An emergency event state dynamic change-based plan publishing method

Abstract

The application discloses a plan publishing method based on dynamic change of emergency state, relates to the technical field of intelligent emergency, and comprises the following steps: activating a pre-constructed knowledge graph framework according to a real-time acquired emergency type identifier, and forming an initialized dynamic knowledge graph; capturing a feedback data stream in real time, guiding the feedback data stream into a state updating layer of the state evolution knowledge graph, and generating an updated state evolution knowledge graph; generating an evacuation instruction based on the updated state evolution knowledge graph, switching to a Mesh network broadcast path to change information through a damaged base station, generating an execution effect, and triggering the activation of the pre-constructed knowledge graph framework again as a new emergency type identifier. The application realizes the dynamic generation of an intelligent countermeasure plan by implanting artificial disturbance simulation nodes in the state evolution knowledge graph and deducing diffusion paths.

Description

Technical Field

[0001] This invention relates to the field of intelligent emergency response technology, and in particular to a method for issuing contingency plans based on the dynamic changes in the state of an emergency. Background Technology

[0002] Currently, knowledge graph-based emergency management is widely used in the field of emergency response. By fusing pre-built event type frameworks with real-time data, static modeling of disaster situations can be achieved. Mainstream methods employ multi-source sensor networks to collect environmental parameters, combine this with Kalman filtering for data fusion, and utilize graph neural networks to update entity relationship weights. In the contingency plan generation stage, path deduction based on reinforcement learning can simulate simple disturbance diffusion processes, while blockchain-enabled command distribution mechanisms improve the reliability of equipment control.

[0003] Existing technologies still have shortcomings: traditional methods rely on predefined rules to update the knowledge graph, which cannot adapt in real time to sudden evolutions caused by human disturbances. When a sudden event damages a base station, static contingency plans cannot automatically switch to broadcast path change instructions via the Mesh network, resulting in delayed evacuation instructions, and the execution effect data is difficult to convert into new event identifiers in real time to trigger iterative updates and optimizations. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for issuing contingency plans based on the dynamic changes in the state of an emergency to solve the problem of delayed dynamic situational awareness and decision-making during the response to an emergency.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for issuing contingency plans based on dynamic changes in the state of an emergency, which includes activating a pre-built knowledge graph framework according to real-time acquired emergency type identifiers to form an initial dynamic knowledge graph;

[0008] Collect environmental parameters and equipment status data of the area where the emergency occurred, inject them into the data fusion layer of the initial dynamic knowledge graph, update entities and relation edges, and generate a state evolution knowledge graph;

[0009] By embedding artificial disturbance simulation nodes into the state evolution knowledge graph, the diffusion path of artificial disturbance simulation nodes along the emergency facility topology is deduced, and countermeasure plans are generated.

[0010] The countermeasure plan is broken down into a set of executable instructions for the equipment, which drives the physical equipment to execute and generates a feedback data stream.

[0011] Real-time capture of feedback data streams; importing the feedback data streams into the state update layer of the state evolution knowledge graph to generate an updated state evolution knowledge graph.

[0012] Based on the updated state evolution knowledge graph, evacuation instructions are generated. By switching from the damaged base station to the Mesh network, path change information is broadcast, and execution results are generated. These results serve as identifiers for new emergency types, triggering the reactivation of the pre-built knowledge graph framework.

[0013] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for forming an initial dynamic knowledge graph are as follows:

[0014] Based on the event type identifier, the corresponding knowledge graph framework is matched from the pre-built knowledge graph framework library to obtain the framework structure containing static entity sets and static relation edges;

[0015] Based on the framework structure, the geographic coordinates and impact radius of the emergency are collected, and emergency facility entities located within the spatial range are extracted from the pre-built emergency resource database and merged with the static entity set to generate an initial entity set.

[0016] Based on the initial entity set, access real-time traffic flow and population density data streams, calculate the dynamic attribute parameters of each entity in the initial entity set, inject the corresponding entity attribute set, and generate an entity set with injected dynamic attributes.

[0017] Based on the injected dynamic attribute entity set, the static relation edges and newly generated relation edges in the knowledge graph framework are dynamically weighted to generate a weighted relation edge set.

[0018] Perform force-directed graph layout and connectivity verification on the weighted relation edge set and the initial entity set, and output the initial dynamic knowledge graph.

[0019] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the knowledge graph framework includes the relationships between event type entities, geographical environment entities, and emergency resource entities.

[0020] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for generating the state evolution knowledge graph are as follows:

[0021] By deploying an IoT sensor network in the event area, multi-source heterogeneous data on environmental parameters, equipment status, and infrastructure status are collected and standardized to generate standardized multi-source heterogeneous data.

[0022] The Kalman filter algorithm is used to fuse and denoise standardized multi-source heterogeneous data to generate high-quality fused data.

[0023] High-quality fused data is injected into the data fusion layer of the initial dynamic knowledge graph, and the entity attributes in the initial dynamic knowledge graph are updated according to the entity type mapping rules.

[0024] Based on the updated entity attributes, the dynamic weights of the relation edges in the initialized dynamic knowledge graph are dynamically calculated, and the relation edges are updated accordingly.

[0025] Based on the updated relation edges and updated entity attributes, a state evolution knowledge graph is generated.

[0026] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the artificial disturbance simulation nodes include emergency resource competition, communication interference sources, and information anomaly source locations.

[0027] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for generating the countermeasure plan are as follows:

[0028] Based on three types of human-induced disturbances—emergency resource competition, communication interference sources, and information anomaly source locations—corresponding multimodal disturbance nodes are created in the state evolution knowledge graph, and static and dynamic attributes are configured for each disturbance node.

[0029] Based on multimodal disturbance nodes, a multi-agent reinforcement learning algorithm is used to deduce the diffusion path of various disturbance nodes along the topology of emergency facilities, and generate disturbance diffusion simulation results.

[0030] A structural causal model is used to conduct counterfactual intervention analysis on the disturbance diffusion simulation results, evaluate the effects of different intervention measures against emergency resource competition, communication interference sources and information anomaly source locations, and generate initial countermeasure plans for emergency resource competition, communication interference sources and information anomaly source locations.

[0031] The initial countermeasure plan is subjected to multi-dimensional feasibility verification and iterative optimization to generate a countermeasure plan.

[0032] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the device's executable instruction set includes signal jammer activation coordinates, scheduling and deployment GPS coordinate sequences, and backup communication channel activation instructions.

[0033] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for generating the feedback data stream are as follows:

[0034] Formal verification and blockchain smart contracts are used to decompose the countermeasure plan into signal jammer activation coordinates, scheduling and control GPS coordinate sequence, and backup communication channel activation instructions, generate an executable instruction sequence, and complete secure distribution verification.

[0035] Based on the distributed sequence of executable instructions, the multimodal physical devices are driven to execute through a federated learning collaborative control model, and execution status data is collected in real time.

[0036] Evidence theory is used to fuse and assess the quality of execution status data, generating a feedback data stream.

[0037] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for generating the updated state evolution knowledge graph are as follows:

[0038] Real-time capture of feedback data streams, dynamic assimilation of the feedback data streams, and fusion with the current state of the state evolution knowledge graph to generate a preliminary updated state;

[0039] Multi-scale feature extraction and multi-source data fusion are performed on the initial updated state to generate an optimized feature vector;

[0040] Based on the optimized feature vectors, the state evolution knowledge graph is optimized in topology and incrementally updated to generate an updated state evolution knowledge graph.

[0041] As a preferred embodiment of the emergency response plan release method based on dynamic changes in emergency event status as described in this invention, the specific steps for re-triggering and activating the pre-built knowledge graph framework are as follows:

[0042] Based on the updated state evolution knowledge graph, spatiotemporal features are extracted through a spatiotemporal graph neural network to generate evacuation instructions;

[0043] In response to the base station damage status, and in conjunction with evacuation instructions, a Mesh broadcast routing table is dynamically constructed. The evacuation instructions and the Mesh broadcast routing table are merged into a unified data packet, and path change information is broadcast through the Mesh network.

[0044] Collect device response data and on-site feedback after broadcasting, generate execution results, and use the execution results as a new emergency type identifier to re-trigger and activate the pre-built knowledge graph framework.

[0045] The beneficial effects of this invention are as follows: by embedding artificially disturbed simulation nodes into the state evolution knowledge graph and deducing the diffusion path, the dynamic generation of intelligent countermeasure plans is realized; for heterogeneous disturbances such as emergency resource competition and communication interference sources, disturbance nodes with static and dynamic attributes are constructed, breaking through the limitations of traditional single entity modeling; based on multi-agent reinforcement learning algorithms, the nonlinear diffusion process of disturbance nodes along the topology of emergency facilities is simulated; a structural causal model is used to evaluate the blocking effect of different intervention measures on the diffusion path, and the optimal countermeasure plan with minimum resource consumption is output. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating the method for issuing contingency plans based on dynamic changes in the state of an emergency.

[0048] Figure 2 A flowchart for initializing the dynamic knowledge graph formation.

[0049] Figure 3 The flowchart for generating state evolution knowledge graphs and countermeasure plans.

[0050] Figure 4 A flowchart for instruction execution and feedback updates. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] Reference Figures 1-4This is one embodiment of the present invention, which provides a method for issuing contingency plans based on dynamic changes in the state of an emergency, including the following steps:

[0055] S1. Activate the pre-built knowledge graph framework based on the real-time acquired event type identifier to form an initial dynamic knowledge graph.

[0056] S1.1: Based on the event type identifier, match the corresponding knowledge graph framework from the pre-built knowledge graph framework library to obtain the framework structure containing static entity sets and static relation edges;

[0057] Specifically, environmental sensor data, video surveillance data, audio data, IoT data, and social network data are collected in real time through a sensor network deployed at the event site, and an intelligent analysis engine automatically generates an emergency type identifier.

[0058] Based on the event type identifier, domain analysis is performed on various events in advance to determine the type of event to be addressed. Response decisions are formulated in advance and stored according to event type. A pre-built knowledge graph framework library is generated. Precise string matching is performed in the pre-built knowledge graph framework library. When a match is successful, the corresponding knowledge graph framework is output, which contains the following predefined elements:

[0059] The static entity set includes event type entities, geographic environment entities, and emergency resource entities;

[0060] Static relation edges contain predefined associations between entities;

[0061] Based on the static entity set and static relation edges, obtain the frame structure containing the static entity set and static relation edges.

[0062] S1.2: Based on the framework structure, collect the geographical coordinates and impact radius of the emergency, extract emergency facility entities located within the spatial range from the pre-built emergency resource database, and merge them with the static entity set to generate an initial entity set;

[0063] Specifically, based on the framework structure, the geographic coordinates and impact radius of the emergency are collected;

[0064] By collecting and standardizing static data from departments such as fire protection, medical care, transportation and materials management in advance, cleaning and integrating the attributes and precise geographic coordinates of various emergency facilities, storing them, generating a pre-built emergency resource database, performing circular spatial range queries in the pre-built emergency resource database, and extracting emergency facility entities that meet the query conditions.

[0065] The extracted emergency facility entities are combined with the static entity set contained in the framework structure, duplicate entities are removed, and an initial entity set is generated.

[0066] It should be noted that the query criteria are that the distance between the facility coordinates and the geographical coordinates of the emergency is less than or equal to the radius of influence.

[0067] S1.3: Based on the initial entity set, access the real-time traffic flow and population density data stream, calculate the dynamic attribute parameters of each entity in the initial entity set, inject the corresponding entity attribute set, and generate the entity set with injected dynamic attributes.

[0068] Specifically, the dynamic attribute parameters of each entity in the initial entity set are written into the attribute fields of the corresponding entity. After the attributes of all entities are updated, an entity set with injected dynamic attributes is generated.

[0069] Based on the initial entity set, real-time traffic flow and population density data streams are accessed, and the dynamic attribute parameters of each entity in the initial entity set are calculated. The expression is as follows:

[0070] ;

[0071] In the formula, Indicates the current entity At the current time Dynamic attribute values ​​at time, Indicates the current time. Indicates the current entity index. This represents an index of road connectivity. Represents the first entity in the initial entity set. The node corresponding to the current entity Represents the first entity in the initial entity set. The actual connectivity of the road associated with the node corresponding to the current entity. This represents the real-time traffic flow coefficient. Indicates the maximum road connectivity. Indicators representing population carrying capacity Represents the first entity in the initial entity set. The actual population load of the node corresponding to the current entity. This represents the real-time population density coefficient. Indicates the maximum population load density. Represents the time decay term. Indicates the time decay coefficient. Indicates the update time interval.

[0072] It should be noted that, and If the dimensions are the same, they can be eliminated directly. and If the dimensions are the same, they can be eliminated directly. and Dimensionless and Dimensionless and The inverse dimensions can be directly eliminated, resulting in the final output. Since it is dimensionless, we maintain dimensional consistency.

[0073] The real-time traffic flow coefficient is derived from the real-time vehicle speed and density data collected by road network sensors, with an example value of 0.93; the real-time population density coefficient is based on the statistics of the number of mobile signaling base station connections, with an example value of 0.85; the time decay coefficient is derived from a preset constant and reflects the data timeliness decay rate, with an example value of 0.1.

[0074] S1.4: Based on the injected dynamic attribute entity set, perform dynamic weight initialization on the static relation edges and newly generated relation edges in the knowledge graph framework to generate a weighted relation edge set;

[0075] Specifically, based on the injected dynamic attribute entity set, the relation edge weight initialization is performed. For static relation edges in the knowledge graph framework, the preset weight values ​​in the knowledge graph framework are directly used for assignment.

[0076] The spatial connection relationship between emergency facility entities and geographic environment entities is newly established through spatial proximity analysis. The Euclidean distance is calculated based on the geographic coordinates of the emergency facility entities and geographic environment entities. The preset weight value is set as the reciprocal of the Euclidean distance to obtain the newly generated relationship edge.

[0077] After assigning weights to all static relation edges and newly generated relation edges, a weighted relation edge set is generated.

[0078] It should be noted that the preset weight values ​​were set through statistical analysis of a large number of historical emergency cases.

[0079] S1.5: Perform force-directed graph layout and connectivity verification on the weighted relation edge set and the initial entity set, and output the initial dynamic knowledge graph.

[0080] Specifically, the Fruchterman-Reingold algorithm is used to arrange the entities, directly converting the weight values ​​of the weighted relation edge set into the magnitude of the attraction between entities, and iteratively adjusting the entity positions until the layout is stable.

[0081] After the layout is completed, connectivity verification is performed. In the graph structure composed of weighted relation edge sets, all emergency resource entities are selected as verification nodes, and the breadth-first search algorithm is applied to check whether there is a valid path between any two emergency resource entities.

[0082] When all emergency resource entity pairs are connected, the output is an initialized dynamic knowledge graph containing the entity location information after layout and weighted relation edges.

[0083] It should be noted that an effective path refers to a path consisting of continuous relation edges between any two emergency resource entities in a weighted relation edge set.

[0084] S1.6: The knowledge graph framework includes the relationships between event type entities, geographic environment entities, and emergency resource entities.

[0085] It should be noted that the event type entity describes the classification of emergencies, the geographic environment entity includes topographic, hydrological and transportation network data, and the emergency resource entity covers personnel, equipment, materials and facilities; the entities include spatial, functional and management relationships.

[0086] S2. Collect environmental parameters and equipment status data of the area where the emergency occurred, inject them into the data fusion layer of the initial dynamic knowledge graph, update entities and relation edges, and generate a state evolution knowledge graph.

[0087] S2.1: Collect multi-source heterogeneous data on environmental parameters, equipment status, and infrastructure status through an IoT sensor network deployed in the event area, and perform standardization processing to generate standardized multi-source heterogeneous data;

[0088] Specifically, ambient temperature, ambient humidity, electrical equipment voltage, and building tilt angle are collected by temperature sensors, humidity sensors, voltage monitors, and structural health monitoring sensors to form raw multi-source heterogeneous data;

[0089] Convert the binary format sensor data to JSON format, convert the pressure values ​​in millimeters of mercury to kilopascals, convert the temperature values ​​in Fahrenheit to Celsius, and add a uniform timestamp to all data;

[0090] After processing, standardized multi-source heterogeneous data is generated.

[0091] S2.2: The Kalman filter algorithm is used to fuse and denoise standardized multi-source heterogeneous data to generate high-quality fused data;

[0092] Specifically, standardized multi-source heterogeneous data generated by temperature sensors, humidity sensors, voltage monitors, and structural health monitoring sensors are used as the observation input vector for the Kalman filter algorithm, where each sensor data corresponds to one dimension of the observation vector;

[0093] The standard recursive process of Kalman filtering is executed through two stages: state prediction and measurement update. The Kalman gain adjusts the weight ratio of the predicted and observed values, performs time series alignment, and outputs high-quality fused data after noise suppression and data fusion.

[0094] S2.3: Inject high-quality fused data into the data fusion layer of the initial dynamic knowledge graph, and update the entity attributes in the initial dynamic knowledge graph according to the entity type mapping rules;

[0095] Specifically, based on predefined entity type mapping rules, the specific values ​​in the high-quality fused data are assigned to the attribute fields of the corresponding entity types;

[0096] The ambient temperature value is assigned to the "road surface temperature" attribute field of the road entity, and the electrical equipment voltage value is assigned to the "output voltage" attribute field of the emergency power supply entity.

[0097] After updating the values ​​of all matching fields, an initial dynamic knowledge graph of the updated entity attributes is generated.

[0098] It should be noted that the predefined entity type mapping rules refer to the standardized correspondence established during the knowledge graph framework construction phase, which specifies the matching logic between different entity types and sensor data parameters. For example, road entities correspond to environmental parameters, and shelter entities correspond to capacity parameters.

[0099] S2.4: Based on the updated entity attributes, perform dynamic weight calculation on the relation edges in the initialized dynamic knowledge graph and update the relation edges;

[0100] ;

[0101] In the formula, Represents the entity at the starting point of the relation edge. With relational edge endpoint entity Between at the current time Time relation weights, Indicates the index of the entity that starts the relation edge. Indicates the index of the endpoint entity of the relation edge. Represents the entity at the starting point of the relation edge. With relational edge endpoint entity Between in time Time relation weights, Represents a constant term. This represents the sensitivity adjustment coefficient for weight differences. Describes the Euclidean norm. Represents the entity at the starting point of the relation edge. The amount of change in the dynamic attribute value, Represents the terminal entity of the relation edge The amount of change in the dynamic attribute value, Represents the time decay term. This represents the time decay coefficient.

[0102] It should be noted that, Dimensionless Dimensionless Dimensionless Dimensionless It is dimensionless, and the final output is... Since it is dimensionless, we maintain dimensional consistency.

[0103] The weight difference sensitivity adjustment coefficient is derived from historical experience calibration values ​​and is used to adjust the influence of entity attribute changes on relation weights. The example value is 0.5. The time decay coefficient is derived from a preset physical decay constant and reflects the time decay rate of historical weights. The example value is 0.05.

[0104] S2.5: Generate a state evolution knowledge graph based on the updated relation edges and updated entity attributes.

[0105] Specifically, the updated entity attribute set is used as the node feature set, and the updated relation edges are used as the connection relationship set, and the topology is integrated; the output after integration is a state evolution knowledge graph containing the latest entity state and relation weights.

[0106] It should be noted that the state evolution knowledge graph directly inherits the framework structure of the initial dynamic knowledge graph, with only the entity attribute values ​​and relation edge weights changing.

[0107] S3. Implant artificial disturbance simulation nodes into the state evolution knowledge graph, deduce the diffusion path of artificial disturbance simulation nodes along the emergency facility topology, and generate countermeasure plans.

[0108] S3.1: The simulated nodes for artificial disturbances include emergency resource competition, communication interference sources, and information anomaly sources.

[0109] It should be noted that emergency resource competition refers to the concentrated area nodes with resource-plundering behavior characteristics simulated in the state evolution knowledge graph, including coordinate location and real-time scale parameters;

[0110] Communication interference sources refer to electromagnetic radiation source nodes that actively block wireless signal transmission in the state evolution knowledge graph, including the type of interfering device and the current interference intensity parameters.

[0111] Information anomaly source points refer to the source nodes that simulate the creation and spread of false information in the state evolution knowledge graph, including parameters such as the type of dissemination channel and the number of people affected in real time.

[0112] S3.2: Based on the three types of human disturbances—emergency resource competition, communication interference sources, and information anomaly source locations—create corresponding multimodal disturbance nodes in the state evolution knowledge graph, and configure static and dynamic attributes for each disturbance node.

[0113] Specifically, add an entity with the node type "emergency resource competition" to the knowledge graph, configure static attributes including preset geographic coordinates, configure dynamic attributes including real-time scale parameters, and generate emergency resource competition nodes;

[0114] Add a new entity with the node type "communication interference source" to the knowledge graph, configure static attributes including the type of interfering device, configure dynamic attributes including the current interference intensity parameter, and generate a communication interference source node;

[0115] Add a new entity with the node type "Information Anomaly Source Location" to the knowledge graph, configure static attributes including the propagation channel type, configure dynamic attributes including the real-time number of people affected, and generate the Information Anomaly Source Location node.

[0116] After the nodes are generated, spatial association edges are established between each disturbed node and the adjacent emergency facility entities.

[0117] S3.3: Based on multimodal disturbance nodes, a multi-agent reinforcement learning algorithm is used to deduce the diffusion path of various disturbance nodes along the emergency facility topology and generate disturbance diffusion simulation results;

[0118] Specifically, each disturbance node acts as an independent intelligent agent, with the emergency resource competition node intelligent agent selecting its direction of movement based on the topological connections of the road entities;

[0119] The communication interference source node intelligent agent expands the interference range along the communication link topology, and the information anomaly source node intelligent agent spreads its influence through social relationships.

[0120] Each agent executes a preset behavior rule at discrete time steps, records changes in movement trajectory or range of influence, and outputs the perturbation diffusion simulation results containing time series trajectory points.

[0121] It should be noted that the pre-defined behavioral rules refer to the deterministic action strategies defined for each type of disturbance node in multi-agent reinforcement learning. Emergency resource competition nodes move along the road entity topology to the nearest resource-dense area, communication interference source nodes expand the interference range along the communication link topology to the signal coverage area, and information anomaly source nodes spread their influence to the associated community center entity through social relationship edges.

[0122] S3.4: Use a structural causal model to conduct counterfactual intervention analysis on the disturbance diffusion simulation results, evaluate the effects of different intervention measures against emergency resource competition, communication interference sources and information anomaly source locations, and generate initial countermeasure plans against emergency resource competition, communication interference sources and information anomaly source locations;

[0123] It should be noted that the pre-training process of the structural causal model is based on historical emergency case data. The specific steps are as follows: extract historical case data that matches the current emergency type identifier from the historical emergency case database to form a training dataset; construct a causal graph structure using the training dataset, where nodes correspond to entity types in the knowledge graph and edges represent potential causal relationships between entities; and calculate the causal strength parameters using the maximum likelihood estimation method to obtain the structural causal model.

[0124] Specifically, intervention operations are applied to three types of disturbances: competition for emergency resources, sources of communication interference, and sources of information anomalies.

[0125] An intervention operation was performed to delete road connection edges to block competition for emergency resources.

[0126] Intervention operations are performed to modify communication frequency band attributes to suppress communication interference sources;

[0127] Apply intervention operations to weaken the propagation of information anomaly source points by adding information anomaly source points to the edge of the information anomaly source points;

[0128] By re-analyzing the disturbance diffusion path after intervention in the structural causal model, comparing the changes in key indicators before and after intervention, and generating an initial countermeasure plan.

[0129] It should be noted that the changes in key indicators refer to whether emergency resources have reached the target area, the percentage reduction in the scope of interference sources, and the decrease in the number of people affected by the spread of information anomaly sources.

[0130] S3.5: Conduct multi-dimensional feasibility verification and iterative optimization of the initial countermeasure plan to generate a countermeasure plan.

[0131] Specifically, a resource matching check is performed on the resource requirement list in the initial countermeasure plan to verify whether the inventory quantity of on-site emergency resource entities meets the resource requirement list of the plan.

[0132] Compare the implementation time of the contingency plan measures with the time window of the disturbance spread rate. If the implementation time of the contingency plan measures is less than the time window of the disturbance spread rate, then the timeliness verification is passed.

[0133] Identify the mutually exclusive relationships between different contingency plans and measures. If no mutually exclusive or contradictory relationships can be identified between different contingency plans and measures, then verify through conflict detection.

[0134] If the verification fails, an optimization operation is performed. If resources are insufficient, an alternative measure that can be met by inventory is used. If timeliness is insufficient, a parallel execution group is added. If there is a conflict between measures, the execution order is adjusted or a compatible solution is used.

[0135] A countermeasure plan was generated after multiple iterations and optimizations.

[0136] Superiorly, compared to conventional emergency management, it achieves dynamic game theory by embedding multimodal human-induced disturbance nodes, simulates the spread path of malicious behavior based on the topology of emergency facilities, and generates precise countermeasure plans by combining counterfactual intervention analysis of structural causal models. This breaks through the traditional approach's inability to quantify the decision-making process of human-induced disturbances and physical environment interaction, and realizes intelligent emergency response with "human-material-technology" triadic collaboration.

[0137] S4. Decompose the countermeasure plan into a set of executable instructions for the equipment, drive the physical equipment to execute, and generate a feedback data stream.

[0138] S4.1: The set of executable instructions for the device includes signal jammer start coordinates, scheduling and control GPS coordinate sequence, and backup communication channel activation instructions.

[0139] Specifically, the signal jammer activation coordinates refer to the set of geographical coordinates of the physical deployment location of the signal jamming equipment, which is clearly specified in the countermeasure plan, and is used to accurately guide the installation and operation of the equipment.

[0140] The GPS coordinate sequence for dispatch and control refers to the dynamic deployment location coordinate chain of dispatch resources arranged in the order of action in the countermeasure plan, which is used to guide patrol routes and deployment points;

[0141] The backup communication channel activation command refers to the operation command that triggers the activation of backup communication equipment in the countermeasure plan, including the equipment number and activation time parameters.

[0142] S4.2: Using formal verification and blockchain smart contracts, the countermeasure plan is decomposed into signal jammer activation coordinates, scheduling and control GPS coordinate sequence, and backup communication channel activation instructions, generating an executable instruction sequence and completing secure distribution verification;

[0143] Specifically, formal verification is performed on the action logic in the countermeasure plan, and timing logic is used to check the spatiotemporal consistency between the signal jammer activation coordinates, the GPS coordinate sequence of the scheduling and control, and the backup communication channel activation command.

[0144] Once verified, the contingency plan is broken down into atomic instructions, and an executable instruction sequence containing instruction content, execution time window, and authorization signature is generated through a blockchain smart contract.

[0145] The executable instruction sequence is distributed to the execution terminal via an encrypted channel. The terminal returns a receipt containing a blockchain timestamp, and the distribution record is compared with the receipt data to complete the security verification.

[0146] S4.3: Based on the distributed executable instruction sequence, drive the multimodal physical device to execute through a federated learning collaborative control model, and collect execution status data in real time;

[0147] It should be noted that the pre-training process of the federated learning collaborative control model is based on the historical operation record database. The specific process is as follows: collect historical execution data streams of multimodal physical devices to construct a distributed training dataset; train a lightweight control model locally on the device; aggregate the model parameters of each device through a gradient weighted average algorithm; and obtain the trained federated learning collaborative control model after iterative optimization.

[0148] Specifically, after receiving the instruction sequence through the federated learning collaborative control model, the signal jammer device activates the device's operating status according to the start coordinate parameters, the dispatch vehicle moves to the deployment position according to the GPS coordinate sequence, and the backup communication device switches the communication channel according to the start instruction parameters.

[0149] During execution, the device's built-in sensors collect real-time monitoring values ​​of the signal jammer voltage, GPS positioning data of the dispatch vehicle, and channel status values ​​of the backup communication equipment, forming an execution status data stream containing timestamps.

[0150] S4.4: Use evidence theory to fuse and assess the quality of execution status data, and generate a feedback data stream.

[0151] Specifically, predefined reliability assessment rules are assigned to three types of data sources: signal jammer voltage monitoring values, dispatch vehicle GPS positioning data, and backup communication equipment channel status values.

[0152] Among them, voltage monitoring values ​​correspond to rule 1, GPS positioning data correspond to rule 2, and channel status values ​​correspond to rule 3;

[0153] The output reliability values ​​of rules 1, 2 and 3 are normalized by Dempster combination operation and combined into a comprehensive reliability distribution.

[0154] Based on the confidence interval range of the comprehensive confidence distribution and the magnitude of the conflict coefficient, quality is marked according to the preset quality grading table, and a feedback data stream with quality marking is output.

[0155] It should be noted that the predefined confidence assessment rules refer to the confidence quantification standards set for the three types of data sources: signal jammer voltage monitoring values, dispatch vehicle GPS positioning data, and backup communication equipment channel status values. Among them, the confidence of voltage monitoring values ​​is assessed based on the preset voltage stability range, the confidence of GPS positioning data is assessed based on the positioning deviation threshold, and the confidence of channel status values ​​is assessed based on the bit error rate range.

[0156] Example: The voltage monitoring value of the signal jammer is defined as being within the range of [12.4V, 12.8V], with a confidence level of 0.8; the GPS positioning data of the dispatch vehicle is defined as having a positioning deviation of ≤10 meters, with a confidence level of 0.9.

[0157] The preset quality grading table refers to a quality level judgment comparison table set according to the evidence theory fusion results. It includes a fixed mapping relationship between the comprehensive confidence interval, the conflict coefficient and the quality label, and is used to automatically output high confidence labels, medium confidence labels and low confidence labels.

[0158] S5. Capture feedback data stream in real time, import the feedback data stream into the state update layer of the state evolution knowledge graph, and generate the updated state evolution knowledge graph.

[0159] S5.1: Real-time capture of feedback data stream, dynamic assimilation of the feedback data stream, fusion with the current state of the state evolution knowledge graph, and generation of preliminary updated state;

[0160] Specifically, the timestamp field in the feedback data stream is precisely matched with the time dimension of the current state in the state evolution knowledge graph;

[0161] The signal jammer voltage monitoring value and dispatch vehicle GPS positioning data in the feedback data stream are updated to the corresponding entity attribute fields according to the predefined entity attribute mapping rules.

[0162] Among them, the voltage monitoring value of the signal jammer updates the "output voltage" attribute of the emergency power supply entity, and the GPS positioning data of the dispatch vehicle updates the "vehicle position" attribute of the road entity.

[0163] During the update process, the system filters data based on high-confidence, medium-confidence, and low-confidence labels from the feedback data stream, integrates only high-confidence label data, and generates a preliminary update status after updating all matching fields.

[0164] It should be noted that the predefined entity attribute mapping rules refer to the standardized correspondence table established during the knowledge graph framework construction phase, which clearly stipulates the matching logic between the device parameter types in the feedback data stream and the entity attribute fields in the state evolution knowledge graph.

[0165] S5.2: Perform multi-scale feature extraction and multi-source data fusion on the initial update state to generate an optimized feature vector;

[0166] Specifically, features are extracted hierarchically according to spatial scale. The average congestion coefficient of road entities is extracted in the kilometer-level grid, and the load rate distribution of shelter entities is extracted in the hundred-meter-level grid.

[0167] Simultaneously, features are extracted hierarchically according to time scales, and the frequency of equipment status changes is statistically analyzed in minute-level windows, while population density trends are analyzed in hour-level windows.

[0168] Spatial scale features and temporal scale features are combined into a multidimensional feature matrix;

[0169] The Dempster combinatorial operation, which employs evidence theory, integrates the feature values ​​of the average congestion coefficient of road entities, the load rate distribution of shelter entities, the frequency of equipment status changes, and the population density trend, and outputs an optimized feature vector with unified dimensions.

[0170] S5.3: Based on the optimized feature vector, perform topological structure optimization and incremental update on the state evolution knowledge graph to generate the updated state evolution knowledge graph.

[0171] Specifically, the location weight parameters of road entities and shelter entities are updated based on the average congestion coefficient and load rate distribution of the optimized feature vectors.

[0172] Adjust the connection strength of the relationship edges between entities based on the frequency of device state changes and population density trends in the optimized feature vectors;

[0173] When a spatial correlation is detected between a high-congestion area and a high-load area, a new collaborative relationship edge is automatically created between the emergency resource entities, provided that the distance between the relevant emergency resource entities meets a preset distance threshold.

[0174] Static entity attributes are extracted, including event type entities, geographic environment entities, and emergency resource entities. Dynamic entity attributes are extracted, including road entities, shelter entities, emergency power supply entities, and human disturbance nodes. All entity attribute update records are generated.

[0175] Extract the relationship edge connection strength adjustment, add cooperative relationship edges and delete invalid relationship edges to generate the topology adjustment result;

[0176] Integrate all entity attribute update records and topology adjustment results to output the updated state evolution knowledge graph.

[0177] It should be noted that, for example: when the distance between the hospital entity and the shelter entity is less than 3 kilometers and both are in high-load areas, create a "hospital-support-shelter" collaborative edge.

[0178] The preset distance threshold is set based on the impact radius parameter of the sudden event; the example value is 3 kilometers.

[0179] S6. Based on the updated state evolution knowledge graph, generate evacuation instructions, switch to the Mesh network via damaged base stations to broadcast path change information, generate execution effects, and use them as identifiers for new emergency types to re-trigger and activate the pre-built knowledge graph framework.

[0180] S6.1: Based on the updated state evolution knowledge graph, spatiotemporal features are extracted through a spatiotemporal graph neural network to generate evacuation instructions;

[0181] Specifically, the spatial association features between road entities and shelter entities are extracted using graph convolutional layers of a spatiotemporal graph neural network;

[0182] The temporal variation features of congestion coefficient and load rate attributes are extracted by the temporal convolutional layer of the spatiotemporal graph neural network.

[0183] After concatenating spatial correlation features and temporal change features, the feature vector representing the regional risk level and the feasibility of evacuation routes is output through a fully connected layer mapping.

[0184] Based on the distribution of entities in the region whose risk level exceeds the preset risk threshold, evacuation instructions containing avoidance coordinates and recommended routes are generated.

[0185] It should be noted that the preset risk threshold is based on the conflict detection setting of multi-dimensional feasibility verification, and the example value is 0.8.

[0186] S6.2: In response to the base station damage status, combine the evacuation command with the Mesh broadcast routing table to dynamically construct the Mesh broadcast routing table, merge the evacuation command with the Mesh broadcast routing table into a unified data packet, and broadcast the path change information through the Mesh network;

[0187] Specifically, the system reads the running state attributes of communication base station entities in the state evolution knowledge graph, and triggers Mesh network switching when the attribute value is "damaged".

[0188] Obtain the precise geographic coordinates of the communication base station entity and the location data of all available Mesh nodes in the state evolution knowledge graph. Calculate the spherical straight-line distance between each Mesh node and the location of the communication base station. When the straight-line distance does not exceed the preset condition, create a routing table entry for the Mesh node containing a unique node identifier field, a geographic coordinate field, a hop number field, and a signal strength field. Sort all matching routing table entries from high to low signal strength to generate a Mesh broadcast routing table.

[0189] The evacuation instructions and Mesh broadcast routing table are merged into a unified data packet in JSON format;

[0190] The Mesh network broadcast protocol distributes uniform data packets to all Mesh nodes registered in the routing table.

[0191] It should be noted that the preset conditions are based on the impact radius parameter of the emergency, with the example value being 500 meters.

[0192] S6.3: Collect device response data and on-site feedback after broadcasting, generate execution results, and use the execution results as a new emergency type identifier to re-trigger and activate the pre-built knowledge graph framework.

[0193] Specifically, device response data is obtained through status confirmation messages returned by Mesh network terminal devices, and on-site feedback is obtained through evacuation status information reported by on-site personnel's mobile terminals.

[0194] The message reception rate parameter in the equipment response data and the proportion of unevacuated personnel in the on-site feedback are combined into an execution effect report in JSON format;

[0195] The execution effect report is mapped to discrete event type codes, generating new incident type identifiers and re-triggering the activation of the pre-built knowledge graph framework.

[0196] A superior approach, compared to conventional emergency communication solutions, utilizes a state evolution knowledge graph to perceive the damaged status of base stations in real time, dynamically constructs a Mesh broadcast routing table based on the coordinates of emergency facility entities, and achieves a breakthrough in the resilience of communication links. It integrates the device response data after Mesh broadcast with on-site feedback to generate execution results and maps them to new emergency type identifiers, reactivating the knowledge graph framework to form a decision-making closed loop. Furthermore, it encapsulates evacuation instructions and network-level Mesh routing parameters into data packet broadcasts, resolving the problem of instruction timing misalignment caused by multi-channel transmission.

[0197] In summary, this invention achieves the dynamic generation of intelligent countermeasure plans by: embedding artificially created disturbance simulation nodes into a state evolution knowledge graph and deducing the diffusion path; constructing disturbance nodes with both static and dynamic attributes to address heterogeneous disturbances such as emergency resource competition and communication interference sources, thus overcoming the limitations of traditional single-entity modeling; simulating the nonlinear diffusion process of disturbance nodes along the emergency facility topology based on a multi-agent reinforcement learning algorithm; and using a structural causal model to evaluate the blocking effect of different intervention measures on the diffusion path, outputting the optimal countermeasure plan with minimum resource consumption.

[0198] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for issuing contingency plans based on dynamic changes in emergency event status, characterized in that: include, The pre-built knowledge graph framework is activated based on the real-time acquired emergency event type identifier to form an initial dynamic knowledge graph. The knowledge graph framework includes the association relationships between event type entities, geographical environment entities, and emergency resource entities. Collect environmental parameters and equipment status data of the area where the emergency occurred, inject them into the data fusion layer of the initial dynamic knowledge graph, update entities and relation edges, and generate a state evolution knowledge graph. In this process, dynamic weight calculation is performed on the relation edges in the initial dynamic knowledge graph based on the updated entity attributes, and the relation edges are updated accordingly. Based on the updated relation edges and updated entity attributes, a state evolution knowledge graph is generated; The data fusion layer is used to assign specific values ​​in the fused data to the attribute fields of the corresponding entity types according to predefined entity type mapping rules; Artificial disturbance simulation nodes are implanted into the state evolution knowledge graph, and the diffusion path of the artificial disturbance simulation nodes along the emergency facility topology is deduced to generate countermeasure plans. The artificial disturbance simulation nodes include emergency resource competition nodes, communication interference source nodes, and information anomaly source node nodes. The implantation refers to creating corresponding multimodal disturbance nodes in the state evolution knowledge graph and configuring static and dynamic attributes for each disturbance node in the multimodal disturbance node. After the node is generated, a spatial association edge is established between each disturbance node in the multimodal disturbance node and the neighboring emergency facility entity. The simulation refers to each disturbance node acting as an independent intelligent agent, where the emergency resource competition node intelligent agent selects its direction of movement based on the topological connection of the road entity; the communication interference source node intelligent agent expands the interference range along the communication link topology; and the information anomaly source stronghold node intelligent agent spreads its influence through social relationship edges; each intelligent agent executes preset behavior rules in discrete time steps, records the movement trajectory or changes in the range of influence, and outputs the disturbance diffusion simulation results containing time series trajectory points. The countermeasure plan is broken down into a set of executable instructions for the equipment, which drives the physical equipment to execute and generates a feedback data stream. The feedback data stream is captured in real time and imported into the state update layer of the state evolution knowledge graph to generate an updated state evolution knowledge graph. The state update layer is used to capture the feedback data stream in real time, dynamically assimilate the feedback data stream, and merge it with the current state of the state evolution knowledge graph to generate a preliminary updated state. Multi-scale feature extraction and multi-source data fusion are performed on the initial updated state to generate an optimized feature vector; Based on the optimized feature vectors, the state evolution knowledge graph is optimized in topology and incrementally updated to generate an updated state evolution knowledge graph. Based on the updated state evolution knowledge graph, evacuation instructions are generated. By switching to the Mesh network from the damaged base station, path change information is broadcast, and execution results are generated. These results serve as new emergency event type identifiers to re-trigger and reactivate the pre-built knowledge graph framework. The execution result report is mapped to discrete event type codes to generate new emergency event type identifiers and re-trigger and reactivate the pre-built knowledge graph framework.

2. The method for issuing contingency plans based on dynamic changes in emergency event status as described in claim 1, characterized in that: The specific steps for generating the countermeasure plan are as follows: Based on three types of human-induced disturbances—emergency resource competition, communication interference sources, and information anomaly source locations—corresponding multimodal disturbance nodes are created in the state evolution knowledge graph, and static and dynamic attributes are configured for each disturbance node. Based on multimodal disturbance nodes, a multi-agent reinforcement learning algorithm is used to deduce the diffusion path of various disturbance nodes along the topology of emergency facilities, and generate disturbance diffusion simulation results. A structural causal model is used to conduct counterfactual intervention analysis on the disturbance diffusion simulation results, evaluate the effects of different intervention measures against emergency resource competition, communication interference sources and information anomaly source locations, and generate initial countermeasure plans for emergency resource competition, communication interference sources and information anomaly source locations. The initial countermeasure plan is subjected to multi-dimensional feasibility verification and iterative optimization to generate a countermeasure plan.

3. The method for issuing contingency plans based on dynamic changes in emergency event status as described in claim 1, characterized in that: The device can execute a set of instructions including signal jammer activation coordinates, scheduling and deployment GPS coordinate sequences, and backup communication channel activation instructions.

4. The method for issuing contingency plans based on dynamic changes in emergency event status as described in claim 1, characterized in that: The specific steps for generating the feedback data stream are as follows: Formal verification and blockchain smart contracts are used to decompose the countermeasure plan into signal jammer activation coordinates, scheduling and control GPS coordinate sequence, and backup communication channel activation instructions, generate an executable instruction sequence, and complete secure distribution verification. Based on the distributed sequence of executable instructions, the multimodal physical devices are driven to execute through a federated learning collaborative control model, and execution status data is collected in real time. Evidence theory is used to fuse and assess the quality of execution status data, generating a feedback data stream.

5. The method for issuing contingency plans based on dynamic changes in emergency event status as described in claim 1, characterized in that: The specific steps for re-triggering the activation of the pre-built knowledge graph framework are as follows. Based on the updated state evolution knowledge graph, spatiotemporal features are extracted through a spatiotemporal graph neural network to generate evacuation instructions; In response to the base station damage status, and in conjunction with evacuation instructions, a Mesh broadcast routing table is dynamically constructed. The evacuation instructions and the Mesh broadcast routing table are merged into a unified data packet, and path change information is broadcast through the Mesh network. Collect device response data and on-site feedback after broadcasting, generate execution results, and use the execution results as a new emergency type identifier to re-trigger and activate the pre-built knowledge graph framework.

Images