Fire extinguishing unmanned aerial vehicle group control method

Abstract

The invention belongs to the technical field of unmanned aerial vehicle group control, and relates to a control method of a fire extinguishing unmanned aerial vehicle group. According to the method, a dynamic fire scene thermodynamic diagram, meteorological data and a fire scene terrain model are collected, and fire scene spreading core kinetic parameters are extracted to calculate a dynamic buffer area; and then predicting meteorological data of a buffer area, generating a combustible flammability index grid map, carrying out weighted fusion on a downsampling result of the dynamic fire scene thermodynamic diagram and the two, and constructing a dynamic fire danger intensity field in a prediction period, thereby solving the problem that multi-dimensional heterogeneous data cannot be effectively integrated in the prior art, and realizing accurate prediction of the dynamic evolution trend of the fire. Meanwhile, state data of the unmanned aerial vehicle group are collected in real time, open fire areas are divided in combination with a fire danger intensity field, reachability of the unmanned aerial vehicles is evaluated, the unmanned aerial vehicle group capable of being dispatched for the second time is screened, traditional strategy defects are overcome, resource allocation is optimized, timely intervention in key areas is ensured, and the overall fire fighting effect is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of drone group control and relates to a control method for a fire-fighting drone group. Background Art

[0002] Forest fires are characterized by sudden onset, rapid spread, and wide-ranging impacts. Traditional firefighting and rescue methods rely primarily on ground forces and large-scale firefighting equipment, which present significant limitations in terms of timely emergency response, operational safety, and adaptability to complex terrain. In recent years, drone technology, with its advantages of high mobility, low cost, and flexible deployment, has been widely adopted in the firefighting sector. Multi-drone collaborative firefighting systems, in particular, enable rapid coverage and efficient response to large fires through clustered operations.

[0003] The core goal of a multi-drone collaborative firefighting system is to achieve accurate fire information perception, optimal allocation of firefighting resources, and efficient coordination of task execution through effective control strategies, thereby improving overall firefighting effectiveness. Drone dispatch control, a key component of the entire system, is responsible for selecting and assigning suitable drones from the existing drone swarm to perform firefighting tasks, serving as the hub connecting situational awareness and firefighting operations.

[0004] However, existing technologies still have the following major flaws in drone dispatching and control: (1) Existing dispatching mechanisms mostly make decisions based on real-time fire intensity and lack comprehensive predictions of the dynamic evolution trend of fires. Fire development is affected by the coupling of multiple factors such as meteorological conditions such as wind speed and direction and combustible characteristics such as vegetation type, distribution density, and moisture content. These factors significantly affect the energy release, spread rate, and expansion direction of fire behavior. However, existing methods often fail to effectively integrate multi-dimensional heterogeneous data such as meteorological and surface combustible data, resulting in insufficient accuracy of fire prediction models, which in turn affects the foresight and scientific nature of dispatching decisions.

[0005] (2) Current dispatch strategies generally adopt the principle of proximity matching based on the spatial distance between drones and the fire boundary, without conducting refined task partitioning and priority assessment of the fire coverage area. This approach ignores the differences in fire intensity, spread trends, and urgency of firefighting within different areas of the fire scene, making it difficult to identify and prioritize high-risk open fire areas or potential spread channels. This leads to irrational resource allocation, failure to meet actual firefighting needs, reduced overall firefighting efficiency, and the potential loss of the optimal intervention opportunity, exacerbating disaster losses. Summary of the Invention

[0006] In view of this, in order to solve the problems raised in the above background technology, a control method for a fire-fighting drone swarm is proposed.

[0007] The purpose of the present invention can be achieved through the following technical solutions: A control method for a fire-fighting drone swarm, comprising: S1, collecting dynamic fire scene thermal maps, meteorological data and fire scene terrain models, and extracting the core dynamic parameters of fire spread from the dynamic fire scene thermal maps, and calculating the dynamic buffer area according to the core dynamic parameters of fire spread; wherein the core dynamic parameters of fire spread include: the fire line advancement velocity vector, the maximum advancement direction of the fire line and the overall spatial expansion trend of the fire scene.

[0008] S2. Predicting meteorological data for a forecast period of the dynamic buffer area based on meteorological data; generating a combustible material flammability index raster map based on combustible material distribution data located in a fire terrain model within the dynamic buffer area using preset combustible material assessment rules.

[0009] S3. Integrate the downsampling results of the dynamic fire scene heat map, the combustible material flammability index raster map and the meteorological data during the forecast period to generate a dynamic fire risk intensity field during the forecast period.

[0010] S4. Collect the status data of the drone group in real time, filter out the candidate set of dispatchable drones according to the preset dispatch trigger rules, use the dynamic fire risk intensity field during the prediction period to divide the open fire area, evaluate the accessibility of the candidate set of dispatchable drones to the open fire area, and determine the drone group that can be dispatched for the second time.

[0011] Compared with the existing technology, the beneficial effects of the present invention are as follows: (1) The present invention constructs a dynamic fire risk intensity field for the forecast period by integrating the downsampling results of the dynamic fire scene thermal map, the combustible flammability index grid map and the meteorological data of the forecast period, which solves the problem that the existing technology fails to effectively integrate multi-dimensional heterogeneous data such as meteorological conditions and combustible characteristics, realizes the accurate prediction of the dynamic evolution trend of the fire, improves the rationality and timeliness of the scheduling decision, and provides a reliable dynamic risk assessment basis for the intervention timing and path planning of the drone swarm.

[0012] (2) The present invention collects the status data of the drone group in real time, screens the schedulable candidate set according to the preset scheduling trigger rules, and divides the open fire area and evaluates the accessibility of drones in combination with the predicted fire risk intensity field. It solves the defect of the traditional proximity matching strategy that lacks refined task partitioning and priority evaluation, realizes the priority identification and disposal of high-risk open fire areas, optimizes the efficiency of fire extinguishing resource allocation, ensures timely intervention in key areas, and effectively improves the overall fire fighting effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 Schematic diagram of the method steps of the present invention.

[0015] Figure 2 This is a schematic diagram of the classification of flammability levels of grid cells according to the ratio of vegetation moisture content to vegetation density in the present invention.

[0016] Figure 3 This is a flow chart of the step contents of S4 in the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] See also Figure 1 As shown, the present invention provides a control method for a fire-fighting drone swarm, including: S1, collecting dynamic fire scene thermal maps, meteorological data and fire scene terrain models, and extracting fire scene spread core dynamic parameters from the dynamic fire scene thermal maps, and calculating dynamic buffer areas based on the fire scene spread core dynamic parameters.

[0019] The core dynamic parameters of fire spread include: fire line advancement velocity vector, fire line maximum advancement direction and overall spatial expansion trend of the fire.

[0020] Furthermore, the method for collecting the dynamic fire scene thermal map, meteorological data and fire scene terrain model is as follows: using a fire scene monitoring terminal equipped with an infrared thermal imager, the dynamic fire scene thermal map is collected in real time through grid path cruising.

[0021] A fire monitoring terminal with integrated three-axis ultrasonic wind speed is used to form a monitoring network to calculate meteorological data of wind speed and direction in real time.

[0022] A fire monitoring terminal equipped with a laser radar is used to scan the surface area of ​​the fire site, and combined with RTK positioning to generate point cloud data to construct a fire site terrain model.

[0023] It should be noted that the fire scene monitoring terminal refers to an equipment terminal that integrates specific sensors such as infrared thermal imagers, three-axis ultrasonic anemometers, lidars, etc., and is used to collect fire scene information. It may include cruise drones equipped with corresponding sensors, ground fixed monitoring stations or mobile monitoring platforms, etc., which work together to form a monitoring network covering the fire area to meet the needs of dynamic data collection.

[0024] The dynamic fire scene heat map is thermal image data that reflects the spatial distribution and dynamic changes of the temperature in the fire area. It depicts the boundary and intensity characteristics of the high-temperature area of ​​the fire scene through the temperature gradient distribution. It serves as the basic data for extracting the core dynamic parameters of the fire spread. The infrared thermal imager can penetrate the smoke and accurately capture the temperature distribution of the high-temperature area. It reflects the spatial form of the fire scene boundary and the changing characteristics in the time dimension in real time, providing a direct temperature field basis for predicting and analyzing the fire spread area. It is the core data source for realizing dynamic fire monitoring.

[0025] The meteorological data include wind direction and wind speed data, which are key meteorological factors affecting the spread of fire. They directly determine the dominant direction, speed and burning intensity of the fire line. For example, strong winds will accelerate the burning rate of combustible materials and promote the rapid expansion of the fire line. Real-time monitoring and prediction of its changes can accurately quantify the impact of meteorological factors on the development of the fire and improve the accuracy of fire risk intensity prediction.

[0026] The fire terrain model includes information on the fire area's surface elevation, slope, aspect, surface relief, and the spatial distribution of combustibles. This model is used to precisely locate the distribution of combustibles within the dynamic buffer zone, providing the foundational data for generating a raster map of the combustible flammability index. Furthermore, it can be used to analyze the impact of terrain elements like slope and aspect on the path of fire spread and the direction of fire front advancement, assessing the impact of terrain on combustion conditions and providing reliable geospatial data support for optimizing drone firefighting routes.

[0027] Furthermore, the specific method for obtaining the core dynamic parameters of fire spread is to extract the danger level tensor from the dynamic fire scene thermal map sequence, and dynamically capture the fire line contour at each time step by calculating the mutation boundary of the fire danger level in the danger level tensor between adjacent time steps.

[0028] It should be noted that the method for extracting the danger level tensor from the dynamic fire scene thermal map sequence is as follows: denoising the thermal map of each time step in the dynamic fire scene thermal map sequence, extracting the thermal value of each pixel as the fire scene temperature intensity; dividing the thermal value into low, medium, high, and extremely high danger levels according to the preset temperature threshold, and converting the thermal value of each pixel into the corresponding danger level quantification value. The preset temperature threshold is set according to the actual monitoring needs. The danger level quantification results of multiple time steps are arranged in time series order and integrated into a three-dimensional data structure containing spatial and temporal dimensions; the spatial dimension corresponds to the grid unit distribution of the fire area, and the temporal dimension corresponds to the acquisition time sequence of the thermal map sequence. Each element value is the danger level quantification value under the time-space coordinate, thereby forming a danger level tensor.

[0029] The mutation boundary represents the spatial boundary where the fire risk level undergoes significant and discontinuous changes in the risk level tensor of adjacent time steps, and the boundary corresponds to the spatial position of the fire line in the fire scene. The spatial dimension data of two consecutive time steps are intercepted from the risk level tensor to obtain two risk level quantization value matrices corresponding to the same fire area. The two risk level quantization value matrices are subjected to grid cell-by-grid cell difference operations to obtain a risk level difference matrix, in which the element value represents the change in the risk level quantization value of the same grid cell in adjacent time steps. The risk level mutation threshold is preset according to the actual combustion characteristics of the fire scene. In the risk level difference matrix, the grid cells whose changes exceed the risk level mutation threshold are screened out, and the spatial coordinates of these grid cells are connected to form a continuous boundary line, which is the mutation boundary of the adjacent time steps.

[0030] The fire line advancement velocity vector is used to describe the vector parameters of the spatial displacement of each discrete point on the fire line contour per unit time, including the magnitude and direction of the advancement velocity.

[0031] In a specific embodiment, the method of normal vector field analysis is: discretize the fire line contour, calculate the normal vector of each point on the contour, and form a normal vector field covering the entire fire line contour; by performing direction clustering or main direction analysis on the normal vector field, screen out the normal vector direction with the highest frequency of occurrence and the largest modulus value, and this direction is the maximum propulsion direction of the fire line.

[0032] The fire scene centroid is calculated based on the fire line contour at each time step to form the centroid migration trajectory. Principal component analysis is performed based on the area change rate of the fire scene contour sequence and the centroid migration trajectory to extract the overall spatial expansion trend of the fire scene.

[0033] The fire centroid refers to the geometric center of the area enclosed by the fire line contour at a single time step, and is a representative coordinate point that characterizes the spatial position of the fire. It is obtained by combining the discrete coordinate points of the fire line contour with the area-weighted center coordinates of the area enclosed by the contour. It is used to track the position changes of the fire centroid at each time step to form a centroid migration trajectory, which can intuitively reflect the overall movement trend of the fire in space and provide basic trajectory data for analyzing the direction of fire expansion.

[0034] The overall spatial expansion trend of the fire scene represents the comprehensive characteristics of the overall spread direction, expansion rate and spatial distribution pattern shown in the time series, and is a quantitative description of the macro-development law of the fire. As a key basis for the calculation of dynamic buffer areas, it can assist in predicting the main impact range of the fire in the future, provide a macro-level reference for the expansion of the fire risk area for drone swarm scheduling, and improve the foresight of fire-fighting resource allocation.

[0035] It should be noted that the principal component analysis is used to extract the most representative trend components from the area change rate and the centroid migration trajectory, to achieve dimensionality reduction and focusing on the overall spatial expansion law of the fire scene, to simplify complex data and highlight the core expansion characteristics.

[0036] Furthermore, the method for obtaining the dynamic buffer area is: analyzing the fire line advancement velocity vector, extracting the instantaneous advancement velocity vector of each discrete point on the fire line contour, and determining the basic buffer width value of each point on the fire line contour.

[0037] It should be noted that the determination of the basic buffer width value is based on the instantaneous advancement speed of each point on the fire line. The greater the speed of the point, the larger the basic buffer width value is usually to match the risk of rapid spread of fire.

[0038] With the center of mass of the fire scene as the origin, draw the main axis along the direction of maximum advancement of the fire line. For each point on the fire line contour, calculate the angle between the point and the main axis; set the angle range and determine the angle threshold based on the angle value.

[0039] It should be noted that the angle threshold is set by generating an angle dataset based on the calculated angle values. This dataset is subjected to frequency distribution statistics and presented as a histogram to identify the main clustering intervals or distribution inflection points. Based on this distribution characteristic, the angle threshold is determined by calculating the mean of the dataset plus or minus a multiple of the standard deviation.

[0040] The basic buffer width value is corrected according to the comparison between the angle value and the angle threshold to obtain a first dynamic buffer width value.

[0041] The specific method for correcting the basic buffer width value based on the comparison between the angle value and the angle threshold is as follows: when the angle value is less than the angle threshold, a correction factor is set, and the correction factor decreases linearly according to the angle value; when the angle value is greater than or equal to the angle threshold, the correction factor is fixed to the minimum reduction multiple; the basic buffer width value is multiplied by the correction factor to obtain the first dynamic buffer width value.

[0042] It should be noted that the correction factor is a coefficient used to dynamically adjust the basic buffer width value based on the comparison result of the angle value and the angle threshold. Its numerical value directly determines the enlargement or reduction ratio of the basic buffer width value, and ultimately affects the calculation result of the first dynamic buffer width value.

[0043] The first dynamic buffer width value is corrected by analyzing the area change rate and centroid migration trajectory of the fire scene contour sequence to obtain the second dynamic buffer width value.

[0044] The specific method for correcting the first dynamic buffer width value is as follows: if the first-order derivative of the area change rate is positive and the absolute value increases, and the center of mass instantaneous velocity modulus increases, the overall amplification coefficient is calculated according to the area increase and the center of mass acceleration amplitude, and the first dynamic buffer width value is multiplied by the overall amplification coefficient; if the first-order derivative of the area change rate approaches zero and the fluctuation is stable, and the center of mass instantaneous velocity modulus is constant, the first dynamic buffer width value remains unchanged; if the first-order derivative of the area change rate is negative, the increase is narrowed or the local area is reduced, and the center of mass instantaneous velocity modulus decreases, the reduction ratio is determined according to the degree of local change, and the first dynamic buffer width value is multiplied by the reduction ratio; the second dynamic buffer width value is obtained through the above modification.

[0045] The second dynamic buffer width value is converted into a dynamic buffer area through normal offset-boundary reconstruction.

[0046] In a specific embodiment, the normal offset-boundary reconstruction method is as follows: first, a normal offset operation is performed, and each discrete point on the fire line contour is used as a reference, and the corresponding second dynamic buffer width value is moved along the normal direction of each point to obtain a number of offset points. For adjacent offset points, it is necessary to determine the transition point on the offset path between the two points by a smooth transition method based on their normal direction and offset distance, to ensure that the preliminary boundary formed after the offset is continuous and has no mutations. Boundary reconstruction is then carried out, and all offset points and transition points are fitted to eliminate local abnormal offset points that may be generated due to uneven distribution of discrete points, and a curve fitting algorithm is used to connect the remaining points into a smooth, closed curve. The area enclosed by the curve is the dynamic buffer area obtained after the normal offset-boundary reconstruction, and its boundary maintains a preset second dynamic buffer width relationship with the fire line contour, and can adapt to the morphological changes of the fire line contour.

[0047] It should be noted that the dynamic buffer area is a dynamically expanded area surrounding the current fire line outline, used to identify key areas where the fire may spread in advance. In subsequent steps, meteorological data forecasts for this area, combustible flammability index assessments, and dynamic fire risk intensity field construction all use the dynamic buffer area as the spatial boundary to ensure that the forecast focuses on the areas where the fire is most likely to spread, improving the pertinence and efficiency of fire risk assessments. In drone swarm scheduling, the division of open fire areas within the dynamic buffer area and the assessment of drone accessibility must be based on this area to ensure that drone resources are prioritized on the key paths of fire spread and improve the accuracy of fire extinguishing responses.

[0048] S2. Predicting meteorological data for a forecast period of the dynamic buffer area based on meteorological data; generating a combustible material flammability index raster map based on combustible material distribution data located in a fire terrain model within the dynamic buffer area using preset combustible material assessment rules.

[0049] In one specific embodiment, the method for acquiring meteorological data for the forecast period comprises: using collected real-time meteorological data combined with the spatial extent of the dynamic buffer zone, using time series analysis to predict meteorological data for a specific future period in the region, thereby generating meteorological data for the forecast period that matches the dynamic buffer zone. The forecast period refers to a specific time interval set to assess the future intensity of the fire risk within the dynamic buffer zone. A reasonable forecast duration is set based on meteorological data, the dynamic changes in the state of combustible materials, and the time window of the actual drone firefighting operation. This is used to balance prediction accuracy with response speed to rapid changes in the fire scene, thereby ensuring the feasibility of the scheduling plan.

[0050] The method for obtaining the combustible material flammability index grid map is as follows: in the fire scene terrain model, combustible material distribution data is extracted based on the contour boundary of the dynamic buffer area, and the combustible material distribution data in the area is divided into preset grid units.

[0051] It should be noted that the size of the preset grid unit can be flexibly adjusted according to the actual terrain complexity and data accuracy requirements.

[0052] The vegetation type of each grid unit is identified, and the vegetation moisture content of each grid unit is obtained through the mapping relationship between vegetation type and moisture content.

[0053] It should be noted that the vegetation moisture content refers to the moisture content per unit area based on vegetation type classification. Its value reflects the inherent water holding characteristics of a specific vegetation type. It is an attribute value under standardized conditions and does not involve the influence of structural or quantitative parameters such as vegetation coverage, biomass density, canopy height or number of vegetation patches.

[0054] The ratio of vegetation coverage area to grid area in each grid cell is the vegetation density ratio.

[0055] It should be noted that the vegetation density ratio is a dynamically calculated measured value, reflecting the actual density of vegetation coverage.

[0056] See Figure 2 As shown, the grid cells are classified into flammability levels according to the vegetation moisture content and vegetation density ratio calculated above through preset thresholds, as follows: Low flammability level: the vegetation moisture content is higher than the preset critical flammability threshold and the vegetation density ratio is lower than the preset critical density threshold.

[0057] Medium flammability level: the vegetation moisture content is higher than the preset critical flammability threshold and the vegetation density ratio is higher than the preset critical density threshold; when the vegetation moisture content is lower than the preset critical flammability threshold and the vegetation density ratio is lower than the preset critical density threshold.

[0058] High flammability level: The moisture content of vegetation is lower than the preset critical flammability threshold and the vegetation density ratio is higher than the preset critical density threshold.

[0059] It should be noted that the vegetation moisture content and vegetation density ratio are key indicators for scientifically dividing flammability levels based on the properties of the combustible materials themselves, and provide a quantitative basis for combustible material risk for subsequent dynamic fire risk intensity prediction and precise scheduling of drone swarms.

[0060] The low flammability level indicates that the moisture is sufficient and sparsely distributed, and the risk of fire spread is low; the medium flammability level covers the intermediate risk states between the high moisture and high density flammable but limited spread conditions and the low moisture and low density flammable but insufficient fuel conditions, avoiding the assessment bias caused by simple dichotomy; the high flammability level indicates that the vegetation is dry and dense, which is very likely to cause rapid fire spread.

[0061] The preset critical flammability threshold represents a numerical threshold that determines whether the moisture content of vegetation has reached a critical flammable state. It is determined using the percentile method by measuring the relationship curve between the ignition point and moisture content of different vegetation types, combined with statistical data on moisture content during vegetation combustion in historical fire cases.

[0062] The preset critical density threshold represents the proportional threshold that determines whether the vegetation cover density reaches the conditions that easily form continuous combustion conditions. It is determined through cluster analysis by statistically analyzing the minimum cover density that can support the continuous spread of fire under different vegetation types, combined with the actual density range of vegetation distribution in the fire terrain model.

[0063] The level threshold refers to the boundary value that distinguishes low, medium, and high flammability levels. The numerical corresponding threshold refers to the mapping relationship that converts qualitative flammability levels into quantitative indices. Based on the weighted analysis of the impact of flammability levels on fire spread speed and combustion intensity, combined with machine learning training based on historical fire data, the critical values ​​for level division and the corresponding index range are determined. For example, the numerical mapping rule can be set to correspond to 1-3 for low flammability levels, 4-6 for medium flammability levels, and 7-10 for high flammability levels.

[0064] The flammability grade grid cells are converted into a combustible material flammability index raster map according to the set grade threshold and the corresponding value threshold.

[0065] It should be noted that the combustible material flammability index grid map is output in a geographic information system compatible format, and the value of each grid cell intuitively represents the flammability level, which is convenient for superimposing on the fire situation map for analysis.

[0066] S3. Integrate the downsampling results of the dynamic fire scene heat map, the combustible material flammability index raster map and the meteorological data during the forecast period to generate a dynamic fire risk intensity field during the forecast period.

[0067] Furthermore, the method for obtaining the dynamic fire risk intensity field during the forecast period is: downsampling the dynamic fire scene thermal map in spatial resolution to generate an energy base grid map of uniform size.

[0068] The combustible flammability index raster map and the energy base raster map were geo-referenced and resampled to ensure accurate matching of grid cells.

[0069] The meteorological data of the forecast period are spatially interpolated to generate a meteorological grid map with the same spatial resolution and coordinate system as the energy base grid map.

[0070] The energy base grid map, the spatially aligned combustible material flammability index grid map, and the meteorological grid map are superimposed grid by grid, and a weighted calculation is performed using a preset weight coefficient to obtain the dynamic fire hazard intensity field during the forecast period.

[0071] The preset weight coefficients are dynamically calibrated based on the real-time status of the fire scene. For example, the energy base grid represents the actual energy distribution of the current fire scene and is the basis for fire risk intensity. It is usually assigned a higher initial weight. When the fire scene is rapidly spreading, the weight of the meteorological grid can be appropriately increased. When the fire scene is stable but the combustible material flammability index is high, the weight of the combustible material flammability index grid can be adjusted first.

[0072] It should be noted that the original data of the dynamic fire scene thermal map may contain a large amount of redundant information due to its high resolution. The downsampling process does not simply compress the data, but adjusts the size and number of grid units while ensuring that the key energy characteristics of the fire scene are not lost, so that they match the spatial scale of the combustible material flammability index grid map and meteorological grid map that will be subsequently fused, ensuring that multi-source data can be calculated under the same grid system.

[0073] The present invention constructs a dynamic fire risk intensity field for the forecast period by integrating the downsampling results of the dynamic fire scene thermal map, the combustible flammability index grid map and the meteorological data of the forecast period. It solves the problem that the existing technology fails to effectively integrate multi-dimensional heterogeneous data such as meteorological conditions and combustible characteristics, realizes the accurate prediction of the dynamic evolution trend of the fire, improves the rationality and timeliness of scheduling decisions, and provides a reliable dynamic risk assessment basis for the intervention timing and path planning of drone swarms.

[0074] See Figure 3 As shown in S4, the status data of the drone group is collected in real time, and the candidate set of schedulable drones is screened out according to the preset scheduling trigger rules. The open fire area is divided using the dynamic fire risk intensity field during the prediction period, and the accessibility of the candidate set of schedulable drones to the open fire area is evaluated to determine the drone group that can be scheduled for secondary scheduling.

[0075] Furthermore, the method for obtaining the candidate set of schedulable drones is: extracting the device identification code, real-time location coordinates, current remaining power, fire-fighting equipment mounting status signal and equipment self-test fault flag of each drone from the drone group status data.

[0076] When a drone meets the equipment self-test fault flag and all submodules are in normal state, and no warning signal is triggered within three consecutive heartbeat cycles, it is marked as the first candidate drone.

[0077] It should be noted that the submodules refer to independent detection units within the drone's self-test system, divided by function. These submodules all indicate signals indicating no abnormalities, such as motor overload, normal positioning signals, and stable communication links. The heartbeat cycle refers to the regular communication interval between the drone and the ground control terminal, used to confirm the device's online status. Warning signals include low battery warnings, device anomaly warnings, communication interruption warnings, and other warnings that may affect mission execution.

[0078] When the first candidate UAV meets the fire extinguishing equipment mounting status signal indicating that it is loaded and the fire extinguishing equipment mounting self-test loop feedback valid ready signal, it is marked as the second candidate UAV.

[0079] The second candidate UAVs are screened by dynamic power threshold to obtain the candidate set of schedulable UAVs.

[0080] The fire-fighting equipment mounting status signal indicates that the drone has detected, through the mounting sensor, that the fire-fighting equipment is correctly installed and ready for launch or deployment. The fire-fighting equipment mounting self-test circuit feedback signal indicates that the equipment's internal self-test circuit has completed testing of the spray mechanism, safety device, and agent status, confirming that the equipment can operate normally as instructed. These two combined features ensure that the drone is not only carrying fire-fighting equipment but also that the equipment is ready for immediate use, preventing firefighting mission failures due to equipment malfunction. This is a key functional requirement for selecting the second candidate drone.

[0081] Furthermore, the process of dynamic power threshold screening is: comparing the current remaining power of the second candidate UAV with the preset scheduling safety threshold, and screening the second candidate UAV whose power is higher than the preset scheduling safety threshold.

[0082] It should be noted that the preset dispatch safety threshold refers to the minimum remaining battery level to ensure that the drone can complete the firefighting mission and return to a safe area. This is calculated based on the drone's endurance, the average energy consumption of historical typical firefighting missions (such as the flight distance to and from the fire scene, hovering time, and power consumption during firefighting operations), and a safety margin. This is used to select drones from the second candidate pool with sufficient remaining battery power to complete the firefighting mission and return, thus preventing drones from losing contact or crashing due to insufficient battery.

[0083] The battery difference between adjacent cycles is calculated based on the battery data of the historical cycles, and the battery difference is compared with the preset battery difference threshold to select the second candidate drone with a battery difference less than the preset threshold.

[0084] It should be noted that the preset battery difference threshold is the critical difference used to determine whether a drone's battery level is stable. It represents the maximum allowable fluctuation in the drone's remaining battery level between successive cycles. This threshold is determined by statistically analyzing the drone's battery consumption fluctuations under normal operating conditions and combining it with battery performance parameters through multiple calculations and calibrations. This threshold is used to screen drones with unstable battery consumption. If the battery difference between successive cycles exceeds this threshold, the drone may be experiencing abnormal power consumption and should be excluded from the schedulable candidate set.

[0085] The power consumption factor is calculated based on the power data of the drone in the historical period, and the power consumption factor is compared with the preset consumption factor threshold. The second candidate drone with a power consumption factor smaller than the preset consumption factor threshold is selected as the dispatchable candidate drone.

[0086] It should be noted that the power consumption factor refers to a parameter that quantifies the power consumption intensity of the drone per unit time or unit distance. It is calculated by associating real-time power changes with flight mileage, flight time, and mission operations, and reflects the energy consumption efficiency of the drone.

[0087] The preset power consumption factor threshold is a critical value used to determine whether a drone's energy efficiency is within the normal range, representing the maximum allowable upper limit of the power consumption factor. This threshold is determined through statistical analysis of energy consumption test data collected from drones of the same model under standard operating conditions, combined with the maximum acceptable energy consumption under extreme operating conditions. This threshold is used to screen drones with abnormal energy efficiency. If a drone's power consumption factor exceeds this threshold, it indicates excessive energy consumption, posing a risk of insufficient battery power during mission execution and should be excluded from the schedulable candidate set.

[0088] Furthermore, the method for dividing the open fire area based on the dynamic fire risk intensity field in the prediction period is: based on the temperature distribution of the dynamic fire risk intensity field in the prediction period, the temperature gradient is calculated by the temperature difference between each grid unit and its adjacent grid units and combined with the grid spacing.

[0089] The combustible density of each grid cell is obtained from the combustible distribution data.

[0090] It should be noted that the combustible density is based on the boundaries of a preset grid unit. The total mass, volume or coverage area of ​​the combustibles in the unit is counted and then divided by the area of ​​the grid unit to obtain the combustible density per unit area.

[0091] The meteorological diffusion parameters of each grid cell are calculated based on the fire spread rate and key meteorological factors matched with the unified grid scale output of the dynamic fire intensity field during the forecast period.

[0092] It should be noted that key meteorological factors refer to core meteorological parameters that significantly influence the rate of fire spread, primarily including wind speed and wind direction. These meteorological diffusion parameters include wind speed correction coefficients and wind direction diffusion weights, a set of parameters quantified based on key meteorological factors that characterize the effect of meteorological conditions on promoting or inhibiting fire spread.

[0093] The temperature gradient, combustible density and meteorological diffusion parameters of each grid cell are weighted and fused to obtain the fire risk energy value of each grid cell.

[0094] Among them, the weight coefficient in the weighted fusion is dynamically adjusted according to the contribution of each parameter to the fire risk energy. The temperature gradient directly reflects the energy driving intensity of the fire spread and is the core representation of the fire risk energy, and is usually given a higher basic weight; the density of combustible materials determines the material basis for the continued burning and spread of the fire. When in the stable combustion stage, the corresponding weight can be appropriately increased; the meteorological diffusion parameter reflects the driving effect of meteorological factors such as wind speed and wind direction on the spread of the fire. When in the rapid expansion period, the corresponding weight can be appropriately increased.

[0095] The fire hazard energy value of each grid cell is compared with the core melting zone threshold obtained by analyzing historical fire data at the same grid size. When the fire hazard energy value is greater than the core melting zone threshold, the grid cell is marked as an open fire unit.

[0096] It should be noted that the core melting zone threshold is based on the critical fire energy value that distinguishes open flame areas from non-open flame areas, obtained through statistical analysis of historical fire scene data at the same grid size. This is determined by collecting the fire energy values ​​of grid cells in confirmed open flame areas from a large number of historical fire scenes and analyzing their distribution characteristics. This standard for determining whether a grid cell is an open flame unit ensures consistency and accuracy in the open flame area demarcation.

[0097] It should be noted that the open fire unit represents the area where the fire is most active and the risk is highest.

[0098] The slope continuity, slope consistency and surface permeability of each grid cell are extracted from the dynamic fire hazard intensity field during the forecast period. The terrain connectivity index is obtained by calculation, and the grid cells with terrain connectivity index greater than the preset threshold are screened.

[0099] It should be noted that the terrain connectivity index represents the ability of terrain conditions to support the continuous spread of fire.

[0100] Based on the spatial topological relationship of the eight-neighborhood, adjacent grid cells with a terrain connectivity index greater than a preset threshold are restrictedly merged to generate connected open fire areas.

[0101] It should be noted that the preset threshold grid cells refer to those with a terrain connectivity index greater than the preset threshold, indicating that the terrain conditions are favorable for the continuous spread of fire. By comparing the terrain connectivity index of each grid cell with the preset threshold, cells with indices exceeding the threshold are screened out. These cells are potential areas for continuous fire spread and provide the basis for subsequent merging to form connected open fire areas.

[0102] The restricted merging is only for the open fire units that meet the two conditions of adjacency and terrain penetration, avoiding the mistaken classification of non-adjacent or terrain-blocked open fire units into the same area, and ensuring the accuracy of the open fire area division.

[0103] Furthermore, the method for obtaining the secondary schedulable drone group is: calculating the shortest Euclidean distance and path weighted distance from the real-time position coordinates of each drone in the schedulable drone candidate set to the open fire area.

[0104] The weight coefficient in the process of calculating the weighted distance of the path is set according to the actual terrain, weather and fire risk intensity.

[0105] The cruising speed is extracted from the drone swarm status data and combined with the shortest Euclidean distance. The theoretical arrival time is calculated, and the expected energy consumption is estimated based on the path weighted distance.

[0106] It should be noted that cruising speed refers to the average speed of a drone in steady flight and is a core parameter reflecting its mobility. It is calculated from real-time data extracted from the drone swarm's status, using position changes and time intervals recorded by the drones' onboard positioning modules. This is used to calculate the theoretical time it takes for a drone to reach an open fire area, a key metric for evaluating drone response speed and directly impacting the timely arrival of rescue efforts in high-risk fire areas.

[0107] It should be noted that the theoretical arrival time is an estimated value under ideal conditions.

[0108] Filter out drones whose theoretical arrival time does not exceed the preset fire rescue time threshold, whose expected energy consumption does not exceed the preset maximum consumable threshold, and whose path weighted distance does not exceed the preset maximum cruising range threshold, and obtain a drone group that can be dispatched secondary.

[0109] It should be noted that the preset fire rescue time threshold refers to the maximum time allowed for drones to reach the open fire area to control the spread of the fire. This threshold is preset by the fire command system based on a combination of the fire risk level of the open fire area in the dynamic fire risk intensity field during the forecast period and statistical data on optimal firefighting response times in historical fires. It is used to select drones that can arrive at the scene before the fire gets out of control, ensuring the timeliness of firefighting operations.

[0110] The preset maximum power consumption threshold is the maximum amount of energy a drone is allowed to consume while performing a mission. It's calculated based on the drone's total battery life and the minimum power required to return to base after completing the mission. This is used to limit a drone's power consumption, preventing excessive power consumption from preventing it from returning home, and is a key indicator for ensuring safe drone recovery.

[0111] The preset maximum range threshold is the maximum distance a drone can fly at full charge, taking into account round-trip missions and redundancy. This is calculated using the drone's factory range specifications, combined with actual payload and safety margins. This is used to screen drones capable of reaching and returning from open flames, ensuring their flight distances are within the safe range to avoid loss of connection or crashes due to exceeding range.

[0112] The present invention collects drone swarm status data in real time, screens the schedulable candidate set according to preset scheduling trigger rules, and divides the open fire area and evaluates the drone accessibility in combination with the predicted fire risk intensity field. It solves the defect of traditional proximity matching strategy that lacks refined task partitioning and priority assessment, realizes the priority identification and disposal of high-risk open fire areas, optimizes the efficiency of fire-fighting resource allocation, ensures timely intervention in key areas, and effectively improves the overall fire fighting effect.

[0113] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0114] Those skilled in the art will appreciate that the algorithmic steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0115] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0116] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0117] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling a group of fire-fighting drones, characterized in that: include: S1. Collect dynamic fire scene thermal maps, meteorological data, and fire scene terrain models, extract the core dynamic parameters of fire spread from the dynamic fire scene thermal maps, and calculate the dynamic buffer area based on the core dynamic parameters of fire spread; The core dynamic parameters of fire spread include: fire front velocity vector, fire front maximum propagation direction and overall fire space expansion trend; S2. Predicting meteorological data for a forecast period of the dynamic buffer area based on meteorological data; generating a combustible material flammability index grid map based on combustible material distribution data located in the fire terrain model within the dynamic buffer area using preset combustible material assessment rules; S3. Fusing the downsampling results of the dynamic fire scene heat map, the combustible material flammability index raster map, and the meteorological data for the forecast period, a dynamic fire risk intensity field for the forecast period is generated; S4. Collect the status data of the drone group in real time, filter out the candidate set of dispatchable drones according to the preset dispatch trigger rules, use the dynamic fire risk intensity field during the prediction period to divide the open fire area, evaluate the accessibility of the candidate set of dispatchable drones to the open fire area, and determine the drone group that can be dispatched for the second time.

2. The method for controlling a swarm of firefighting drones according to claim 1, wherein: Method for collecting the dynamic fire scene thermal map, meteorological data and fire scene terrain model: Using a fire monitoring terminal equipped with an infrared thermal imager, dynamic fire scene thermal maps are collected in real time through grid-based path patrol; A fire monitoring terminal with integrated triaxial ultrasonic wind speed is used to form a monitoring network, which can calculate meteorological data of wind speed and direction in real time. A fire monitoring terminal equipped with a laser radar is used to scan the surface area of ​​the fire site, and combined with RTK positioning to generate point cloud data to construct a fire site terrain model.

3. The method for controlling a swarm of firefighting drones according to claim 1, wherein: The specific method for obtaining the core dynamic parameters of fire spread is as follows: Extract the danger level tensor from the dynamic fire scene heat map sequence, and dynamically capture the fire line contour at each time step by calculating the sudden change boundary of the fire danger level in the danger level tensor between adjacent time steps; The fire line's advancing velocity vector is calculated by the spatial offset of the fire line's contours in adjacent time steps, and the fire line's maximum advancing direction is determined using normal vector field analysis. The fire scene centroid is calculated based on the fire line contour at each time step to form the centroid migration trajectory. Principal component analysis is performed based on the area change rate of the fire scene contour sequence and the centroid migration trajectory to extract the overall spatial expansion trend of the fire scene.

4. The method for controlling a swarm of firefighting drones according to claim 1, wherein: The method for obtaining the dynamic buffer area is: Analyze the fire line's advancing velocity vector, extract the instantaneous advancing velocity vector of each discrete point on the fire line's contour, and determine the basic buffer width value of each point on the fire line's contour; With the center of mass of the fire scene as the origin, draw the main axis along the direction of maximum advancement of the fire line. For each point on the fire line contour, calculate the angle between the point and the main axis. Set the angle range and determine the angle threshold according to the angle value; The basic buffer width value is corrected according to the comparison between the angle value and the angle threshold value to obtain a first dynamic buffer width value; The first dynamic buffer width value is corrected by analyzing the area change rate and centroid migration trajectory of the fire scene contour sequence to obtain the second dynamic buffer width value; The second dynamic buffer width value is converted into a dynamic buffer area through normal offset-boundary reconstruction.

5. The method for controlling a swarm of firefighting drones according to claim 1, wherein: The method for obtaining the combustible material flammability index grid map is as follows: Extract combustible material distribution data based on the contour boundary of the dynamic buffer area in the fire terrain model, and divide the combustible material distribution data in the area into preset grid units; Identify the vegetation type of each grid unit and obtain the vegetation moisture content of each grid unit through the mapping relationship between vegetation type and moisture content; The ratio of vegetation coverage area to grid area in each grid cell is the vegetation density ratio; Based on the ratio of vegetation moisture content to vegetation density calculated above, the grid cells are classified into flammability levels using preset thresholds as follows: Low flammability level: the moisture content of vegetation is higher than the preset critical flammability threshold and the vegetation density ratio is lower than the preset critical density threshold; Medium flammability level: the moisture content of vegetation is higher than the preset critical flammability threshold and the vegetation density ratio is higher than the preset critical density threshold; when the moisture content of vegetation is lower than the preset critical flammability threshold and the vegetation density ratio is lower than the preset critical density threshold; High flammability level: The moisture content of vegetation is lower than the preset critical flammability threshold and the vegetation density ratio is higher than the preset critical density threshold; The flammability grade grid cells are converted into a combustible material flammability index raster map according to the set grade threshold and the corresponding value threshold.

6. The method for controlling a swarm of firefighting drones according to claim 1, wherein: The method for obtaining the dynamic fire risk intensity field during the forecast period is as follows: Downsample the spatial resolution of the dynamic fire scene thermal map to generate an energy base grid map of uniform size; Geo-reference and resample the combustible flammability index raster map and the energy base raster map to ensure accurate matching of grid cells; Perform spatial interpolation on the meteorological data during the forecast period to generate a meteorological grid map with the same spatial resolution and coordinate system as the energy base grid map; The energy base grid map, the spatially aligned combustible material flammability index grid map, and the meteorological grid map are superimposed grid by grid, and a weighted calculation is performed using a preset weight coefficient to obtain the dynamic fire hazard intensity field during the forecast period.

7. The method for controlling a swarm of firefighting drones according to claim 1, wherein: The method for obtaining the candidate set of dispatchable drones is as follows: Extract each drone's device identification code, real-time location coordinates, current remaining battery power, firefighting equipment mounting status signal, and device self-test fault flag from the drone swarm status data; When a drone meets the equipment self-test fault flag and all submodules are in normal state, and no warning signal is triggered within three consecutive heartbeat cycles, it is marked as the first candidate drone; When the first candidate UAV meets the fire extinguishing equipment mounting status signal indicating that it is loaded, and the fire extinguishing equipment mounting self-test loop feedback valid ready signal, it is marked as the second candidate UAV; The second candidate UAVs are screened by dynamic power threshold to obtain the candidate set of schedulable UAVs.

8. The method for controlling a swarm of firefighting drones according to claim 7, wherein: The process of dynamic power threshold screening is as follows: Compare the current remaining power of the second candidate UAV with the preset scheduling safety threshold, and select the second candidate UAV whose power exceeds the preset scheduling safety threshold; Calculate the battery difference between adjacent cycles based on the historical drone battery data, compare the battery difference with the preset battery difference threshold, and select the second candidate drone with a battery difference less than the preset threshold; The power consumption factor is calculated based on the power data of the drone in the historical period, and the power consumption factor is compared with the preset consumption factor threshold. The second candidate drone with a power consumption factor smaller than the preset consumption factor threshold is selected as the dispatchable candidate drone.

9. The method for controlling a swarm of firefighting drones according to claim 5, wherein: The method for dividing the open fire area based on the dynamic fire risk intensity field during the forecast period is as follows: Based on the temperature distribution of the dynamic fire intensity field during the forecast period, the temperature gradient is calculated by the temperature difference between each grid cell and its adjacent grid cells and the grid spacing. Obtain the combustible density of each grid cell from the combustible distribution data; Based on the fire spread rate and key meteorological factors matched to the unified grid scale output of the dynamic fire intensity field during the forecast period, the meteorological diffusion parameters of each grid cell are calculated; The fire risk energy value of each grid cell is obtained by weighted fusion of the temperature gradient, combustible density and meteorological diffusion parameters of each grid cell; The fire danger energy value of each grid cell is compared with the core melting zone threshold value obtained by analyzing historical fire scene data at the same grid size. When the fire danger energy value is greater than the core melting zone threshold value, the grid cell is marked as an open fire unit. Extract the slope continuity, aspect consistency, and surface permeability of each grid cell from the dynamic fire intensity field during the forecast period, calculate the terrain connectivity index, and select grid cells with a terrain connectivity index greater than a preset threshold. Based on the spatial topological relationship of the eight-neighborhood, adjacent grid cells with a terrain connectivity index greater than a preset threshold are restrictedly merged to generate connected open fire areas.

10. The method for controlling a swarm of firefighting drones according to claim 9, wherein: The method for obtaining the secondary dispatchable drone group is as follows: Calculate the shortest Euclidean distance and path weighted distance from the real-time location coordinates of each drone in the candidate set of dispatchable drones to the open fire area; The cruising speed is extracted from the drone swarm status data and combined with the shortest Euclidean distance to calculate the theoretical arrival time. The expected energy consumption is estimated based on the weighted distance of the path. Filter out drones whose theoretical arrival time does not exceed the preset fire rescue time threshold, whose expected energy consumption does not exceed the preset maximum consumable threshold, and whose path weighted distance does not exceed the preset maximum cruising range threshold, and obtain a drone group that can be dispatched secondary.

Images