Civil air defense emergency commanding and dispatching method and system based on unmanned aerial vehicle

Abstract

The invention discloses a civil air defense emergency command and dispatch method and system based on an unmanned aerial vehicle, and relates to the technical field of emergency command and control. According to the invention, data fusion is carried out through Kalman filtering to obtain a three-dimensional disaster situation entropy map; a disaster core area is identified based on an entropy threshold value, a dynamic weight centroid is calculated, and adaptive task partitioning is realized through a weighted Voronoi graph; a millimeter wave radar and binocular vision are fused to construct a three-dimensional grid map, a safe air route is generated through heuristic sampling and air route smoothing under flight constraints, and a collision avoidance mechanism is integrated to guarantee navigation safety; dynamically selecting a communication link, and matching the similarity of historical scheduling schemes or generating a scheduling scheme through a GRU model; optimizing the objective function, outputting an optimal scheduling instruction in combination with endurance and material matching constraints, and triggering closed-loop dynamic re-partitioning through periodic entropy re-evaluation until the disaster situation is relieved; a traditional one-way command mode is broken through, a perception-decision-execution full-chain closed loop is formed, and rescue timeliness and safety are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of emergency command and control, and particularly relates to an air defense emergency command and dispatch method and system based on a UAV. BACKGROUND

[0002] With the acceleration of urbanization and the densification of high-risk industrial facilities, emergency rescue of complex disasters (such as chemical explosion, nuclear leakage associated fire) faces severe challenges. Traditional disaster assessment relies on manual reconnaissance, which has problems such as data lag, limited coverage, and large subjective errors, which can easily lead to secondary casualties. Especially in the multi-factor coupled scene of toxic gas diffusion and building collapse, real-time global situation awareness capability is needed for rescue force dispatching and path planning. Current satellites and UAVs can provide multi-source data, but lack a cross-platform dynamic fusion mechanism, and existing partition algorithms are difficult to adapt to disaster entropy mutations, and the efficiency of cluster collaboration is limited by the static task allocation model.

[0003] In the prior art, air defense emergency relies on fixed monitoring points and manual patrol, which has problems such as insufficient coverage of blind areas and high response delay; a single UAV only has video return function, lacks multi-source data fusion capability and autonomous collaboration mechanism; the command system relies on preset plans and cannot dynamically adapt to disaster site mutations; communication links are easily subject to electromagnetic interference or physical damage, resulting in command interruption. SUMMARY

[0004] (I) Technical problems solved

[0005] To solve the problems in the related art, the present application provides an air defense emergency command and dispatch method based on a UAV to overcome the above technical problems existing in the prior art.

[0006] (II) Technical solutions

[0007] To solve the above technical problems, the present application is realized by the following technical solutions:

[0008] S1, taking fire intensity, toxic gas concentration and building damage degree as state variables, based on Kalman filtering, satellite data, UAV data and ground data are fused to obtain a real-time disaster entropy map;

[0009] S2, based on the entropy threshold and the UAV capability matrix set, the real-time disaster entropy map is partitioned by calculating the weight centroid and boundary conflict detection to obtain an adaptive task area;

[0010] The UAV is allocated to the adaptive task area to obtain the allocated task area;

[0011] S3, based on the allocated task area, real-time environment modeling and route planning operations are performed to obtain a final route;

[0012] S4, collecting real-time disaster data according to a current network state; if there is historical disaster data with a similarity greater than a similarity threshold to the real-time disaster data, selecting a scheduling scheme of the historical disaster as a UAV macro-scheduling scheme; otherwise, obtaining a UAV macro-scheduling scheme through a GRU UAV macro-scheduling scheme model;

[0013] S5, constructing a target function according to the UAV macro-scheduling scheme, a final flight path and a constraint condition; finding a scheduling instruction to minimize the target function to obtain an optimal scheduling instruction; and executing the UAV macro-scheduling scheme through the optimal scheduling instruction until the disaster ends;

[0014] The application generates a three-dimensional disaster entropy map by fusing satellite, UAV and ground data; dynamically partitions high-entropy areas by the entropy value and UAV capability, and finely surveys the high-entropy areas by explosion-proof UAVs; then constructs a navigation map planning safe flight path by millimeter wave radar and binocular vision; adaptively switches through a communication link, and generates a scheduling instruction by multiplexing a historical scheme or a GRU model; finally, optimizes a target function by a sparrow algorithm to form a perception-decision-execution closed loop, breaks through the traditional one-way command mode, and comprehensively improves the timeliness and safety of golden rescue.

[0015] Preferably, the S1 comprises the following steps:

[0016] S11, obtaining satellite data through satellite remote sensing;

[0017] Obtaining UAV data by deploying a UAV cluster with multiple sensors to collect data;

[0018] Obtaining ground data by reporting positioning information through a handheld terminal of a ground rescue personnel;

[0019] S12, converting the satellite data, the UAV data and the ground data into a standard UTM grid to obtain the satellite data, the UAV data and the ground data in a unified coordinate system;

[0020] Configuring a confidence weight set for the satellite data, the UAV data and the ground data in the unified coordinate system to obtain a confidence weight set; the confidence weight set comprises a confidence of the satellite data, the UAV data and the ground data in the unified coordinate system;

[0021] Establishing a state space model; the state space model takes fire intensity, toxic gas concentration and building damage degree as state variables;

[0022] Obtaining a real-time disaster entropy map by a dynamic weighted fusion algorithm based on Kalman filtering, a confidence distribution set and a state space model; the real-time disaster entropy map divides a disaster location into multiple grids, and each grid comprises a three-dimensional entropy value;

[0023] The present application generates a grid disaster entropy map by combining Kalman filtering with fire, toxic gas and building damage three-dimensional state variables through unified coordinate conversion and dynamic confidence weight distribution, thereby improving multi-source data collaborative accuracy and realizing quantitative and accurate positioning of disaster risks.

[0024] Preferably, the S2 comprises the following steps:

[0025] S21, loading the endurance, payload and sensor type of each currently available unmanned aerial vehicle to obtain an unmanned aerial vehicle capability matrix set; the unmanned aerial vehicle capability matrix set comprises the data of the endurance, payload and sensor type of each currently available unmanned aerial vehicle;

[0026] S22, setting an entropy threshold value; taking the continuous grids in the real-time disaster entropy map that exceed the entropy threshold value as a disaster core area to obtain a disaster core area set;

[0027] Matching the endurance, payload and sensor type in each unmanned aerial vehicle capability matrix in the unmanned aerial vehicle capability matrix set to obtain an unmanned aerial vehicle capability coefficient set;

[0028] Taking the entropy value of each disaster core area in the disaster core area set as a gravitational weight, combining the unmanned aerial vehicle capability coefficient set, and attracting the available unmanned aerial vehicles to move to the disaster core area with a greater gravitational weight to obtain a dynamic weight centroid;

[0029] S23, taking the dynamic weight centroid as a base point, performing polygon region division on the real-time disaster entropy map to obtain a polygon region set; the polygon region set comprises a high-entropy region and a low-entropy region;

[0030] Resubdividing the high-entropy region in the polygon region set to obtain a reduced high-entropy region set; and enlarging the low-entropy region in the polygon region set to obtain an enlarged region set;

[0031] Optimizing the region boundary of each region in the reduced high-entropy region set and the enlarged region set to obtain an adaptive task region;

[0032] Setting a task priority rule; according to the task priority rule, assigning a polygon region in the adaptive task region to each unmanned aerial vehicle to obtain an assigned task region;

[0033] The present application identifies a disaster core area based on an entropy threshold value, generates a dynamic weight centroid in combination with an unmanned aerial vehicle capability coefficient, drives a weighted Voronoi diagram partition: finely divides a high-entropy region (improving reconnaissance density), merges and enlarges a low-entropy region (optimizing transportation efficiency), and realizes adaptive task allocation through a boundary negotiation mechanism, thereby improving unmanned aerial vehicle resource matching accuracy and response speed.

[0034] Preferably, the S3 comprises the following steps:

[0035] S31, obtain an obstacle point cloud map through millimeter wave radar penetration smoke detection, and obtain fine obstacle data through binocular vision fine identification;

[0036] Superimpose the obstacle point cloud map data and the fine obstacle data to obtain superimposed map data; add a danger level label in the superimposed map data to obtain a three-dimensional grid map with a label added;

[0037] S32, set the current position of the unmanned aerial vehicle as a starting point, the center point of the polygon region in the optimized region set as an ending point, and set the flight constraint condition of the unmanned aerial vehicle to obtain route initialization data;

[0038] Based on the three-dimensional grid map with a label added and the route initialization data, perform heuristic sampling, dynamic step adjustment and route smoothing optimization operation to obtain a final route;

[0039] S33, formulate a cluster collision avoidance mechanism; and perform disaster relief according to the final route based on the cluster collision avoidance mechanism;

[0040] The application constructs a three-dimensional grid map with a danger label by fusing millimeter wave radar (penetrating smoke) and binocular vision (identifying small obstacles), and generates a safe route under flight constraints by combining heuristic sampling and route smoothing optimization algorithm; the cluster collision avoidance mechanism realizes second-level response, improves the safety of complex environment navigation, and multiplies the efficiency of rescue.

[0041] Preferably, the S4 comprises the following steps:

[0042] S41, set a communication link selection strategy; collect current network state data, and obtain a real-time communication link according to the current network state data in combination with the communication link selection strategy;

[0043] S42, collect real-time disaster data through the real-time communication link; set a disaster data similarity threshold; collect a large number of historical unmanned aerial vehicle scheduling schemes to obtain a historical unmanned aerial vehicle scheduling scheme set; match the disaster data in the historical unmanned aerial vehicle scheduling scheme set with the real-time disaster data, and use the corresponding historical unmanned aerial vehicle scheduling scheme as the unmanned aerial vehicle scheduling scheme when the similarity is greater than or equal to the disaster data similarity threshold;

[0044] When the similarity is greater than or equal to the disaster data similarity threshold, input the real-time disaster data into a GRU unmanned aerial vehicle scheduling scheme model to obtain an unmanned aerial vehicle scheduling scheme;

[0045] The application guarantees data transmission through a communication link dynamic switching strategy, and generates a new scheduling instruction based on a disaster similarity threshold matching a historical scheduling scheme or a GRU model, thereby improving decision efficiency and compressing resource allocation response time.

[0046] Preferably, the scheduling scheme model in S42 is realized by the following steps:

[0047] S421, a GRU model is constructed, and a learning rate of the GRU model training is set;

[0048] S422, historical disaster data and a UAV scheduling scheme of the historical disaster data are collected to obtain historical data;

[0049] S423, the GRU model is trained using the historical data, a genetic algorithm is used to find an optimal learning rate of the GRU model in the training process, and an optimal solution is obtained; the optimal solution is used as the learning rate of the GRU model, and a GRU UAV macro scheduling scheme model is obtained;

[0050] The application optimizes the learning rate of the GRU model by using the genetic algorithm, iteratively screens the optimal solution by constructing a chromosome population, trains the scheduling model with the historical disaster data, improves the prediction accuracy, compresses the scheduling scheme generation time, and significantly enhances the timeliness and reliability of complex disaster decision-making.

[0051] Preferably, the training process in S423 uses a genetic algorithm to find an optimal learning rate of the GRU model to obtain an optimal solution, which includes the following steps:

[0052] S4231, a model accuracy threshold of the GRU model is set as f1, and a model accuracy is set as f2;

[0053] S4232, a chromosome population is constructed, a size of the chromosome population is set, a maximum number of iterations of training is set as h1, and a current number of iterations of training is set as h2;

[0054] S4233, the learning rate in the chromosome population is substituted into the GRU model, the GRU model is trained using the historical data, a model accuracy is obtained, and the chromosomes in the chromosome population are subjected to crossover and mutation operations according to the model accuracy, to obtain a chromosome population subjected to the crossover and mutation operations;

[0055] S4234, S4233 is repeated, and when f2 is greater than or equal to f1 or h2 is greater than or equal to h1, the iteration is stopped, and an optimal solution is obtained;

[0056] The application adaptively screens an optimal learning rate of the GRU by using the genetic algorithm, iteratively performs crossover and mutation operations with the chromosome population, outputs the optimal solution when the accuracy threshold or the maximum number of iterations is reached, improves the model training efficiency, improves the scheduling decision accuracy, and effectively avoids the blindness of manual parameter adjustment.

[0057] Preferably, the S5 comprises the following steps:

[0058] S51, set the constraint conditions as the endurance of the unmanned aerial vehicle, the matching of the material demand and the path safety coefficient; construct a target function according to the unmanned aerial vehicle scheduling scheme combined with the constraint conditions, and the target function is to minimize the response time and minimize the resource consumption;

[0059] S52, find the scheduling instruction to minimize the target function combined with the sparrow population algorithm, and obtain the optimal scheduling instruction;

[0060] S53, the unmanned aerial vehicle executes the scheduling scheme through the optimal scheduling instruction; set the collection period; reevaluate the disaster entropy value under each collection period, and when the disaster entropy value changes, return to S1 to trigger the re-partitioning of the whole system until the disaster entropy value is lower than the disaster entropy value threshold;

[0061] The application optimizes the scheduling instruction based on the sparrow population algorithm, minimizes the response time and resource consumption under the constraints of endurance, material matching and the like, triggers the dynamic closed-loop feedback through the periodic entropy reevaluation, makes the system adapt to the disaster evolution, improves the rescue efficiency and maximizes the resource utilization.

[0062] Preferably, the S52 comprises the following steps:

[0063] S521, construct a sparrow population, set the size of the sparrow population, and set the maximum optimization iteration number;

[0064] According to the scheduling instruction, the initial positions of the sparrow population are randomly set, and the initial position set of the sparrow population is obtained;

[0065] S522, perform an iteration operation on the initial position set of the sparrow population, the smaller the target function is, the better the position is; in each iteration process, the target function value of each position in the initial position set of the sparrow population is calculated according to the target function, the positions of each sparrow in the initial position set of the sparrow population are updated from high to low according to the target function value, and the best sparrow individual position in the sparrow population and the global best sparrow position are obtained in each iteration process;

[0066] S523, repeat S522, when the maximum optimization iteration number is reached, stop the iteration, and take the global best sparrow position as the optimal scheduling instruction;

[0067] The application efficiently optimizes the scheduling instruction through the sparrow population algorithm; the sparrow position iteration update is guided by the target function value, the global optimal solution is quickly converged within the maximum iteration number, the scheduling scheme generation speed is improved, the resource allocation is highly matched, and the disaster relief response accuracy is significantly strengthened.

[0068] The application discloses an emergency command and dispatching system based on a UAV (Unmanned Aerial Vehicle), which is used for realizing the emergency command and dispatching method based on the UAV, and comprises a multi-source disaster information sensing and fusion module, an adaptive task partitioning and UAV distribution module, a three-dimensional environment modeling and route planning module, an intelligent dispatching decision and communication link management module and a dynamic optimization and closed-loop response module.

[0069] The multi-source disaster information sensing and fusion module is used for collecting fire, toxic gas concentration and building damage data through satellite remote sensing, a UAV cluster and a ground terminal; after unifying a coordinate system, multi-source data is dynamically fused based on a confidence weight and a Kalman filter to generate a real-time disaster information entropy map; and the map quantifies disaster risks in a grid three-dimensional entropy value.

[0070] The adaptive task partitioning and UAV distribution module is used for identifying a disaster core area and calculating a dynamic weight center in combination with a UAV capability matrix and an entropy value threshold; a weighted Voronoi diagram is used to divide a polygon task area, a high-entropy area is finely split to ensure intensive reconnaissance, and a low-entropy area is combined and expanded; a boundary conflict detection and negotiation mechanism is used to optimize the partitioning, and finally, UAVs are distributed to adaptive areas according to a task priority rule.

[0071] The three-dimensional environment modeling and route planning module is used for constructing a three-dimensional grid map with a danger label by fusing a millimeter wave radar and a binocular vision; a smooth route is generated by heuristic sampling and dynamic step adjustment under flight constraints, with the starting point being the current position of the UAV and the ending point being the center of the task area, and a cluster collision avoidance mechanism is integrated.

[0072] The intelligent dispatching decision and communication link management module is used for collecting real-time disaster information data by dynamically selecting a communication link, matching a historical dispatching scheme similarity, directly reusing the historical scheme if the similarity exceeds a threshold, otherwise generating a new scheme by a GRU dispatching model; the GRU model uses a genetic algorithm to optimize a learning rate, is trained by historical disaster data, and outputs a UAV macroscopic dispatching instruction.

[0073] The dynamic optimization and closed-loop response module takes minimizing a response time and resource consumption as an objective function, combines constraints to build an optimization model, searches for an optimal dispatching instruction by using a sparrow swarm algorithm, iteratively updates a position to minimize the objective function, and outputs a global optimal solution; after the UAV executes the instruction, the disaster information entropy is re-evaluated at a period, and if the entropy value changes, the whole system is re-partitioned until the entropy value is lower than a safety threshold.

[0074] (Three) beneficial effects

[0075] The application has the following beneficial effects:

[0076] The present application generates a three-dimensional disaster entropy map by fusing satellite, unmanned aerial vehicle and ground data; based on entropy value and unmanned aerial vehicle capability dynamic partition, the high-entropy area is finely surveyed by explosion-proof unmanned aerial vehicle; then a navigation map is constructed by millimeter wave radar and binocular vision to plan a safe flight path; through adaptive switching of communication link, historical scheme or GRU model is reused to generate scheduling instructions; finally, the target function is optimized by sparrow algorithm to form a closed loop of perception-decision-execution, break through the traditional one-way command mode, and comprehensively improve the timeliness and safety of golden rescue.

[0077] The present application realizes multi-dimensional breakthrough in the field of emergency rescue; through Kalman filtering of multi-source heterogeneous data, the information island problem in traditional disaster monitoring is effectively solved; satellite data is easily disturbed by clouds, positioning deviation caused by inconsistency between local coordinate system of unmanned aerial vehicle and UTM projection on the ground, and subjective error of artificial reporting; the system dynamically allocates reliability for satellite, unmanned aerial vehicle and ground data to generate a three-dimensional entropy map, improves the quantitative accuracy of disaster, and provides a high reliability basis for rescue decision.

[0078] The present application creates an adaptive partition mechanism based on gravitational weight for the problems of disaster dynamic evolution and mismatch of unmanned aerial vehicle capability; after the disaster core area is determined by entropy threshold, the dynamic weight centroid is generated by combining the unmanned aerial vehicle capability coefficient to drive the weighted Voronoi diagram partition; fine splitting of high-entropy area and merging and expansion of low-entropy area are realized; boundary conflict detection and negotiation mechanism reduces the time consumption of partition adjustment; unmanned aerial vehicle type and response time are matched according to disaster grade, which improves the reconnaissance response speed of explosion risk area.

[0079] In the aspect of flight path planning in complex environment, the present application constructs a three-dimensional grid map by fusing millimeter wave radar and binocular vision, and solves the problems of smoke obstruction and small obstacle recognition; millimeter wave radar penetrates smoke to detect large obstacles, binocular camera recognizes small threats such as high-voltage cables, and generates a navigation map with danger labels; combined with heuristic sampling and dynamic step optimization algorithm and constraint mechanism, the smoothness of flight path is improved, and the response time of obstacle avoidance is shortened; cluster collision avoidance mechanism ensures vertical safety interval and reduces collision risk.

[0080] The present application constructs an intelligent decision-making closed loop system to significantly optimize resource scheduling efficiency; communication link dynamic switching strategy guarantees the reachability of instructions; after the learning rate of GRU scheduling model is optimized by genetic algorithm, the historical scheme reuse rate for similar disaster is improved, and the scheduling scheme is more accurate; sparrow population algorithm takes minimizing response time and resource consumption as the goal, searches for the optimal scheduling instruction under the constraints of endurance and material matching, reduces the overall rescue resource consumption, and triggers dynamic repartition through periodic entropy reevaluation, forms a "perception-decision-execution-optimization" closed loop, and the disaster entropy value is reduced to the safety threshold.

[0081] Of course, implementing any product of the application does not necessarily require achieving all the advantages described above at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0082] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings described below only show some of the embodiments of the application, and for those skilled in the art, other drawings can be obtained from these drawings without creative labor.

[0083] Figure 1 A flowchart of a civil air defense emergency command and dispatch method based on a UAV according to the present application;

[0084] Figure 2 A module diagram of a civil air defense emergency command and dispatch system based on a UAV according to the present application. DETAILED DESCRIPTION

[0085] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, not all the embodiments. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0086] In the description of the present application, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inner" and the like indicate the orientation or positional relationship, which are only for the convenience of describing the application and simplifying the description, and do not indicate or imply that the indicated component or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the application.

[0087] Embodiment one:

[0088] Please refer to Figure 1 The present application discloses a civil air defense emergency command and dispatch method based on a UAV, comprising the following steps:

[0089] S1, taking fire intensity, toxic gas concentration and building damage degree as state variables, based on Kalman filtering, satellite data, UAV data and ground data are fused to obtain a real-time disaster entropy map;

[0090] The S1 comprises the following steps:

[0091] S11, satellite remote sensing is used to obtain satellite data (fire point distribution, road damage); the satellite remote sensing identifies the fire point distribution (temperature > 300℃ area is marked as red) through the infrared band, and analyzes the road damage condition through the visible light band.

[0092] The unmanned aerial vehicle cluster is deployed to collect data by multiple sensors, and unmanned aerial vehicle data is obtained; the multiple sensors include a multispectral camera, a laser methane detector, and a gamma ray sensor; the multispectral camera simultaneously collects visible light and infrared images (wavelength 400-1700 nm) to identify open fires and hidden fire sources, the laser methane detector scans combustible gas leaks (detection accuracy 1 ppm, distance 100 m), and the gamma ray sensor monitors nuclear radiation intensity (range 0.1-100 μSv / h, automatically alarms when exceeding the standard);

[0093] The ground rescue personnel report positioning information through a handheld terminal to obtain ground data; the terminal device is a Beidou / GPS dual-mode positioning, and the ground data includes the location of trapped personnel, information of on-site dangerous sources (such as chemical barrel leakage), and video / picture supplementary evidence;

[0094] S12, the satellite data (WGS84 coordinate system), unmanned aerial vehicle data (local rectangular coordinate system), and ground data (UTM projection) are converted to a standard UTM grid to obtain satellite data, unmanned aerial vehicle data, and ground data in a unified coordinate system;

[0095] The satellite data, unmanned aerial vehicle data, and ground data in the unified coordinate system are assigned a confidence weight to obtain a confidence weight set; the confidence weight set includes the confidence of the satellite data, unmanned aerial vehicle data, and ground data in the unified coordinate system; for example, the weight of the satellite data is 0.7 (easily affected by cloud interference), the weight of the unmanned aerial vehicle data is 0.9 (high accuracy at close range), and the weight of the ground data is 0.8 (may have subjective errors);

[0096] A state space model is established; the state space model takes fire intensity, toxic gas concentration, and building damage degree as state variables;

[0097] A real-time disaster condition entropy value map is obtained by combining the confidence distribution set and the state space model based on a dynamic weighted fusion algorithm of Kalman filtering; the real-time disaster condition entropy value map divides the disaster condition location into multiple grids, and each grid includes a three-dimensional entropy value; the three-dimensional entropy value includes a fire entropy value (0-1.0, >0.8 indicates an explosion risk), a toxic gas entropy value (0-1.0, >0.7 reaches a lethal concentration), and a structure entropy value (0-1.0, >0.9 indicates collapse);

[0098] S2, based on an entropy value threshold and an unmanned aerial vehicle capability matrix set, a weight centroid and boundary conflict detection are calculated to divide the real-time disaster condition entropy value map to obtain an adaptive task area;

[0099] Unmanned aerial vehicles are assigned to the areas in the adaptive task area to obtain the assigned task area;

[0100] S2 includes the following steps:

[0101] S21, load the current available UAVs of the UAVs' endurance (minutes, such as 120 min), payload (kg, such as 5 kg gas mask box), and sensor types (such as toxic gas detector, life detector), to obtain a UAV capability matrix set; the UAV capability matrix set contains the data of the endurance, payload, and sensor types of each current available UAV;

[0102] S22, set an entropy value threshold; take the continuous grids in the real-time disaster entropy value map that exceed the entropy value threshold as the disaster core area, to obtain a disaster core area set;

[0103] Match the endurance, payload, and sensor types in each UAV capability matrix in the UAV capability matrix set to obtain a UAV capability coefficient set; for example, the coefficient of an explosion-proof UAV in a toxic gas area = 1.2, and the coefficient of a common UAV = 0.8;

[0104] Take the entropy value of each disaster core area in the disaster core area set as the gravitational weight, and combine the UAV capability coefficient set to attract the available UAVs to move to the disaster core area with a greater gravitational weight, to obtain a dynamic weight centroid;

[0105] S23, take the dynamic weight centroid as the base point to divide the real-time disaster entropy value map into a polygon area set; the polygon area set includes high-entropy areas and low-entropy areas;

[0106] Subdivide the high-entropy areas in the polygon area set to obtain a reduced high-entropy area set; ensure fine reconnaissance, and enlarge the low-entropy areas in the polygon area set to obtain an enlarged area set;

[0107] Optimize the region boundary of each area in the reduced high-entropy area set and the enlarged area set to obtain an adaptive task area; detect the overlap or gap (such as triggering a conflict when the boundary distance of two UAVs is less than 50 m); adjust the boundary through a negotiation mechanism: the UAV with strong capability takes over the high-entropy conflict area, and the UAV with weak capability takes over the adjacent low-entropy area;

[0108] Set a task priority rule; according to the task priority rule, assign each UAV a polygon area in the adaptive task area, to obtain the assigned task area;

[0109] Flowchart: A [entropy value map], B [calculate weight centroid], C [generate weighted Voronoi diagram], D [region boundary optimization], E [conflict detection], F {conflict?}, Yes > G [redistribute boundary], F No > H [generate final partition];

[0110] If the task priority rule includes disaster level, response priority and UAV type requirement; if red (entropy value > 0.9), immediate response, explosion-proof + toxic gas detection; yellow (0.6-0.9, response within 5 minutes, standard reconnaissance type; green (<0.6), response within 10 minutes, transportation type;

[0111] S3, based on the allocated task area, real-time environment modeling and route planning operation are performed to obtain a final route;

[0112] The S3 includes the following steps:

[0113] S31, obstacle point cloud data is obtained by penetrating smoke detection through a millimeter wave radar; the UAV starts a 77GHz millimeter wave radar (with the ability to penetrate smoke / dust), emits a frequency-modulated continuous wave (FMCW) at a frequency of 10Hz, scans a radius of 300 meters, marks large obstacles such as building debris and vehicles (positioning accuracy ±0.5 meters), and generates obstacle point cloud data;

[0114] Fine obstacle data is obtained by fine identification through binocular vision; a binocular camera with a baseline distance of 12cm (2000 pixels) is turned on, a stereo matching algorithm is executed, a left camera identifies small obstacles such as high-voltage cables and steel bars, and a right camera calculates the depth of the obstacles (accuracy ±5cm), and dynamic obstacles such as moving vehicles and falling objects are specially marked;

[0115] The obstacle point cloud data and the fine obstacle data are superimposed to obtain superimposed map data; a danger level label is added to the superimposed map data to obtain a three-dimensional grid map with labels added; for example, a red grid is a prohibited entry area (such as a toxic gas diffusion area), a yellow grid is a high-risk area (less than 5 meters from the obstacle), and a green grid is a safe passage;

[0116] S32, the current position of the UAV is taken as a starting point, the center point of the polygon area in the optimized regional set is taken as an end point, and UAV flight constraint conditions (maximum climb angle 30°, minimum turning radius 15 meters) are set to obtain route initialization data;

[0117] Based on the three-dimensional grid map with labels added and the route initialization data, heuristic sampling, dynamic step adjustment and route smoothing optimization operations are performed to obtain a final route;

[0118] S33, a cluster collision avoidance mechanism is developed; for example, based on the interval keeping of ADSB broadcast (horizontal > 50m, vertical > 30m), sudden obstacles trigger emergency hovering (response time <0.2s);

[0119] Based on the cluster collision avoidance mechanism, disaster relief is performed according to the final route;

[0120] S4, collecting real-time disaster data according to the selected communication link according to the current network state; if there is historical disaster data with a similarity greater than a similarity threshold to the real-time disaster data, selecting the dispatching scheme of the historical disaster as the unmanned aerial vehicle macroscopic dispatching scheme; otherwise, obtaining the unmanned aerial vehicle macroscopic dispatching scheme through a GRU unmanned aerial vehicle macroscopic dispatching scheme model;

[0121] The S4 includes the following steps:

[0122] S41, setting a communication link selection strategy; when the network state is 5G signal > 90 dBm, the first selected link is a 5G private network, the backup link is a mesh self-organizing network, and the emergency link is a Beidou short message; when the network state is G signal ≤ 90 dBm, the first selected link is a mesh self-organizing network, the backup link is LoRa, and the emergency link is Beidou; when electromagnetic interference is received, the first selected link is frequency hopping communication, the backup link is Beidou, and the emergency link is LoRa;

[0123] Collecting current network state data, and obtaining real-time communication links according to the current network state data and the communication link selection strategy;

[0124] S42, collecting real-time disaster data through the real-time communication link; setting a disaster data similarity threshold; collecting a large number of historical unmanned aerial vehicle dispatching schemes to obtain a historical unmanned aerial vehicle dispatching scheme set; matching the disaster data in the historical unmanned aerial vehicle dispatching scheme set with the real-time disaster data, and using the corresponding historical unmanned aerial vehicle dispatching scheme as the unmanned aerial vehicle dispatching scheme when the similarity is greater than or equal to the disaster data similarity threshold;

[0125] When the similarity is greater than or equal to the disaster data similarity threshold, inputting the real-time disaster data into a GRU unmanned aerial vehicle dispatching scheme model to obtain the unmanned aerial vehicle dispatching scheme; for example, using an unmanned aerial vehicle to deliver 8 respirators to area A;

[0126] The dispatching scheme model in the S42 is implemented through the following steps:

[0127] S421, constructing a GRU model and setting a learning rate for training the GRU model;

[0128] S422, collecting historical disaster data and unmanned aerial vehicle dispatching schemes for the historical disaster data to obtain historical data;

[0129] S4233, training the GRU model using the historical data, finding an optimal learning rate of the GRU model using a genetic algorithm during the training process, obtaining an optimal solution, and using the optimal solution as the learning rate of the GRU model to obtain a GRU unmanned aerial vehicle macroscopic dispatching scheme model;

[0130] The step of finding an optimal learning rate of the GRU model using a genetic algorithm during the training process in the S53 to obtain an optimal solution includes the following steps:

[0131] S341, set the model accuracy threshold of the GRU model as f1, and the model accuracy as f2;

[0132] S342, construct a chromosome population, set the size of the chromosome population as e, and the chromosome population is represented as d={d1, d2,..., d i ,...,d e}, wherein q i represents the i-th chromosome in the chromosome population, each chromosome represents a learning rate of the GRU model, set the maximum number of iterations of training as h1, and the current number of iterations of training as h2;

[0133] S343, substitute the learning rate in the chromosome population into the GRU model, and train the GRU model using historical data to obtain a model accuracy, and according to the model accuracy, perform crossover and mutation operations on the chromosomes in the chromosome population to obtain a chromosome population subjected to the crossover and mutation operations;

[0134] S344, repeat S343, and when f2 is greater than or equal to f1 or h2 is greater than or equal to h1, stop iteration to obtain an optimal solution;

[0135] S5, construct a target function according to the unmanned aerial vehicle macroscopic scheduling scheme, the final flight path and the constraint condition; find a scheduling instruction to minimize the target function to obtain an optimal scheduling instruction; and execute the unmanned aerial vehicle macroscopic scheduling scheme through the optimal scheduling instruction until the disaster ends;

[0136] The S5 includes the following steps:

[0137] S51, set the constraint condition as the endurance of the unmanned aerial vehicle, the matching of the material demand and the path safety factor; construct a target function according to the unmanned aerial vehicle scheduling scheme combined with the constraint condition, and the target function is to minimize the response time and minimize the resource consumption; the target function is as follows,

[0138] F=w1T+w2C;

[0139] wherein F represents the target function, w1 and w2 represent the set weight of the response time and the weight of the resource consumption respectively, and T and C represent the response time and the resource consumption respectively;

[0140] S52, find a scheduling instruction to minimize the target function combined with the sparrow population algorithm to obtain an optimal scheduling instruction;

[0141] The S52 includes the following steps:

[0142] S521, construct a sparrow population, set the size of the sparrow population as p, and the sparrow population is represented as p={p1, p2,..., p i ,...,p q}, wherein p i represents the i-th sparrow in the sparrow population; set the maximum number of optimization iterations;

[0143] According to the scheduling instruction, the initial position of the sparrow population is randomly set, and the initial position set of the sparrow population is obtained as f={f1, f2,..., f i ,...,f d}, wherein f i represents the position of the i-th sparrow in the sparrow population, that is, the i-th scheduling instruction;

[0144] S522, iteration operation is performed on the initial position set of the sparrow population, and the smaller the target function is, the better the position is; in each iteration process, the target function value of each position in the initial position set of the sparrow population is calculated according to the target function, the position of each sparrow in the initial position set of the sparrow population is updated according to the target function value from high to low, and the best sparrow individual position in the sparrow population and the global best sparrow position are obtained in each iteration process;

[0145] S523, repeat S522, when the maximum number of optimization iterations is reached, stop iteration, and take the global best sparrow position as the optimal scheduling instruction;

[0146] S53, the unmanned aerial vehicle executes the scheduling scheme through the optimal scheduling instruction; set the collection period; reevaluate the disaster entropy value under each collection period, and return to S1 when the disaster entropy value changes, triggering the whole system to rezone until the disaster entropy value is lower than the minimum risk threshold.

[0147] Embodiment two:

[0148] Please refer to Figure 2 , a civil air defense emergency command scheduling system based on unmanned aerial vehicles, for realizing the above-mentioned civil air defense emergency command scheduling method based on unmanned aerial vehicles, comprising a multi-source disaster perception and fusion module, an adaptive task zoning and unmanned aerial vehicle allocation module, a three-dimensional environment modeling and route planning module, an intelligent scheduling decision and communication link management module, and a dynamic optimization and closed-loop response module;

[0149] The multi-source disaster perception and fusion module is used to collect fire, toxic gas concentration, and building damage data through satellite remote sensing, unmanned aerial vehicle cluster, and ground terminal; after unifying the coordinate system, the multi-source data is dynamically fused based on confidence weight and Kalman filter to generate a real-time disaster entropy value map; the map quantifies disaster risk in a grid three-dimensional entropy value;

[0150] The adaptive task partitioning and UAV allocation module is used for identifying a disaster core area and calculating a dynamic weight centroid in combination with a UAV capability matrix and an entropy threshold value; a polygon task area is divided based on a weighted Voronoi diagram, a high-entropy area is finely split to ensure intensive reconnaissance, and a low-entropy area is merged and expanded; the partitioning is optimized through a boundary conflict detection and negotiation mechanism, and finally the UAVs are allocated to the adaptive areas according to a task priority rule;

[0151] The three-dimensional environment modeling and route planning module is used for constructing a three-dimensional grid map with a danger label by fusing a millimeter wave radar and a binocular vision; a smooth route is generated by heuristic sampling and dynamic step adjustment under flight constraints, with the current position of the UAV as a starting point and the center of the task area as an ending point, and a swarm collision avoidance mechanism is integrated;

[0152] The intelligent scheduling decision and communication link management module collects real-time disaster data by dynamically selecting a communication link, matches a historical scheduling scheme similarity, and directly reuses the historical scheme if the similarity exceeds a threshold value; otherwise, a new scheme is generated by a GRU scheduling model; the GRU model uses a genetic algorithm to optimize a learning rate, is trained by historical disaster data, and outputs a UAV macro-scheduling instruction;

[0153] The dynamic optimization and closed-loop response module constructs an optimization model in combination with constraints, taking minimization of response time and resource consumption as an objective function; a sparrow swarm algorithm is used to search for an optimal scheduling instruction, the sparrow position is initialized, and the position is iteratively updated to minimize the objective function, and a global optimal solution is output; after the UAV executes the instruction, the disaster entropy is re-evaluated at a period, and if the entropy changes, the whole system is re-partitioned until the entropy is lower than a safety threshold value.

[0154] In the description of the present specification, the description of the terms "one embodiment", "an example", "a specific example" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the invention. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0155] The preferred embodiments of the above disclosed invention are only used to help explain the invention. The preferred embodiments do not describe all the details and limit the invention to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the present specification. The embodiments are selected and described in detail in the present specification in order to better explain the principles and practical applications of the invention, so that those skilled in the art can well understand and utilize the invention.

Claims

1. A civil air defense emergency command and dispatch method based on drones, characterized in that: The following steps are involved: S1. Using fire intensity, toxic gas concentration, and building damage as state variables, the Kalman filter is used to fuse satellite data, drone data, and ground data to obtain a real-time disaster entropy map. S2. Based on the entropy threshold and the UAV capability matrix set, the real-time disaster entropy map is partitioned by calculating the weighted centroid and boundary conflict detection to obtain the adaptive mission area; Allocate drones to areas in the adaptive mission area to obtain allocated mission areas; S3. Based on the assigned mission area, perform real-time environment modeling and route planning operations to obtain the final route; S4. Select a communication link based on the current network status to collect real-time disaster data; If there is historical disaster data whose similarity with the real-time disaster data is greater than the similarity threshold, the scheduling plan of the historical disaster is selected as the drone macro scheduling plan; otherwise, the drone macro scheduling plan is obtained through the GRU drone macro scheduling plan model; S5. Construct an objective function based on the UAV macro-scheduling plan, the final route, and the constraints; Find the scheduling instructions to minimize the objective function and obtain the optimal scheduling instructions; execute the drone macro scheduling plan through the optimal scheduling instructions until the disaster ends.

2. The method for civil air defense emergency command and dispatch based on drone according to claim 1, characterized in that: S11. Obtain satellite data through satellite remote sensing; By deploying a multi-sensor drone cluster to collect data, drone data is obtained; Rescue personnel on the ground report positioning information through handheld terminals and obtain ground data; S12, converting the coordinate system of the satellite data, the UAV data, and the ground data to a standard UTM grid to obtain the satellite data, the UAV data, and the ground data in a unified coordinate system; Assign confidence weights to the satellite data, UAV data, and ground data after unifying the coordinate system to obtain a confidence weight set; Build a state-space model; The state space model uses fire intensity, toxic gas concentration and building damage as state variables; A dynamic weighted fusion algorithm based on Kalman filtering is combined with the confidence distribution set and state space model to obtain a real-time disaster entropy map.

3. The method for civil air defense emergency command and dispatch based on drone according to claim 1, characterized in that: The S2 comprises the following steps: S21. Load the flight endurance, payload, and sensor type of the currently available drones to obtain a drone capability matrix set; the drone capability matrix set includes data on the flight endurance, payload, and sensor type of each currently available drone; S22, setting an entropy threshold; taking the continuous grids exceeding the entropy threshold in the real-time disaster entropy map as the disaster core area, and obtaining a disaster core area set; The endurance, payload, and sensor type in each UAV capability matrix in the UAV capability matrix set are matched to obtain the UAV capability coefficient set; The entropy value of each disaster core area in the disaster core area set is used as the gravity weight. Combined with the UAV capability coefficient set, the available UAVs are pulled to move towards the disaster core area with a larger gravity weight to obtain the dynamic weighted centroid. S23, dividing the real-time disaster entropy map into polygonal regions based on the dynamic weight centroid to obtain a polygonal region set; the polygonal region set includes a high entropy region and a low entropy region; The high entropy area in the polygonal area concentration is redivided to obtain a reduced high entropy area set; the low entropy area in the polygonal area concentration is expanded to obtain an expanded area set; Optimize the region boundaries of each region in the reduced high entropy region set and the expanded region set to obtain the adaptive task region; A task priority rule is set; according to the task priority rule, a polygonal area in the adaptive task area is allocated to each UAV to obtain the allocated task area.

4. The method for civil air defense emergency command and dispatch based on drone according to claim 1, characterized in that: The S3 includes the following steps: S31. Use millimeter-wave radar to penetrate smoke and detect obstacles to obtain a point cloud map; use binocular vision to perform fine recognition and obtain fine obstacle data; Overlaying the obstacle point cloud data with the refined obstacle data to obtain overlaid map data; adding a hazard level label to the overlaid map data to obtain a labeled three-dimensional grid map; S32, using the current position of the UAV as the starting point and the center point of the polygonal area where the optimized area is concentrated as the end point, setting the UAV flight constraints, and obtaining the route initialization data; Based on the labeled 3D grid map and route initialization data, heuristic sampling, dynamic step size adjustment, and route smoothing optimization are performed to obtain the final route. S33. Develop a cluster collision avoidance mechanism; based on the cluster collision avoidance mechanism, carry out disaster relief according to the final route.

5. The method for civil air defense emergency command and dispatch based on drone according to claim 1, characterized in that: The S4 comprises the following steps: S41, setting a communication link selection strategy; collecting current network status data, and obtaining a real-time communication link based on the current network status data and the communication link selection strategy; S42. Collect real-time disaster data through a real-time communication link; set a disaster data similarity threshold; collect a large number of historical drone scheduling plans to obtain a historical drone scheduling plan set; match the disaster data in the historical drone scheduling plan set with the real-time disaster data; when the similarity is greater than or equal to the disaster data similarity threshold, use the corresponding historical drone scheduling plan as the drone scheduling plan; When the similarity is ≥ the disaster data similarity threshold, the drone scheduling plan is obtained by inputting the real-time disaster data into the GRU drone scheduling plan model.

6. The method for civil air defense emergency command and dispatch based on drone according to claim 5, characterized in that: The scheduling solution model in S42 is implemented by the following steps: S421, constructing a GRU model and setting a learning rate for the GRU model training; S422, collecting historical disaster data and a UAV dispatching plan for the historical disaster data to obtain historical data; S423. Use historical data to train the GRU model. During the training process, use a genetic algorithm to find the optimal learning rate of the GRU model to obtain the optimal solution. Use the optimal solution as the learning rate of the GRU model to obtain the GRU UAV macro scheduling solution model.

7. The method for civil air defense emergency command and dispatch based on drone according to claim 6, characterized in that: In the training process in S423, the genetic algorithm is used to find the optimal learning rate of the GRU model, and obtaining the optimal solution includes the following steps: S4231. Set the model accuracy threshold of the GRU model to f1 and the model accuracy to f2; S4232, construct a chromosome population, set the size of the chromosome population, set the maximum number of training iterations to h1, and the current number of training iterations to h2; S4233, substituting the learning rate in the chromosome population into the GRU model, and using historical data to train the GRU model to obtain model accuracy, and performing crossover and mutation operations on the chromosomes in the chromosome population based on the model accuracy to obtain a chromosome population subjected to crossover and mutation operations; S4234. Repeat S4233. When f2≥f1 or h2≥h1, stop the iteration and obtain the optimal solution.

8. The method for civil air defense emergency command and dispatch based on drone according to claim 1, characterized in that: The S5 comprises the following steps: S51. Set the constraints as drone endurance, material demand matching, and path safety factor; construct an objective function based on the drone scheduling plan and the constraints, where the objective function is to minimize response time and resource consumption; S52, combining the sparrow population algorithm to find a scheduling instruction to minimize the objective function and obtain the optimal scheduling instruction; S53. The UAV executes the scheduling plan through the optimal scheduling instruction; sets the collection cycle; re-evaluates the disaster entropy value in each collection cycle. When the disaster entropy value changes, it returns to S1 and triggers the re-partitioning of the entire system until the disaster entropy value is lower than the disaster entropy value threshold.

9. The method for civil air defense emergency command and dispatch based on drone according to claim 8, characterized in that: The S52 includes the following steps: S521. Construct a sparrow population and set the size of the sparrow population; set the maximum number of optimization iterations; randomly setting the initial positions of the sparrow population according to the scheduling instruction to obtain an initial position set of the sparrow population; S522, performing an iterative operation on the initial position set of the sparrow population, wherein the smaller the objective function, the better the position; in each round of iteration, calculating the objective function value of each position in the initial position set of the sparrow population according to the objective function, and updating the position of each sparrow in the initial position set of the sparrow population in descending order of the objective function values, and obtaining the best individual sparrow position in the sparrow population and the global best sparrow position in each round of iteration; S523. Repeat S522. When the maximum number of optimization iterations is reached, stop the iteration and use the global best sparrow position as the optimal scheduling instruction.

10. A civil air defense emergency command and dispatch system based on drones, characterized in that: A method for implementing a civil air defense emergency command and dispatch method based on a drone as described in any one of claims 1 to 9, wherein the system includes a multi-source disaster perception and fusion module, an adaptive task partitioning and drone allocation module, a three-dimensional environment modeling and route planning module, an intelligent scheduling decision and communication link management module, and a dynamic optimization and closed-loop response module.

Images