A method and system for dynamic allocation of heterogeneous multi-UAV firefighting tasks

Abstract

This invention relates to the field of unmanned aerial vehicle (UAV) technology, and particularly to a method and system for dynamic allocation of heterogeneous multi-UAV firefighting tasks. The system acquires real-time wind direction data over the fire site through a wind field detection unit, driving a formation of detection UAVs to perform dynamic position rotation based on wind direction. After completing its current detection task, a detection UAV on the upwind side automatically moves along the wind direction to take over the position of the original leeward detection UAV. The original leeward detection UAV then moves against the wind to replenish its upwind position, forming a continuous cyclical rotation. This mechanism ensures that the outermost edge of the leeward direction is always under the cross-coverage of multiple detection UAVs, ensuring continuous and accurate monitoring of the fire line in the direction of fastest fire spread. Furthermore, it avoids the loss of sensor accuracy or platform structural damage to a single detection UAV due to prolonged exposure to high-temperature smoke and gas in the leeward direction, significantly improving the reliability and endurance of the formation in continuous iterative firefighting tasks.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method and system for dynamically allocating heterogeneous multi-UAV firefighting tasks. Background Technology

[0002] With the increasing frequency of fires in complex environments such as high-rise buildings, petrochemical industrial parks, and forests and grasslands, drones, with their advantages of maneuverability, wide field of vision, and ability to penetrate dangerous areas, have become an important aerial force for firefighting and rescue. Heterogeneous multi-drone systems, by carrying different types of sensors and firefighting payloads, can achieve coordinated operations of fire scene reconnaissance, situational awareness, and fire suppression deployment. However, existing methods for allocating and scheduling drone firefighting tasks still face the following prominent technical bottlenecks in practical applications: In a fire environment, the intense heat radiation and high-temperature smoke generated by flames spread downwind with the airflow. Current task allocation methods typically deploy detection drones in fixed positions, neglecting the influence of wind direction on the fire's spread and the damage to the drone platforms caused by persistent high-temperature smoke. On the one hand, the fire line on the downwind side expands fastest and has the highest risk of reignition, making it a priority monitoring target; however, due to interference from hot smoke, the accuracy of detection drone sensors in this area will significantly decrease, and the drones may even be damaged due to overheating. Current technology cannot dynamically rotate the position of detection drones based on real-time wind direction, resulting in the outermost downwind fire line being under constant monitoring weakness, leading to a lack of critical situational data and severely impacting the quality of firefighting decisions.

[0003] Existing methods mostly employ a static allocation strategy of "one-time planning, batch execution," which assigns firefighting tasks to a number of drones at once based on the fire outline at the initial moment of the mission. However, fires evolve dynamically; after the outer flames are extinguished, the fire radius continues to shrink, and the location of new fire lines constantly changes. Existing systems lack an iterative closed-loop mechanism of "extinguishing-relocation-re-extinguishing," failing to progressively peel away the outermost area layer by layer according to real-time changes in the fire. This results in some drones being unable to extinguish fires that were previously unavailable, while the new outermost area lacks sufficient fire coverage, leading to low firefighting efficiency.

[0004] Even after the main open flames are extinguished, smoldering and latent high temperatures may still exist on the surface of the burned area, making it highly susceptible to reignition under wind conditions. Existing methods typically rely on manual inspections using thermal imagers, but given the large burned area and complex terrain, manual inspections are inefficient and pose safety risks. Even with some solutions deploying infrared pods for scanning, they only reach the level of detection and alarm, lacking a closed-loop response capability from autonomous location to immediate extinguishing to re-testing and confirmation. This results in the long-term presence of lingering fire hazards, severely impacting the final effectiveness of firefighting operations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for dynamic allocation of heterogeneous multi-UAV firefighting tasks, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A heterogeneous multi-UAV firefighting task dynamic allocation system includes the following modules: The multi-source sensing module is used to extract the outline of the entire fire area in real time and dynamically capture the local airspace wind field; The inter-machine communication and mutual sensing module is used to build a fully connected data link between the UAV cluster and the ground station, and to realize the relative spatial positioning between any two UAVs in the detection formation. The dynamic allocation decision module is used to fuse and calculate the fire field contour and wind field data output by the multi-source perception module, decompose the fire fighting requirements into an iterative task queue and select the most suitable execution platform from the heterogeneous drone cluster, as well as resolve the conflict of overlapping vision and spatial boundary crossing in multi-drone collaborative operation. The task execution module is used to receive task instructions issued by the dynamic allocation decision module and execute corresponding scanning, fire extinguishing and monitoring operations; In one possible implementation, the multi-source sensing module includes: The ring imaging unit is used to achieve real-time imaging of the fire burning area with a 360-degree panoramic view and to extract the vectorized edge of the fire line. The wind field detection unit is used to inject wind field physical quantities into the fire spread prediction engine in real time, enabling the system to have the ability to predict the acceleration of the fire head advancing downwind, and at the same time provide data support for the dynamic position rotation of the detection drone formation based on wind direction.

[0007] In one possible implementation, the inter-machine communication and mutual sensing module includes: The communication unit is used to ensure low-latency and reliable transmission of flight control commands, situational awareness data, and task allocation commands between the UAV swarm and the ground station.

[0008] The mutual sensing unit is used to continuously output the relative distance and relative orientation between any two drones in a formation without relying on GPS satellite signals.

[0009] In one possible implementation, the dynamic allocation decision module includes: The fire prediction unit is used to drive the system to peel away the outermost edge of the fire layer by layer in an iterative shrinking manner. Task splitting units are used to transform the macro fire situation into a queue of single-point clearing tasks to be executed iteratively; The planning unit injects the virtual safety boundary as a hard constraint for navigation into the flight control system of the firefighting drone; at the same time, it drives the detection drone formation to perform dynamic position rotation based on real-time wind direction data to ensure that the outermost area downwind is always under optimal detection coverage.

[0010] In one possible implementation, the task execution module includes: The observation drone performs a 360-degree panoramic scan of the fire site at the beginning of the mission to quickly locate the outermost protruding area with the largest current radius. After each iteration of extinguishing the fire, it rescans to update the fire site outline and locate the new outermost protruding area, while sending a relocation signal to the detection drone formation. After receiving the repositioning signal from the observation drone formation, the detection drones focus on performing detailed detection on the outermost prominent area. Firefighting drone formations are used to carry out high-intensity, concentrated firefighting operations on the outermost, most prominent area with the largest current radius. The observation drone formation is used to autonomously extinguish any abnormally high temperatures detected.

[0011] One possible implementation includes the following steps: Step 1: Observe the UAV's global scan and outermost radius positioning; Step 2: Dynamic rotation mechanism of multi-detection drone swarm deployment and wind field-driven operation; Step 3: Firefighting drones concentrate on extinguishing the outermost and largest radius area. Step 4: Iterative relocation and layer-by-layer stripping loop; Step 5: Observe the temperature difference detection and autonomous fire repair of the entire burned area using the drone.

[0012] In one possible implementation, in step two, the wind field detection units of the six probe drones continuously report the instantaneous wind direction and wind speed at their respective locations, assuming the current wind direction vector is... The detection drone pointing northwest, located on the windward side (southeast), has a clear field of view and is unaffected by hot smoke. After completing its current detection task, the system automatically instructs it to move along the wind direction to replace the detection drone originally located on the leeward side (northwest). The original detection drone on the leeward side (northwest) moves against the wind to fill the vacated detection position on the windward side. This rotation forms a closed loop, and its position update rules are as follows: (Upwind drone) (Downwind drone) Where Δt is the rotation time interval. The current average wind speed, Using the wind direction as a unit vector, through the above rotation, the outermost edge of the downwind side is always under the cross coverage of multiple detection drones. At the same time, it avoids the sensor accuracy of a single detection drone from decreasing or the platform being damaged due to continuous exposure to high-temperature flue gas environment. When the wind field direction changes, the system automatically adjusts the rotation direction to always ensure that the downwind detection array is taken over by a healthy drone from the upwind side.

[0013] In one possible implementation, in step four, the iterative process is repeated. After each round of suppression, the outermost radius of the fire is compressed inward by one layer. The fire spread prediction unit calculates the maximum radial distance of the current fire outline after each iteration. :

[0014] Where n is the iteration round number, after two consecutive iterations The decrease approaches zero, that is:

[0015] in, When the system determines that the fire spread has been completely contained and the outermost contour of the fire area is no longer measurably expanding outward as confirmed by the panoramic scan of the observation drone, the outer layer stripping stage ends and the system moves into the remaining fire internal cleanup stage.

[0016] In one possible implementation, in step five, observe the drone activating its mounted cooled mid-wave infrared thermal imager and read the temperature values ​​of each pixel in each frame of the thermal image below. The current ambient background temperature, corrected by the airborne barometric altimeter. Perform dynamic differential calculations to obtain temperature difference values. ,when Right now It was identified as a potential hazard point due to abnormally high temperatures.

[0017] In one possible implementation, in step five, when an observation drone detects an abnormally high temperature point, because the drone itself carries small dry powder fire extinguishing bombs, the system immediately activates the drone's onboard visual servo locking module. This module uses a monocular visual ranging algorithm to calculate the three-dimensional deviation of the abnormal point relative to the drone's coordinate system. When the locking deviation satisfies: At that moment, the flight control system automatically executed the hovering and diving launch command, dropping a small dry powder fire extinguishing bomb at the abnormal point. The observation drone immediately used the infrared thermal imager to re-measure the coordinate point, confirming that the temperature had dropped below the 56-degree Celsius warning line, and completed the autonomous closed-loop handling of the potential hazard point.

[0018] Beneficial effects compared to existing technologies: 1. In this solution, the system acquires real-time wind direction data over the fire area through a wind field detection unit, driving the detection drone formation to perform dynamic position rotation based on wind direction. After completing its current detection task, the detection drone on the upwind side automatically moves along the wind direction to take over the position of the original leeward detection drone, while the original leeward detection drone moves along the upwind direction to supplement the upwind position, forming a continuous cyclical rotation. This mechanism ensures that the outermost edge of the leeward direction is always under the cross coverage of multiple detection drones, ensuring that the fire line in the direction of fastest fire spread is continuously and accurately monitored; on the other hand, it avoids the sensor accuracy degradation or platform structure damage caused by a single detection drone being in the high-temperature smoke flow downwind for a long time, significantly improving the reliability and endurance of the formation in continuous iterative firefighting missions. 2. In this scheme, the system uses observation drones to perform a 360-degree panoramic scan, selecting the outermost protruding area with the largest radial distance from the current fire as the highest priority target. Subsequently, a formation of detection drones is dispatched to conduct detailed inspections of this area and transmit fire line characteristics back to the decision module. Based on this, a formation of firefighting drones concentrates on extinguishing the target area. After one round of extinguishing, the observation drones immediately rescan the fire area, locate the new outermost protruding area, and initiate the next iteration. This layer-by-layer stripping strategy continuously compresses the fire boundary inward. When the reduction in the maximum radial distance of the fire area in two consecutive rounds is less than the convergence threshold, the fire spread is considered completely contained. After the fire is controlled, the observation drones use a cooled mid-wave infrared thermal imager to perform a gridded scan and temperature measurement of all burned areas. Abnormally high-temperature points with a temperature difference ≥30℃ are automatically locked, and small fire extinguishing bombs are deployed for autonomous fire suppression. After confirming the temperature has fallen back, the loop is closed. The entire process requires no manual intervention, achieving fully autonomous closed-loop firefighting from the outer perimeter of the fire to the complete elimination of remaining fires, significantly improving firefighting efficiency and safety under complex fire conditions. Attached Figure Description

[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the system framework of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0021] Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention can also be implemented in various different forms, and therefore the present invention is not limited to the embodiments described below. The technical solution in this application embodiment is to solve the problems mentioned in the background art, and the overall idea is as follows: Example: Please refer to Figures 1 to 2 As shown in the figure, this embodiment introduces a method and system for dynamic allocation of heterogeneous multi-UAV firefighting tasks. The system and method are used together to achieve the gradual shrinking of the fire field and the autonomous closed-loop disposal of remaining fires by driving the detection UAVs to rotate dynamically according to wind direction and the firefighting UAVs to extinguish the fire in layers.

[0022] A heterogeneous multi-UAV firefighting task dynamic allocation system includes a multi-source sensing module, an inter-UAV communication and mutual sensing module, a dynamic allocation decision module, and a task execution module; The multi-source sensing module is used to extract the overall contour of the fire area in real time and dynamically capture the local airspace wind field; this module includes a ring imaging unit and a wind field detection unit.

[0023] The ring-shaped imaging unit, deployed on each of the reconnaissance drones in the formation, consists of six infrared and visible light dual-light cameras evenly distributed around the horizontal circumference of the drone's belly. Adjacent cameras are spaced 60 degrees apart, and each camera has a horizontal field of view of 60 degrees. After simultaneous exposure of all six cameras, the onboard image processor performs panoramic stitching and edge detection, outputting a time-stamped set of fire scene contour feature points. This unit is used to achieve real-time, 360-degree panoramic imaging of the fire burning area without blind spots and to extract vectorized fire line edges.

[0024] The wind field detection unit, deployed on the top of the back of each reconnaissance and observation drone, incorporates a three-dimensional ultrasonic anemometer. In this embodiment, the sampling frequency is 100 Hz, the measurement resolution is 0.1 m / s, and it outputs the instantaneous wind speed scalar and wind direction azimuth of the airspace. This unit is used to inject wind field physical quantities into the fire spread prediction engine in real time, enabling the system to predict the acceleration of the downwind fire front and providing data support for the dynamic position rotation of the reconnaissance drone formation based on wind direction.

[0025] The inter-drone communication and mutual sensing module is used to establish a fully connected data link between the UAV swarm and the ground station, and to achieve relative spatial positioning between any two UAVs within the swarm. This module includes a communication unit and a mutual sensing unit.

[0026] The communication unit, deployed across all UAV platforms, employs a decentralized mesh network topology. Its communication protocol is based on an improved version of the IEEE 802.11s standard and supports frequency hopping for interference mitigation. It is divided into three independent channels based on transmission priority: a real-time control layer, a situational awareness layer, and a mission command layer. This unit ensures low-latency and reliable transmission of flight control commands, situational awareness data, and mission allocation commands between the UAV swarm and the ground station.

[0027] The mutual sensing unit is deployed at the end of the arm of each reconnaissance drone. Each drone is symmetrically equipped with four ultra-wideband positioning and communication integrated nodes. In this embodiment, the operating frequency band is 6.5 GHz, the ranging accuracy is ±0.1 meter, and the data refresh rate is 200 Hz. It calculates the relative three-dimensional spatial coordinates of any two reconnaissance drones in the formation in real time using a time difference of arrival algorithm. This unit is used to continuously output the relative distance and relative azimuth between any two drones in the reconnaissance drone formation without relying on GPS satellite signals.

[0028] The dynamic allocation decision module is used to fuse and calculate the fire field contour and wind field data output by the multi-source sensing module. It decomposes firefighting requirements into an iterative task queue and selects the most suitable execution platform from the heterogeneous UAV cluster. It also resolves overlapping field-of-view conflicts and spatial boundary violations in multi-UAV collaborative operations. This module includes a fire spread prediction unit, a task decomposition and priority calibration unit, and a conflict resolution and trajectory planning unit.

[0029] The fire prediction unit, incorporating the Rothermel wildfire spread correction model, spatially superimposes and numerically couples the wind speed vector uploaded by the wind field detection unit with the fire line edge feature point set output by the ring imaging unit. It calculates the outward expansion velocity of each discrete point on the fire contour in real time and marks the outermost protruding area with the largest current radius. This unit drives the system to iteratively shrink and peel away the outermost edge of the fire layer by layer.

[0030] The task splitting unit receives the coordinates of the outermost prominent area from the fire spread prediction unit, marks this area as the highest priority task, and matches the required heterogeneous resource feature vectors for this task. This unit is used to transform the macro-fire situation into a queue of iteratively executed single-point clearing tasks.

[0031] The planning unit integrates the relative distance data of the detection drones output by the mutual induction unit with the image feature point cloud output by the ring imaging unit to calculate the field of view overlap coefficient, and generates normal escape trajectories for detection drones that are determined to be redundant and overlapping; the virtual safety boundary is injected into the flight control system of the fire-fighting drone as a navigation hard constraint; at the same time, the detection drone formation is driven to perform dynamic position rotation based on real-time wind direction data to ensure that the outermost area downwind is always under optimal detection coverage.

[0032] The task execution module receives task instructions from the dynamic allocation decision module and executes corresponding scanning, firefighting, or monitoring operations. This module includes observation drones, detection drone formations, firefighting drone formations, and observation drones.

[0033] The observation drone is equipped with the same ring imaging unit and wind field detection unit as the detection drone, and is additionally equipped with a full-domain scanning mode. It is used to perform a 360-degree panoramic scan of the fire site at the beginning of the mission to quickly locate the outermost protruding area with the largest current radius. After each iteration of extinguishing the fire, it rescans to update the fire site outline and locate the new outermost protruding area, while sending a repositioning signal to the detection drone formation.

[0034] The detection drone swarm, each equipped with a ring imaging unit, wind field detection unit, and mutual induction unit, is deployed in an equilateral quadrilateral formation over the fire area for continuous monitoring, with two drones serving as a dynamic mobile reserve. This swarm receives repositioning signals from the observation drones, focuses on performing detailed detection of the outermost prominent area, and transmits fire line characteristic data back to the dynamic allocation decision module. The detection drone swarm performs dynamic position rotation based on real-time wind direction: after completing its current detection task, the drone located upwind automatically moves downwind to take over the position of the original downwind drone, while the original downwind drone moves upwind to form a cyclical rotation. This ensures that the outermost leeward edge is always under the overlapping coverage of multiple detection drones, while preventing individual drones from being damaged by continuous exposure to high-temperature smoke and gas flow downwind.

[0035] Firefighting drone formations are used to carry out high-intensity concentrated fire suppression on the outermost prominent area with the largest current radius; and to supplement and clean up residual fire points after the main fire has been peeled away layer by layer.

[0036] The drone formation was observed, with each drone equipped with a cooled mid-wave infrared thermal imager and a small fire extinguishing bomb mount. The thermal imager has a thermal sensitivity better than 0.03 degrees Celsius. It is used to perform gridded scanning temperature measurement on the entire burned area after all visible flames have been peeled away layer by layer and the fire has stopped spreading outward, in order to identify abnormally high temperature points on the ground and potential smoldering hazards. It is used to implement immediate autonomous extinguishing of the detected abnormally high temperature points.

[0037] The following describes in detail a method for dynamically allocating heterogeneous multi-UAV firefighting tasks provided by this invention, based on the functional modules of the aforementioned system. This method corresponds to the specific processing steps executed by each module in the aforementioned system during runtime, and the specific steps are as follows: Step 1: Observe the UAV's global scan and outermost radius positioning After the mission was initiated, the ground station dispatched an observation drone to an altitude of 150 meters above the fire site via the mission command layer of the communication unit. The observation drone activated its ring imaging unit to perform a 360-degree panoramic scan. After simultaneous exposure by six cameras, the images were stitched together by the onboard image processor to create a panoramic image of the entire fire area. The image processing algorithm extracted the outermost edge contour of the burning area and calculated the radial distance of each discrete point on the contour relative to the geometric centroid of the fire. :

[0038] in, For the outline of the first The horizontal coordinates of a discrete point Let be the horizontal coordinates of the centroid of the fire's outline polygon. The algorithm iterates through all discrete points, selects the arc region corresponding to the largest radial distance, and marks this region as the outermost protruding area of ​​the current fire, i.e., the area with the largest radius and the highest threat of spread.

[0039] The observation drone simultaneously acquires current airspace wind direction data through the wind field detection unit. In this embodiment, the wind direction is northwest (i.e., the wind is blowing in a northwest direction). The observation drone transmits this wind field data and the coordinates of the outermost protruding area back to the fire spread prediction unit of the dynamic allocation decision module through the situational awareness layer of the communication unit. The fire spread prediction unit, combined with the wind direction data, determines that the outermost protruding area is currently located on the northwest side of the fire field, and this area is expanding outward at maximum speed driven by the wind, requiring priority handling.

[0040] Step 2: Dynamic rotation mechanism of multi-detection UAV swarm deployment and wind farm-driven operation After receiving the coordinates of the outermost prominent area transmitted back by the observation UAV, the task decomposition and priority calibration unit of the dynamic allocation decision module immediately dispatches six reconnaissance UAVs to the area to perform detailed detection. The six reconnaissance UAVs are deployed in a preset formation at different locations above the outermost prominent area, with six of them deployed over the fire site.

[0041] The circular imaging units of six reconnaissance drones simultaneously perform high-precision edge scanning of the area, extracting their respective fire line feature point sets. The data is then fed back to the dynamic allocation decision module. The conflict resolution and trajectory planning unit fuses and registers the six sets of feature points, and uses an iterative nearest-point algorithm to unify the multi-point cloud data into the same coordinate system, generating the accurate fire line profile of the current outermost protruding area.

[0042] During the detailed detection process, the mutual sensing unit continuously monitors the relative distance between the six detection drones. When the relative distance between any two aircraft is less than 25 meters and the overlap coefficient of their field of vision is... At that time, a normal escape command is issued to the lower-priority detection drones with later numbers, causing them to translate outwards until... .

[0043] The wind field detection units of the six reconnaissance drones continuously report the instantaneous wind direction and speed at their respective locations. In this embodiment, the wind direction is northwest, meaning the flames and hot smoke continue to spread to the northwest downwind area. Based on real-time wind direction data, the conflict resolution and trajectory planning unit of the dynamic allocation decision module implements a wind field-driven dynamic position rotation mechanism for the reconnaissance drone formation. Let the current wind direction vector be Pointing northwest. The detection drone located upwind (southeast) has a clear view and is unaffected by hot smoke due to its upwind position. After completing its current detection task, the system automatically instructs it to move northwestward to replace the drone originally located leeward (northwest). The drone originally leeward (northwest) moves upwind (southeast) to fill the vacated detection position on the upwind side. This rotation forms a closed loop, with the position update rules as follows: (Upwind drone) (Downwind drone) Where Δt is the rotation time interval, which is thirty seconds in this embodiment. The current average wind speed, This is a unit vector representing wind direction. Through the above rotation, the outermost edge of the leeward side is always under the overlapping coverage of multiple detection drones, while avoiding sensor accuracy degradation or platform damage to a single detection drone due to continuous exposure to high-temperature flue gas. When the wind direction changes, the system automatically adjusts the rotation direction to ensure that the leeward detection array is always taken over by a healthy drone from the upwind side.

[0044] Step 3: Firefighting drones concentrate on extinguishing the fire in the outermost, largest radius area. The fire spread prediction unit of the dynamic allocation decision module integrates the precise fire line feature point set transmitted back by six detection drones to calculate the coordinates of the optimal extinguishing agent delivery center point in the outermost protruding area and marks this point as the current highest priority target. The task splitting and priority marking unit selects drone models from the twelve firefighting drones in the task execution module that meet the following conditions: the shortest horizontal projection distance from the hovering position to the target center point. The objective function used for selection is the same as in Example S2.

[0045] After multiple constraint screenings, the system selected three heavy-duty firefighting drones to carry out this fire suppression mission. The conflict resolution and trajectory planning unit planned differentiated optimal flight paths for the three drones. After reaching a safe distance of 15 meters from the outer edge of the target area, the three drones formed a triangular formation and simultaneously launched fire extinguishing bombs to concentrate fire suppression efforts on the outermost, most prominent area with the largest current radius. After the fire extinguishing bombs hit, the radiant heat flux in that area dropped sharply from its peak to below 40 kilowatts per square meter, the fire was extinguished, and that area was no longer considered part of the outermost range of the fire.

[0046] Step 4: Iterative Relocation and Layer-by-Layer Stripping Loop After completing a concentrated fire suppression operation, the dynamic allocation decision module sends a rescan command to the observation drone via the communication unit. The observation drone then ascends again to a position 150 meters above the fire, performs a new 360-degree panoramic scan, and recalculates the radial distance of each discrete point on the current fire outline relative to the fire's centroid. The new arc segment region with the largest radial distance is selected, which is the new outermost protruding region.

[0047] The observation drone transmits the coordinates of the newly repositioned outermost prominent area back to the dynamic allocation decision module through the situational awareness layer. The system repeats step two, dispatching six reconnaissance drones to conduct detailed inspections of the new outermost area, and dynamically rotating the reconnaissance drone formation based on wind direction. Step three then dispatches three heavy-duty firefighting drones to concentrate on extinguishing the fire in the new outermost area.

[0048] The above iterative process is repeated, and with each round of suppression, the outermost radius of the fire is compressed inward by one layer. The fire spread prediction unit calculates the maximum radial distance of the current fire outline after each iteration. :

[0049] Where n is the iteration round number. After two consecutive iterations... The decrease approaches zero, that is:

[0050] in, In this embodiment, the preset convergence threshold is set to one meter. When the panoramic scan of the observation drone confirms that the outermost contour of the fire site no longer expands outward in a measurable manner, the system determines that the spread of the fire has been completely contained, the outer layer stripping stage ends, and the system moves into the residual fire internal cleaning stage.

[0051] In each iteration, the wind-direction-based dynamic rotation mechanism for detection drones remains in effect. Taking a northwest wind in this embodiment as an example, after the detection drone located on the upwind side (southeast) of the fire area completes its detection of the outermost region, the system instructs it to move northwest to replace the original detection drone in the northwest position. The original detection drone in the northwest position then moves southeast to fill the upwind position, forming a continuous cyclical rotation. This mechanism ensures that in each round of iterative fire suppression, the outermost edge on the leeward side is always under the accurate detection of the detection drone in optimal condition.

[0052] Step 5: Observe the entire burned area using drone temperature difference detection and autonomous fire suppression. Once the fire was completely contained, the dynamic allocation decision module recalled all detection and firefighting drones and dispatched three observation drones to hover 50 meters above the burned area. Each of the three observation drones was assigned a scanning responsibility zone and performed a line-by-line grid-based temperature measurement of the entire burned area.

[0053] Observe the drone activating its cooled mid-wave infrared thermal imager and read the temperature value of each pixel in each frame of the thermal image below. The current ambient background temperature, corrected by the airborne barometric altimeter. Perform dynamic differential calculations. Temperature difference values. ,when Right now It was identified as a potential hazard point due to abnormally high temperatures.

[0054] When an observation drone detects an abnormally high temperature point, the system immediately activates its onboard visual servo locking module, as the drone carries small dry powder fire extinguishing bombs. This module uses a monocular visual ranging algorithm to calculate the three-dimensional deviation of the abnormal point relative to the drone's coordinate system. The locking mechanism locks the point when the deviation meets the following conditions: At that moment, the flight control system automatically executed the hovering and diving launch command, dropping a small dry powder fire extinguishing bomb at the anomaly point. After the hit, the observation drone immediately used an infrared thermal imager to re-measure the coordinates of the point, confirming that the temperature had dropped below the 56-degree Celsius warning line, thus completing the autonomous closed-loop handling of the hazard point.

[0055] After three observation drones completed point-by-point inspection and fire suppression along the entire scanned responsibility zone, they performed a second panoramic infrared stitching scan of the entire burned area to obtain the final thermal distribution map of the entire fire site. If the thermal distribution map contains no pixels with temperature difference values... If all locations are below 56 degrees Celsius, the risk of reignition is considered completely eliminated, and the system sends a mission closed-loop message to the ground control station, allowing the entire fleet of drones to return autonomously.

[0056] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A heterogeneous multi-UAV firefighting task dynamic allocation system, characterized in that, Includes the following modules: The multi-source sensing module is used to extract the outline of the entire fire area in real time and dynamically capture the local airspace wind field; The inter-machine communication and mutual sensing module is used to build a fully connected data link between the UAV cluster and the ground station, and to realize the relative spatial positioning between any two UAVs in the detection formation. The dynamic allocation decision module is used to fuse and calculate the fire field contour and wind field data output by the multi-source perception module, decompose the fire fighting requirements into an iterative task queue and select the most suitable execution platform from the heterogeneous drone cluster, as well as resolve the conflict of overlapping vision and spatial boundary crossing in multi-drone collaborative operation. The task execution module is used to receive task instructions issued by the dynamic allocation decision module and execute the corresponding scanning, fire extinguishing and monitoring operations.

2. The heterogeneous multi-UAV firefighting task dynamic allocation system as described in claim 1, characterized in that, The multi-source sensing module includes: The ring imaging unit is used to achieve real-time imaging of the fire burning area with a 360-degree panoramic view and to extract the vectorized edge of the fire line. The wind field detection unit is used to inject wind field physical quantities into the fire spread prediction engine in real time, enabling the system to have the ability to predict the acceleration of the fire head advancing downwind, and at the same time provide data support for the dynamic position rotation of the detection drone formation based on wind direction.

3. The heterogeneous multi-UAV firefighting task dynamic allocation system as described in claim 1, characterized in that, The inter-machine communication and mutual sensing module includes: The communication unit is used to ensure low-latency and reliable transmission of flight control commands, situational awareness data, and task allocation commands between the UAV cluster and the ground station. The mutual sensing unit is used to continuously output the relative distance and relative orientation between any two drones in a formation without relying on GPS satellite signals.

4. The heterogeneous multi-UAV firefighting task dynamic allocation system as described in claim 1, characterized in that, The dynamic allocation decision module includes: The fire prediction unit is used to drive the system to peel away the outermost edge of the fire layer by layer in an iterative shrinking manner. Task splitting units are used to transform the macro fire situation into a queue of single-point clearing tasks to be executed iteratively; The planning unit injects the virtual safety boundary as a hard constraint for navigation into the flight control system of the firefighting drone; at the same time, it drives the detection drone formation to perform dynamic position rotation based on real-time wind direction data to ensure that the outermost area downwind is always under optimal detection coverage.

5. A heterogeneous multi-UAV firefighting task dynamic allocation system as described in claim 1, characterized in that, The task execution module includes: The observation drone performs a 360-degree panoramic scan of the fire site at the beginning of the mission to quickly locate the outermost protruding area with the largest current radius. After each iteration of fire suppression is completed, it rescans to update the fire site outline and locate the new outermost protruding area, while sending a relocation signal to the detection drone formation. After receiving the repositioning signal from the observation drone formation, the detection drones focus on performing detailed detection on the outermost prominent area. Firefighting drone formations are used to carry out high-intensity, concentrated firefighting operations on the outermost, most prominent area with the largest current radius. The observation drone formation is used to autonomously extinguish any abnormally high temperatures detected.

6. A method for dynamically allocating heterogeneous multi-UAV firefighting tasks using a heterogeneous multi-UAV firefighting task dynamic allocation system as described in any one of claims 1 to 5, characterized in that, Includes the following steps: Step 1: Observe the UAV's global scan and outermost radius positioning; Step 2: Dynamic rotation mechanism of multi-detection drone swarm deployment and wind field-driven operation; Step 3: Firefighting drones concentrate on extinguishing the outermost, largest radius area. Step 4: Iterative relocation and layer-by-layer stripping loop; Step 5: Observe the temperature difference detection and autonomous fire repair of the entire burned area using the drone.

7. The method for dynamic allocation of heterogeneous multi-UAV firefighting tasks as described in claim 1, characterized in that, In step two, the wind field detection units of the six reconnaissance drones continuously report the instantaneous wind direction and wind speed at their respective locations. Let the current wind direction vector be... The detection drone pointing northwest, located on the windward side (southeast), has a clear field of view and is unaffected by hot smoke. After completing its current detection task, the system automatically instructs it to move along the wind direction to replace the detection drone originally located on the leeward side (northwest). The original detection drone on the leeward side (northwest) moves against the wind to fill the vacated detection position on the windward side. This rotation forms a closed loop, and its position update rules are as follows: (Upwind drone) (Downwind drone) Where Δt is the rotation time interval. The current average wind speed, Using the wind direction as a unit vector, through the above rotation, the outermost edge of the downwind side is always under the cross coverage of multiple detection drones. At the same time, it avoids the sensor accuracy of a single detection drone from decreasing or the platform being damaged due to continuous exposure to high-temperature flue gas environment. When the wind direction changes, the system automatically adjusts the rotation direction to always ensure that the downwind detection array is taken over by a healthy drone from the upwind side.

8. The method for dynamic allocation of heterogeneous multi-UAV firefighting tasks as described in claim 1, characterized in that, In step four, the iterative process is repeated. After each round of extinguishing, the outermost radius of the fire is compressed inward by one layer. After each iteration, the fire spread prediction unit calculates the maximum radial distance of the current fire outline. : Where n is the iteration round number, after two consecutive iterations The decrease approaches zero, that is: in, When the system determines that the fire spread has been completely contained and the panoramic scan of the observation drone confirms that the outermost contour of the fire site no longer expands outward in a measurable manner, the outer layer stripping stage ends and the system moves into the remaining fire internal cleanup stage.

9. The method for dynamic allocation of heterogeneous multi-UAV firefighting tasks as described in claim 1, characterized in that, In step five, observe the drone activating its mounted cooled mid-wave infrared thermal imager and read the temperature values ​​of each pixel in each frame of the thermal image below. The current ambient background temperature, corrected by the airborne barometric altimeter. Perform dynamic differential calculations to obtain temperature difference values. ,when Right now It was identified as a potential hazard point due to abnormally high temperatures.

10. The method for dynamic allocation of heterogeneous multi-UAV firefighting tasks as described in claim 1, characterized in that, In step five, when an observation drone detects an abnormally high temperature point, because the drone carries small dry powder fire extinguishing bombs, the system immediately activates the drone's onboard visual servo locking module. This module uses a monocular visual ranging algorithm to calculate the three-dimensional deviation of the abnormal point relative to the drone's coordinate system. When the locking deviation meets the following conditions: At that moment, the flight control system automatically executed the hovering and diving launch command, dropping a small dry powder fire extinguishing bomb at the abnormal point. The observation drone immediately used the infrared thermal imager to re-measure the coordinate point, confirming that the temperature had dropped below the 56-degree Celsius warning line, and completed the autonomous closed-loop handling of the potential hazard point.

Images