AI-driven resource scheduling method and system for intelligent fire fighting

Abstract

The invention relates to the technical field of public service data processing, and provides an AI-driven resource scheduling method and system for intelligent fire fighting. The method comprises the following steps: acquiring first data through an intelligent sensor, performing multi-source data fusion, generating a first tensor formed by a spatial graph through a neural network and modeling processing, and extracting node features and edge features from the first tensor; performing linear transformation on the node features and the edge features based on the learnable weight matrix and the activation function to determine a risk propagation matrix of the fire behavior; generating an alternative strategy based on the risk propagation matrix and the resource data, and minimizing the total loss of the alternative strategy to generate a resource allocation strategy; and determining a scheduling strategy through Nash equilibrium based on the resource allocation strategy of the adjacent regions and a preset inter-region cooperation protocol, and sending a generated scheduling instruction to the target terminal. Through multi-source data fusion and space-time modeling, accurate risk prediction is realized, the fire spreading risk and emergency response cost are effectively reduced, and the efficiency and effect of fire-fighting resource scheduling are improved.

Description

Technical Field

[0001] The present application relates to the field of public service data processing technology, and more specifically, to an AI-driven resource scheduling method and system for smart firefighting. Background Art

[0002] With the development of smart firefighting, promoting the informatization and intelligent transformation of firefighting and building an IoT-based fire supervision model have become increasingly important. Traditional firefighting systems rely heavily on manual inspections and post-event handling, resulting in delayed responses and uneven resource allocation. For example, after a fire breaks out, dispatchers must manually analyze information such as the fire location, fire station distribution, and road traffic. This results in inefficient decision-making and can easily lead to missed opportunities for rescue.

[0003] Existing technologies, through the deep integration of the Internet of Things, big data, AI, and 5G, provide a technical foundation for smart firefighting. For example, sensors monitor the status of firefighting facilities and environmental parameters in real time, combined with AI for dynamic risk assessment. However, due to the multitude of data types collected by various sensors, it is difficult to extract precise features. Furthermore, the complex external environment and firefighting resources make it difficult to efficiently and accurately determine firefighting methods and schedule resources, resulting in relatively low efficiency and effectiveness in firefighting resource scheduling. Summary of the Invention

[0004] This application provides an AI-driven resource scheduling method and system for smart firefighting, which can at least to some extent solve the problem of low efficiency and effectiveness of firefighting resource scheduling.

[0005] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the present application.

[0006] According to one aspect of the present application, an AI-driven resource scheduling method for smart firefighting is provided, comprising: acquiring first data through an intelligent sensor, wherein the first data includes environmental data and satellite data; performing multi-source data fusion on the first data, and generating a first tensor composed of a spatial map through neural network and modeling processing; extracting node features and edge features from the first tensor and the spatial map, performing linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determining a risk propagation matrix of the fire situation; generating alternative resource allocation strategies based on the risk propagation matrix and the acquired resource data, and minimizing the total loss of the alternative strategies to generate a resource allocation strategy; determining a scheduling strategy through Nash equilibrium based on the resource allocation strategies of adjacent areas and a preset inter-regional cooperation protocol; determining a target terminal related to the scheduling strategy based on a preset fire protection system, and sending a scheduling instruction generated by the scheduling strategy to the target terminal.

[0007] In the present application, based on the aforementioned scheme, the multi-source data fusion is performed on the first data, and a first tensor composed of a spatial map is generated through neural network and modeling processing, including: based on a preset spatiotemporal convolution kernel, three-dimensional convolution processing is performed on the first data to generate second data; dynamic weight coefficients are generated based on the attention mechanism, and the second data is processed by the dynamic weight coefficients to generate third data; satellite data in the first data is modeled to generate a geographic topology, and the geographic topology is processed through a graph neural network to generate a spatial map; based on the third data and the spatial map, a first tensor composed of a spatial map is generated.

[0008] In the present application, based on the aforementioned scheme, node features and edge features are extracted from the first tensor and the spatial map, and the node features and edge features are linearly transformed based on the learnable weight matrix and the activation function to determine the risk propagation matrix of the fire situation, including: based on the preset graph attention network, the information of the neighbor nodes in the spatial map is aggregated to generate node features; the edge features between the neighbor nodes are extracted from the first tensor and the spatial map, and the node features and the edge features are spliced ​​to generate a first feature; the first feature is linearly transformed based on the first learnable weight matrix and the first activation function to generate a second feature; the second feature is linearly transformed based on the second learnable weight matrix and the second activation function to obtain the risk propagation matrix of the fire situation.

[0009] In the present application, based on the aforementioned scheme, the alternative strategies for resource allocation are generated based on the risk propagation matrix and the acquired resource data, and the total loss of the alternative strategies is minimized to generate a resource allocation strategy, including: generating an alternative strategy for resource allocation based on the risk propagation matrix and the acquired resource data; determining the risk loss based on the risk parameters of the building at a set time in the risk propagation matrix; determining the resource mobilization cost based on the time-varying cost function corresponding to the resources in the alternative strategy at each time; determining the timeout penalty item based on the time when the resources in the alternative strategy arrive at the building; constructing a total loss based on the risk loss, the resource mobilization cost and the timeout penalty item; minimizing the total loss by adjusting the alternative strategy, and outputting the final resource allocation strategy.

[0010] In the present application, based on the above solution, the risk loss is determined based on the risk parameters of the building at the set time in the risk propagation matrix, including: determining the risk loss based on the risk parameters of the building at the set time in the risk propagation matrix for:

[0011]

[0012] Among them, t and T represent the total duration of the moment and strategy prediction respectively. represents the preset discount factor, represents the risk parameter of building i at time t in the risk propagation matrix, i and N represent the identification and total number of buildings in the risk propagation matrix respectively. Represents the indicator function.

[0013] In the present application, based on the aforementioned scheme, the resource allocation strategy based on adjacent areas and the preset inter-regional cooperation protocol determine the scheduling strategy through Nash equilibrium, including: obtaining the resource allocation strategy of adjacent areas and the preset inter-regional cooperation protocol; determining the utility function based on the resource allocation strategy of adjacent areas and the preset inter-regional cooperation protocol; iterating the utility function through the alternating direction multiplier method to determine the scheduling strategy.

[0014] In the present application, based on the aforementioned scheme, the preset fire protection system determines the target terminal related to the scheduling strategy, and sends the scheduling instruction generated by the scheduling strategy to the target terminal, including: based on the preset fire protection system, determining the target terminal related to the scheduling strategy; generating a scheduling instruction based on the scheduling strategy and the target terminal; and sending the scheduling instruction to the target terminal.

[0015] According to one aspect of the present application, an AI-driven resource scheduling system for smart firefighting is provided, comprising:

[0016] an acquiring unit, configured to acquire first data through a smart sensor, wherein the first data includes environmental data and satellite data;

[0017] a coupling unit, configured to perform multi-source data fusion on the first data, and generate a first tensor consisting of a spatial atlas through neural network and modeling processing;

[0018] a feature unit, configured to extract node features and edge features from the first tensor and the spatial map, perform linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determine a risk propagation matrix of the fire situation;

[0019] a strategy unit, configured to generate alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimize the total loss of the alternative strategies to generate a resource allocation strategy;

[0020] The scheduling unit is used to determine the scheduling strategy through Nash equilibrium based on the resource allocation strategy of adjacent areas and the preset inter-area cooperation protocol;

[0021] The instruction unit is used to determine the target terminal related to the dispatching strategy based on the preset fire protection system, and send the dispatching instruction generated by the dispatching strategy to the target terminal.

[0022] According to one aspect of the present application, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for AI-driven resource scheduling for smart firefighting as described in the above embodiments is implemented.

[0023] According to one aspect of the present application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the AI-driven resource scheduling method for smart firefighting as described in the above-mentioned embodiments.

[0024] According to one aspect of the present application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to implement the AI-driven resource scheduling method for smart firefighting provided in the various optional implementations described above.

[0025] The technical solution of this application obtains first data through intelligent sensors, performs multi-source data fusion on the first data, and generates a first tensor composed of a spatial map through neural network and modeling processing; extracts node features and edge features from the first tensor and the spatial map, performs linear transformation on the node features and edge features based on a learnable weight matrix and activation function, and determines the risk propagation matrix of the fire situation; generates alternative resource allocation strategies based on the risk propagation matrix and the obtained resource data, and minimizes the total loss of the alternative strategies to generate a resource allocation strategy; determines the scheduling strategy through Nash equilibrium based on the resource allocation strategies of adjacent regions and a preset inter-regional cooperation agreement; determines the target terminal related to the scheduling strategy based on the preset fire protection system, and sends the scheduling instructions generated by the scheduling strategy to the target terminal. The above process achieves accurate risk prediction through multi-source data fusion and spatiotemporal modeling, combines dynamic resource allocation with cross-regional collaborative scheduling, and constructs a closed-loop system from data perception to instruction execution, effectively reducing the risk of fire spread and emergency response costs, and improving the efficiency and effectiveness of fire protection resource scheduling.

[0026] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0028] Figure 1 The following schematically illustrates a flow chart of an AI-driven resource scheduling method for smart firefighting in one embodiment of the present application.

[0029] Figure 2 The flowchart of generating a first tensor in one embodiment of the present application is schematically shown.

[0030] Figure 3 A schematic diagram of an AI-driven resource scheduling system for smart firefighting in one embodiment of the present application is shown schematically.

[0031] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0032] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0033] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner.In the following description, many specific details are provided so as to provide a full understanding of the embodiments of the present application. However, it will be appreciated by those skilled in the art that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps etc. can be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0034] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0035] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, while others may be combined or partially combined. Therefore, the actual execution order may vary depending on the actual situation.

[0036] The implementation details of the technical solution of this application are described in detail below:

[0037] Figure 1 The flowchart of the AI-driven resource scheduling method for smart firefighting according to one embodiment of the present application is shown. Figure 1 As shown, the AI-driven resource scheduling method for smart firefighting includes at least steps S110 to S160, which are described in detail as follows:

[0038] S110: Acquire first data through a smart sensor, where the first data includes environmental data and satellite data.

[0039] In one embodiment of the present application, smart sensors are deployed in the firefighting system within environments such as building interiors, public places, and forested areas to comprehensively perceive environmental conditions. These sensors possess high sensitivity and accuracy, enabling real-time collection of environmental data such as temperature, humidity, smoke concentration, and gas composition, providing the firefighting system with first-hand environmental information. Furthermore, some smart sensors are capable of receiving satellite data, such as satellite remote sensing images and meteorological satellite data, further broadening the dimensionality of data sources.

[0040] In addition, the first data in this embodiment also includes: various sensor data, historical fire cases obtained from the database, geographic topology data, meteorological data, etc.

[0041] In one embodiment of the present application, data collected by smart sensors, including environmental data and satellite data, is transmitted in real time to the fire protection system's data center via a wireless or wired network. During transmission, the data undergoes preliminary preprocessing, such as data cleaning, noise removal, and format conversion, to ensure data accuracy and consistency, and improve the efficiency and accuracy of subsequent data analysis.

[0042] In one embodiment of this application, distributed storage, cloud computing, and other technologies are used to classify, archive, and back up real-time data, ensuring data security and accessibility. Furthermore, data query, analysis, and visualization capabilities are provided, enabling firefighters to quickly access the information they need, providing strong support for developing dispatch strategies and responding to fires.

[0043] The above process uses intelligent sensors to acquire real-time environmental and satellite data, achieving comprehensive awareness of the firefighting environment. Multi-source data fusion technology integrates this data from various sources and formats to generate a first tensor consisting of a spatial map. This tensor not only contains rich spatiotemporal information but also reflects the complex interdependencies between data points, providing a comprehensive and accurate data foundation for subsequent risk analysis and resource scheduling.

[0044] S120: Perform multi-source data fusion on the first data, and generate a first tensor consisting of a spatial map through neural network and modeling processing.

[0045] In one embodiment of the present application, multi-source data fusion is performed on the first data. In the process of generating a first tensor composed of a spatial map through neural network and modeling processing, real-time first data from smart sensors are first collected. These data cover multiple sources such as environmental data and satellite data, and have the characteristics of multi-source heterogeneity. In order to make full use of these data, multi-source data fusion technology is adopted to integrate data from different sources and in different formats to form a unified and complete data set. Then, the fused data is deeply modeled and processed using a neural network. The neural network can automatically learn the complex patterns and spatiotemporal correlations in the data, thereby extracting more valuable features and information and generating a spatiotemporal coupled first tensor. This tensor not only contains rich spatiotemporal information, but also reflects the complex correlations between data, providing a solid foundation for subsequent risk propagation modeling and resource scheduling optimization.

[0046] like Figure 2 As shown, in one embodiment of the present application, multi-source data fusion is performed on the first data, and a first tensor consisting of a spatial map is generated through neural network and modeling processing, including:

[0047] S210, performing three-dimensional convolution processing on the first data based on a preset spatiotemporal convolution kernel to generate second data;

[0048] S220, generating a dynamic weight coefficient based on the attention mechanism, and processing the second data using the dynamic weight coefficient to generate third data;

[0049] S230: Modeling the satellite data in the first data to generate a geographic topology, and processing the geographic topology through a graph neural network to generate a spatial atlas;

[0050] S240 : Generate a first tensor consisting of a spatial atlas based on the third data and the spatial atlas.

[0051] In one embodiment of the present application, a learnable spatiotemporal convolution kernel is used to perform three-dimensional convolution processing on a time series of the mth type of first data to generate second data. This is done by capturing the local spatiotemporal features of the first data, i.e., the temporal and spatial variation patterns of the first data.

[0052] In one embodiment of the present application, a dynamic weight coefficient is generated through an attention mechanism. The attention mechanism can dynamically adjust the weights of different data sources based on the characteristics of the input data to reflect the confidence level of the mth type of data source, enabling the model to focus on more important information. Based on the dynamic weight coefficient generated by the attention mechanism, the second data is weighted using the dynamic weight coefficient to generate the third data.

[0053] In one embodiment of the present application, the satellite data in the first data is modeled to generate a geographic topology, which is then processed through a graph neural network to generate a spatial map. The graph neural network is used to embed the geographic topology into spatial adjacency relationships, generating a three-dimensional spatial knowledge map. By combining the impact of geospatial information on fire risk, the graph neural network captures the dependencies in the spatial structure. At the same time, a balance coefficient is introduced to balance the influence of different components, ensuring that the model maintains stability when integrating multi-source data.

[0054] Specifically, during the graph neural network processing, the representation of the current node is updated by aggregating information from neighboring nodes, thereby encoding the structure and features of the graph. This process consists of three steps: message generation, aggregation, and update. In each layer of the network, a node receives messages from neighboring nodes, which are generated based on the neighboring nodes' features and edge weights. The node then aggregates these messages, using common aggregation functions such as summation, averaging, or max pooling. Finally, the node combines the aggregated messages with its own features and updates its state through a nonlinear transformation.

[0055] In one embodiment of the present application, based on the third data and the spatial atlas, a first tensor consisting of a spatial atlas is generated. Summarizing the above process, the generated first tensor for:

[0056]

[0057] Wherein, m and M represent the category identifier and the total number of categories of the first data respectively. represents the dynamic weight coefficient, represents three-dimensional convolution processing, represents the time series corresponding to the first data of the mth category, represents a learnable spatiotemporal convolution kernel, represents the balance coefficient, Representing geographic topology. The first tensor generated in this embodiment integrates the local spatiotemporal patterns, dynamic weights, and geographic topological constraints of different sensor data to generate a feature representation that comprehensively reflects the fire risk.

[0058] The first tensor determined in this embodiment is the dynamic risk feature tensor, which integrates information from multiple data sources, including sensor data, historical fire cases, geographic topology data, and meteorological data, providing a basis for subsequent risk assessment.

[0059] S130: Extract node features and edge features from the first tensor and the spatial map, perform linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determine a risk propagation matrix of the fire situation.

[0060] In one embodiment of the present application, node features and edge features are extracted from the first tensor and the spatial map. Node features include the physical properties of the building, historical risk data, and the like, while edge features include the spatial relationships between buildings, risk propagation paths, and the like. These features are linearly transformed using a learnable weight matrix and an activation function. The learnable weight matrix can automatically adjust the importance of features based on different tasks and data, while the activation function can introduce nonlinear factors to enhance the expressive power of the model. Finally, the risk propagation matrix of the fire situation is determined using the linearly transformed node and edge features. This matrix reflects the probability and intensity of risk propagation between buildings, providing an important foundation for subsequent risk propagation modeling and resource scheduling optimization.

[0061] In one embodiment of the present application, node features and edge features are extracted from the first tensor and the spatial map, and linear transformation is performed on the node features and edge features based on a learnable weight matrix and an activation function to determine a risk propagation matrix of the fire situation, including:

[0062] Based on a preset graph attention network, information of neighbor nodes in the spatial graph is aggregated to generate node features;

[0063] Extracting edge features between the neighboring nodes from the first tensor and the spatial graph, and concatenating the node features and the edge features to generate a first feature;

[0064] Performing a linear transformation on the first feature based on a first learnable weight matrix and a first activation function to generate a second feature;

[0065] A linear transformation is performed on the second feature based on a second learnable weight matrix and a second activation function to obtain a risk propagation matrix of the fire situation.

[0066] In one embodiment of the present application, information about neighboring nodes in the spatial graph is aggregated based on a pre-defined graph attention network to generate node features. A graph attention network (GAT) is used to aggregate information about neighboring nodes i and j in the spatial graph. This mechanism enables node features to capture local information about the graph structure and generates node features by aggregating information about their neighboring nodes.

[0067] In one embodiment of the present application, edge features between neighboring nodes are extracted from the first tensor and the spatial graph. These edge features may include building spacing, connectivity, and wind direction factors, which are crucial for quantifying risk propagation paths. Subsequently, the node features and edge features are concatenated to generate a first feature that comprehensively considers the interactions between nodes and the properties of edges, thereby more accurately quantifying risk propagation paths.

[0068] In one embodiment of the present application, a linear transformation is performed on the first features based on a first learnable weight matrix and a first activation function to generate a second feature. Specifically, the concatenated features are linearly transformed using the first learnable weight matrix, biased, and a ReLU activation function is applied. By introducing nonlinearity through multiple layers of linear transformations and activation functions, the model is able to learn complex nonlinear relationships.

[0069] In one embodiment of the present application, a linear transformation is performed on the second feature based on a second learnable weight matrix and a second activation function to obtain a fire risk propagation matrix. Specifically, a linear transformation is performed on the transformed feature using the second learnable weight matrix, and a bias is added and a sigmoid function is applied to obtain the fire risk propagation matrix.

[0070] Based on the above process, the risk propagation matrix of fire situation is generated for:

[0071]

[0072] in, Represents the information of neighbor nodes i and j respectively, Represents the edge features between neighbor nodes, denote the first learnable weight matrix and the second learnable weight matrix respectively, represent the bias terms, By calculating the fire risk propagation matrix, we can intuitively understand the spatiotemporal diffusion path of the fire in the building complex, providing an important basis for the dispatch of firefighting resources.

[0073] The above process extracts node and edge features from the first tensor and the spatial graph. These features reflect the spatial relationships between buildings and the risk propagation paths. Linear transformations of these features, based on a learnable weight matrix and activation function, automatically learn the complex relationships between these features, accurately determining the fire risk propagation matrix and improving the accuracy of risk propagation probability calculations. Based on the risk propagation probability, this can be used to prioritize firefighting resource dispatch, prioritizing higher-risk areas and improving firefighting efficiency.

[0074] S140 , generating alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimizing the total loss of the alternative strategies to generate a resource allocation strategy.

[0075] In this embodiment, the risk propagation matrix is ​​used to evaluate the probability and intensity of risk propagation in different buildings or areas, thereby determining which areas require more resources to deal with potential risks. Combined with the acquired resource data, such as the number and status of fire trucks, drones, and rescue teams, multiple alternative resource allocation strategies are generated. These alternative strategies will consider different resource allocation schemes to ensure that a quick and effective response can be made when a risk occurs. These alternative strategies are evaluated and the total loss of each strategy is calculated. The total loss includes multiple aspects such as response time or resource consumption. By minimizing the total loss, the optimal resource allocation strategy can be selected. This optimal strategy is used as the final resource allocation strategy to guide the actual resource scheduling and allocation work to ensure that a quick and effective response can be made when a risk occurs and to minimize the loss.

[0076] In one embodiment of the present application, generating alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimizing the total loss of the alternative strategies to generate a resource allocation strategy includes:

[0077] generating alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data;

[0078] Determining risk losses based on risk parameters of the building at a set time in the risk propagation matrix;

[0079] Determining resource mobilization costs based on a time-varying cost function corresponding to resources in the alternative strategy at each time;

[0080] Determining a timeout penalty item based on the time when the resources in the alternative strategy arrive at the building;

[0081] A total loss is calculated based on the risk loss, the resource mobilization cost, and the overtime penalty item;

[0082] By adjusting the alternative strategies, the total loss is minimized and a final resource allocation strategy is output.

[0083] In one embodiment of the present application, an alternative resource allocation strategy is generated based on the risk propagation matrix and the acquired resource data. Optionally, a deep reinforcement learning algorithm is used to generate the alternative strategy, which is a mapping from time to resource sets, determining when and where to deploy which resources.

[0084] For example, in the reinforcement learning-based strategy generation process, the current resource allocation state and the environment state (e.g., risk propagation state) are represented as a state vector. Based on the current state and the action-value function table, the action with the largest action-value function is selected, i.e., the resource allocation strategy. A reward function is designed to reflect the performance of the resource allocation strategy, such as minimizing total loss. The action-value function table is updated based on the reward after executing the action and the new state. This process is repeated until the action-value function table converges or the maximum number of iterations is reached, resulting in a strategy.

[0085] In one embodiment of the present application, based on the alternative strategy, the total loss consisting of risk loss, resource mobilization cost and timeout penalty is determined. Specifically, in this process, the risk loss is determined based on the risk parameters of the building at the set time in the risk propagation matrix. for:

[0086]

[0087] Among them, t and T represent the total duration of the moment and strategy prediction respectively. represents the preset discount factor, represents the risk parameter of building i at time t in the risk propagation matrix, i and N represent the identification and total number of buildings in the risk propagation matrix respectively. represents an indicator function, indicating whether building i is not covered at time t. If building i is not protected by any resource at time t, then ;otherwise, .

[0088] In one embodiment of the present application, the resource mobilization cost is determined based on the time-varying cost function corresponding to the resources in the alternative strategy at each time. for:

[0089]

[0090] Among them, k and K represent the identification and total number of firefighting resources respectively. Represents a resource The time-varying cost function at time t; Is an indicator function, indicating that at time t the resource Is it active? If the resource At time t, it is being used (e.g., a fire truck is on its way to a fire scene), then ;otherwise .

[0091] In one embodiment of the present application, the timeout penalty item is determined based on the time when the resource in the alternative strategy arrives at the building. for:

[0092]

[0093] in, represents the time when resources arrive at building i, that is, the time interval from the occurrence of fire to the first arrival of resources at building i; Represents the response time threshold, which represents the maximum allowed time interval from the occurrence of the fire to the first arrival of resources at building i. If the time it takes for resources to arrive at building i exceeds , then the building will be penalized.

[0094] In one embodiment of the present application, based on the alternative strategy, the total loss is determined by the sum of the risk loss, resource mobilization cost, and timeout penalty. By adjusting the alternative strategy to minimize the total loss, a final resource allocation strategy is output. The risk loss considers the risk loss posed by uncovered buildings at different points in time; the resource mobilization cost calculates the activity cost of all resources at different points in time; and the timeout penalty penalizes when the time it takes for resources to reach a building exceeds a response time threshold.

[0095] The above process generates multiple alternative resource allocation strategies based on the risk propagation matrix and acquired resource data, and selects the optimal strategy by minimizing total losses. Reinforcement learning algorithms can adapt to dynamically changing environments, such as the spread of fire and resource consumption, to generate more robust and flexible scheduling solutions. This effectively reduces cumulative risk losses, resource mobilization costs, and timeout penalties, thereby improving the efficiency and effectiveness of resource scheduling. It can generate the optimal resource allocation strategy while meeting scheduling cost constraints and response time thresholds. This ensures efficient and effective resource allocation, allowing resources to be used to their maximum potential in the shortest possible time.

[0096] S150, based on the resource allocation strategies of the adjacent regions and the preset inter-regional cooperation protocol, a scheduling strategy is determined through Nash equilibrium.

[0097] In an embodiment of the present application, resource allocation strategies formulated by adjacent regions based on their own risk conditions and resources are obtained. These strategies reflect the resource deployment tendencies of each region under independent decision-making. At the same time, a preset inter-regional cooperation agreement is included, which clarifies rules such as resource sharing, mutual assistance mechanisms, and conflict resolution principles, providing a framework for cross-regional collaboration. The multi-regional resource allocation problem is modeled as a game scenario, in which each region is regarded as a rational participant, aiming to maximize its own utility. By defining a utility function that comprehensively considers local losses, cross-regional cooperation benefits, and resource conflict penalties, the benefits of each region under different strategy combinations are quantitatively evaluated. With the help of the Nash equilibrium concept in game theory, a stable scheduling strategy is iteratively solved. This strategy ensures that given the strategies of other regions, no region can obtain higher benefits by unilaterally adjusting its strategy, thereby achieving a balance between global optimality and regional individual rationality.

[0098] In one embodiment of the present application, based on the resource allocation strategies of adjacent regions and a preset inter-regional cooperation protocol, a scheduling strategy is determined through Nash equilibrium, including:

[0099] Obtain resource allocation strategies for adjacent regions and pre-set inter-regional collaboration agreements;

[0100] Determine the utility function based on the resource allocation strategy of adjacent regions and the preset inter-regional cooperation agreement;

[0101] The utility function is iterated by an alternating direction multiplier method to determine a scheduling strategy.

[0102] In one embodiment of the present application, resource allocation policies of neighboring regions are obtained; these policies describe how other regions allocate their resources, and an inter-region collaboration protocol A is obtained, which defines how regions collaborate and share resources.

[0103] In one embodiment of the present application, the utility function is determined based on the resource allocation strategies of adjacent regions and the preset inter-regional cooperation agreement. In this process, for each region i, its utility function is calculated based on the given strategy (the resource allocation strategy of the region) and the strategies of other regions (the set of resource allocation strategies of other regions except region i). for:

[0104]

[0105] in, represents the utility function of region i, where is the resource allocation strategy of region i, is the set of resource allocation strategies for other regions except region i; Represents the weight coefficient of local loss; represents the local loss of region i, which means the loss suffered by region i itself during the resource scheduling process; represents the weight coefficient of cross-regional collaboration benefits, represents the cross-regional collaboration benefits, represents the weight coefficient of the resource conflict penalty term, represents the conflict penalty for resources being used by multiple regions. By measuring the benefits of each region under different policy combinations, the utility function comprehensively considers the local losses of region i, the benefits of cross-region collaboration, and the conflict penalty for resources being used by multiple regions.

[0106] In one embodiment of the present application, the utility function is iterated by the alternating direction multiplier method to determine the scheduling strategy. The utility function is iterated by the alternating direction multiplier method to obtain the Nash equilibrium strategy. The Nash equilibrium strategy satisfies the resource allocation strategy for each region. Its corresponding other regional policies are fixed as When The strategy that reaches the maximum value ultimately achieves strategy optimization.

[0107] With other participants' strategies fixed, no participant can unilaterally change their strategy to improve their own benefits. A Nash equilibrium strategy achieves optimal global resource allocation because no region has an incentive to deviate from it. By solving a Nash equilibrium strategy, we can achieve optimal global resource allocation, improving overall resource utilization efficiency and system performance. This helps to more effectively schedule and allocate resources and reduce losses in emergency situations such as fires.

[0108] S160: Based on a preset fire protection system, determine a target terminal related to the dispatching strategy, and send a dispatching instruction generated by the dispatching strategy to the target terminal.

[0109] In the embodiments of the present application, a comprehensive analysis of the specific fire situation, such as fire size, direction of spread, and affected area, is conducted based on the generated dispatch strategy. This analysis is combined with the target terminal's attribute information, such as the type of firefighting equipment (fire truck, drone, firefighting robot, etc.), functional characteristics (firefighting, rescue, reconnaissance, etc.), current status (location, resource reserves, availability, etc.), and historical task execution status, to accurately determine the target terminal most suitable for executing the current dispatch task. Once the target terminal is determined, the dispatch strategy generates detailed dispatch instructions, including key information such as the specific task content (firefighting, rescue, alert, etc.), route of action, required resources, and expected completion time, using its internal, stable and reliable communication network, in real time and accurately transmit them to the target terminal.

[0110] During the transmission process, technical means such as encrypted transmission and data verification can be used to ensure the security and integrity of the dispatch instructions and prevent them from being tampered with or lost. Once the target terminal receives the dispatch instruction, it will immediately initiate the corresponding task execution process and quickly respond to the fire, thus ensuring that the fire protection system can respond to the fire quickly and effectively, minimizing the damage caused by the fire.

[0111] In one embodiment of the present application, based on a preset fire protection system, determining a target terminal associated with the dispatching strategy, and sending a dispatching instruction generated by the dispatching strategy to the target terminal includes:

[0112] Based on a preset fire protection system, determining a target terminal related to the dispatching strategy;

[0113] generating a scheduling instruction based on the scheduling policy and the target terminal;

[0114] The scheduling instruction is sent to the target terminal.

[0115] In one embodiment of the present application, target terminals required to perform specific tasks are identified based on the generated dispatch policy. These target terminals may be fire trucks, drones, rescue teams, or other firefighting equipment. By comprehensively considering the location, scale, and development trends of the fire, as well as the current status of each terminal (such as location, resource reserves, and response time), the target terminal most suitable for the current task is precisely matched.

[0116] After determining the target terminal, detailed dispatch instructions are generated based on the dispatch strategy and the target terminal's characteristics. These instructions include, but are not limited to, the specific content of the mission (e.g., firefighting, rescue, alert), the route of action, the required resources (e.g., the amount of water, foam, rescue equipment), and the expected completion time. The generation of dispatch instructions is designed to ensure that the target terminal can clearly and accurately understand the mission requirements and respond promptly accordingly.

[0117] Generated dispatch instructions are sent to the target terminal in real time via the communication network, ensuring secure and reliable transmission of instructions while minimizing delays in their delivery. Upon receiving the dispatch instructions, the target terminal initiates the corresponding task execution process, such as sending a fire truck to the fire scene, launching a drone for reconnaissance, or assembling a rescue team. This ensures that the firefighting system can respond quickly and effectively to the fire, minimizing damage caused by the fire.

[0118] The above process, based on the preset firefighting system, determines the target terminals associated with the dispatch strategy and sends dispatch instructions to them. This ensures the timeliness and accuracy of resource scheduling. The determination of target terminals takes into account the terminals' current status and task execution capabilities. The issuance of dispatch instructions ensures that the terminals can respond and execute the dispatch tasks promptly, thus achieving rapid and effective dispatch of firefighting resources and ultimately achieving strategy optimization.

[0119] The technical solution of this application obtains first data through intelligent sensors, performs multi-source data fusion on the first data, and generates a first tensor composed of a spatial map through neural network and modeling processing; extracts node features and edge features from the first tensor and the spatial map, performs linear transformation on the node features and edge features based on a learnable weight matrix and activation function, and determines the risk propagation matrix of the fire situation; generates alternative resource allocation strategies based on the risk propagation matrix and the obtained resource data, and minimizes the total loss of the alternative strategies to generate a resource allocation strategy; determines the scheduling strategy through Nash equilibrium based on the resource allocation strategies of adjacent regions and a preset inter-regional cooperation agreement; determines the target terminal related to the scheduling strategy based on the preset fire protection system, and sends the scheduling instructions generated by the scheduling strategy to the target terminal. The above process achieves accurate risk prediction through multi-source data fusion and spatiotemporal modeling, combines dynamic resource allocation with cross-regional collaborative scheduling, and constructs a closed-loop system from data perception to instruction execution, effectively reducing the risk of fire spread and emergency response costs, and improving the efficiency and effectiveness of fire protection resource scheduling.

[0120] The following introduces an embodiment of the AI-driven resource scheduling system for smart firefighting of the present application, which can be used to execute the AI-driven resource scheduling method for smart firefighting in the above-mentioned embodiment of the present application. It can be understood that the AI-driven resource scheduling system for smart firefighting can be a computer program (including program code) running on a computer device. For example, the AI-driven resource scheduling system for smart firefighting is an application software; the AI-driven resource scheduling system for smart firefighting can be used to execute the corresponding steps in the method provided in the embodiment of the present application. For details not disclosed in the embodiment of the AI-driven resource scheduling system for smart firefighting of the present application, please refer to the embodiment of the AI-driven resource scheduling method for smart firefighting described above in the present application.

[0121] Figure 3 A block diagram of an AI-driven resource scheduling system for smart firefighting according to an embodiment of the present application is shown.

[0122] Reference Figure 3 As shown, according to one embodiment of the present application, an AI-driven resource scheduling system for smart firefighting includes:

[0123] An acquiring unit 310 is configured to acquire first data through a smart sensor, wherein the first data includes environmental data and satellite data;

[0124] A coupling unit 320 is configured to perform multi-source data fusion on the first data and generate a first tensor consisting of a spatial atlas through neural network and modeling processing;

[0125] A feature unit 330 is configured to extract node features and edge features from the first tensor and the spatial map, perform linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determine a risk propagation matrix of the fire situation;

[0126] A strategy unit 340 is configured to generate alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimize the total loss of the alternative strategies to generate a resource allocation strategy;

[0127] The scheduling unit 350 is configured to determine a scheduling strategy through Nash equilibrium based on resource allocation strategies of adjacent regions and a preset inter-region cooperation protocol;

[0128] The instruction unit 360 is configured to determine a target terminal associated with the dispatching strategy based on a preset fire protection system, and send a dispatching instruction generated by the dispatching strategy to the target terminal.

[0129] In the present application, based on the aforementioned scheme, the multi-source data fusion is performed on the first data, and a first tensor composed of a spatial map is generated through neural network and modeling processing, including: based on a preset spatiotemporal convolution kernel, three-dimensional convolution processing is performed on the first data to generate second data; dynamic weight coefficients are generated based on the attention mechanism, and the second data is processed by the dynamic weight coefficients to generate third data; satellite data in the first data is modeled to generate a geographic topology, and the geographic topology is processed through a graph neural network to generate a spatial map; based on the third data and the spatial map, a first tensor composed of a spatial map is generated.

[0130] In the present application, based on the aforementioned scheme, node features and edge features are extracted from the first tensor and the spatial map, and the node features and edge features are linearly transformed based on the learnable weight matrix and the activation function to determine the risk propagation matrix of the fire situation, including: based on the preset graph attention network, the information of the neighbor nodes in the spatial map is aggregated to generate node features; the edge features between the neighbor nodes are extracted from the first tensor and the spatial map, and the node features and the edge features are spliced ​​to generate a first feature; the first feature is linearly transformed based on the first learnable weight matrix and the first activation function to generate a second feature; the second feature is linearly transformed based on the second learnable weight matrix and the second activation function to obtain the risk propagation matrix of the fire situation.

[0131] In the present application, based on the aforementioned scheme, the alternative strategies for resource allocation are generated based on the risk propagation matrix and the acquired resource data, and the total loss of the alternative strategies is minimized to generate a resource allocation strategy, including: generating an alternative strategy for resource allocation based on the risk propagation matrix and the acquired resource data; determining the risk loss based on the risk parameters of the building at a set time in the risk propagation matrix; determining the resource mobilization cost based on the time-varying cost function corresponding to the resources in the alternative strategy at each time; determining the timeout penalty item based on the time when the resources in the alternative strategy arrive at the building; constructing a total loss based on the risk loss, the resource mobilization cost and the timeout penalty item; minimizing the total loss by adjusting the alternative strategy, and outputting the final resource allocation strategy.

[0132] In the present application, based on the above solution, the risk loss is determined based on the risk parameters of the building at the set time in the risk propagation matrix, including: determining the risk loss based on the risk parameters of the building at the set time in the risk propagation matrix for:

[0133]

[0134] Among them, t and T represent the total duration of the moment and strategy prediction respectively. represents the preset discount factor, represents the risk parameter of building i at time t in the risk propagation matrix, i and N represent the identification and total number of buildings in the risk propagation matrix respectively. Represents the indicator function.

[0135] In the present application, based on the aforementioned scheme, the resource allocation strategy based on adjacent areas and the preset inter-regional cooperation protocol determine the scheduling strategy through Nash equilibrium, including: obtaining the resource allocation strategy of adjacent areas and the preset inter-regional cooperation protocol; determining the utility function based on the resource allocation strategy of adjacent areas and the preset inter-regional cooperation protocol; iterating the utility function through the alternating direction multiplier method to determine the scheduling strategy.

[0136] In the present application, based on the aforementioned scheme, the preset fire protection system determines the target terminal related to the scheduling strategy, and sends the scheduling instruction generated by the scheduling strategy to the target terminal, including: based on the preset fire protection system, determining the target terminal related to the scheduling strategy; generating a scheduling instruction based on the scheduling strategy and the target terminal; and sending the scheduling instruction to the target terminal.

[0137] The technical solution of this application obtains first data through intelligent sensors, performs multi-source data fusion on the first data, and generates a first tensor composed of a spatial map through neural network and modeling processing; extracts node features and edge features from the first tensor and the spatial map, performs linear transformation on the node features and edge features based on a learnable weight matrix and activation function, and determines the risk propagation matrix of the fire situation; generates alternative resource allocation strategies based on the risk propagation matrix and the obtained resource data, and minimizes the total loss of the alternative strategies to generate a resource allocation strategy; determines the scheduling strategy through Nash equilibrium based on the resource allocation strategies of adjacent regions and a preset inter-regional cooperation agreement; determines the target terminal related to the scheduling strategy based on the preset fire protection system, and sends the scheduling instructions generated by the scheduling strategy to the target terminal. The above process achieves accurate risk prediction through multi-source data fusion and spatiotemporal modeling, combines dynamic resource allocation with cross-regional collaborative scheduling, and constructs a closed-loop system from data perception to instruction execution, effectively reducing the risk of fire spread and emergency response costs, and improving the efficiency and effectiveness of fire protection resource scheduling.

[0138] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing an electronic device according to an embodiment of the present application is shown.

[0139] It should be noted that the computer system of the electronic device in this embodiment is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0140] In this embodiment, the computer system includes a central processing unit 401, which can perform various appropriate actions and processes based on programs stored in a read-only memory 402 or programs loaded from a storage unit 408 into a random access memory 403, such as executing the AI-driven resource scheduling method for smart firefighting described in the above embodiment. The random access memory 403 also stores various programs and data required for system operation. The central processing unit 401, the read-only memory 402, and the random access memory 403 are connected to each other via a bus 404. An input / output interface 405 is also connected to the bus 404.

[0141] The following components are connected to the input / output interface 405: an input section 406 including a keyboard, mouse, and the like; an output section 407 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage section 408 including devices such as a hard disk; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 410 as needed, so that computer programs read from the removable media can be installed in the storage section 408 as needed.

[0142] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium, the computer program including a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 409 and / or installed from a removable medium 411. When the computer program is executed by the central processing unit 401, the various functions defined in the system of the present application are performed.

[0143] It should be noted that the computer-readable medium described in the embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present application, a computer-readable signal medium may include a data signal transmitted in baseband or as part of a carrier wave, which carries a computer-readable computer program. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0145] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. In some cases, the names of these units do not constitute limitations on the units themselves.

[0146] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described above.

[0147] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments, or may exist independently without being incorporated into the electronic device. The computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device implements the AI-driven resource scheduling method for smart firefighting described in the above embodiments.

[0148] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the application, the features and functions of two or more modules or units described above can be concretized in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0149] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0150] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art that are not disclosed herein.

[0151] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An AI-driven resource scheduling method for smart firefighting, characterized in that: include: Acquire first data through a smart sensor, wherein the first data includes environmental data and satellite data; Performing multi-source data fusion on the first data, and generating a first tensor consisting of a spatial map through neural network and modeling processing; Extracting node features and edge features from the first tensor and the spatial map, performing linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determining a risk propagation matrix of the fire situation; generating alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimizing the total loss of the alternative strategies to generate a resource allocation strategy; Based on the resource allocation strategies of adjacent regions and the preset inter-regional cooperation agreement, the scheduling strategy is determined through Nash equilibrium; Based on a preset fire protection system, a target terminal related to the dispatching strategy is determined, and a dispatching instruction generated by the dispatching strategy is sent to the target terminal.

2. The AI-driven resource scheduling method for smart firefighting according to claim 1 is characterized in that: Perform multi-source data fusion on the first data, and generate a first tensor consisting of a spatial atlas through neural network and modeling processing, including: Performing three-dimensional convolution processing on the first data based on a preset spatiotemporal convolution kernel to generate second data; generating a dynamic weight coefficient based on an attention mechanism, and processing the second data using the dynamic weight coefficient to generate third data; Modeling the satellite data in the first data to generate a geographic topology, and processing the geographic topology through a graph neural network to generate a spatial atlas; Based on the third data and the spatial atlas, a first tensor consisting of a spatial atlas is generated.

3. The AI-driven resource scheduling method for smart firefighting according to claim 1 is characterized in that: Extracting node features and edge features from the first tensor and the spatial map, performing linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determining a risk propagation matrix of the fire situation, including: Based on a preset graph attention network, information of neighbor nodes in the spatial graph is aggregated to generate node features; Extracting edge features between the neighboring nodes from the first tensor and the spatial graph, and concatenating the node features and the edge features to generate a first feature; Performing a linear transformation on the first feature based on a first learnable weight matrix and a first activation function to generate a second feature; A linear transformation is performed on the second feature based on a second learnable weight matrix and a second activation function to obtain a risk propagation matrix of the fire situation.

4. The AI-driven resource scheduling method for smart firefighting according to claim 1 is characterized in that: Based on the risk propagation matrix and the acquired resource data, alternative strategies for resource allocation are generated, and the total loss of the alternative strategies is minimized to generate a resource allocation strategy, including: generating alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data; Determining risk losses based on risk parameters of the building at a set time in the risk propagation matrix; Determining resource mobilization costs based on a time-varying cost function corresponding to resources in the alternative strategy at each time; Determining a timeout penalty item based on the time when the resources in the alternative strategy arrive at the building; A total loss is calculated based on the risk loss, the resource mobilization cost, and the overtime penalty item; By adjusting the alternative strategies, the total loss is minimized and a final resource allocation strategy is output.

5. The AI-driven resource scheduling method for smart firefighting according to claim 4 is characterized in that: Based on the risk parameters of the building at a set time in the risk propagation matrix, the risk loss is determined, including: Determine the risk loss based on the risk parameters of the building at a set time in the risk propagation matrix for: ； Among them, t and T represent the total duration of the moment and strategy prediction respectively. represents the preset discount factor, represents the risk parameter of building i at time t in the risk propagation matrix, i and N represent the identification and total number of buildings in the risk propagation matrix respectively. Represents the indicator function.

6. The AI-driven resource scheduling method for smart firefighting according to claim 1 is characterized in that: Based on the resource allocation strategies of adjacent regions and the preset inter-regional collaboration protocol, the scheduling strategy is determined through Nash equilibrium, including: Obtain resource allocation strategies for adjacent regions and pre-set inter-regional collaboration agreements; Determine the utility function based on the resource allocation strategy of adjacent regions and the preset inter-regional cooperation agreement; The utility function is iterated by an alternating direction multiplier method to determine a scheduling strategy.

7. The AI-driven resource scheduling method for smart firefighting according to claim 1 is characterized in that: Determining a target terminal related to the dispatching strategy based on a preset fire protection system, and sending a dispatching instruction generated by the dispatching strategy to the target terminal, including: Based on a preset fire protection system, determining a target terminal related to the dispatching strategy; generating a scheduling instruction based on the scheduling policy and the target terminal; The scheduling instruction is sent to the target terminal.

8. An AI-driven resource scheduling system for smart firefighting, characterized by: include: an acquiring unit, configured to acquire first data through a smart sensor, wherein the first data includes environmental data and satellite data; a coupling unit, configured to perform multi-source data fusion on the first data, and generate a first tensor consisting of a spatial atlas through neural network and modeling processing; a feature unit, configured to extract node features and edge features from the first tensor and the spatial map, perform linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determine a risk propagation matrix of the fire situation; a strategy unit, configured to generate alternative strategies for resource allocation based on the risk propagation matrix and the acquired resource data, and minimize the total loss of the alternative strategies to generate a resource allocation strategy; The scheduling unit is used to determine the scheduling strategy through Nash equilibrium based on the resource allocation strategy of adjacent areas and the preset inter-area cooperation protocol; The instruction unit is used to determine the target terminal related to the dispatching strategy based on the preset fire protection system, and send the dispatching instruction generated by the dispatching strategy to the target terminal.

9. The AI-driven resource scheduling system for smart firefighting according to claim 7 is characterized in that: Perform multi-source data fusion on the first data, and generate a first tensor consisting of a spatial atlas through neural network and modeling processing, including: Performing three-dimensional convolution processing on the first data based on a preset spatiotemporal convolution kernel to generate second data; generating a dynamic weight coefficient based on an attention mechanism, and processing the second data using the dynamic weight coefficient to generate third data; Modeling the satellite data in the first data to generate a geographic topology, and processing the geographic topology through a graph neural network to generate a spatial atlas; Based on the third data and the spatial atlas, a first tensor consisting of a spatial atlas is generated.

10. The AI-driven resource scheduling system for smart firefighting according to claim 7 is characterized in that: Extracting node features and edge features from the first tensor and the spatial map, performing linear transformation on the node features and edge features based on a learnable weight matrix and an activation function, and determining a risk propagation matrix of the fire situation, including: Based on a preset graph attention network, information of neighbor nodes in the spatial graph is aggregated to generate node features; Extracting edge features between the neighboring nodes from the first tensor and the spatial graph, and concatenating the node features and the edge features to generate a first feature; Performing a linear transformation on the first feature based on a first learnable weight matrix and a first activation function to generate a second feature; A linear transformation is performed on the second feature based on a second learnable weight matrix and a second activation function to obtain a risk propagation matrix of the fire situation.

Images