Fire-fighting unmanned aerial vehicle autonomous patrol system and method based on Internet of Things

Abstract

The invention discloses a fire-fighting unmanned aerial vehicle autonomous patrol system and method based on the Internet of Things, and relates to the technical field of intelligent fire fighting, and the method comprises the steps: converging fixed Internet of Things sensors, unmanned aerial vehicle airborne sensor sensing data and unmanned aerial vehicle cluster state data, processing the data into a standardized data set, dividing park monitoring nodes, and calculating a comprehensive risk value; constructing a weighted directed dynamic risk map by combining the node space relationship, the distance and the environment prevailing wind direction; simulating risk propagation based on a graph, identifying key nodes to form a key check set, determining a routine inspection set, and generating a differentiated task set; calculating task priority and execution cost, establishing a collaborative task allocation model with maximized system overall efficiency, solving the collaborative task allocation model, and generating a task allocation scheme; the system comprises corresponding function modules, realizes accurate risk identification and intelligent task distribution, improves the patrol efficiency, and guarantees the fire safety of the park.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent fire-fighting technology, in particular to a fire-fighting unmanned aerial vehicle autonomous patrol system and method based on the Internet of Things. BACKGROUND

[0002] With the expansion of the park scale and the improvement of fire safety demand, the traditional fire-fighting patrol method has been difficult to meet the precise, real-time and efficient protection requirements. The existing park fire-fighting patrol relies on manual inspection combined with fixed Internet of Things sensor monitoring, which has obvious limitations: manual inspection is low in efficiency and limited in coverage, and it is difficult to meet the all-weather patrol demand of large parks, and it is poor in real-time and cannot capture dynamic risks in time; although the fixed sensor can collect data in real time, the monitoring range is fixed, the data is easy to be in the "island" state, and there is a lack of multi-source data integration analysis, resulting in one-sided risk assessment. Some patrol schemes using unmanned aerial vehicles still mainly rely on manual control or simple path planning, and have not realized risk-driven task allocation: on the one hand, the risk assessment has not constructed a dynamic model combined with environmental factors such as dominant wind direction, which cannot simulate the evolution of risk propagation and is difficult to accurately identify key risk areas; on the other hand, the task allocation does not take into account factors such as the remaining power of the unmanned aerial vehicle and the task cost, which cannot maximize the overall efficiency of the system, resulting in a lag in the response of high-priority tasks and low resource utilization. In summary, the existing technology lacks an integrated solution of multi-source data fusion, dynamic risk modeling and intelligent task collaboration, and there is a need for a fire-fighting patrol system and method that can autonomously identify risks and accurately allocate tasks to improve the park's fire safety protection capability. SUMMARY

[0003] The purpose of the present application is to provide a fire-fighting unmanned aerial vehicle autonomous patrol system and method based on the Internet of Things to solve the problems in the prior art.

[0004] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a fire-fighting unmanned aerial vehicle autonomous patrol system and method based on the Internet of Things, The method comprises: Step S1: Real-time aggregation of the perception data of the fixed Internet of Things sensors and the mobile unmanned aerial vehicle onboard sensors distributed in the park, as well as the state data of the unmanned aerial vehicle cluster, and forming a standardized data set after processing; based on the data set, the park space is divided into a plurality of monitoring nodes, and the comprehensive risk value of each node is calculated; Step S2: Construct a directed edge according to the spatial relationship between the nodes to form a node network; calculate the weight of each directed edge in combination with the distance factor and the environmental dominant wind direction factor, thereby constructing a weighted directed dynamic risk map; Step S3: Based on the weighted directed dynamic risk map, simulate the risk evolution process over a future period of time using a risk propagation model; based on the simulation results, identify nodes that meet preset risk conditions as key nodes to form a set of key inspection targets, and simultaneously determine a set of routine inspection areas; based on these two sets, generate a differentiated task set containing tasks of different priorities. Step S4: Calculate priority scores for each task in the differentiated task set, and comprehensively consider the state of the executing UAV and the task cost to establish a collaborative task allocation model with the goal of maximizing the overall system efficiency; by solving the model, generate the final task allocation scheme.

[0005] Furthermore, step S1 includes: Step S1-1: Receive multi-source heterogeneous sensing data from fixed IoT sensors and UAV-borne sensors in real time, as well as status data from the UAV flight control system, to form a dataset. ; Step S1-2: Divide the park into several independent monitoring areas, each area is defined as a node in the dynamic risk map, and all nodes constitute the node set V; Step S1-3: Based on the dataset According to the node temperature Smoke concentration Data, computing nodes Dynamic risk increment The calculation formula is: ; and then combine the nodes static risk base value compute nodes Comprehensive risk value at current time t The calculation formula is: ;in, and These are weighting coefficients, and ; and As a weighting factor, and These are preset thresholds for temperature and smoke concentration, respectively.

[0006] Furthermore, step S2 includes: Step S2-1: Based on the spatial relationship between nodes, for any two adjacent nodes... With nodes Build a path from node Pointing to node directed edges The directed edges constitute a set F of directed edges; Step S2-2: Based on the current prevailing wind direction angle WD, from the node Pointing to node Direction angle Given the preset wind influence factor k, the wind assist coefficient g (WD) is calculated using the following formula: ; and then combine the nodes With nodes Euclidean distance between centers and distance decay function Calculate the weight of the directed edge. The calculation formula is: Among them, the prevailing wind direction angle WD is the angle measured clockwise with due north as the reference 0 degrees. For about A monotonically decreasing function; Step S2-3: Integrate the node set V, the directed edge set F, the node risk value set R(t), and the edge weight set W to construct the weighted directed dynamic risk graph G(t) =<V,F,R(t),W> .

[0007] Furthermore, step S3 includes: Step S3-1: Based on the dynamic risk map A risk propagation model is used to simulate the propagation process of risk among nodes over a future period Δt; specifically, for nodes... Its simulated risk value at time t+Δt The calculation is as follows: Where η is the risk propagation damping coefficient. This indicates that for all nodes The existence of slave nodes is satisfied Pointing to node directed edges Current risk value With directed edge weight Summing the products; Step S3-2: Based on the simulation results, identify nodes that meet any of the following conditions and mark them as key nodes to form a set of key inspection targets. : a) Simulated risk value Exceeding the first risk threshold Nodes; b) Current risk value The simulated risk value did not exceed the threshold. The increment exceeds the second risk threshold Nodes; c) In the risk propagation path, the sum of its in-degree weights exceeds the third risk threshold. Nodes; At the same time, nodes that are not marked as critical nodes and have a current risk value Below the routine inspection threshold The nodes are included in the routine inspection area set. ; Step S3-3: Based on the aforementioned set of key verification targets and routine inspection areas Generate the corresponding task list; where, is Each key node generates a high-priority key verification task. and for Generate a low-to-medium priority routine inspection task covering all nodes in the set. The and Together they constitute the differentiated task set. .

[0008] Furthermore, step S4 includes: Step S4-1: For the differentiated task set Tasks in Calculate a priority score Among them, based on key verification tasks Simulated risk values ​​of corresponding key nodes Compared with the current risk value Determine key verification tasks Priority score The calculation formula is: Where a and b represent weights; for routine inspection tasks Its priority score Set to a fixed base value; Step S4-2: For drones With the task Based on drones Current location to task Flight distance to target location and drones Remaining battery power Calculation cost The specific formula is as follows: ,in, This is the cost normalization coefficient; Step S4-3: Establish a task allocation model with the goal of maximizing the overall execution efficiency of the system. The optimization objective function H of the model is expressed as: Where M represents the total number of drones participating in task allocation, and K represents the total number of tasks in the differentiated task set. As a decision variable, when the task Assigned to drones The value is 1 when the condition is met, and 0 otherwise; μ and ν are weighting coefficients; when solving the objective function H, the following constraints must be satisfied: a) Each task can be assigned to a maximum of one drone: ; b) The total time spent on each task assigned to a drone shall not exceed the maximum flight time that its remaining battery power can support; c) High-priority key verification tasks must be assigned; Step S4-4: Solve the task allocation model and obtain the optimal solution. The value of will be 1. The corresponding task Assigned to the corresponding drone This leads to the final task allocation plan.

[0009] The system includes: a data processing and risk value calculation module, a dynamic risk map construction module, a differentiated task set generation module, and a collaborative task allocation module; The data processing and risk value calculation module is used to collect in real time the perception data of fixed IoT sensors and mobile drone airborne sensors distributed in the park, as well as the status data of the drone swarm, and process them to form a standardized dataset. Based on the dataset, the park space is divided into several monitoring nodes, and the comprehensive risk value of each node is calculated. The dynamic risk graph construction module is used to construct directed edges based on the spatial relationships between nodes to form a node network; it combines the distance factors between nodes and the prevailing wind direction factors to calculate the weight of each directed edge, thereby constructing a weighted directed dynamic risk graph. The differentiated task set generation module is used to simulate the risk evolution process over a future period of time based on the weighted directed dynamic risk map and a risk propagation model; based on the simulation results, nodes that meet preset risk conditions are identified as key nodes to form a set of key inspection targets, and a set of routine inspection areas is determined; based on these two sets, a differentiated task set containing tasks of different priorities is generated. The collaborative task allocation module is used to calculate priority scores for each task in the differentiated task set, and to establish a collaborative task allocation model with the goal of maximizing the overall system efficiency by comprehensively considering the state of the executing UAV and the task cost; by solving the model, the final task allocation scheme is generated.

[0010] The data processing and risk value calculation module includes: The multi-source data receiving unit is used to receive multi-source heterogeneous sensing data from fixed IoT sensors, UAV onboard sensors, and status data from the UAV flight control system in real time, forming a dataset. ; The monitoring node division unit is used to divide the park into several independent monitoring areas. Each area is defined as a node in the dynamic risk map, and all nodes constitute the node set V. The comprehensive risk value calculation unit is used for calculation based on datasets. According to the node temperature Smoke concentration Data, computing nodes Dynamic risk increment The calculation formula is: ; and then combine the nodes static risk base value compute nodes Comprehensive risk value at current time t The calculation formula is: ;in, and These are weighting coefficients, and ; and As a weighting factor, and These are preset thresholds for temperature and smoke concentration, respectively.

[0011] The dynamic risk mapping construction module includes: Directed edge construction unit, used to construct edges based on the spatial relationship between nodes, for any two adjacent nodes. With nodes Build a path from node Pointing to node directed edges The directed edges constitute a set F of directed edges; The edge weight calculation unit is used to calculate the edge weight based on the current prevailing wind direction angle WD and the node. Pointing to node Direction angle Given the preset wind influence factor k, the wind assist coefficient g (WD) is calculated using the following formula: ; and then combine the nodes With nodes Euclidean distance between centers and distance decay function Calculate the weight of the directed edge. The calculation formula is: Among them, the prevailing wind direction angle WD is the angle measured clockwise with due north as the reference 0 degrees. For about A monotonically decreasing function; The dynamic risk graph integration unit is used to integrate the node set V, the directed edge set F, the node risk value set R(t), and the edge weight set W to construct the weighted directed dynamic risk graph G(t).<V,F,R(t),W> .

[0012] The differentiated task set generation module includes: The risk propagation simulation unit is used to simulate the dynamic risk map. A risk propagation model is used to simulate the propagation process of risk among nodes over a future period Δt; specifically, for nodes... Its simulated risk value at time t+Δt The calculation is as follows: Where η is the risk propagation damping coefficient. This indicates that for all nodes The existence of slave nodes is satisfied Pointing to node directed edges Current risk value With directed edge weight Summing the products; The key node and inspection area identification unit is used to identify nodes that meet any of the following conditions based on the simulation results, and mark them as key nodes to form a set of key inspection targets. : a) Simulated risk value Exceeding the first risk threshold Nodes; b) Current risk value The simulated risk value did not exceed the threshold. The increment exceeds the second risk threshold Nodes; c) In the risk propagation path, the sum of its in-degree weights exceeds the third risk threshold. Nodes; At the same time, nodes that are not marked as critical nodes and have a current risk value Below the routine inspection threshold The nodes are included in the routine inspection area set. ; The differentiated task generation unit is used to generate tasks based on the set of key verification targets. and routine inspection areas Generate the corresponding task list; where, is Each key node generates a high-priority key verification task. and for Generate a low-to-medium priority routine inspection task covering all nodes in the set. The and Together they constitute the differentiated task set. .

[0013] The collaborative task allocation module includes: The task priority calculation unit is used for the differentiated task set. Tasks in Calculate a priority score Among them, based on key verification tasks Simulated risk values ​​of corresponding key nodes Compared with the current risk value Determine key verification tasks Priority score The calculation formula is: Where a and b represent weights; for routine inspection tasks Its priority score Set to a fixed base value; Task cost calculation unit, used for drones With the task Based on drones Current location to task Flight distance to target location and drones Remaining battery power Calculation cost The specific formula is as follows: , where γ is the cost normalization coefficient; The task allocation model construction unit is used to establish a task allocation model with the goal of maximizing the overall execution efficiency of the system. The optimization objective function H of the model is expressed as: Where M represents the total number of drones participating in task allocation, and K represents the total number of tasks in the differentiated task set. As a decision variable, when the task Assigned to drones The value is 1 when the condition is met, and 0 otherwise; μ and ν are weighting coefficients; when solving the objective function H, the following constraints must be satisfied: a) Each task can be assigned to a maximum of one drone: ; b) The total time spent on each task assigned to a drone shall not exceed the maximum flight time that its remaining battery power can support; c) High-priority key verification tasks must be assigned; The task allocation model solving unit is used to solve the task allocation model and obtain the optimal solution. The value of will be 1. The corresponding task Assigned to the corresponding drone This leads to the final task allocation plan.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention aggregates perception data from fixed IoT sensors, UAV-borne sensors, and UAV swarm status data to form a standardized dataset. It combines static risk baselines with dynamic risk increments to calculate the comprehensive risk value of nodes, avoiding the shortcomings of isolated data from traditional fixed sensors and one-sided data from manual inspections, thus making the fire risk assessment of the park more comprehensive and objective. 2. By constructing a weighted directional dynamic risk map based on the spatial relationship of nodes, distance attenuation characteristics, and prevailing wind direction, and simulating the future risk evolution process through a risk propagation model, it can accurately identify key nodes where simulated risks exceed thresholds, risk increments are abnormal, and in-degree weights exceed standards. This solves the problems of traditional inspections being unable to predict risk spread and the lagging location of key hidden dangers, providing clear targets for key inspections. 3. By calculating task priorities and combining the drone's flight distance and remaining battery power to calculate task costs, a task allocation model is constructed with the goal of maximizing the overall system efficiency. This model satisfies the constraints of "one drone per task, battery power matching time, and high priority must be allocated," avoiding the problems of manual operation and resource waste in traditional drone patrols. It ensures that high-risk tasks are responded to first, improving the resource utilization and patrol efficiency of drone swarms. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of the Internet of Things-based autonomous patrol system for fire fighting drones according to the present invention; Figure 2 This is a schematic diagram of the dynamic risk map construction module of the Internet of Things-based autonomous patrol system for fire-fighting drones according to the present invention; Figure 3 This is a flowchart illustrating the Internet of Things-based autonomous patrol method for fire-fighting drones according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example: Figure 1 , Figure 2 and Figure 3 As shown, this invention provides a technical solution: an autonomous patrol system and method for fire-fighting drones based on the Internet of Things. With a certain area of ​​40,000 The application scenario is an industrial park (200m×200m) containing warehouse areas, office areas, production workshops, etc. Three fire-fighting drones are deployed, equipped with infrared thermal imagers and smoke sensors, as well as 20 fixed IoT sensors, including DS18B20 temperature sensors and MQ-2 smoke sensors.

[0018] The methods include: Step S1: In real time, the perception data from fixed IoT sensors and mobile drone-borne sensors distributed within the park, as well as the status data of the drone swarm, are collected, processed, and formed into a standardized dataset; based on the dataset, the park space is divided into several monitoring nodes, and the comprehensive risk value of each node is calculated. Step S2: Construct directed edges based on the spatial relationships between nodes to form a node network; calculate the weight of each directed edge by combining the distance between nodes and the prevailing wind direction in the environment, thereby constructing a weighted directed dynamic risk map. Step S3: Based on the weighted directed dynamic risk map, simulate the risk evolution process over a future period of time using a risk propagation model; based on the simulation results, identify nodes that meet preset risk conditions as key nodes to form a set of key inspection targets, and simultaneously determine a set of routine inspection areas; based on these two sets, generate a differentiated task set containing tasks of different priorities. Step S4: Calculate priority scores for each task in the differentiated task set, and comprehensively consider the state of the executing UAV and the task cost to establish a collaborative task allocation model with the goal of maximizing the overall system efficiency; by solving the model, generate the final task allocation scheme.

[0019] Furthermore, step S1 includes: Step S1-1: Receive real-time data from fixed sensors and drone status data. Fixed sensor data includes: warehouse area sensor data at temperature T=60℃ and smoke concentration S=0.15mg / m³; office area sensor data at T=28℃ and S=0.05mg / m³. Drone status data includes: drone Remaining battery power =80%, flight speed , Remaining battery power =70%, Remaining battery power =90%; forming a dataset ; Step S1-2: Divide the park into 100 monitoring nodes using a 20m×20m grid, forming a node set. ,in As a node in the warehouse area, it stores flammable materials and has a high fire risk level; It is a production workshop node with production equipment, and the fire risk level is medium. It is a node in the office area, mainly consisting of office facilities, and has a low fire risk level. Step S1-3: Static risk baseline of nodes A risk level-specific value is set for its respective area, specifically for warehouse area nodes. Due to the storage of flammable materials, the risk of fire is high. All are 0.6; production workshop nodes The fire risk is moderate due to the operation of equipment. All are 0.4; Office area nodes Since it mainly consists of office facilities, the fire risk is low. Both are 0.2; Other calculation parameters: weighting coefficients , Weighting factors , Temperature threshold Smoke concentration threshold ; (1) Warehouse area node ( =0.6) Calculate the dynamic risk increment : Overall risk value : ; (2) Production workshop node ( =0.4) Calculate the dynamic risk increment : Overall risk value : ; (3) Office area node V51 (B 51 =0.2) Calculate the dynamic risk increment : Overall risk value : .

[0020] Furthermore, step S2 includes: Step S2-1: Based on the spatial adjacency relationship of nodes, construct directed edges for adjacent nodes: Node (Coordinates (40, 160)) and nodes (coordinates (60, 160)), node With nodes Construct bidirectional directed edges at coordinates (40, 140) to form a set F of directed edges; Step S2-2: Set parameters: prevailing wind direction WD=0° (north wind, due north is 0°), wind force influence factor k=0.3, distance attenuation function. ,in The Euclidean distance between the node centers; For directed edges ( → Calculate: Direction angle : point to The direction is due east. =270°; Wind assist coefficient g (WD): ;distance : and center distance =20m, f( ) Edge weight : =0.833×1=0.833; Step S2-3: Integrate the node set V, directed edge set F, node risk value set R(t), and edge weight set W, and construct a dynamic risk graph G(t) using the Neo4j graph database.<V,F,R(t),W> .

[0021] Furthermore, step S3 includes: Step S3-1: Set parameters: time Δt = 30 min, risk propagation damping coefficient η = 0.2; for production workshop nodes The associated directed edges are ( =0.8), ( =0.75), and =0.432、 =0.22: Calculate the simulated risk value at time t+Δt. : ; Step S3-2: Set the threshold: First risk threshold =0.35, Second Risk Threshold =0.1, Third Risk Threshold =1.5, Routine Inspection Threshold =0.2; Identify key nodes and build : Condition a (simulated risk value exceeds) ): Warehouse node of =0.41, associated edge ( =0.82), ( =0.78), calculated as follows , included ; Condition b (risk increment exceeds) ):node of =0.23, =0.35, increment 0.12 > 0.1, included ; Condition c (the sum of in-degree weights exceeds) ):node The sum of the weights of the in-degree edges = , included ; Final key verification target set ; Identification Office Area Node The R(t) values ​​are all [0.08, 0.18] < 0.2, therefore .

[0022] Step S3-3: High-priority tasks: for Each node generates key verification tasks. (Verification) ), (Verification) ), (Verification) Low to medium priority tasks: Generate routine inspection tasks (cover ); forming a differentiated task set .

[0023] Furthermore, step S4 includes: Step S4-1: Set weights a=0.6 and b=0.4, and focus on verifying the key tasks. The priority score is: , =0.3: ; ; ; ; Step S4-2: Set the cost normalization factor =0.01: arrive Flight distance , , ; arrive Flight distance , , ; arrive Flight distance , , . Step S4-3: The optimization objective function F of the model is expressed as: Set M=3 (total number of drones) and K=4 (total number of missions). =0.8、 =0.2, constraint conditions: A maximum of one drone can be assigned to each mission: ; The drone's operating time is less than or equal to the remaining battery power's operating time. implement (20min)+ (40min) = 60min ≤ 120min =80% support time); High-priority tasks must be assigned: . Step S4-4: Solve to obtain the optimal solution: =1、 =1、 =1、 =1, final solution: Execute first (Verification) ), then execute (cover ); implement (Verification) ); implement (Verification) ). The system includes: a data processing and risk value calculation module, a dynamic risk map construction module, a differentiated task set generation module, and a collaborative task allocation module; The data processing and risk value calculation module is used to collect in real time the perception data of fixed IoT sensors and mobile drone airborne sensors distributed in the park, as well as the status data of the drone swarm, and process them to form a standardized dataset. Based on the dataset, the park space is divided into several monitoring nodes, and the comprehensive risk value of each node is calculated. The dynamic risk graph construction module is used to construct directed edges based on the spatial relationships between nodes to form a node network; it combines the distance factors between nodes and the prevailing wind direction factors to calculate the weight of each directed edge, thereby constructing a weighted directed dynamic risk graph. The differentiated task set generation module is used to simulate the risk evolution process over a future period of time based on the weighted directed dynamic risk map and a risk propagation model; based on the simulation results, nodes that meet preset risk conditions are identified as key nodes to form a set of key inspection targets, and a set of routine inspection areas is determined; based on these two sets, a differentiated task set containing tasks of different priorities is generated. The collaborative task allocation module is used to calculate priority scores for each task in the differentiated task set, and to establish a collaborative task allocation model with the goal of maximizing the overall system efficiency by comprehensively considering the state of the executing UAV and the task cost; by solving the model, the final task allocation scheme is generated.

[0024] The data processing and risk value calculation module includes: The multi-source data receiving unit is used to receive multi-source heterogeneous sensing data from fixed IoT sensors, UAV onboard sensors, and status data from the UAV flight control system in real time, forming a dataset. ; The monitoring node division unit is used to divide the park into several independent monitoring areas. Each area is defined as a node in the dynamic risk map, and all nodes constitute the node set V. The comprehensive risk value calculation unit is used for calculation based on datasets. According to the node temperature Smoke concentration Data, computing nodes Dynamic risk increment The calculation formula is: ; and then combine the nodes static risk base value compute nodes Comprehensive risk value at current time t The calculation formula is: ;in, and These are weighting coefficients, and ; and As a weighting factor, and These are preset thresholds for temperature and smoke concentration, respectively.

[0025] The dynamic risk mapping construction module includes: Directed edge construction unit, used to construct edges based on the spatial relationship between nodes, for any two adjacent nodes. With nodes Build a path from node Pointing to node directed edges The directed edges constitute a set F of directed edges; The edge weight calculation unit is used to calculate the edge weight based on the current prevailing wind direction angle WD and the node. Pointing to node Direction angle Given the preset wind influence factor k, the wind assist coefficient g (WD) is calculated using the following formula: ; and then combine the nodes With nodes Euclidean distance between centers and distance decay function Calculate the weight of the directed edge. The calculation formula is: Among them, the prevailing wind direction angle WD is the angle measured clockwise with due north as the reference 0 degrees. For about A monotonically decreasing function; The dynamic risk graph integration unit is used to integrate the node set V, the directed edge set F, the node risk value set R(t), and the edge weight set W to construct the weighted directed dynamic risk graph G(t).<V,F,R(t),W> .

[0026] The differentiated task set generation module includes: The risk propagation simulation unit is used to simulate the dynamic risk map. A risk propagation model is used to simulate the propagation process of risk among nodes over a future period Δt; specifically, for nodes... Its simulated risk value at time t+Δt The calculation is as follows: Where η is the risk propagation damping coefficient. This indicates that for all nodes The existence of slave nodes is satisfied Pointing to node directed edges Current risk value With directed edge weight Summing the products; The key node and inspection area identification unit is used to identify nodes that meet any of the following conditions based on the simulation results, and mark them as key nodes to form a set of key inspection targets. : a) Simulated risk value Exceeding the first risk threshold Nodes; b) Current risk value The simulated risk value did not exceed the threshold. The increment exceeds the second risk threshold Nodes; c) In the risk propagation path, the sum of its in-degree weights exceeds the third risk threshold. Nodes; At the same time, nodes that are not marked as critical nodes and have a current risk value Below the routine inspection threshold The nodes are included in the routine inspection area set. ; The differentiated task generation unit is used to generate tasks based on the set of key verification targets. and routine inspection areas Generate the corresponding task list; where, is Each key node generates a high-priority key verification task. and for Generate a low-to-medium priority routine inspection task covering all nodes in the set. The and Together they constitute the differentiated task set. .

[0027] The collaborative task allocation module includes: The task priority calculation unit is used for the differentiated task set. Tasks in Calculate a priority score Among them, based on key verification tasks Simulated risk values ​​of corresponding key nodes Compared with the current risk value Determine key verification tasks Priority score The calculation formula is: Where a and b represent weights; for routine inspection tasks Its priority score Set to a fixed base value; Task cost calculation unit, used for drones With the task Based on drones Current location to task Flight distance to target location and drones Remaining battery power Calculation cost The specific formula is as follows: ,in, This is the cost normalization coefficient; The task allocation model construction unit is used to establish a task allocation model with the goal of maximizing the overall execution efficiency of the system. The optimization objective function H of the model is expressed as: Where M represents the total number of drones participating in task allocation, and K represents the total number of tasks in the differentiated task set. As a decision variable, when the task Assigned to drones The value is 1 when the condition is met, and 0 otherwise; μ and ν are weighting coefficients; when solving the objective function H, the following constraints must be satisfied: a) Each task can be assigned to a maximum of one drone: ; b) The total time spent on each task assigned to a drone shall not exceed the maximum flight time that its remaining battery power can support; c) High-priority key verification tasks must be assigned; The task allocation model solving unit is used to solve the task allocation model and obtain the optimal solution. The value of will be 1. The corresponding task Assigned to the corresponding drone This leads to the final task allocation plan.

[0028] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A fire-fighting unmanned aerial vehicle autonomous patrol method based on the Internet of Things, characterized in that: Step S1: Real-time aggregation of the perception data of the fixed Internet of Things sensors and mobile unmanned aerial vehicle onboard sensors distributed in the park, as well as the state data of the unmanned aerial vehicle cluster, processing to form a standardized data set; Based on the data set, the park space is divided into several monitoring nodes, and the comprehensive risk value of each node is calculated; Step S2: According to the spatial relationship between nodes, a directed edge is constructed to form a node network; the distance factor between nodes and the environmental dominant wind direction factor are combined to calculate the weight of each directed edge, thereby constructing a weighted directed dynamic risk map; Step S3: Based on the weighted directed dynamic risk map, the risk evolution process in the future period of time is simulated through a risk propagation model; according to the simulation result, the nodes that meet the preset risk condition are identified as key nodes to form a key check target set, and a routine inspection area set is determined; based on the two sets, a differentiated task set containing different priority tasks is generated; Step S4: Calculate the priority score for each task in the differentiated task set, and comprehensively consider the state of the executing unmanned aerial vehicle and the task cost to establish a collaborative task allocation model with the goal of maximizing the overall system performance; by solving the model, the final task allocation scheme is generated. 2.The IoT-based fire-fighting unmanned aerial vehicle autonomous patrol method according to claim 1, characterized in that: The step S1 includes: Step S1-1: Real-time receiving multi-source heterogeneous sensing data from fixed Internet of Things sensors, unmanned aerial vehicle on-board sensors, and state data from unmanned aerial vehicle flight control systems to form a data set ; Step S1-2: Divide the park into several independent monitoring areas, each area is defined as a node of the dynamic risk map, and all nodes constitute a node set V; Step S1-3: Based on the dataset According to the temperature of the node Smoke concentration Data, computing nodes Dynamic risk increment The calculation formula is: ; and then combine the nodes static risk base value compute nodes Comprehensive risk value at current time t The calculation formula is: ;in, and These are weighting coefficients, and ; and As a weighting factor, and These are preset thresholds for temperature and smoke concentration, respectively. 3.The IoT-based fire-fighting unmanned aerial vehicle autonomous patrol method according to claim 1, characterized in that: The step S2 includes: Step S2-1: According to the spatial relationship between nodes, for any two adjacent nodes corresponding to two adjacent spaces and , a directed edge from node to node is constructed, and the directed edge constitutes a directed edge set F; ​ Step S2-2: Based on the current prevailing wind direction angle WD, from the node Pointing to node Direction angle Given the preset wind influence factor k, the wind assist coefficient g (WD) is calculated using the following formula: ; and then combine the nodes With nodes Euclidean distance between centers and distance decay function Calculate the weight of the directed edge. The calculation formula is: Among them, the prevailing wind direction angle WD is the angle measured clockwise with due north as the reference 0 degrees. For about A monotonically decreasing function; Step S2-3: Integrate the node set V, the directed edge set F, the node risk value set R(t), and the edge weight set W to construct the weighted directed dynamic risk map G(t) = <V, F, R(t), W>. 4.The IoT-based fire-fighting unmanned aerial vehicle autonomous patrol method according to claim 1, characterized in that: The step S3 includes: Step S3-1: based on the dynamic risk map , simulate the propagation process of the risk among the nodes in a future period of time Δt using a risk propagation model; specifically, for a node , the simulated risk value of the node at time t+Δt is calculated as follows: , where η is a risk propagation damping coefficient, represents the sum of the products of the current risk value of the node and the directed edge weight , provided that there exists a directed edge from the node pointing to the node ​​​ Step S3-2: According to the simulation result, a node satisfying any one of the following conditions is identified and marked as a key node, constituting a key check target set : a) simulated risk value exceeding a first risk threshold of nodes; b) current risk value Not over threshold, but simulated risk value Delta of risk value exceeds second risk threshold Node; c) in the risk propagation path, the sum of the in-degree weights of its nodes exceeds a third risk threshold of nodes; Meanwhile, nodes that are not marked as critical nodes and have a current risk value below a routine inspection threshold are included in a routine inspection region set ; Step S3-3: generating a corresponding task list based on the set of key check targets and the set of routine inspection areas ; wherein, for each key node in , a high-priority key check task is generated , and for , a medium-low-priority routine inspection task covering all nodes in the set is generated , the and together constitute the differentiated task set . 5.The IoT-based fire-fighting drone autonomous patrol method according to claim 1, characterized in that: The step S4 includes: Step S4-1: calculating a priority score for the tasks in the differentiated task set ; wherein the priority score of a key-checking task is determined based on the simulated risk value of the corresponding key node of the key-checking task , and the current risk value , and the calculation formula is: , wherein a and b represent weights; for routine inspection tasks, the priority score is set as a fixed basic value;​​​​​​ Step S4-2: For the UAV with the task , the flight distance from the current position of the UAV to the target position of the task , and the remaining power of the UAV are calculated , and the cost is calculated , specifically as follows: , wherein is a cost normalization coefficient; Step S4-3: a task allocation model is established, which aims to maximize the overall performance of the system. The optimization objective function H of the model is expressed as: wherein M represents the total number of UAVs participating in task allocation, K represents the total number of tasks in the differentiated task set, is a decision variable, and takes the value of 1 when the task is allocated to the UAV , otherwise 0; μ and ν are weight coefficients; when solving the objective function H, the following constraint conditions need to be met: a) each task is assigned to at most one drone: ; b) The total time of the task allocated to each unmanned aerial vehicle must not exceed the maximum flight time supported by the remaining power; c) High-priority key check tasks must be allocated; Step S4-4: Solve the task allocation model and obtain the optimal solution. The value of will be 1. The corresponding task Assigned to the corresponding drone This leads to the final task allocation plan.

6. The fire-fighting unmanned aerial vehicle autonomous patrol system based on the Internet of Things is used for executing the fire-fighting unmanned aerial vehicle autonomous patrol method based on the Internet of Things in any one of claims 1-5, characterized in that: The system includes: Data processing and risk value calculation module, dynamic risk map construction module, differentiated task set generation module, and collaborative task allocation module; The data processing and risk value calculation module is used to real-time aggregation of the perception data of the fixed Internet of Things sensors and mobile unmanned aerial vehicle onboard sensors distributed in the park, as well as the state data of the unmanned aerial vehicle cluster, processing to form a standardized data set; Based on the data set, the park space is divided into several monitoring nodes, and the comprehensive risk value of each node is calculated; The dynamic risk map construction module is used to construct a directed edge according to the spatial relationship between nodes to form a node network; the distance factor between nodes and the environmental dominant wind direction factor are combined to calculate the weight of each directed edge, thereby constructing a weighted directed dynamic risk map; The differentiated task set generation module is configured to simulate a risk evolution process in a future period of time based on the weighted directed dynamic risk graph through a risk propagation model, identify nodes meeting a preset risk condition as key nodes according to a simulation result to form a key inspection target set, determine a routine inspection area set, and generate a differentiated task set containing tasks of different priorities based on the two sets. The collaborative task allocation module is configured to calculate a priority score for each task in the differentiated task set, consider a state of an execution unmanned aerial vehicle and a task cost, and establish a collaborative task allocation model with a target of maximizing a system overall performance; and generate a final task allocation scheme by solving the model. 7.The IoT-based fire-fighting drone autonomous patrol system according to claim 6, characterized in that: The data processing and risk value calculation module includes: A multi-source data receiving unit is configured to receive multi-source heterogeneous sensing data from fixed Internet of Things sensors, unmanned aerial vehicle (UAV) on-board sensors, and state data from a UAV flight control system in real time to form a data set ; A monitoring node division unit is configured to divide the park into a plurality of independent monitoring areas, each area being defined as a node of the dynamic risk graph, and all nodes constituting a node set V. The comprehensive risk value calculation unit is used for calculation based on datasets. According to the node temperature Smoke concentration Data, computing nodes Dynamic risk increment The calculation formula is: ; and then combine the nodes static risk base value compute nodes Comprehensive risk value at current time t The calculation formula is: ;in, and These are weighting coefficients, and ; and As a weighting factor, and These are preset thresholds for temperature and smoke concentration, respectively. 8.The IoT-based fire-fighting drone autonomous patrol system according to claim 6, characterized in that: The dynamic risk graph construction module includes: A directed edge construction unit is configured to construct a directed edge from a node corresponding to a first space to a node corresponding to a second space according to a spatial relationship between the first and second spaces corresponding to a first space , and construct a directed edge from a node corresponding to a second space to the node corresponding to the first space, the directed edge constituting a directed edge set F The edge weight calculation unit is used to calculate the edge weight based on the current prevailing wind direction angle WD and the node. Pointing to node Direction angle Given the preset wind influence factor k, the wind assist coefficient g (WD) is calculated using the following formula: ; and then combine the nodes With nodes Euclidean distance between centers and distance decay function Calculate the weight of the directed edge. The calculation formula is: Among them, the prevailing wind direction angle WD is the angle measured clockwise with due north as the reference point. For about A monotonically decreasing function; A dynamic risk graph integration unit is configured to integrate the node set V, the directed edge set F, the node risk value set R(t), and the edge weight set W to construct the weighted directed dynamic risk graph G(t) = <V, F, R(t), W>. 9.The IoT-based fire-fighting drone autonomous patrol system according to claim 6, characterized in that: The differentiated task set generation module includes: The risk propagation simulation unit is used to simulate the dynamic risk map. A risk propagation model is used to simulate the propagation process of risk among nodes over a future period Δt; specifically, for nodes... Its simulated risk value at time t+Δt The calculation is as follows: Where η is the risk propagation damping coefficient. This indicates that for all nodes The existence of slave nodes is satisfied Pointing to node directed edges Current risk value With directed edge weight Summing the products; The key node and inspection area identification unit is configured to identify a node satisfying any one of the following conditions according to the simulation result, and mark the node as a key node to form a key check target set : a) simulated risk value exceeding a first risk threshold of nodes; b) current risk value Not over threshold, but delta risk value of second risk threshold of nodes; c) in the risk propagation path, the sum of the in-degree weights of its nodes exceeds a third risk threshold of the nodes; Meanwhile, nodes that are not marked as critical nodes and have a current risk value below a routine inspection threshold are included in a routine inspection region set ; A differentiated task generation unit is configured to generate a task list based on the set of key check targets and the set of routine inspection areas ; wherein a high-priority key check task is generated for each key node in , and a medium-low-priority routine inspection task covering all nodes in is generated , the and together constitute the differentiated task set . 10.The IoT-based fire-fighting drone autonomous patrol system according to claim 6, characterized in that: The collaborative task allocation module includes: The task priority calculation unit is used for the differentiated task set. Tasks in Calculate a priority score Among them, based on key verification tasks Simulated risk values ​​of corresponding key nodes Compared with the current risk value Determine key verification tasks Priority score The calculation formula is: Where a and b represent weights; for routine inspection tasks Its priority score Set to a fixed base value; Task cost calculation unit, used for drones With the task Based on drones Current location to task Flight distance to target location and drones Remaining battery power Calculation cost The specific formula is as follows: , where γ is the cost normalization coefficient; The task allocation model construction unit is configured to establish a task allocation model with the maximum system overall execution performance as a target, and an optimization objective function H of the model is represented as: wherein M represents a total number of unmanned aerial vehicles participating in task allocation, K represents a total number of tasks in the differentiated task set, is a decision variable, and is 1 when the task is allocated to the unmanned aerial vehicle , and is 0 otherwise; μ and ν are weight coefficients; and the following constraint conditions need to be met when the objective function H is solved. a) each task is assigned to at most one drone: ; b) A total time consumption of a task allocated to each unmanned aerial vehicle must not exceed a maximum flight time supported by a remaining electric quantity of the unmanned aerial vehicle; c) A key inspection task of a high priority must be allocated. The task allocation model solving unit is used to solve the task allocation model and obtain the optimal solution. The value of will be 1. The corresponding task Assigned to the corresponding drone This leads to the final task allocation plan.

Images