Smart city emergency response method and system combined with artificial intelligence

Abstract

The present invention provides a smart city emergency response method and system combined with artificial intelligence. First, the method performs situational awareness of urban emergency events, generates an event situation map containing spatiotemporal evolution characteristics and impact levels, calls a pre-trained policy migration model to match historical cases, outputs a preliminary response policy set, dynamically adapts and adjusts the preliminary response policy set based on real-time update characteristics of the situation map, generates an adapted response policy set, coordinates and analyzes the preliminary response policy set with the urban emergency resource network, generates a target coordination plan that matches resource supply with policy requirements, outputs an emergency response instruction containing execution priority according to the target coordination plan, and sends it to the corresponding node to trigger a multi-agent linkage response, thereby improving the scientificity and efficiency of the smart city emergency response.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a smart city emergency response method and system combined with artificial intelligence. Background Art

[0002] In the construction and development of smart cities, urban emergency response capabilities are crucial for ensuring safe and stable urban operations. Traditional urban emergency response methods rely primarily on manual judgment and pre-defined emergency plans. When an emergency occurs in a city, emergency commanders must assess the severity and scope of the incident based on on-site reports and their own experience, and then select an appropriate emergency response plan.

[0003] However, these traditional approaches have numerous drawbacks. For one thing, human judgment based on experience is subjective and limited, and different commanders may have different understandings and judgments of the same event, compromising the accuracy and timeliness of emergency response decisions. Furthermore, pre-established fixed emergency plans are unable to cope with the complexity and dynamic nature of urban emergencies. Urban emergencies are often sudden, diverse, and uncertain, with their spatiotemporal evolution and impact levels constantly changing over time. Fixed emergency plans cannot be flexibly adjusted to reflect real-time changes in events, making it prone to mismatches between response strategies and actual event situations. This reduces the effectiveness of emergency responses and may even lead to wasted resources and delays in emergency response. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a smart city emergency response method combined with artificial intelligence, the method comprising:

[0005] Conduct situational awareness of urban emergency events and generate event situation maps that include spatiotemporal evolution characteristics and impact levels;

[0006] Calling the pre-trained strategy transfer model to perform historical case matching on the event situation map, and outputting a preliminary response strategy set associated with the current event situation;

[0007] Dynamically adapt and adjust the preliminary response strategy set based on the real-time update feature of the event situation map to generate an adapted response strategy set adapted to the current event evolution;

[0008] Conducting collaborative scheduling analysis on the adaptive response strategy set and the city emergency resource network to generate a target collaborative solution that matches resource supply with strategy requirements;

[0009] An emergency response instruction including an execution priority is output according to the target collaboration plan, and the emergency response instruction is sent to a corresponding emergency handling node to trigger a multi-agent linkage response operation.

[0010] On the other hand, an embodiment of the present invention also provides a smart city emergency response system combined with artificial intelligence, including a processor and a machine-readable storage medium, the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present invention generates an event situation map containing spatiotemporal evolution characteristics and impact levels through situational awareness, calls a pre-trained policy migration model for historical case matching, and outputs a preliminary response policy set, making full use of the experience and wisdom of historical emergency cases, improving the efficiency and accuracy of emergency response policy formulation, and dynamically adapting and adjusting the preliminary response policy set based on the real-time update feature of the event situation map to generate an adaptive response policy set that adapts to the current event evolution. It can respond to the dynamic changes of urban emergency events in real time, ensuring that the emergency response strategy always matches the actual situation of the event, and effectively solves the problem that traditional fixed emergency plans are difficult to adapt to event changes. The adaptive response policy set is coordinated with the urban emergency resource network for scheduling analysis to generate a target coordination plan that matches resource supply with policy requirements, achieving efficient allocation and rational utilization of emergency resources, avoiding waste and idleness of resources, and finally outputting an emergency response instruction containing execution priority according to the target coordination plan and sending it to the corresponding emergency disposal node to trigger a multi-agent linkage response operation, thereby improving the coordination and efficiency of the emergency response and significantly improving the scientificity, timeliness and effectiveness of the smart city emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a schematic diagram of the execution flow of the smart city emergency response method combined with artificial intelligence provided by an embodiment of the present invention.

[0013] Figure 2 Schematic diagram of exemplary hardware and software components of a smart city emergency response system combined with artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a smart city emergency response method combined with artificial intelligence provided by an embodiment of the present invention. The smart city emergency response method combined with artificial intelligence is introduced in detail below.

[0015] Step S110: Conduct situation awareness of urban emergency events and generate an event situation map including spatiotemporal evolution characteristics and impact levels.

[0016] In smart city systems, urban emergencies can occur at any time, encompassing a variety of types, including natural disasters, public safety incidents, and public health incidents. Situational awareness aims to provide a comprehensive and in-depth understanding of the current status, development trends, and potential impacts of urban emergencies. Event situation mapping provides an intuitive and structured presentation of this information.

[0017] Take, for example, a major public health emergency. These events often spread rapidly, impact a wide area, and persist for a long time, posing a serious threat to a city's social order, economic development, and the health of its residents. Accurate situational awareness in the early stages of an incident can help relevant departments promptly grasp the scale, transmission path, and development of the epidemic, enabling them to formulate scientific and rational prevention and control strategies.

[0018] Step S111: collecting real-time monitoring information of urban emergency events, wherein the real-time monitoring information includes status data of the core area of ​​the event, status data of the surrounding associated areas, and city-level infrastructure associated data.

[0019] To achieve situational awareness of urban emergency events, it is first necessary to collect comprehensive and accurate real-time monitoring information. This real-time monitoring information comes from multiple levels and fields and can reflect the actual situation of the event from different angles.

[0020] Event core area status data provides a detailed description of the core area where the event occurred. In a public health incident, a core area may be the source of an outbreak, such as a specific community, hospital, or workplace. This area's status data includes real-time statistics on the number of infected people, the type and severity of patients' symptoms, and the use of medical resources. By monitoring and analyzing this data, we can understand the speed of the epidemic's spread in the core area, the characteristics of the infected population, and the pressure on the medical system.

[0021] Peripheral area status data focuses on the environment and population conditions surrounding the core area. Due to the nature of human mobility and transmission, peripheral areas are often susceptible to events in the core area. This data includes information on the flow of people in the surrounding area, the frequency of suspected cases, and the sanitation of public places. By collecting and analyzing this data, we can promptly identify epidemic trends and implement preventative measures to prevent further spread.

[0022] City-level infrastructure linked data covers the operational status of various urban infrastructure at the time of an incident. During a public health incident, the proper functioning of a city's medical facilities, transportation systems, communications networks, and other infrastructure is crucial for epidemic prevention, control, and response. Linked data for medical facilities includes information on the number of hospital beds, the availability of medical equipment, and the number and distribution of medical staff. Linked data for transportation systems includes information on public transportation operations and road conditions. Linked data for communications networks includes information on network coverage, signal strength, and data transmission speeds. By monitoring and analyzing city-level infrastructure linked data, we can assess the infrastructure's ability to support incidents, identify potential issues promptly, and implement adjustments and optimizations.

[0023] To gather this real-time monitoring information, sensor networks, IoT devices, and intelligent monitoring systems can be used to collect real-time environmental data and equipment operating status. Medical information systems and epidemic monitoring platforms can be used to obtain medical data. Traffic monitoring cameras and mobile device positioning systems can be used to monitor the flow of people. Furthermore, data sharing mechanisms can be established with relevant departments and institutions to integrate resources and ensure the comprehensiveness and accuracy of this information.

[0024] Step S112: The real-time monitoring information is divided and processed, and the event impact range is divided into a core impact layer, an indirect impact layer and a potential impact layer according to the spatial scale. Each impact layer corresponds to a monitoring data collection density of different granularity.

[0025] After collecting real-time monitoring information, for more targeted analysis and response, it is necessary to rationally divide the scope of the event's impact. Based on different spatial scales, the scope of the event's impact is divided into a core impact layer, an indirect impact layer, and a potential impact layer. Each impact layer has different characteristics and impact levels, necessitating different granularities of monitoring data collection density.

[0026] The core impact layer is the core area where an event occurs, directly impacted and affected by the event. In public health events, this layer may be a high-risk area for an outbreak, such as a community or medical facility with a high concentration of confirmed cases. This area has a high infection rate and rapid spread of the epidemic, posing a direct threat to the city's public health security. Therefore, for this core impact layer, a high-density monitoring data collection method is necessary to ensure timely capture of every detail and change. For example, the number of monitoring sites can be increased, the monitoring frequency can be increased, and detailed health monitoring and tracking can be conducted for every resident, providing real-time information on the number of infections, symptom changes, and treatment progress.

[0027] The indirect impact layer is the area surrounding the core impact layer. While relatively less affected by the event, it still presents a certain risk. In a public health incident, the indirect impact layer might include commercial areas, schools, residential areas, and other areas surrounding the core impact layer. People in this area may have some contact with the core impact layer or be potentially affected by the spread of the epidemic. Therefore, for the indirect impact layer, the density of monitoring data collection can be relatively low, but a certain level of monitoring intensity must still be maintained. For example, regular personnel screening and hygiene monitoring of public places can be conducted, focusing on personnel flow and the frequency of suspected cases.

[0028] The potential impact layer is located farther from the core of an incident, currently less affected by the incident but potentially presenting risks. In a public health incident, this layer might be other parts of a city or surrounding cities. People in this area have less direct contact with the core impact layer, but they could still be affected as the epidemic spreads. Therefore, for the potential impact layer, monitoring data collection can be more intensive, but macro-level monitoring and analysis are required, focusing on information such as population mobility trends and epidemic transmission models, to promptly identify potential risks and implement appropriate measures.

[0029] By dividing the scope of event impact and setting the density of monitoring data collection at different granularities, monitoring resources can be used more efficiently and the pertinence and accuracy of monitoring can be improved.

[0030] Step S113: extracting the change trend characteristics of the state data in each impact layer in the time dimension, wherein the change trend characteristics include the characteristic value rising rate, fluctuation frequency and stability index, and generating a time evolution feature sequence.

[0031] After stratifying the impact of an event, it's necessary to further analyze the temporal trends of the state data within each impact layer. By extracting trend characteristics such as the rate of increase of eigenvalues, fluctuation frequency, and stability indicators, we can understand the dynamics of the event within each impact layer.

[0032] The eigenvalue rate of increase reflects the growth rate of state data over a period of time. In public health events, for example, the number of infected people, the eigenvalue rate of increase can indicate the speed of the epidemic's spread. If the eigenvalue rate of increase in the number of infected people is fast, it indicates that the epidemic is spreading faster within the affected area, and more stringent prevention and control measures are needed; conversely, if the eigenvalue rate of increase is slow, it indicates that the epidemic is under control to a certain extent.

[0033] Fluctuation frequency describes the temporal fluctuations in status data. In public health events, the frequency of infection fluctuations may be related to factors such as the epidemic's spread and the effectiveness of prevention and control measures. A high frequency of fluctuation indicates uncertainty in the spread of the epidemic, potentially influenced by multiple factors, and the need for enhanced monitoring and analysis. A low frequency of fluctuation indicates a relatively stable epidemic.

[0034] The stability index measures the stability of status data. In public health events, the stability index reflects the degree of epidemic control within the impact layer. A high stability index indicates that the epidemic is well controlled within the impact layer and the status data is relatively stable. Conversely, a low stability index indicates significant uncertainty in the development of the epidemic and the need for further strengthening of prevention and control measures.

[0035] By extracting and analyzing the characteristic value rise rate, fluctuation frequency, and stability index of the state data within each impact layer, a time evolution feature sequence is generated. This time evolution feature sequence can intuitively show the development and changes of events over time within different impact layers.

[0036] Step S114: Analyze the correlation relationship between the status data of different influence layers, calculate the conduction delay parameter of the core influence layer to the indirect influence layer and the diffusion attenuation parameter of the indirect influence layer to the potential influence layer, and generate a spatial correlation feature matrix.

[0037] In addition to focusing on the temporal trends of status data within each impact layer, it's also necessary to analyze the correlations between status data across different impact layers. In a public health event, the development of the epidemic in the core impact layer will affect the indirect impact layer, which in turn will further affect the potential impact layer. By analyzing these correlations, we can understand the spatial spread of the event.

[0038] The transmission delay parameter from the core impact layer to the indirect impact layer reflects the time required for an event in the core impact layer to develop a significant impact in the indirect impact layer. This transmission delay parameter is influenced by various factors, such as the speed of population movement, the characteristics of the transmission pathway, and the implementation of prevention and control measures. In a public health event, if there is frequent movement of people between the core impact layer and the indirect impact layer, and the transmission pathway is relatively convenient, the transmission delay parameter may be small, indicating that the epidemic can be transmitted quickly from the core impact layer to the indirect impact layer. Conversely, if strict prevention and control measures are implemented, restricting the movement of people, the transmission delay parameter may be large.

[0039] The diffusion attenuation parameter of the indirect impact layer to the potential impact layer describes the degree of attenuation of an event as it spreads from the indirect impact layer to the potential impact layer. This diffusion attenuation parameter is related to factors such as the prevention and control measures in the indirect impact layer, the population density in the potential impact layer, and environmental conditions. In a public health event, if the indirect impact layer implements effective prevention and control measures, such as enhanced testing, isolation, and treatment, or if the potential impact layer has a low population density and low mobility, then the diffusion attenuation parameter may be large, indicating that the extent of the epidemic spread from the indirect impact layer to the potential impact layer will be relatively small.

[0040] By analyzing the correlation between state data across different impact layers, we calculate the transmission delay parameter and diffusion attenuation parameter to generate a spatial correlation feature matrix. This spatial correlation feature matrix clearly demonstrates the spatial correlation between different impact layers. Each element in the matrix represents the degree of correlation and transmission characteristics between different impact layers. By analyzing and processing this matrix, we can predict the propagation trend of events across different impact layers.

[0041] Step S115: Construct an event situation map based on the time evolution feature sequence and the spatial correlation feature matrix. The nodes of the event situation map represent key status indicators of different impact layers, and the edges represent the time transmission relationship and spatial diffusion relationship between indicators. The node attributes include spatiotemporal evolution characteristics, and the edge attributes include impact level parameters.

[0042] Based on the previously extracted temporal evolution feature sequences and the generated spatial correlation feature matrix, we can construct an event situation map. This is an intuitive and visual tool that integrates and presents the temporal and spatial evolution features of an event and its impact hierarchical relationships.

[0043] In the event situation map, nodes represent key status indicators at different impact levels. In public health events, these key status indicators can include the number of infections, the number of suspected cases, and the use of medical resources. Each node contains the evolutionary characteristics of the indicator in time and space. The node attributes provide an intuitive understanding of the changing trends and characteristics of the key status indicator.

[0044] Edges represent the temporal transmission and spatial diffusion relationships between indicators. In public health events, edges can represent the spread of infections across different impact levels or the allocation of medical resources across different regions. Edge attributes include impact level parameters, such as transmission delay and diffusion attenuation. These attributes can be used to understand the degree of correlation and transmission characteristics between indicators.

[0045] By constructing an event situation map, we can integrate and visualize complex information about an event, helping decision-makers quickly understand the overall picture and development trends. Furthermore, the event situation map can serve as a dynamic tool, continuously adjusted and improved over time and with updated data, providing real-time guidance for ongoing incident response.

[0046] Step S120: calling the pre-trained strategy transfer model to perform historical case matching on the event situation map, and outputting a preliminary response strategy set associated with the current event situation.

[0047] After generating the event situation map, a pre-trained policy transfer model can be used to quickly identify appropriate emergency response strategies by matching historical cases. The policy transfer model is trained on a large number of historical emergency incident cases. It learns the relationship between different event situations and corresponding response strategies. By analyzing and matching the current event situation map, it outputs a preliminary set of response strategies associated with the current event situation.

[0048] Taking public health events as an example, similar epidemics may have occurred in the past. By analyzing and summarizing these historical cases, we can derive some effective response strategies. The strategy migration model compares the current public health event situation map with historical cases, identifies the most similar historical cases, and extracts the corresponding response strategies as the initial response strategy set.

[0049] Step S121: Input the event situation graph into the graph encoding layer of the strategy migration model, vectorize the node attributes and edge attributes in the event situation graph, and generate a graph feature vector containing spatiotemporal evolution characteristics and impact level information.

[0050] The graph encoding layer of the policy migration model is the first step in processing the event situation graph. At this layer, the node and edge attributes in the event situation graph need to be vectorized to facilitate model processing and analysis.

[0051] Node attributes capture the spatiotemporal evolution of key status indicators at different impact levels, such as the rate of increase, fluctuation frequency, and stability of the characteristic value of the number of infected people. Edge attributes capture the impact-level parameters of the temporal transmission and spatial diffusion relationships between indicators, such as transmission delay and diffusion attenuation. Through vectorized representation, these attributes are converted into graph feature vectors.

[0052] The process of vectorization involves integrating and transforming node and edge attribute information so that it can be processed in the model in vector form. In public health events, the rate of increase in the number of infected people, a characteristic value in the node attribute, can be mapped to a dimension of the vector, with the fluctuation frequency and stability index also mapped to different dimensions. Similar processing is performed for the conduction delay parameter and diffusion attenuation parameter in the edge attributes. This results in a graph feature vector that contains both spatiotemporal evolution characteristics and information on the level of impact.

[0053] Step S122: extracting a case situation map from a historical emergency case library through the case retrieval module of the policy migration model, wherein the historical emergency case library contains situation maps of handled emergency events and corresponding response policy sets.

[0054] After obtaining the graph feature vector, the case retrieval module of the policy transfer model begins its work. This module extracts case situation graphs from the historical emergency case library. The historical emergency case library is a database that stores a large amount of information related to handled emergency incidents, including each incident's situation graph and corresponding response strategy set.

[0055] In public health incidents, a historical emergency case database may contain historical emergency case maps of similar epidemics, along with the various prevention and control measures and emergency response strategies implemented at the time. The case retrieval module searches and matches historical emergency case databases based on the map's feature vectors, looking for case situation maps that share similar spatiotemporal evolution characteristics and impact level information with the current event's situation map. This approach allows valuable lessons and strategies to be learned from historical cases.

[0056] Step S123: Calculate the graph similarity between the graph feature vector of the current event situation graph and the graph feature vector of each case situation graph, wherein the graph similarity is obtained by a weighted combination of the node attribute cosine similarity and the edge attribute edit distance.

[0057] After finding a possible matching case situation graph, we need to calculate the graph similarity between the graph feature vector of the current event situation graph and the graph feature vectors of each case situation graph. Graph similarity is a metric that measures the degree of similarity between the current event situation graph and the case situation graph. By calculating graph similarity, we can identify the historical case that is most similar to the current event situation.

[0058] Calculating graph similarity requires considering the similarity of both node and edge attributes. Node attribute cosine similarity is calculated by comparing the attribute vectors of the corresponding nodes in the current event situation graph and the case situation graph. It measures the degree of directional similarity between the two node attribute vectors. Higher cosine similarity indicates more similar attributes between the two nodes.

[0059] The edge attribute edit distance is obtained by comparing and analyzing the edge sets of the current event situation map and the case situation map. It measures the degree of difference between the edges in the current event situation map and the case situation map. The smaller the edit distance, the more similar the edges between the current event situation map and the case situation map.

[0060] After obtaining the node attribute cosine similarity and edge attribute edit distance, they need to be weighted and combined according to the set rules to obtain the graph similarity. This rule is dynamically adjusted based on the results of historical case matching. Generally, the weight of node attribute cosine similarity is relatively high because node attributes can better reflect the key characteristics of the event.

[0061] Step S1231: extract the attribute vectors of the corresponding nodes in the current event situation map and the case situation map, calculate the similarity between each pair of corresponding node attribute vectors, and take the comprehensive result of the similarity of all nodes as the node attribute cosine similarity.

[0062] When calculating the cosine similarity of node attributes, we first extract the attribute vectors of the corresponding nodes in the current event and case situation maps. These attribute vectors contain information such as the spatiotemporal evolution characteristics of the nodes. In public health events, for the infection count node in the core impact layer, its attribute vector may contain information such as the rate of increase of the characteristic value of the number of infections, the frequency of fluctuation, and the stability index.

[0063] Next, the similarity between each pair of corresponding node attribute vectors is calculated. This similarity is calculated based on a certain comparison method between vectors, by comparing features such as their direction and length. For example, the cosine similarity method can be used to calculate the cosine of the angle between two vectors. The closer the cosine value is to 1, the more similar the two vectors are.

[0064] Finally, the similarity of all nodes is comprehensively calculated, such as taking the average, to obtain the node attribute cosine similarity. This node attribute cosine similarity can reflect the overall similarity between the current event situation map and the case situation map in terms of node attributes.

[0065] Step S1232: Align the edge sets of the current event situation graph and the case situation graph, identify the edges with the same association relationship in the two graphs as matching edges, and identify the unmatched edges as difference edges.

[0066] When calculating the edge attribute edit distance, the edge sets of the current event situation graph and the case situation graph need to be aligned. The edge set contains the relationship information between different nodes in the event situation graph. The purpose of the alignment process is to find edges with the same relationship and edges with different relationships in the two graphs.

[0067] By carefully comparing the edge sets of the two graphs, edges with identical associations are identified as matching edges. For example, if the attribute information of the conductive relationship edge between the core impact layer and the indirect impact layer in the current event situation graph and the case situation graph is similar, then this edge is a matching edge. Unmatched edges are considered difference edges, reflecting the differences in edge attributes between the current event situation graph and the case situation graph.

[0068] Step S1233: Calculate the edit distance of the difference edges. The edit distance includes the complexity of the edge addition operation, deletion operation, and attribute modification operation. The edit distance is processed in a certain way to obtain the edge attribute edit distance.

[0069] For the identified difference edges, we need to calculate their edit distance. Edit distance is a metric that measures the degree of difference between the difference edges in the current event situation graph and the case situation graph. It includes the complexity of edge addition, deletion, and attribute modification operations.

[0070] In a public health event, if there is an edge in the case situation map representing the spread of the number of infections from the core impact layer to the indirect impact layer, but there is no such edge in the current event situation map, then an add operation is required; conversely, if there is an edge in the current event situation map but not in the case situation map, then a delete operation is required; if the association relationship between two edges is the same but the attributes are different, such as different conduction delay parameters, then an attribute modification operation is required.

[0071] The complexity of these operations is comprehensively calculated to obtain the edit distance of the difference edges. To facilitate subsequent calculations and comparisons, the edit distance needs to be processed, such as mapping it to a specific range to obtain the edge attribute edit distance. This edge attribute edit distance reflects the similarity between the current event situation map and the case situation map in terms of edge attributes.

[0072] Step S1234: Determine the weight distribution rules for node attribute cosine similarity and edge attribute edit distance, where the weight of node attribute cosine similarity is higher than the weight of edge attribute edit distance. The weight distribution rules are dynamically adjusted based on the feedback results of historical case matching accuracy.

[0073] After obtaining the node attribute cosine similarity and edge attribute edit distance, we need to determine their weight distribution rules. Because node attributes contain more key information and can more directly reflect the characteristics of the event, the weight of node attribute cosine similarity is usually higher than the weight of edge attribute edit distance.

[0074] The weight assignment rules are not fixed but are dynamically adjusted based on historical case matching accuracy. If, at a certain stage, increasing the weight of node attribute cosine similarity improves case matching accuracy, then its weight is increased accordingly. Conversely, if adjusting the weight of edge attribute edit distance improves matching accuracy, then its weight is adjusted accordingly. This dynamic adjustment enables the policy migration model to more accurately match cases.

[0075] Step S1235: Calculate the graph similarity by weighted combination, that is, combine the relevant values ​​of the node attribute cosine similarity and the edge attribute edit distance according to the weight distribution rule to obtain the comprehensive graph similarity between the current event and the case.

[0076] After determining the weighting rules for node attribute cosine similarity and edge attribute edit distance, graph similarity can be calculated through a weighted combination. Since edge attribute edit distance reflects the degree of difference and has a different direction from node attribute cosine similarity, it is necessary to transform the edge attribute edit distance so that it has the same comparison direction as node attribute cosine similarity.

[0077] According to the weight distribution rule, the node attribute cosine similarity and the converted edge attribute edit distance are multiplied by their respective weights, and then the results are combined to obtain the comprehensive graph similarity between the current event and the case. This comprehensive graph similarity can reflect the overall similarity between the current event situation graph and the case situation graph.

[0078] Step S124: Filter historical cases whose graph similarity meets the set conditions as matching cases, and extract the response strategy sets corresponding to the matching cases as the candidate strategy sets.

[0079] Based on the calculated comprehensive graph similarity, historical cases whose graph similarity meets the set criteria are selected as matching cases. This criteria is a standard set based on actual conditions and experience. Only historical cases with graph similarity greater than this standard are considered matching cases similar to the current event.

[0080] For each matched case, the corresponding response strategy set is extracted as a candidate strategy set. These response strategies have been proven effective in similar emergency situations and have a certain reference value. In public health events, the response strategies for matched cases may include strengthening personnel management, increasing medical resource investment, and carrying out publicity and education measures.

[0081] Step S125: De-duplication of policy types and extraction of core measures are performed on the candidate policy set, key execution steps of different policy types are retained, and a preliminary response policy set associated with the current event situation is generated.

[0082] After obtaining the candidate policy set, further processing is required. First, policy type deduplication is performed, removing duplicate policy types from the candidate policy set. In public health events, there may be multiple similar personnel control policies, such as restricting personnel movement and setting checkpoints. Only one representative policy type should be retained.

[0083] Next, we extract core measures, identifying the key execution steps within each policy type. For personnel management strategies, these might include setting up checkpoints, conducting temperature checks, and restricting personnel movement. For medical resource allocation strategies, these might include increasing hospital beds, deploying medical equipment, and organizing medical staff.

[0084] By retaining the key execution steps of different policy types, a preliminary response policy set associated with the current incident situation is generated. This preliminary response policy set is more streamlined and effective, and can provide more targeted guidance for subsequent emergency response.

[0085] Step S130: dynamically adapt and adjust the preliminary response strategy set based on the real-time update characteristics of the event situation map to generate an adapted response strategy set adapted to the current event evolution.

[0086] Event dynamics are constantly evolving, so the initial response strategy set needs to be dynamically adapted based on the real-time updates of the event dynamics graph. For example, in the case of a public health incident, as the epidemic progresses, the number of infected people may increase or decrease, the scope of transmission may expand or contract, and the demand for medical resources may change. Based on these real-time updates, the initial response strategy set is adjusted to ensure it adapts to the evolving event.

[0087] Step S131: monitor the changes in node attributes of the event situation map in real time, extract nodes whose attribute value changes exceed the set limit as sensitive nodes, and record the attribute change time and change direction of the sensitive nodes.

[0088] To promptly detect changes in event status, it is necessary to monitor changes in node attributes in the event status map in real time. For each node's attribute value, a limit is set. When the change in a node's attribute value exceeds this limit, the node is identified as a sensitive node.

[0089] During a public health incident, if the number of infected people in a core impact layer increases by more than a set threshold within a short period of time, the node is considered a sensitive node. The attribute change time and direction of the sensitive node are also recorded. The attribute change time records the specific moment when the node attribute changes, while the change direction indicates whether the attribute value increases or decreases.

[0090] Step S132: Analyze the position of the sensitive node in the event situation map and the associated edge attributes, determine the potential impact range of the sensitive node change on other nodes, and generate the impact transmission path.

[0091] Once a sensitive node is discovered, its position in the event landscape and the attributes of its associated edges need to be analyzed. The location of the sensitive node determines its importance and influence in the event landscape. The attributes of its associated edges reflect the relationship between the sensitive node and other nodes.

[0092] In a public health event, if a sensitive node is the infected person node in the core impact layer, its associated edge connects to the suspected case node in the indirect impact layer. By analyzing the transmission delay parameter and diffusion attenuation parameter of the associated edge, the potential impact of changes in the sensitive node on other nodes can be determined. For example, based on the transmission delay parameter, it is possible to predict how long it will take for an increase in the number of infected people to affect the suspected case count in the indirect impact layer.

[0093] Based on the analysis results, an impact transmission path is generated. This path starts from a sensitive node and passes through a series of affected nodes. This path clearly shows the process and scope of the impact of changes in sensitive nodes on other nodes.

[0094] Step S1321: Locate the hierarchical affiliation of the sensitive node in the event situation map and determine whether it belongs to the core impact layer, indirect impact layer or potential impact layer. Sensitive nodes at different levels have different initial impact strengths.

[0095] When analyzing the impact of sensitive nodes, the first step is to identify their hierarchical level within the event landscape. Sensitive nodes at different levels have different initial impacts. In public health events, sensitive nodes in the core impact layer, due to their location at the heart of the epidemic, have a relatively greater impact on other nodes; sensitive nodes in the indirect impact layer have the second highest impact; and sensitive nodes in the potential impact layer have relatively minimal impact.

[0096] By determining the hierarchical level of a sensitive node, we can more accurately assess its impact on other nodes. For example, if a sensitive node belongs to the core impact layer, then when analyzing its impact on the indirect impact layer and potential impact layer, its larger initial impact should be fully considered.

[0097] Step S1322: traverse all associated edges of the sensitive node, and calculate the influence intensity of the sensitive node change transmitted to the adjacent nodes through each associated edge based on the conduction delay parameter and diffusion attenuation parameter in the edge attributes.

[0098] After locating the hierarchical affiliation of a sensitive node, it is necessary to traverse all of its associated edges. Each associated edge contains a transmission delay parameter and a diffusion attenuation parameter. Based on these parameters, the impact strength of the sensitive node change transmitted to adjacent nodes through each associated edge is calculated.

[0099] In public health events, the transmission delay parameter determines how long it takes for changes at sensitive nodes to propagate to adjacent nodes, while the diffusion attenuation parameter determines how much the impact is weakened during the transmission process. For example, for an edge connecting a sensitive node in the core impact layer and an adjacent node in the indirect impact layer, if both the transmission delay parameter and the diffusion attenuation parameter are small, the impact of the change at the sensitive node on the adjacent nodes will be greater.

[0100] By performing the above calculation on each associated edge, the influence intensity of the sensitive node change transmitted to the adjacent nodes through different associated edges is obtained.

[0101] Step S1323: Determine the influence strength of the adjacent nodes, retain the adjacent nodes whose influence strength exceeds the set standard as first-level influence nodes, and record the conductive path from the sensitive node to the first-level influence node.

[0102] After determining the impact of changes on sensitive nodes on adjacent nodes, we need to determine the impact intensity received by adjacent nodes. A standard is set, and only adjacent nodes with an impact intensity exceeding this standard are considered to be significantly affected by the sensitive node and are designated as first-level impact nodes.

[0103] During a public health incident, if the impact intensity received by a neighboring node exceeds a set standard, then that node is considered a first-level impact node. At the same time, the transmission path from the sensitive node to the first-level impact node is recorded. This path is part of the impact transmission path and records how the impact of the sensitive node is transmitted to the first-level impact node.

[0104] Step S1324: Using the first-level influence node as a new starting point, repeat the above-mentioned associated edge traversal and influence strength calculation process to identify the second-level influence node and the corresponding conductive path part, and stop traversal until the influence strength is lower than the set standard.

[0105] Using the first-level influence node as the new starting point, repeat the previous process of traversing the associated edges and calculating the influence strength. For all associated edges of the first-level influence node, calculate the influence strength of the first-level influence node's change transmitted to its adjacent nodes through each associated edge based on the transmission delay parameter and diffusion attenuation parameter in the edge attributes.

[0106] Similarly, the influence strength received by these adjacent nodes is judged, and the adjacent nodes whose influence strength exceeds the set standard are retained as secondary influence nodes, and the conduction path part from the primary influence node to the secondary influence node is recorded.

[0107] This process is repeated continuously, starting with the second-level impact node and continuing to identify the third-level impact nodes and their corresponding transmission paths until the impact intensity falls below the set threshold. By doing this, the impact transmission path is gradually expanded to identify all nodes affected by the sensitive node.

[0108] Step S1325: Connect all conduction path parts according to the conduction order to form a tree-like influence conduction path starting from the sensitive node and passing through multiple levels of influence nodes. Each node in the tree-like influence conduction path is marked with the received influence intensity and conduction delay time.

[0109] After all the influencing nodes and the corresponding conduction path parts are identified, all the conduction path parts are connected in accordance with the conduction order to form a tree-like influence conduction path starting from the sensitive node and passing through multiple levels of influencing nodes.

[0110] In the tree-like impact transmission path, each node is labeled with the received impact intensity and transmission delay time. The impact intensity indicates the degree to which the node is affected by the sensitive node, and the transmission delay time indicates how long it takes for the sensitive node's influence to be transmitted to the node. By annotating this information, the changes in the event situation and the impact propagation process can be more clearly demonstrated.

[0111] Step S133: Based on the impact transmission path and the sensitive node change characteristics, identify the policy clauses in the preliminary response policy set that may fail, where the policy clauses that may fail refer to clauses whose policy execution conditions do not match the current sensitive node change characteristics.

[0112] Based on the generated impact transmission paths and sensitive node change characteristics, the initial response strategy set is analyzed to identify potential invalid policy clauses. Potentially invalid policy clauses are clauses whose execution conditions do not match the current sensitive node change characteristics.

[0113] In a public health incident, if a policy clause in the initial response strategy is designed for a slow increase in the number of infections, but the current sensitive node (the infection node in the core impact layer) is characterized by a rapid increase in the number of infections, then this policy clause may become ineffective. By analyzing the impact transmission path and the changing characteristics of sensitive nodes, these potentially ineffective policy clauses can be identified.

[0114] Step S134: Adaptively modify the policy clauses that may become invalid to obtain modified policy clauses. The modification methods include adjusting the policy execution strength, changing the policy scope, or adding auxiliary execution measures to generate modified policy clauses.

[0115] For identified policy clauses that may become invalid, adaptive modifications are required. Modifications can include adjusting the enforcement strength of the policy, changing the scope of the policy, or adding auxiliary enforcement measures.

[0116] During a public health incident, if a policy provision restricts movement, but the rapid increase in infections necessitates strengthened personnel controls, policy enforcement can be adjusted, such as by adding checkpoints or increasing the frequency of temperature checks. If a policy's scope originally targeted only the core impact layer, but the transmission pathways reveal that the indirect impact layer is also significantly affected, the policy's scope can be adjusted to include the indirect impact layer as well. If a policy is found to be ineffective during implementation, supplementary enforcement measures can be added, such as publicity and education campaigns to raise public awareness of prevention and control.

[0117] Through these modifications, we obtain revised policy clauses that can better adapt to the evolution of current events.

[0118] Step S135: Integrate the modified policy clauses with the policy clauses in the initial response policy set that have not expired, supplement the dynamic adjustment trigger conditions for policy execution, and generate an adapted response policy set that adapts to the current event evolution.

[0119] The revised policy terms are integrated with the policy terms in the initial response policy set that have not expired. The integration process is to merge and optimize the two to ensure the integrity and effectiveness of the policy.

[0120] At the same time, dynamic adjustment trigger conditions for policy execution are supplemented. Dynamic adjustment trigger conditions are set based on the changing situation of the event. When these conditions are met, further adjustments to the policy are required. In public health events, when the growth rate of the number of infected people exceeds a certain rate, adjustments to the policy enforcement strength are triggered; when the epidemic spreads to new areas, adjustments to the policy scope are triggered.

[0121] By integrating and supplementing the dynamic adjustment trigger conditions, an adaptive response strategy set that adapts to the current event evolution is generated. This adaptive response strategy set can better cope with the development and changes of events.

[0122] Step S140: performing collaborative scheduling analysis on the adaptive response strategy set and the city emergency resource network to generate a target collaborative solution that matches resource supply with strategy requirements.

[0123] After the adaptive response strategy set is formulated, it needs to be coordinated and analyzed with the city's emergency resource network. The city's emergency resource network includes various emergency resources, such as medical equipment, rescue personnel, and material reserves. Through coordinated scheduling analysis, we ensure that resource supply meets the strategy requirements and generate a targeted coordination plan that matches resource supply and strategy requirements.

[0124] Step S141: parsing the resource requirement characteristics of each strategy in the adaptation response strategy set, wherein the resource requirement characteristics include the required resource type, resource skill requirements, resource spatial distribution range, and resource available time range.

[0125] First, we need to analyze the resource requirements of each strategy in the adaptive response strategy set. In public health incidents, different strategies may require different types of resources. For example, a strategy to strengthen medical treatment may require resources such as medical equipment and medical personnel; a strategy to control personnel may require resources such as police force and protective equipment.

[0126] Resource skill requirements focus on the professional skills and capabilities of resource providers. In public health incidents, medical staff need to have relevant medical knowledge and clinical experience, and rescue workers need to have first aid skills and emergency response capabilities.

[0127] The spatial distribution of resources takes into account the geographical location and distribution of resources within the city. In public health incidents, medical resources may be distributed across different hospitals and medical institutions, and resources need to be rationally allocated based on strategic needs.

[0128] The resource availability timeframe refers to the time period during which resources can be used. During public health events, some resources may only be available during specific time periods, and resource usage needs to be properly arranged based on the execution time of the policy.

[0129] By analyzing the resource demand characteristics of each strategy in the adaptive response strategy set, the specific resource requirements of each strategy can be clarified.

[0130] Step S142: constructing a city emergency resource network model, wherein the nodes of the resource network model represent different types of emergency resource units, the edges represent the coordination capabilities between resource units, and the node attributes include the current status of the resources and the schedulable time.

[0131] In order to conduct collaborative scheduling analysis, it is necessary to build an urban emergency resource network model. This urban emergency resource network model abstracts and represents various emergency resource units in the city, and demonstrates the collaborative capabilities and connections between resources through the relationships between nodes and edges.

[0132] In the urban emergency resource network model, nodes represent different types of emergency resource units. In public health incidents, these resource units can include hospitals, medical equipment suppliers, rescue teams, and more. Each node contains the current status and dispatchable time of the resource unit. The current status of the resource can include the quantity, quality, and usage of the resource; the dispatchable time indicates the time period when the resource can be deployed.

[0133] Edges represent the collaborative capabilities between resource units. During public health incidents, different resource units need to collaborate, such as when hospitals and medical equipment suppliers coordinate supplies, or when rescue teams and hospitals transfer patients. Edge attributes can reflect information such as the collaborative efficiency and communication channels between resource units.

[0134] By constructing an urban emergency resource network model, the complex information of urban emergency resources can be integrated and visualized, helping decision makers understand the distribution and coordination of resources.

[0135] Step S143: Match the resource demand characteristics of the strategy with the node attributes of the resource network model, and calculate the degree to which each resource unit meets the resource demand of the strategy. The degree of matching is determined based on the resource type fit, skill requirement satisfaction, spatial accessibility, and time range overlap.

[0136] The resource requirements of the strategy are matched with the node attributes of the resource network model, and the degree to which each resource unit meets the resource requirements of the strategy is calculated. The degree of match is a comprehensive indicator that takes into account the influence of multiple factors.

[0137] Resource type compatibility refers to the match between the resource unit type and the resource type required by the policy. In a public health incident, if the policy requires medical equipment, and a resource unit is a hospital with relevant medical equipment, then resource type compatibility is high. If the resource unit type does not match the resource type required by the policy at all, then resource type compatibility is low.

[0138] Skill requirement satisfaction refers to whether the skills of the personnel in a resource unit meet the skill requirements of the strategy. In a public health incident, if the strategy requires medical staff to possess certain professional skills, and the medical staff in a particular hospital possess those skills, then the skill requirement satisfaction is high. If the skills of the personnel in a resource unit do not match the skill requirements of the strategy, then the skill requirement satisfaction is low.

[0139] Spatial accessibility refers to the ease with which resource units can reach the strategic impact area. In a public health incident, if the strategic impact area is in the core impact layer and a hospital is close to the core impact layer with convenient transportation, then spatial accessibility is high. If the resource unit is far away from the strategic impact area and transportation is inconvenient, then spatial accessibility is low.

[0140] Time range overlap refers to the overlap between a resource unit's schedulable time range and the policy's execution time range. In public health events, if the policy's execution time range is a certain time period and a hospital is schedulable within that time period, then the time range overlap is high. If the resource unit's schedulable time range does not overlap with the policy's execution time range, then the time range overlap is low.

[0141] By comprehensively considering factors such as resource type compatibility, skill requirement satisfaction, spatial accessibility, and time range overlap, we can calculate the degree to which each resource unit meets the strategic resource requirements. This degree of match serves as an important basis for resource scheduling and allocation.

[0142] Step S1431: Determine whether the type of the resource unit is consistent with the resource type required by the policy. If it is completely consistent, the resource type compatibility is high. If it is partially consistent, the compatibility level is determined based on the type association. If it is completely inconsistent, the compatibility level is low.

[0143] When calculating resource type compatibility, the first step is to determine whether the resource unit type is consistent with the resource type required by the policy. In a public health incident, if the policy requires medical equipment and the resource unit happens to be a medical equipment supplier, then the resource type compatibility is high.

[0144] In the case of partial alignment, for example, if the policy requires multiple resource types but the resource unit only provides a subset, the level of fit is determined based on type association. Type association can consider the relevance and importance of resource types. For example, if the policy requires both medical devices and pharmaceuticals, and the resource unit only provides medical devices, resource type fit can be determined based on the importance of medical devices in the policy and their relationship to pharmaceuticals.

[0145] If the type of resource unit is completely inconsistent with the resource type required by the strategy, such as the strategy requires medical equipment, but the resource unit is a fire station, then the resource type compatibility is low.

[0146] Step S1432: extract the skill tag set of the resource unit and the skill requirement set of the strategy, compare the overlap between the two sets, and use the ratio of the overlap to the strategy skill requirement set as the skill requirement satisfaction.

[0147] To calculate skill requirement satisfaction, we need to extract the resource unit's skill tag set and the strategy's skill requirement set. In public health incidents, the resource unit's skill tag set might include medical staff's professional skills and rescue workers' first aid skills, while the strategy's skill requirement set specifies the skills required for strategy execution.

[0148] Compare the overlap between the resource unit's skill tag set and the strategy's skill requirements, calculate the proportion of the overlap to the strategy's skill requirements, and use this proportion as the skill requirement satisfaction. For example, if the strategy's skill requirements have 10 skill tags, and the resource unit's skill tag set overlaps with the strategy's skill requirements by 6, then the skill requirement satisfaction is 60%.

[0149] Step S1433: Based on the current position of the resource unit and the position of the policy action area, calculate the convenience of the resource unit reaching the policy action area, combine the mobility of the resource unit, determine the arrival time, and determine the spatial accessibility level based on the difference between the arrival time and the execution start time required by the policy.

[0150] Based on the resource unit's current location and the location of the policy's area of ​​effect, calculate how convenient it is for the resource unit to reach the policy's area of ​​effect. In public health incidents, factors such as traffic conditions and distance can be considered to assess convenience. If the resource unit is close to the policy's area of ​​effect and traffic is smooth, then the convenience is high; conversely, if the distance is far and traffic is congested, then the convenience is low.

[0151] Combined with the mobility of resource units, such as the driving speed of vehicles and the movement speed of personnel, the arrival time of resource units to the strategic action area is determined.

[0152] The spatial accessibility level is determined based on the difference between the arrival time and the policy's required execution start time. If the arrival time is before the policy's required execution start time and the difference is small, the spatial accessibility level is high; if the arrival time is after the execution start time or the difference is large, the spatial accessibility level is low.

[0153] Step S1434: Compare the schedulable time range of the resource unit with the execution time range of the policy, and calculate the ratio of the overlapping time period between the two to the policy execution time range as the time range overlap degree.

[0154] Compare the schedulable time range of the resource unit with the policy's execution time range. Find the overlapping time period between the two and calculate the proportion of the overlapping time period to the policy's execution time range as the time range overlap.

[0155] In a public health event, if the policy execution time range is from 9 a.m. to 5 p.m., and the resource unit's schedulable time range is from 10 a.m. to 4 p.m., then the overlapping time period is from 10 a.m. to 4 p.m., a total of 6 hours, and the policy execution time range is 8 hours, so the time range overlap is 75%.

[0156] Step S1435: Comprehensively score the resource type compatibility, skill requirement satisfaction, spatial accessibility level, and time range overlap in the set order of importance, and obtain the degree of match of each resource unit to the strategic resource requirements through the comprehensive score.

[0157] The resource type fit, skill requirement satisfaction, spatial accessibility level, and time range overlap are comprehensively scored in a set order of importance. This order of importance can be determined based on actual conditions and experience. For example, resource type fit may be relatively more important.

[0158] Each factor is scored. For example, a high level of resource type fit is scored as 10 points, and a low level is scored as 1 point; skill requirement satisfaction can be scored according to the proportion, such as 70% is scored as 7 points; a high level of spatial accessibility is scored as 10 points, and a low level is scored as 1 point; time range overlap can be scored according to the proportion, such as 80% is scored as 8 points.

[0159] Then combine these scores in order of importance. For example, the weight of resource type fit is 0.4, the weight of skill requirement satisfaction is 0.3, the weight of spatial accessibility level is 0.2, and the weight of time range overlap is 0.1. Then the degree of match = resource type fit score × 0.4 + skill requirement satisfaction score × 0.3 + spatial accessibility level score × 0.2 + time range overlap score × 0.1.

[0160] Through the above comprehensive scoring method, the matching degree of each resource unit to the strategic resource requirements can be obtained.

[0161] Step S144: Filter out a set of candidate resource units that meet the policy requirements based on the matching degree, analyze the coordination capabilities between the resource units in the candidate resource unit set, and identify resource combinations that meet the set coordination efficiency conditions.

[0162] Based on the calculated matching degree, resource units that meet the set matching requirements are selected as candidate resource units. This requirement is a standard set based on actual conditions and experience. Only resource units with a matching degree exceeding this standard are considered resource units that meet the policy requirements.

[0163] For each candidate resource unit set, analyze the collaborative capabilities between them. During public health incidents, different resource units will need to collaborate, such as between hospitals and medical equipment suppliers for material allocation, or between rescue teams and hospitals for patient transfers. Collaborative capabilities can be assessed by factors such as communication channels, collaborative experience, and information sharing between resource units.

[0164] Based on their collaborative capabilities, identify resource combinations that meet the specified collaborative efficiency criteria. These criteria can be set based on actual conditions and experience. For example, the collaborative efficiency of a resource combination must reach a set level to enable it to complete strategic tasks within a specified timeframe. By identifying resource combinations that meet these criteria, resource utilization efficiency and emergency response effectiveness can be improved.

[0165] Step S145: associate each strategy with the corresponding resource combination that meets the set collaborative efficiency conditions, mark the scheduling priority and collaborative execution process of the resource combination, and generate a target collaborative solution that matches resource supply with strategy requirements.

[0166] Each strategy is associated with a corresponding resource combination that meets the specified collaborative efficiency conditions. In a public health incident, a strategy to strengthen medical treatment can be associated with a resource combination that includes hospitals, medical equipment suppliers, and medical staff; a strategy for personnel management can be associated with a resource combination that includes police, protective equipment, and community workers.

[0167] Identify the resource portfolio's scheduling priority and collaborative execution process. Scheduling priority can be determined based on the importance and urgency of the strategy. Resource portfolios for key strategies can have a higher scheduling priority. The collaborative execution process clarifies the specific tasks and execution order of each resource unit in the resource portfolio. For example, in a medical treatment strategy, medical equipment suppliers must deliver equipment to the hospital within a specified timeframe, and medical staff must follow a defined process to diagnose and treat patients.

[0168] By associating strategies with resource combinations and annotating scheduling priorities and collaborative execution processes, a targeted collaborative plan can be generated that matches resource supply with strategy requirements. This targeted collaborative plan ensures that resources are rationally allocated and utilized during emergencies, improving the efficiency and effectiveness of emergency response.

[0169] Step S150: Output an emergency response instruction including an execution priority according to the target collaboration plan, and send the emergency response instruction to a corresponding emergency handling node to trigger a multi-agent linkage response operation.

[0170] After generating a targeted collaborative plan, it needs to be converted into specific emergency response instructions and sent to the corresponding emergency response nodes to trigger a multi-agent coordinated response. Emergency response instructions are specific action guidelines for implementing emergency response, clarifying the tasks and execution order of each emergency response node.

[0171] Based on the scheduling priorities of the resource combinations in the target coordination plan and the execution requirements of the strategies, the execution priority of the emergency response instructions corresponding to each strategy is determined. Instructions with high execution priority need to be executed first to ensure the timeliness and effectiveness of the emergency response.

[0172] Send emergency response instructions, including execution priorities, to the corresponding emergency response nodes. Emergency response nodes can be hospitals, fire stations, community service centers, and other institutions and departments involved in emergency response. By sending emergency response instructions, these nodes are notified to initiate the corresponding emergency response actions.

[0173] When an emergency response node receives an emergency response command, it triggers a multi-agent coordinated response. Different emergency response nodes collaborate according to the command's requirements to jointly respond to the city's emergency. In public health incidents, hospitals are responsible for treating patients, community service centers are responsible for personnel management and information collection, and fire stations are responsible for material transportation. This multi-agent coordinated response creates a powerful emergency response force, improving the city's ability to respond to emergencies. Furthermore, throughout the emergency response process, the event situation map is continuously monitored for updates, and strategies and resources are dynamically adjusted based on real-time conditions to ensure the effectiveness and adaptability of the emergency response.

[0174] During the data collection process, various privacy protection and anti-leakage technologies are employed for privacy-sensitive data, such as personal health information and movement trajectory data. For health information, encryption algorithms are used to ensure data security during transmission and storage. Furthermore, strict access control is implemented for data, ensuring only authorized personnel have access. For movement trajectory data, anonymization technology is used to remove personal identifying information from the data, retaining only the necessary information relevant to the event. Furthermore, a comprehensive security audit mechanism has been established to record and monitor data access and use, promptly identifying and addressing anomalies and preventing the leakage and misuse of private data.

[0175] In terms of AI model construction and training, the strategy transfer model includes essential modules such as the graph encoding layer and the case retrieval module. The graph encoding layer is responsible for vectorizing the event situation graph, while the case retrieval module is responsible for matching cases from the historical emergency case library. The model layers are connected through data flow and processing. The input data is the event situation graph, which is processed by the graph encoding layer and converted into a graph feature vector. This is then input into the case retrieval module for matching, outputting a preliminary set of response strategies associated with the current event situation.

[0176] When training the policy transfer model, we first collect a large number of historical emergency incident cases, including event situation maps and corresponding response strategies. This data is then preprocessed, including data cleaning and feature extraction. The preprocessed data is then divided into a training set and a test set. The model is trained using the training set, and by continuously adjusting the model parameters, it learns the relationship between event situation and response strategy. During training, cross-validation and other methods are used to evaluate model performance, and the model is optimized and improved based on the evaluation results. The trained model is then tested on the test set to verify its accuracy and generalization ability. Through these steps and parameter settings, we ensure that the policy transfer model can effectively match historical cases.

[0177] Figure 2 A schematic diagram illustrates exemplary hardware and software components of an artificial intelligence-integrated smart city emergency response system 100 that can implement the concepts of the present application, as provided in some embodiments of the present application. For example, the processor 120 can be used in the artificial intelligence-integrated smart city emergency response system 100 to perform the functions described in the present application.

[0178] The AI-integrated smart city emergency response system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the AI-integrated smart city emergency response method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0179] For example, the smart city emergency response system 100 incorporating artificial intelligence may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in various forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the smart city emergency response system 100 incorporating artificial intelligence may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application may be implemented according to these program instructions. The smart city emergency response system 100 incorporating artificial intelligence also includes an I / O interface 150 between the computer and other input and output devices.

[0180] For ease of explanation, only one processor is described in the smart city emergency response system 100 in combination with artificial intelligence. However, it should be noted that the smart city emergency response system 100 in combination with artificial intelligence in this application may also include multiple processors, so the steps performed by one processor described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the smart city emergency response system 100 in combination with artificial intelligence performs step A and step B, it should be understood that step A and step B may also be performed jointly by two different processors or performed individually in one processor. For example, the first processor performs step A and the second processor performs step B, or the first processor and the second processor perform steps A and B together.

[0181] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the smart city emergency response method combined with artificial intelligence as described above is implemented.

[0182] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A smart city emergency response method combined with artificial intelligence, characterized in that: The method comprises: Conduct situational awareness of urban emergency events and generate event situation maps that include spatiotemporal evolution characteristics and impact levels; Calling the pre-trained strategy transfer model to perform historical case matching on the event situation map, and outputting a preliminary response strategy set associated with the current event situation; Dynamically adapt and adjust the preliminary response strategy set based on the real-time update feature of the event situation map to generate an adapted response strategy set adapted to the current event evolution; Conducting collaborative scheduling analysis on the adaptive response strategy set and the city emergency resource network to generate a target collaborative solution that matches resource supply with strategy requirements; Outputting an emergency response instruction including an execution priority according to the target coordination scheme, and sending the emergency response instruction to the corresponding emergency handling node to trigger a multi-agent linkage response operation; The pre-trained strategy transfer model is called to perform historical case matching on the event situation map, and output a preliminary response strategy set associated with the current event situation, including: Inputting the event situation graph into the graph encoding layer of the strategy migration model, vectorizing the node attributes and edge attributes in the event situation graph, and generating a graph feature vector containing spatiotemporal evolution characteristics and impact level information; Extracting a case situation map from a historical emergency case library through a case retrieval module of the policy migration model, wherein the historical emergency case library contains situation maps of handled emergency events and corresponding response strategy sets; Calculate the graph similarity between the graph feature vector of the current event situation graph and the graph feature vector of each case situation graph, where the graph similarity is obtained by a weighted combination of the node attribute cosine similarity and the edge attribute edit distance; Filter historical cases whose graph similarity meets the preset conditions as matching cases, and extract the response strategy sets corresponding to the matching cases as the candidate strategy sets; Deduplication of policy types and extraction of core measures are performed on the candidate policy set, key execution steps of different policy types are retained, and a preliminary response policy set associated with the current event situation is generated; The aforementioned situational awareness of urban emergency events generates an event situation map containing spatiotemporal evolution characteristics and impact levels, including: Collecting real-time monitoring information of urban emergency events, including status data of the core area of ​​the event, status data of surrounding related areas, and related data of city-level infrastructure; The real-time monitoring information is divided and processed, and the event impact range is divided into a core impact layer, an indirect impact layer, and a potential impact layer according to the spatial scale, and each impact layer corresponds to a monitoring data collection density of different granularity; Extract the change trend characteristics of the state data in each impact layer in the time dimension, the change trend characteristics include the characteristic value rising rate, fluctuation frequency and stability index, and generate a time evolution feature sequence; Analyze the correlation between the status data of different influence layers, calculate the conduction delay parameters of the core influence layer to the indirect influence layer and the diffusion attenuation parameters of the indirect influence layer to the potential influence layer, and generate a spatial correlation feature matrix; An event situation map is constructed based on the time evolution feature sequence and the spatial correlation feature matrix. The nodes of the event situation map represent key status indicators of different impact layers, and the edges represent the time transmission relationship and spatial diffusion relationship between indicators. The node attributes include spatiotemporal evolution characteristics, and the edge attributes include impact level parameters.

2. The smart city emergency response method combined with artificial intelligence according to claim 1 is characterized in that: The calculating of the graph similarity between the graph feature vector of the current event situation graph and the graph feature vectors of each case situation graph includes: Extract the attribute vectors of the corresponding nodes in the current event situation map and the case situation map, calculate the cosine similarity of each pair of corresponding node attribute vectors, and take the average of the cosine similarities of all nodes as the node attribute cosine similarity; Align the edge sets of the current event situation graph and the case situation graph, identify the edges with the same association relationship in the two graphs as matching edges, and identify the unmatched edges as difference edges; Calculate the edit distance of the difference edge, where the edit distance includes the sum of the number of edge addition operations, deletion operations, and attribute modification operations, and normalize the edit distance to obtain the edge attribute edit distance; Determine the weighting rules for node attribute cosine similarity and edge attribute edit distance, where the weighting of node attribute cosine similarity is higher than the weighting of edge attribute edit distance. The weighting rules are dynamically adjusted based on the feedback results of historical case matching accuracy. The graph similarity is calculated using a weighted combination formula, which is a weighted summation of the complement of the node attribute cosine similarity and the edge attribute edit distance according to the weight distribution rule to obtain the comprehensive graph similarity between the current event and the case.

3. The smart city emergency response method combined with artificial intelligence according to claim 1 is characterized in that: The method of dynamically adapting and adjusting the preliminary response strategy set based on the real-time update feature of the event situation map to generate an adapted response strategy set adapted to the current event evolution includes: Real-time monitoring of node attribute changes in the event situation map, extracting nodes whose attribute value changes exceed the preset threshold as sensitive nodes, and recording the attribute change timestamp and change direction of the sensitive nodes; Analyze the location of sensitive nodes in the event situation map and the attributes of associated edges, determine the potential impact of changes in sensitive nodes on other nodes, and generate impact transmission paths; Identify, based on the impact transmission path and the sensitive node change characteristics, policy clauses in the preliminary response policy set that may become invalid, wherein the policy clauses that may become invalid refer to clauses whose policy execution conditions do not match the current sensitive node change characteristics; Adaptively modify policy clauses that may become invalid to obtain modified policy clauses. Modification methods include adjusting policy enforcement strength, changing policy scope, or adding auxiliary enforcement measures to generate modified policy clauses. The modified policy clauses are integrated with the policy clauses in the initial response policy set that have not expired, and the dynamic adjustment trigger conditions of policy execution are supplemented to generate an adaptive response policy set that adapts to the evolution of current events.

4. The smart city emergency response method combined with artificial intelligence according to claim 3 is characterized in that: The analysis of the position and associated edge attributes of sensitive nodes in the event situation map, determining the potential impact range of sensitive node changes on other nodes, and generating impact transmission paths includes: Locate the hierarchical affiliation of sensitive nodes in the event situation map, and determine whether they belong to the core impact layer, indirect impact layer, or potential impact layer. Sensitive nodes at different levels have different initial impact weights. Traverse all associated edges of the sensitive node and calculate the impact strength of the sensitive node change transmitted to the adjacent nodes through each associated edge based on the conduction delay parameter and diffusion attenuation parameter in the edge attributes; A threshold is applied to the impact strength received by adjacent nodes, and adjacent nodes whose impact strength exceeds the preset conditions are retained as first-level impact nodes. The transmission path segments from sensitive nodes to first-level impact nodes are recorded. Taking the first-level influence node as the new starting point, repeat the above-mentioned associated edge traversal and influence strength calculation process to identify the second-level influence node and the corresponding conduction path segment until the influence strength is lower than the preset condition and the traversal is stopped; All conduction path segments are connected in a conduction order to form a tree-like influence conduction path starting from a sensitive node and passing through multiple levels of influence nodes. Each node in the tree-like influence conduction path is marked with the received influence intensity and conduction delay time.

5. The smart city emergency response method combined with artificial intelligence according to claim 1 is characterized in that: The method of performing collaborative scheduling analysis on the adaptive response strategy set and the urban emergency resource network to generate a target collaborative solution that matches resource supply with strategy requirements includes: Analyzing the resource requirement characteristics of each strategy in the adaptive response strategy set, wherein the resource requirement characteristics include the required resource type, resource skill requirements, resource spatial distribution range, and resource availability time window; Constructing a city emergency resource network model, where nodes represent different types of emergency resource units, edges represent coordination capability parameters between resource units, and node attributes include the current state of resources and schedulable time; Match the resource requirements of the strategy with the node attributes of the resource network model, and calculate the degree to which each resource unit meets the resource requirements of the strategy. The degree of matching is determined based on resource type compatibility, skill requirement satisfaction, spatial accessibility, and time window overlap. Filter out candidate resource units that meet the strategic requirements based on the matching degree, analyze the coordination capability parameters between the resource units in the candidate resource unit set, and identify resource combinations that meet the preset coordination efficiency conditions; Bind each strategy to the corresponding resource combination that meets the preset collaborative efficiency conditions, mark the scheduling priority and collaborative execution process of the resource combination, and generate a target collaborative plan that matches resource supply with strategy requirements.

6. The smart city emergency response method combined with artificial intelligence according to claim 5 is characterized in that: The calculation of the matching degree of each resource unit to the strategic resource requirement includes: Determine whether the resource unit type is consistent with the resource type required by the strategy. If it is completely consistent, the resource type compatibility is the highest level. If it is partially consistent, the compatibility level is determined according to the type association rules. If it is completely inconsistent, the compatibility level is the lowest level. Extract the skill tag set of the resource unit and the skill requirement set of the strategy, and calculate the ratio of the number of elements in the intersection of the two sets to the number of elements in the strategy skill requirement set as the skill requirement satisfaction; Based on the current location coordinates of the resource unit and the coordinates of the policy action area, the shortest path length of the resource unit to the policy action area is calculated. Combined with the movement speed parameter of the resource unit, the arrival time is determined. The spatial accessibility level is determined based on the time difference between the arrival time and the execution start time required by the policy. Align the schedulable time window of the resource unit with the execution time window of the policy on the time axis, and calculate the ratio of the overlapping duration of the schedulable time window and the execution time window to the total duration of the policy execution time window as the time window overlap; After the resource type fit, skill requirement satisfaction, spatial accessibility level and time window overlap are quantitatively scored according to the preset order of importance, the matching degree of each resource unit to meet the strategic resource needs is obtained through weighted summation.

7. The smart city emergency response method combined with artificial intelligence according to claim 1 is characterized in that: Outputting an emergency response instruction including an execution priority according to the target coordination scheme, and sending the emergency response instruction to a corresponding emergency handling node to trigger a multi-agent linkage response operation, includes: Extracting the scheduling priority, collaborative execution process, resource combination and corresponding emergency response nodes of each strategy from the target collaborative plan to generate a strategy execution element set; Analyze the dependency relationship between the strategies in the strategy execution element set, determine the execution order of the strategies according to the scheduling priority and the dependency relationship, and generate a strategy execution sequence; Generate an initial emergency response instruction set including a strategy identifier, a resource combination identifier, an emergency response node identifier, and execution operation content based on each strategy in the strategy execution sequence and its corresponding collaborative execution process, resource combination, and emergency response node; The initial emergency response instruction set is classified according to the emergency handling node identifier, the execution order of the instructions received by each emergency handling node is determined according to the policy execution sequence, and the coordinated timing relationship of the instructions between different emergency handling nodes is determined to form a coordinated emergency response instruction set; According to the emergency handling node identifier in the collaborative emergency response instruction set, the corresponding emergency response instruction is sent to the corresponding emergency handling node to trigger a multi-agent linkage response operation.

8. A smart city emergency response system combined with artificial intelligence, characterized by: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the smart city emergency response method combined with artificial intelligence as described in any one of claims 1 to 7 above.

Images