting task content into multi-dimensional semantic entities. Detailed attributes and relationships are defined for each semantic entity. For example, the task type entity includes attributes such as task name, task description, and task execution flow; the task area entity includes attributes such as geographic coordinate range, terrain, and building distribution; and the target object entity includes attributes such as the target object's location, characteristics, operation interface, and historical service records. From these multi-dimensional semantic entities, the sequential constraints between tasks are identified, such as: maintenance can only be carried out after inspection is completed, and data processing can only be carried out after environmental monitoring is completed. By analyzing the task types and execution flows in the task model, corresponding constraints are automatically identified to ensure that the execution order of tasks meets requirements during task allocation. Furthermore, by analyzing the task area entity... The system identifies overlapping areas within the geographic coordinate range of the entity and dynamically divides spatiotemporal buffer zones based on task priority and type to avoid conflicts and interference between drones. For example, if multiple patrol tasks require the use of the same type of drone, the system can automatically group these tasks into a task group and prioritize the allocation of drones with the corresponding equipment or skills during task assignment. The system also identifies task groups that require drones of the same type or skills and collaborative task packages that can be merged for execution. For example, multiple patrol points on the same route can be merged into a collaborative task package. By analyzing the task objectives and task areas in the task model, the system automatically identifies these collaborative task packages and assigns them to the same drone or a group of drones during task assignment, improving task execution efficiency. Based on the identification results, the system presets multiple task templates for multi-dimensional semantic entities.

[0069] The system identifies and encodes target objects, converting their characteristic information into unique identifiers and storing them in a database. This identification and encoding follows specific rules, such as combining the target object's location coordinates and characteristic information to generate a unique code. This facilitates the identification and management of target objects during subsequent task allocation and execution. For complex tasks or large-scale task areas, the system decomposes tasks into multiple sub-tasks based on task type and task area characteristics, according to preset rules such as geographical region division and task type division. Each sub-task has an independent task model, including sub-task type, sub-task area, sub-task time, sub-task priority, and sub-task objective. For example, in logistics delivery tasks, a large-area delivery task is decomposed into multiple small-area delivery sub-tasks; in environmental monitoring tasks, a large-scale monitoring task is decomposed into multiple monitoring point monitoring sub-tasks. This better adapts to the parallel operation capabilities of drone swarms and improves task execution efficiency.

[0070] S2. Obtain the status information of the drone swarm, including but not limited to the drones' battery level (remaining battery percentage), payload (currently carried cargo or equipment weight), flight status (e.g., flight speed, flight altitude, flight attitude), flight capabilities (e.g., maximum flight speed, maximum flight altitude, endurance), communication status (communication link signal strength, communication latency), and health status (e.g., motor failure probability, battery health). Based on the task model and the drone swarm's status information, use a multi-objective optimization algorithm to allocate tasks, comprehensively considering multiple objectives such as task execution efficiency, task completion quality, drone energy consumption, and task risk, and assign tasks to drones suitable for performing the task; including the following steps:

[0071] When a task request for identifying and encoding a specific object is received, the system obtains the complete digital profile of that object, analyzes its physical characteristics, operational interfaces, and historical service records, and matches it with a preset task template based on the analysis results. Personalized task parameters are then generated. For example, for a wind turbine blade inspection task, the system automatically adapts to the specific model's required detection mode, including parameters such as detection distance, detection angle, and detection frequency. For agricultural plant protection tasks, the system generates corresponding plant protection parameters based on the crop type and growth stage, such as pesticide spraying amount and spraying height. A multi-objective optimization algorithm is used to allocate tasks, prioritizing high-priority tasks to drones with higher execution efficiency to reduce task completion time. Furthermore, based on the drone's payload capacity, equipment type, and skill level, tasks are further optimized. Priority and low-priority tasks are assigned to suitable drones; for example, for tasks requiring high-resolution cameras, drones equipped with high-resolution cameras are given priority. The system monitors task execution and drone status in real time, periodically sending task progress reports and its own status information to the system, and dynamically adjusting the task allocation scheme based on real-time information. If an anomaly occurs during task execution, such as insufficient drone battery or the appearance of new obstacles in the task area, an anomaly handling mechanism will be automatically activated, reassigning the task to a drone currently without a task or adjusting the task path to avoid obstacles. Task priorities are dynamically adjusted based on real-time feedback during task execution; for example, if a high-priority task cannot be completed on time due to environmental changes or drone malfunction, the system will automatically increase the priority of other related tasks to ensure the achievement of the overall task objective.

[0072] Based on real-time feedback data, the system identifies problems and potential optimization points during task execution. For example, if a task takes longer than expected, the system analyzes the reasons, which could include unreasonable task path planning, complex task area environment, or insufficient drone performance. Based on the identified problems and optimization points, the system automatically adjusts the parameters and processes of the task template. For example, if the task path planning is unreasonable, the system will replan the path to avoid complex terrain or obstacles, improving task execution efficiency. If task parameters such as detection distance and spraying height need adjustment, the system will dynamically update the parameters based on real-time data. For example, in a wind turbine blade inspection task, if the blade... If the detection results for certain parts of the image are inaccurate, the system will adjust the detection distance and angle to ensure the accuracy of the detection results. By continuously collecting and analyzing task execution data, the system dynamically optimizes task templates to ensure that the task templates can adapt to different task scenarios and environmental changes. For example, the system automatically adjusts task templates based on seasonal changes and dynamic changes in the task area to ensure efficient task execution. During task allocation, priority is given to assigning tasks to drones in good health, and backup task plans are reserved for drones that may malfunction. For example, if the fault prediction model predicts that the battery of a drone may malfunction, a backup drone will be arranged in advance to take over the task to ensure the continuity and reliability of the task.

[0073] The system uses a fault prediction model to predict the health status of drones in real time, including motor failure probability, battery health, and sensor performance. If a drone malfunction is predicted, an early warning will be issued. During task allocation, drones in good health are prioritized for assignment. For drones prone to malfunction, backup task plans will be generated in advance. For example, if the fault prediction model predicts a potential battery failure in a drone, a backup drone will be assigned to take over the task. The backup drone will dynamically adjust its task path and parameters based on real-time data to ensure task continuity and reliability. If a drone malfunctions during task execution, the system will automatically reassign the task to another available drone. The system will dynamically adjust the task allocation plan based on the drone's current status and task priority to ensure successful task execution. The drone's health status is updated in real-time to the task allocation system, ensuring that task allocation decisions are always based on the latest health status information. For example, if a drone requires maintenance after completing a task, the system will automatically mark it as unavailable until maintenance is completed and it returns to a healthy state.

[0074] The beneficial effects achieved by the above content are as follows: by establishing an ontology knowledge base and multi-dimensional semantic entities for low-altitude economic tasks, the sequential constraints, overlapping areas, and dynamic division of spatiotemporal buffer zones between tasks are identified, effectively avoiding conflicts and interference between UAVs and improving the safety and efficiency of task execution; and by identifying task groups and collaborative task packages of UAVs with similar equipment or skills, the task allocation strategy is further optimized, improving the collaborative operation capability of UAV swarms.

[0075] S3. Establish a ground calibration system to calibrate the UAV's navigation system and improve navigation accuracy. For example, by setting multiple calibration points at known locations, the UAV can correct its navigation system errors during flight by comparing its relative positions with these calibration points. Utilize satellite antennas to receive satellite signals, providing high-precision positioning information for the UAV and enabling accurate navigation. For example, install satellite antennas on the UAV to receive signals from multiple satellites and obtain the UAV's positioning information through satellite positioning algorithms. Establish a communication network, including communication links between UAVs and between UAVs and the ground control center, to achieve information sharing and collaborative control command transmission between UAVs, such as Wi-Fi, 4G / 5G, and self-organizing networks, ensuring communication stability and reliability.

[0076] S4. Based on the task allocation results and the real-time status information of the UAVs, formulate a collaborative control strategy. This strategy includes, but is not limited to, flight path planning, task execution order, communication coordination, and conflict avoidance. Flight path planning, based on the environmental information of the task area and the flight capabilities of the UAVs, plans the optimal flight path for each UAV to avoid collisions and duplicate flights. Task execution order is determined by the task priority and the UAV status, rationally arranging the order in which the UAVs execute tasks. Communication coordination ensures that UAVs can transmit information such as task status information and environmental information in a timely and accurate manner. The conflict avoidance mechanism monitors the position and flight trajectory of the UAVs in real time, and when a potential conflict is detected, adjusts the flight path or speed of the UAVs in a timely manner to avoid collisions. This ensures that the UAV swarm can complete its tasks efficiently and safely. This includes the following steps:

[0077] A multi-agent cooperative communication mechanism is introduced, defining a communication agent for each UAV. Through a communication network, UAVs can share information and make collaborative decisions in real time. When any UAV detects a new obstacle, it transmits the real-time obstacle information to neighboring UAVs. The neighboring UAVs then dynamically adjust their flight paths and task execution strategies based on this information to avoid collisions with obstacles, achieving intelligent collaborative control of the swarm. When multiple UAVs are performing patrol tasks on the same route or in the same area, their flight paths and task execution order are dynamically adjusted. The UAVs share patrol results in real time through the communication network, ensuring efficient task execution. For example, in a maintenance task, one UAV… The first drone is responsible for inspecting and locating fault points. Other drones dynamically adjust their flight paths based on the inspection results, carrying maintenance tools and equipment to the fault points for repair. During the mission, drones share mission status and repair progress in real time through the communication network to avoid duplicate inspections. The communication link parameters, such as bandwidth, latency, and signal strength, are dynamically adjusted according to the mission priority and the drone's communication status to ensure the communication needs of high-priority tasks. When a communication link interruption or weak signal is detected, the system will automatically switch to a backup communication mode or re-establish the communication link. For example, when the 4G signal is weak, the system will automatically switch to Wi-Fi or self-organizing network mode to ensure communication continuity.

[0078] S5. Equip the drone with multiple sensors to collect real-time operational data, including but not limited to motor speed, battery voltage, flight attitude, temperature, and vibration data; use convolutional neural networks to build a fault prediction model, extract and learn features from the operational data, and predict the types and probabilities of possible drone faults; establish a health management mechanism to perform timely maintenance and upkeep of the drone based on the fault prediction results; when the fault prediction model predicts a fault risk in the drone, take corresponding maintenance measures according to the fault type and risk level, such as replacing the battery, repairing the motor, and calibrating the sensors; establish a health record for the drone, recording its operational data and maintenance history to provide a reference for subsequent health management.

[0079] The beneficial effects achieved by the above are as follows: By using a multi-objective optimization algorithm to comprehensively analyze task priority, UAV status, and task execution efficiency during the task allocation process, high-priority tasks are ensured to be executed first; the task allocation scheme is dynamically adjusted based on real-time feedback, enhancing the system's adaptability and reliability; and priority is given to assigning tasks to UAVs in good health, while reserving backup task schemes for UAVs that may malfunction, further improving the reliability and continuity of task execution; at the same time, by continuously collecting and analyzing task execution data and dynamically optimizing task templates, the system ensures that task templates can adapt to different task scenarios and environmental changes, improving the overall performance of UAV swarms in low-altitude economic tasks.

[0080] Example 1: Suppose a power company uses a swarm of drones to conduct routine inspections and emergency fault handling of high-voltage transmission lines in mountainous areas.

[0081] Implementation steps: Upon receiving the task of inspecting power transmission lines in mountainous areas, a pilot drone is used to visually inspect 50 kilometers of the line, identify 27 towers and assign them unique object identification codes; the system automatically recognizes the order constraint of first inspecting the entire line and then focusing on key areas.

[0082] Twelve drones were deployed and are in good condition. Historical records of the transmission towers were analyzed, and targeted inspection templates were matched for towers with previous insulator damage. Eight drones were assigned to routine inspections, three to key inspections, and one as a backup. A ground calibration system was deployed in mountainous areas to improve positioning accuracy. Drones were equipped with satellite antennas, and satellite-enhanced positioning was activated when the signal was weak in canyons. A combination of 4G and self-organizing networks was used to ensure continuous communication.

[0083] During operation, abnormal overheating of the insulator was detected. The event priority was immediately escalated, a backup drone was deployed for confirmation, and a maintenance alert was sent to the ground. A low-altitude flight path was planned for the drones. Upon detecting the fault, multiple drones communicated collaboratively; one confirmed the fault, while the others adjusted their routes to avoid it. The maintenance drone proceeded to handle the issue, and collisions were avoided by dynamically dividing the spatiotemporal buffer zone. The drone status was monitored in real time, and a fault prediction model issued an alert: a drone's motor was malfunctioning. Its mission was immediately adjusted, and it was scheduled to return to base early for maintenance.

[0084] Example 2: Assume that a large-scale modern farm uses a swarm of drones to simultaneously carry out pest and disease control, crop monitoring, and growth assessment.

[0085] Implementation steps: The farm issues tasks for wheat pest control (high priority), corn monitoring (medium priority), and soybean modeling (low priority); drones use visual detection to identify crop boundaries and disease patches, assigning object identification codes to each field. The farm deploys a ground calibration system to provide centimeter-level positioning, drones are equipped with satellite antennas to assist in precise positioning, and a 5G private network provides full coverage to ensure real-time data sharing.

[0086] Fourteen drones, including those for plant protection, monitoring, and scanning, were deployed to analyze field records, match key prevention and control templates for fields that had previously suffered from diseases, and automatically generate personalized operation parameters; all six plant protection drones were deployed for wheat prevention and control, while the monitoring drones operated simultaneously.

[0087] During the operation, signs of nutrient deficiency in the corn were detected. Priorities were dynamically adjusted, and agricultural drones were deployed to supplement the operation. The monitoring drone first scanned and generated a prescription map. The agricultural drones sprayed according to the map's variables. The scanning drone then assessed the operation quality, delineated dynamic safety zones, and negotiated passage priorities to avoid collisions. Through real-time collection of drone operation data, a fault prediction model issued a warning: a pump in one agricultural drone was malfunctioning, and post-operation maintenance was recommended.

[0088] Example 3: Assuming a sudden flood in the city, use a swarm of drones for emergency supplies delivery and personnel search.

[0089] Implementation steps: The emergency system issues three tasks: delivery, search and assessment. Drones rapidly conduct visual inspections of the disaster area, mapping the flooded area, identifying safe points and population gathering points, and assigning emergency object identification codes. A portable ground calibration system is deployed, using satellite antennas for navigation and communication, establishing a mesh self-organizing network to cover the disaster area, with satellite communication used for remote data transmission.

[0090] Twenty drones of various types were deployed on standby to analyze the needs of resettlement sites and match precise delivery templates. Four logistics drones were prioritized for delivering medical kits, and four search drones were assigned to focus on searching high-probability areas.

[0091] Upon discovering trapped personnel during the operation, priority was immediately prioritized, and nearby logistics drones were deployed to deliver rescue supplies. A fault prediction model issued an early warning: some drones faced risks under severe weather conditions, prompting advance task replacements. Flight paths were dynamically updated based on flood changes, and search drones employed a coordinated, grid-like approach, automatically focusing on targets upon discovery. Airspace was managed through altitude-level separation to avoid conflicts. Enhanced monitoring was implemented in extreme environments, and the fault prediction model assessed the overall risk to the cluster, recommending regular rotation and rest, and establishing rapid maintenance points to ensure continuous operational capability. Based on this, the mission successfully delivered 1.2 tons of supplies, located 47 trapped personnel, and suffered no drone losses throughout the entire operation.

[0092] Working principle: A task model is established based on the requirements of low-altitude economic missions. Environmental and target object information of the mission area is acquired through visual detection, and an ontology knowledge base is built to identify constraints and collaborative task packages between tasks. The status information of the UAV swarm is acquired, and multi-objective optimization algorithms are used to dynamically allocate tasks by comprehensively considering factors such as task priority, UAV status, and task execution efficiency. The task execution status is monitored in real time, and the task allocation scheme is adjusted based on feedback. A communication network is established to ensure information sharing and collaborative control. A multi-agent collaborative communication mechanism is introduced to share obstacle information and task status in real time, dynamically adjust flight paths and task execution order, and ensure efficient task execution. By carrying multiple sensors to collect UAV operation data in real time, a fault prediction model is used to predict UAV faults in advance and take maintenance measures to ensure the health status of UAVs and the continuity of missions.

[0093] Example 4:

[0094] In complex application scenarios for the low-altitude economy (such as high-density urban logistics, refined power grid inspection, and emergency disaster relief), the operational efficiency of drone swarms no longer depends solely on endurance. Instead, it is constrained by the nonlinear coupling of multiple physical constraints, including communication link stability, environmental perception clarity, and mission time windows. Traditional linear weighted allocation methods struggle to cope with the challenges posed by the complex electromagnetic environment and variable weather conditions at low altitudes. This can easily lead to high-value missions being assigned to drones with high battery power but nearing disconnection or those with high battery power but poor perception, resulting in mission failures or even safety incidents. To address this, this embodiment constructs a nonlinear coupling evaluation model based on real-time flight data to achieve globally optimal allocation of mission resources.

[0095] In step S2, to support the accurate operation of the subsequent multi-objective optimization algorithm, the system first needs to establish a high-frequency, reliable data acquisition link on the UAV's onboard terminal. The real-time flight data includes at least: the current remaining battery percentage (…). ), current communication signal-to-noise ratio ( ) and current environmental perception confidence ( The acquisition of the above data is not a simple register read, but rather involves rigorous filtering and unification of physical dimensions.

[0096] Get the current remaining battery percentage ( When performing high-maneuver maneuvers (such as rapid acceleration and wind-resistant hovering), the drastic fluctuations in motor load current during drone operation can cause a false drop in battery terminal voltage, leading to errors when directly using the voltage method to calculate battery capacity. Therefore, in this embodiment, the BMS system incorporates a high-precision current sampling resistor and a fuel gauge chip, employing a hybrid estimation strategy combining the Ah-Integration method and the Open Circuit Voltage (OCV) method. The processor samples the battery pack's charging and discharging current at a frequency of 50Hz and, combined with real-time cell temperature data, uses the Arrhenius equation to compensate for and correct the battery's chemical activity. The system internally maintains a dynamic battery aging model (SOH), updating the battery's effective total capacity in real time. The final calculated... It is a double-precision floating-point number normalized to between 0.00 and 1.00, eliminating the effects of transient voltage fluctuations. This value accurately reflects the material basis for the UAV to perform its mission under the current physical conditions.

[0097] Current communication signal-to-noise ratio (SNR) The radio frequency (RF) link status is a key indicator for quantifying the controllability of a drone. The broadband communication module onboard the drone (e.g., a 5GCPE module supporting the 3GPP standard or a Mesh self-organizing network radio with a proprietary protocol) continuously monitors the wireless link status at the RF physical layer. The communication baseband processor converts the received time-domain signal into a frequency-domain signal using a Fast Fourier Transform (FFT) and calculates the ratio of the received reference signal power to the in-band interference noise power spectral density in real time. To avoid drastic fluctuations in the instantaneous signal-to-noise ratio due to multipath effects caused by urban building obstruction, this embodiment introduces a 500-millisecond moving average filter. The processor performs weighted smoothing on all sampled values ​​within this time window, outputting a stable signal-to-noise ratio. The value is uniformly expressed in decibels (dB). This physical quantity characterizes the stability of the link between the UAV receiving ground control commands and transmitting mission payload data, and is the basis for various criteria to prevent the UAV from losing contact during mission execution.

[0098] Current environmental perception confidence level ( The parameter introduced in this invention is used to characterize environmental adaptability risk. The front-end perception system of an unmanned aerial vehicle (UAV) typically includes binocular vision sensors, LiDAR, or optical flow modules for real-time localization and mapping. However, in areas with strong direct sunlight, dense fog, low illumination at night, or weak textures (such as calm water surfaces or glass curtain walls), the feature extraction capability of the perception system significantly decreases. To quantify this risk, the onboard vision processor outputs real-time state diagnostic data while running visual odometry. The system extracts the number of effective feature points in the current frame image (such as the number of FAST corner points), the matching inlier rate of feature points between consecutive frames, and the reprojection error during the back-end optimization process. This embodiment constructs a nonlinear mapping function to comprehensively map the above indicators to probability values ​​ranging from 0.00 to 1.00. .when When the value approaches 1, it indicates that the current environment has clear texture, the drone's positioning is accurate, and the risk of collision is extremely low; when... When the value is below 0.4, it indicates that the drone is in a semi-blinded state, and the risk of collision during missions increases exponentially.

[0099] After acquiring the real-time data verified by the physical layer, the core scheduling engine of the task allocation system (which can run on ground edge computing nodes or cluster leaders) will call the calculation formula to quantitatively score the suitability of each UAV for a specific task to be assigned.

[0100] The formula is detailed below:

[0101] ;

[0102] The following explains the meaning, function, and technical effect of each physical term in the formula:

[0103] In the formula This project introduces a soft gating mechanism based on the Sigmoid function. Specifically, This is the energy weighting coefficient (dimensionless constant, preferably in the range of 0.5-0.8), dynamically set according to the mission type. For long-endurance logistics missions, this weight is higher. Denominator This is one of the core designs of this invention. It is a preset communication quality benchmark threshold (unit: dB), representing the minimum physical threshold required to maintain a reliable data link (e.g., set to 15 dB). This is a signal sensitivity adjustment coefficient (a constant, preferably in the range of 0.5-1.0), used to control the steepness of the gating.

[0104] In actual low-altitude operations, the impact of communication quality on mission capability is not linear, but rather exhibits a waterfall effect. That is, when the signal-to-noise ratio (SNR) is above a certain threshold, communication is normal, and mission capability mainly depends on battery power; however, once the SNR falls below the threshold, the packet loss rate rises sharply until communication is lost. At this point, regardless of the battery level, the drone cannot be controlled to perform its mission. This formula effectively simulates this physical process through the construction of an exponential function:

[0105] when At that time, the exponential term When the denominator approaches 0, the value of the term approaches 1, and at this point, the value of the term is approximately equal to... It exhibits linear characteristics dominated by electrical quantity.

[0106] when Close to or below At this point, the exponential term increases rapidly, causing the denominator to rise exponentially. This causes the entire fraction to plummet and approach 0 within a very short signal decay range.

[0107] This design can automatically eliminate risky nodes that have sufficient power but are on the verge of communication interruption, effectively preventing the possibility of assigning critical tasks to drones that are about to lose contact in signal blind spots.

[0108] In the formula This item is used to quantify the potential risk costs of performing the task, and is considered a deduction item. The risk penalty weighting coefficient (a dimensionless constant) reflects the system's emphasis on security. The item directly quantifies the uncertainty of perception, i.e. the degree of blind flight. The item introduces a time-based urgency constraint. It is the estimated time that the path planning algorithm estimates in real time based on the current mission range. It is the maximum battery life predicted by the BMS based on the current battery level.

[0109] This illustrates the physical law that risk accumulates and amplifies over time. When near At this point, the ratio approaches 1, meaning the drone must exhaust all its power to barely complete the mission, leaving no time margin for error. Under these extreme conditions, if environmental perception is also poor ( (If the product of the two is relatively large), it will produce a huge penalty value, significantly reducing or even making the value lower. It becomes a negative value.

[0110] In situations where fuel is critically low (time is of the essence), excellent weather conditions (clear perception) are required to continue flight; if the environment is harsh and fuel is insufficient, the mission must be abandoned immediately. Through this mathematical model, the system can automatically identify and avoid high-risk allocation schemes for extreme ranges in adverse environments, significantly improving the safety of swarm operations.

[0111] Throughout the entire lifecycle of task execution, the above The calculation is not performed all at once, but rather refreshed in real time at a frequency of 1Hz to 5Hz. The system has a preset stripping threshold (e.g., 0.35).

[0112] When a drone performing a mission is affected by a sudden change in the environment (such as suddenly flying into the shadow of a tall building), Sudden drop, or sudden exposure to dense smoke (deterioration), leading to its If the data is below the stripping threshold for N consecutive cycles (e.g., 3 seconds), the system determines that the machine can no longer safely maintain the current task.

[0113] At this point, the task stripping and reallocation mechanism is automatically triggered:

[0114] The airborne system immediately freezes the current mission process, records the mission breakpoint (such as the boundary coordinates of the scanned area and the remaining quantity of undelivered materials), and broadcasts a mission capability downgrade request to the cluster network.

[0115] After receiving the request, the task scheduling center immediately iterates through other idle or low-priority drones in the cluster, substitutes their real-time status data, and calculates the task for that breakpoint. .

[0116] System selection The candidate drone with the highest value that exceeds the succession threshold (e.g., 0.6) is sent a task relay instruction packet.

[0117] Step 4: The original drone executes a downgraded risk avoidance strategy (such as increasing altitude to search for signals or returning to base), and the new drone flies to the breakpoint to seamlessly take over the mission.

[0118] Through this closed-loop control driven by a physical model, the present invention ensures that each task is always carried out by the UAV with the optimal physical state at the time, achieving an effective balance between swarm efficiency and safety in low-altitude economic scenarios.

[0119] Example 5: Distributed Cooperative Obstacle Avoidance Control System Based on Semantic Fragments in a Weak Communication Environment

[0120] In scenarios where hundreds of drones are operating simultaneously or in remote mountainous areas or post-disaster emergency situations where there is a lack of public network coverage, the traditional collaborative mode based on sharing raw image streams or high-density point clouds will instantly occupy limited spectrum resources, leading to network congestion and high latency, which in turn will cause cluster collisions.

[0121] Therefore, in this embodiment, during the execution phase of the collaborative control strategy, the UAV's onboard communication monitoring module senses the network status in real time. When the communication network bandwidth is detected to be lower than a preset safety value (e.g., 2Mbps), the system automatically switches to semantic collaborative mode. In this mode, the raw data collected by the UAV's front-end perception system (binocular camera, LiDAR) is no longer directly transmitted, but is instead input to the onboard high-performance AI edge computing unit (such as NPU or FPGA).

[0122] Edge computing units run lightweight object detection and semantic segmentation algorithms (such as YOLOv8-Nano or MobileNet-SSD) to perform real-time semantic abstraction of the environment. The core task of the algorithm is to extract obstacle entities that have a substantial impact on flight safety and transform them into structured, lightweight semantic fragment data packets. These packets undergo bit-level compression and contain only the following core fields:

[0123] Obstacle category identifiers: These are encoded using 8-bit unsigned integers. For example, 0x01 represents static buildings, 0x02 represents dynamic drones, 0x03 represents high-voltage power line towers, and 0x04 represents trees. These IDs strictly correspond to the ontology knowledge base pre-installed in each drone.

[0124] (2) Three-dimensional centroid coordinates of obstacles: including floating point coordinates of the X, Y, and Z axes relative to a unified geographic coordinate system (or cluster relative coordinate system).

[0125] (3) Obstacle boundary vector information: This is crucial for describing the spatial occupancy of obstacles. This embodiment does not transmit complex surface meshes, but instead transmits the length, width, and height vectors describing the minimum bounding box of the obstacle, as well as quaternions describing its attitude. For dynamic obstacles, its velocity vector is also included.

[0126] (4) Timestamp and confidence level: used by the receiving end for time synchronization and data fusion weighting.

[0127] The lightweight semantic fragment data packets employ a low-bandwidth encoding method, enabling transmission in low signal-to-noise ratio or narrowband environments. Compared to transmitting a single image frame (hundreds of KB), a semantic fragment data packet is only a few tens of bytes in size. This data compression allows drone swarms to maintain high-frequency status broadcasts even on extremely low-speed links such as LoRa, ZigBee, or narrowband radios.

[0128] After a neighboring drone (the receiver) receives the broadcast semantic fragment data packets via the communication network, it initiates the holographic reconstruction process. This process relies on an ontology knowledge base pre-stored in the drone's non-volatile memory. This base stores detailed 3D mesh models of various standard obstacles and their physical properties (such as hardness, reflectivity, and danger radius).

[0129] Specifically, the refactoring process is as follows:

[0130] The receiving end parses the obstacle category identifier code (such as 0x03) in the data packet and retrieves the standard three-dimensional model of the high-voltage tower from the ontology knowledge base.

[0131] Based on the three-dimensional centroid coordinates in the data packet, the standard model is placed at the corresponding location in the local virtual space maintained in the receiver's memory.

[0132] By using the boundary vector information in the data packet, the standard model is scaled, rotated, and stretched so that its geometry in virtual space is highly consistent with the real object detected by the sending end.

[0133] Through this mechanism, although the receiving drone does not receive any image pixels or actually see any obstacles, the scene seen by the sending end is reproduced in the virtual world of its logical computing core. As different semantic fragments are continuously incorporated, each drone can construct a dynamically updated three-dimensional holographic mapping model covering a radius of several kilometers.

[0134] Based on the reconstructed 3D holographic mapping model described above, this embodiment employs an improved Artificial Potential Field (APF) method to generate cooperative control commands. In traditional APF, the repulsive force field is often calculated based on the sensor's original ranging data, making it susceptible to noise interference. In this embodiment, however, the force field is calculated based on the reconstructed, clean virtual entity.

[0135] In the airborne flight control computer, the system sets each virtual obstacle in the holographic model as a high potential energy point to generate a repulsive field; and sets the mission target point as a low potential energy point to generate a gravitational field.

[0136] For each reconstructed virtual obstacle The repulsive potential function it generates Defined as:

[0137] ;

[0138] in, Current location of the drone Nearest point to virtual obstacle surface The Euclidean distance; The radius of influence of the obstacle (e.g., 30 meters); This is the repulsive force gain coefficient; The relative speed between the drone and the obstacle. This is a regulating factor. When a drone approaches an obstacle at high speed, the repulsive field dynamically strengthens. Specifically, This is the repulsive force gain coefficient, and its value is a constant greater than 0. This is a positive real number (e.g., 2), used to adjust the nonlinear effect of relative velocity on the repulsive field strength. It should be noted that the above repulsive potential function... Only Effective at time; when When this occurs, it is considered to be outside the area of ​​influence of the obstacle. .

[0139] To obtain the force vector required for control, the system uses the negative gradient method to transform the potential field function into a control force.

[0140] First, define the gravitational vector generated by the target point. Let the gravitational potential function be... ,in This is the gravitational gain coefficient (a constant greater than 0). Here are the three-dimensional coordinates of the mission target point. The gravitational vector is obtained by taking the negative gradient of the gravitational potential function.

[0141] ;

[0142] Secondly, regarding the first A virtual obstacle, which generates a repulsive force vector. The above repulsive potential function The negative gradient, i.e. .

[0143] The flight control computer calculates in real time the vector sum of all virtual repulsive forces (from the reconstructed environment) and gravitational forces (from the target point) acting on the UAV at its current position:

[0144] ;

[0145] The resultant force vector The values ​​are directly mapped to acceleration or angular velocity control commands, which are then input to the underlying attitude controller (PID controller) to drive the motors and adjust the flight attitude. In the formula, This represents the virtual resultant force vector acting on the drone; This represents the total number of virtual obstacles reconstructed by the drone from semantic fragments at the current moment. The index number of the virtual obstacle; The gravitational vector generated for the target point; For the first A repulsive force vector generated by a virtual obstacle.

[0146] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for task allocation and collaborative control of unmanned aerial vehicle (UAV) swarms for the low-altitude economy, characterized in that, Includes the following steps: S1. Based on the requirements of low-altitude economic missions, establish a mission model, including mission type, mission area, mission time, mission priority, and mission objectives; perform visual inspection on the mission area to obtain environmental information and target object information within the mission area; S2. Obtain the status information of the drone cluster, including but not limited to the drone's battery level, payload, flight status, flight capability, communication status, and health status. Based on the task model and the drone cluster's status information, use a multi-objective optimization algorithm to allocate tasks and assign tasks to drones suitable for performing the task. S3. Establish a ground calibration system to calibrate the UAV's navigation system; use satellite antennas to receive satellite signals to provide positioning information for the UAV; establish a communication network, including communication links between UAVs and communication links between UAVs and the ground control center, to realize information sharing and transmission of collaborative control commands between UAVs; S4. Based on the task allocation results and the real-time status information of the UAV, formulate a collaborative control strategy. The collaborative control strategy includes, but is not limited to, flight path planning, task execution sequence, communication coordination, and conflict avoidance.

2. The method for task allocation and collaborative control of UAV swarms for low-altitude economy as described in claim 1, characterized in that, It also includes the following steps: S5. Multiple sensors are mounted on the drone to collect real-time operating data, including but not limited to motor speed, battery voltage, flight attitude, temperature and vibration data. A fault prediction model is established using a convolutional neural network to extract and learn features from operational data, and to predict the types and probabilities of possible faults in the UAV. Establish a health management mechanism to perform timely maintenance and upkeep of drones based on fault prediction results; When the fault prediction model predicts that the UAV has a fault risk, corresponding maintenance measures are taken according to the fault type and risk level. Establish health records for drones, recording their operational data and maintenance history.

3. The method for task allocation and collaborative control of UAV swarms for low-altitude economy according to claim 2, characterized in that, In S1, a mission model is established based on the requirements of low-altitude economic missions, including the following steps: Based on the task model, an ontology knowledge base for low-altitude economic tasks is established, the task content is converted into multi-dimensional semantic entities, and detailed attributes and relationships are defined for each semantic entity. Identify the sequential constraints between tasks from multi-dimensional semantic entities; By analyzing the geographical coordinate range of entities in the task area, overlapping areas are identified; and spatiotemporal buffer zones are dynamically divided according to task priority and task type to avoid conflicts and interference between drones. Identify task groups that require similar equipment or drones with the same skills, as well as collaborative task packages that can be combined for execution; Based on the recognition results, multiple task templates are preset for multi-dimensional semantic entities.

4. The method for task allocation and collaborative control of UAV swarms for low-altitude economy according to claim 3, characterized in that, In S2, tasks are allocated based on the task model and the state information of the UAV cluster using a multi-objective optimization algorithm, including the following steps: When a task request is received to identify and encode a specific object, the complete digital profile of the specific object is obtained, and the physical characteristics, operation interface and historical service records of the specific object are parsed. Based on the parsing results, a preset task template is matched, and personalized task parameters are generated; A multi-objective optimization algorithm is used to allocate tasks. Based on task priority, high-priority tasks are assigned to drones with higher task execution efficiency, thereby reducing task completion time. Based on the drone's payload capacity, equipment type, and skill level, medium-priority and low-priority tasks are assigned to appropriate drones. The system monitors task execution and drone status in real time, regularly sends task progress reports and its own status information to the system, and dynamically adjusts the task allocation plan based on real-time information. If an anomaly occurs during task execution, the system will automatically activate the anomaly handling mechanism to reassign tasks to drones that are currently not on a task or adjust the task path. The task priority is dynamically adjusted based on real-time feedback during task execution.

5. The method for task allocation and collaborative control of UAV swarms for low-altitude economy according to claim 4, characterized in that, In S2, tasks are allocated using a multi-objective optimization algorithm based on the task model and the state information of the UAV cluster. This also includes the following steps: Based on the real-time feedback data, identify problems and potential optimization points that arise during task execution; Based on the identified problems and optimization points, the parameters and processes of the task template are automatically adjusted; By continuously collecting and analyzing task execution data, we dynamically optimize task templates to ensure that they can adapt to different task scenarios and environmental changes. During the task allocation process, priority is given to assigning tasks to drones in good health, and backup task plans are reserved for drones that may malfunction.

6. The method for task allocation and collaborative control of UAV swarms for low-altitude economy according to claim 5, characterized in that, In S2, tasks are allocated using a multi-objective optimization algorithm based on the task model and the state information of the UAV cluster. This also includes the following steps: The health status of drones is predicted in real time based on a fault prediction model. If a fault is predicted in a drone, an early warning will be issued. During the task allocation process, priority is given to assigning tasks to drones in good health; for drones that may malfunction, backup task plans will be generated in advance. The health status of the drone is updated in real time to the task allocation system, ensuring that task allocation decisions are always based on the latest health status information.

7. The method for task allocation and cooperative control of UAV swarms for low-altitude economy according to claim 6, characterized in that, In S4, a collaborative control strategy is formulated based on the task allocation results and the real-time status information of the UAV, including the following steps: A multi-agent cooperative communication mechanism is introduced, defining a communication agent for each UAV, enabling real-time information sharing and collaborative decision-making among UAVs through a communication network; When any drone detects a new obstacle, it transmits the real-time obstacle information to neighboring drones. The neighboring drones then dynamically adjust their flight paths and mission execution strategies based on this information to avoid collisions with obstacles, thus achieving intelligent collaborative control of the swarm. When multiple drones are conducting patrol missions on the same route or in the same area, the flight paths and mission execution order of the drones are dynamically adjusted, and the drones share the patrol results in real time through the communication network to ensure the efficient execution of the mission. Based on the mission priority and the UAV's communication status, the parameters of the communication link are dynamically adjusted to ensure the communication needs of high-priority missions; when a communication link interruption or weak signal is detected, the system will automatically switch to a backup communication mode or re-establish the communication link.

8. The method for task allocation and collaborative control of UAV swarms for low-altitude economy according to claim 7, characterized in that, S1 also includes the following steps: The target object is identified and encoded by converting its feature information into a unique identifier and storing it in the database. Based on the task type and task area characteristics, the task is decomposed into multiple subtasks according to preset rules; each subtask has an independent task model, including subtask type, subtask area, subtask time, subtask priority and subtask objective.

9. The method for task allocation and cooperative control of UAV swarms for low-altitude economy according to claim 5, characterized in that, In step S2, the allocation of tasks based on the task model and the state information of the UAV cluster using a multi-objective optimization algorithm specifically includes the following steps: The real-time flight data of the UAV during the execution of the mission is acquired, and the real-time flight data includes at least: the current remaining battery percentage, the current communication signal-to-noise ratio, and the current environmental perception confidence level. Based on the real-time flight data, the task execution matching index of the UAV for the task is calculated using a preset calculation formula; the calculation formula is: ； In the formula, The matching degree index is used to determine the task execution. The current remaining battery percentage; The signal-to-noise ratio of the current communication signal; This is a preset communication quality benchmark threshold; The current environment perception confidence level is used to characterize the probability of the UAV's visual sensor accurately extracting environmental features; The estimated remaining time required to perform the task described above; This represents the maximum remaining flight time of the drone. is the base of the natural logarithm; is the signal sensitivity adjustment coefficient, which is a constant greater than 0; Energy weighting coefficient, This is the risk penalty weighting coefficient, and and All are dimensionless constants; The task allocation scheme is dynamically weighted and adjusted based on the calculated task execution matching index. When the task execution matching index is lower than a preset threshold, the currently allocated task is automatically removed and a reallocation mechanism is triggered.

10. The method for task allocation and cooperative control of UAV swarms for low-altitude economy according to claim 7, characterized in that, In step S4, the step of formulating a collaborative control strategy based on the task allocation results and the real-time status information of the UAV further includes the following steps: A distributed semantic fragment reconstruction mechanism is established, which is triggered when the bandwidth of the communication network between the drones is lower than a preset security value. The UAV extracts features from the collected environmental image data and transforms the extracted key features into lightweight semantic fragment data packets. The lightweight semantic fragment data packets contain only obstacle category identification codes, obstacle three-dimensional coordinate center points, and obstacle boundary vector information. The drone broadcasts the lightweight semantic fragment data packet through the communication network; The adjacent UAV receives the lightweight semantic fragment data packet and, in conjunction with the locally pre-stored ontology knowledge base, reconstructs a three-dimensional holographic mapping model of the surrounding environment in the local virtual space. Based on the reconstructed 3D holographic mapping model, the adjacent UAVs use the artificial potential field method to calculate the obstacle avoidance resultant force, achieving collaborative obstacle avoidance and dynamic path correction without transmitting the original image data.

Images