A method, device and medium for generating a crowd movement trajectory in a disaster

Abstract

The application discloses a disaster-time crowd movement trajectory generation method, device and medium, first, the geographical entity set in the region, the current disaster point information, the current shelter information and the current position information of the residents are acquired, and the information is input into the pre-trained trajectory generation model together with the geographical entity set. Then, the travel log of the residents is generated according to the identity type of the residents, and then the travel log is mapped to the geographical entity set, and an initial target point set can be obtained. On this basis, the preliminary target point set is further screened, and a more accurate travel target point set can be obtained. Through the current disaster point information, the shelter information, the current position of the residents and the travel target point set, the movement trajectory of the residents during the disaster can be obtained. The embodiment of the application can realize accurate prediction of crowd flow in the disaster scene, thereby improving the quality and efficiency of disaster-time emergency response.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of disaster management technology, and in particular to a method, device, and medium for generating population movement trajectories during a disaster. Background Technology

[0002] In the field of disaster management, large language models (LLMs) can be used to accurately predict and simulate the movement trajectories of people during disasters. However, directly applying LLMs to generate crowd movement trajectories in disaster scenarios still faces challenges. On the one hand, LLMs may have relatively limited understanding of geographical knowledge, which may lead to geographically illogical inconsistencies in the generated trajectories. On the other hand, crowd movement in disaster scenarios is often influenced by a variety of complex factors, such as the specific location of the disaster site, the distribution of shelters, and current traffic conditions, all of which need to be fully considered during the trajectory generation process.

[0003] Therefore, achieving accurate prediction of population movement in disaster scenarios and improving the quality of emergency response during disasters is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0005] This application provides a method for generating population movement trajectories during disasters, which can accurately predict population flow in disaster scenarios, thereby improving the quality and efficiency of emergency response during disasters.

[0006] In a first aspect, embodiments of this application provide a method for generating population movement trajectories during a disaster, comprising: acquiring a set of geographic entities within a region, current disaster-affected area information, current shelter information, and current location information of residents; inputting the current disaster-affected area information, the current shelter information, the current location information, and the set of geographic entities into a pre-trained trajectory generation model; generating a travel log for the residents based on their identity type; mapping the travel log to the set of geographic entities to obtain an initial target location set; filtering the initial target location set to obtain a travel target location set; and obtaining the resident movement trajectory during the disaster based on the current disaster-affected area information, the current shelter information, the current location information, and the travel target location set.

[0007] In conjunction with the first aspect, in one embodiment of this application, the training steps of the pre-trained trajectory generation model include: acquiring a travel log data set and filtering the travel log data set to obtain a standard log data set; constructing an initial trajectory generation model, inputting the standard log data set into the initial trajectory generation model for iterative training, adjusting the first parameter of the initial trajectory generation model to obtain an intermediate trajectory generation model; acquiring historical disaster-affected point information and historical shelter information within the region; training the intermediate trajectory generation model based on the historical disaster-affected point information and the historical shelter information, adjusting the second parameter of the intermediate trajectory generation model to obtain the pre-trained trajectory generation model.

[0008] In conjunction with the first aspect, in one embodiment of this application, generating the resident's travel log based on the resident's identity type includes: constructing a resident identity type based on the resident's travel patterns within the region; generating the resident's travel intention based on the identity type, the travel intention including travel time, current location, destination location, and purpose; and serializing the travel intention to obtain the resident's travel log.

[0009] In conjunction with the first aspect, in one embodiment of this application, mapping the travel log to the set of geographic entities to obtain an initial set of target locations includes: extracting the target location of the trip from the travel log; determining the location type corresponding to the target location; and obtaining the initial set of target locations based on the location type and the set of geographic entities.

[0010] In conjunction with the first aspect, in one embodiment of this application, the geographic entity set includes multiple location information; the step of filtering the initial target location set to obtain a travel target location set includes: vectorizing the travel intention to obtain an intention embedding vector; vectorizing the location information to obtain a location embedding vector; calculating the cosine similarity between the intention embedding vector and the location embedding vector to obtain a similarity result; selecting a candidate target location set from the initial target location set based on the similarity result; and filtering the candidate target location set based on the current location and a preset balance factor to obtain the travel target location set.

[0011] In conjunction with the first aspect, in one embodiment of this application, the candidate target location set includes multiple target travel locations; the step of filtering the candidate target location set according to the current location and a preset balance factor to obtain a travel target location set includes: calculating the spatial distance between the current location and the multiple target travel locations; scoring the multiple target travel locations according to the preset balance factor and the spatial distance to obtain a travel score sequence; and filtering the candidate target location set according to the travel score sequence to obtain the travel target location set.

[0012] In conjunction with the first aspect, in one embodiment of this application, obtaining the resident's movement trajectory during a disaster based on the current disaster-affected area information, the current shelter information, the current location information, and the set of travel destinations includes: determining a target location from the set of travel destinations based on the current disaster-affected area information, the current shelter information, and the current location information; and obtaining the resident's movement trajectory during a disaster based on the target location and the current location information.

[0013] In conjunction with the first aspect, in one embodiment of this application, obtaining the resident movement trajectory during a disaster based on the target location and the current location information includes: calculating a set of paths for the target location and the current location information based on a path planning algorithm; and selecting a target path from the set of paths as the resident movement trajectory during a disaster.

[0014] Secondly, embodiments of this application provide an electronic device, including: at least one processor; at least one memory for storing at least one program; and when at least one of the programs is executed by at least one of the processors, implementing the disaster-time population movement trajectory generation method as described above.

[0015] Thirdly, embodiments of this application provide a computer-readable storage medium storing a processor-executable computer program, which, when executed by a processor, is used to implement the disaster-time population movement trajectory generation method described above.

[0016] This application provides a method for generating population movement trajectories during disasters. First, it comprehensively collects key information such as a set of geographic entities within the region, current disaster-affected area information, current shelter information, and residents' current location information. This information, along with the set of geographic entities, is input into a pre-trained trajectory generation model. Then, the model executes the following process: Based on residents' identity types (including different identity attributes), the model generates residents' travel logs, fully considering the possible movement paths of various types of residents in a disaster situation. Next, the model accurately matches this travel log with geographic entities to generate an initial set of destinations, all based on residents' possible travel needs and geographical realities. On this basis, the model meticulously filters the initial set, eliminating destinations that are inconsistent with reality or logically illogical, thereby determining a more accurate set of travel destinations. Finally, the model integrates current disaster-affected area information, shelter information, residents' current locations, and the filtered travel destinations (i.e., the set of travel destinations) to draw the residents' movement trajectories during the disaster. This application embodiment can achieve accurate prediction of population movement in disaster scenarios, thereby improving the quality and efficiency of emergency response during disasters. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for generating population movement trajectories during a disaster provided in an embodiment of this application;

[0018] Figure 2 This is a schematic diagram of the training steps of the trajectory generation model provided in the embodiments of this application;

[0019] Figure 3 This is provided by the embodiments of this application. Figure 1 The detailed flowchart of step 140;

[0020] Figure 4 This is provided by the embodiments of this application. Figure 3 The detailed flowchart of step 350;

[0021] Figure 5 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] It should be noted that although the flowchart shows a logical order, in some cases, the steps shown or described may be performed in a different order than that shown in the flowchart. The terms "first," "second," etc., used in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the structures, proportions, sizes, etc., depicted in the drawings are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the implementation conditions of this application. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, without affecting the effects and purposes achieved by this application, should still fall within the scope of the technical content disclosed in this application. Similarly, the terms such as "upper," "lower," "left," "right," "middle," and "one" used in this specification are only for clarity of description and are not used to limit the scope of implementation of this application. Changes or adjustments in their relative relationships, without substantially altering the technical content, should also be considered within the scope of implementation of this application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0025] In related technological fields, the simulation and prediction of population movement behavior during disasters generally suffer from insufficient adaptability to specific disaster scenarios. Specifically, these technologies often struggle to accurately predict dynamic changes in populations under extreme weather or emergency conditions. Especially when disasters actually occur, population behavior patterns are profoundly influenced by multiple complex factors, such as flood conditions, the availability of shelter in buildings, and group psychological effects (such as herd behavior). Current technological frameworks are insufficient in comprehensively considering these key factors. In other words, existing technological means cannot fully simulate the complex decision-making processes and dynamic changes faced by individuals in disaster environments.

[0026] In modern urban environments, population movement patterns play a crucial role in urban planning, traffic management, and emergency response due to population density. Especially during frequent extreme weather events (such as torrential rain, blizzards, and typhoons), a deep understanding of the complex movement patterns of people in cities during these disasters, and the ability to accurately predict these patterns, will greatly improve the efficiency of emergency response, thereby effectively reducing casualties and property damage.

[0027] In the field of disaster management, accurately predicting and simulating the movement trajectories of people during disasters is crucial for developing effective emergency response strategies. However, traditional trajectory prediction methods often rely on large amounts of historical trajectory data, which is not always feasible in many real-world scenarios, especially in disaster-prone areas where the availability of historical data may be limited by various factors, such as incomplete data records and the rarity of disaster events. It is worth noting that in recent years, with the rapid development of artificial intelligence technology, particularly the rise of LLMs (Limited Learning Models), new approaches have been provided to address this problem. LLMs have demonstrated powerful capabilities in contextual learning and few-shot learning, meaning that even without a large amount of training data, the model can generate reasonable outputs based on limited input information. This characteristic makes LLMs a powerful tool for trajectory generation tasks, especially in disaster scenarios where available historical trajectory data is very limited. However, despite these advantages, directly applying LLMs to generate population movement trajectories in disaster scenarios still faces challenges. On the one hand, LLMs may have relatively limited understanding of geographical knowledge, which may lead to geographically illogical inconsistencies in the generated trajectories. On the other hand, population movement in disaster scenarios is often influenced by a variety of complex factors, such as the specific location of the disaster site, the distribution of shelters, and the current traffic conditions. These factors all need to be fully considered during trajectory generation. Therefore, achieving accurate prediction of population movement in disaster scenarios and improving the quality of emergency response during disasters is a technical problem that urgently needs to be solved.

[0028] In view of this, embodiments of this application provide a method, electronic device, and computer-readable storage medium for generating population movement trajectories during disasters. First, key information such as a comprehensive collection of geographic entities within the region, current disaster-affected area information, current shelter information, and residents' current location information is gathered and input into a pre-trained trajectory generation model along with the geographic entity set. Subsequently, the model executes the following process: Based on residents' identity types (including different identity attributes), the model generates residents' travel logs, fully considering the possible action paths of various types of residents in a disaster situation. Next, the model accurately matches this travel log with geographic entities to generate an initial set of target locations, all based on residents' possible travel needs and geographical realities. On this basis, the model meticulously filters the initial set, eliminating destinations that are inconsistent with reality or logically illogical, thereby determining a more accurate set of travel target locations. Finally, the model integrates the current disaster-affected area information, shelter information, residents' current locations, and the filtered travel destinations (i.e., the set of travel target locations) to plot the residents' movement trajectories during the disaster. The embodiments of this application can achieve accurate prediction of population movement in disaster scenarios, thereby improving the quality and efficiency of emergency response during disasters.

[0029] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0030] Reference Figure 1 , Figure 1 This is a flowchart of a method for generating population movement trajectories during a disaster, provided in an embodiment of this application. The method includes, but is not limited to, steps 110 to 150.

[0031] Step 110: Obtain the set of geographic entities within the region, current disaster-affected area information, current shelter information, and current location information of residents, and input the current disaster-affected area information, current shelter information, current location information, and set of geographic entities into the pre-trained trajectory generation model;

[0032] Step 120: Generate the resident's travel log based on the resident's identity type;

[0033] Step 130: Map the travel logs to a set of geographic entities to obtain the initial set of target locations;

[0034] Step 140: Filter the initial set of target locations to obtain the set of travel target locations;

[0035] Step 150: Based on the current disaster site information, current shelter information, current location information, and travel destination set, obtain the residents' movement trajectory during the disaster.

[0036] Steps 110 to 150 will be described in detail below.

[0037] In a feasible embodiment, the set of geographic entities typically includes all Points of Interest (POIs) within the region. Each POI data record carries rich information, including at least a unique identifier (ID) to clearly distinguish each POI; type information, such as library, internet company, restaurant, etc., to indicate the basic attributes of the POI; and the name of the POI for easy user identification and retrieval. Furthermore, each POI data also records its precise geographic coordinates, including longitude and latitude, providing the POI's accurate location on the Earth's surface. Additionally, the data may include descriptions of the POI's functional attributes, such as opening hours and facility details. It is worth noting that a user trajectory can typically be represented as a stay at a set of POIs (i.e., a set of locations of specific interest). Specifically, for an individual identity P and geographic factor κ, the user's trajectory for one day is represented as χ(P,κ), where κ includes geographic influences such as weather, holidays, and disasters. χ(P,κ)={x 1 ,x 2 ...,x n}, where each x represents a user's stay x. i= <l i ,t i ,e i >, where l, t, and e represent a specific POI, the time of movement, and the purpose of the trip, respectively. χ={ <l 1 ,t 1 ,e 1 >, <l 2 ,t 2 ,e 2 >,..., <l N ,t N ,e N >}.

[0038] In one feasible embodiment, the selection of disaster-aware Points of Interest (POIs)—locations of specific importance during a disaster or in disaster risk management—can involve collecting a range of key information to accurately grasp the disaster situation and provide strong support for population movement prediction. Specifically, this information mainly covers two aspects: disaster-affected area information and shelter information. Disaster-affected area information requires detailed records of the specific geographical location of the disaster, including precise coordinates such as latitude and longitude, as well as the scope of the disaster's impact. This helps to clarify the affected area and severity of the disaster. Furthermore, the scope and characteristics of the impact may differ for different types of disasters (such as earthquakes, floods, and fires). Therefore, information such as the type and level of the disaster can also be collected to more accurately assess its impact on population movement. Shelter information mainly includes the geographical coordinates and names of shelters, which are crucial for determining safe gathering places for people after a disaster. In addition, the capacity of shelters, facility conditions, and surrounding traffic conditions are also factors that cannot be ignored, as they directly affect the efficiency and safety of population movement.

[0039] In one feasible embodiment, based on the obtained disaster site and shelter information, urban population distribution data can be further combined for comprehensive analysis. It is understood that population distribution data reveals the population density and distribution in different areas of a city, serving as a crucial basis for inferring dynamic travel patterns of people in the context of a disaster. By integrating this data with disaster site and shelter information, the flow trends of people after a disaster can be predicted more accurately.

[0040] In one feasible embodiment, the set of geographic entities can be obtained from sources such as Geographic Information System (GIS) data and Open Street Map (OSM). Current disaster-affected area information may include the specific location of the disaster, its impact range, and the type of disaster. This data may come from disaster monitoring agencies, emergency management departments, or on-site reports. Current shelter information may include the shelter's geographical location, capacity, and facilities. This information is typically provided by emergency management departments or relevant agencies. Residents' current location information can be obtained through mobile phone location data, social media check-in data, or public transportation card usage records.

[0041] In a feasible embodiment, after acquiring the set of geographic entities within the region, current disaster-affected point information, current shelter information, and current location information of residents, and completing data processing and cleaning, a text vector computation task can be initiated. The POI data is then stored in a database suitable for storing vector data (such as a Chroma vector database) for subsequent use. Data processing and cleaning can at least include steps such as data deduplication, data format standardization, data integrity and accuracy verification, and handling of missing data. Text vector computation can at least include steps such as text preprocessing, vectorization, and dimensionality reduction. Specifically, text preprocessing refers to converting the POI's name, type, and other information into a format suitable for vector computation. This may include steps such as word segmentation, stop word removal, and stemming. Vectorization refers to using word embedding techniques (such as Word2Vec, BERT, etc.) to convert text data into vector representations. These vectors can capture the semantic relationships between texts. As for dimensionality reduction, if the vector dimension is too high, dimensionality reduction techniques (such as PCA, t-SNE, etc.) can be considered to reduce computational complexity and storage space. After these POI data are stored in the Chroma vector database, they can be used for various subsequent operations, such as: Similarity search: Based on the input query vector (such as a resident's current location vector), the database is searched for similar POI vectors to quickly locate nearby disaster sites, shelters, etc. Spatial analysis: Spatial analysis is performed using the distance relationships between vectors, such as assessing the distribution of disaster sites and the coverage of shelters. Emergency response decision-making: By combining other relevant information (such as population distribution, traffic conditions, etc.), decision support is provided for emergency response and rescue operations.

[0042] In one feasible embodiment, such as Figure 2 As shown, the training steps of the pre-trained trajectory generation model may include at least steps 210 to 240.

[0043] Step 210: Obtain the travel log data set and filter the travel log data set to obtain the standard log data set;

[0044] Step 220: Construct an initial trajectory generation model. Input the standard log data set into the initial trajectory generation model for iterative training. Adjust the first parameter of the initial trajectory generation model to obtain an intermediate trajectory generation model.

[0045] Step 230: Obtain information on historical disaster sites and historical shelters within the region;

[0046] Step 240: Train the intermediate trajectory generation model based on historical disaster site information and historical shelter information, and adjust the second parameter of the intermediate trajectory generation model to obtain the pre-trained trajectory generation model.

[0047] In one feasible embodiment, travel log datasets can be used to describe residents' movement and activity patterns in the city. These datasets typically contain key information such as residents' origins, destinations, travel times, and modes of transportation, comprehensively reflecting residents' daily travel behavior. Specifically, by analyzing the origin and destination information in the travel log datasets, popular travel areas in the city can be identified, such as commercial centers, transportation hubs, and residential areas. For example, if a certain area has a high number of both origins and destinations, it indicates that the area is an important node for residents' daily travel, possibly a commercial center or transportation hub. Combining origin, destination, and mode of transportation information, residents' travel routes can be analyzed to understand their primary modes of transportation and routes chosen at different times. For example, during weekday morning and evening rush hours, residents may be more inclined to choose public transportation such as subways or buses, while during off-peak hours they may choose driving or cycling. By analyzing the time information in the travel log datasets, the frequency and distribution characteristics of residents' travel at different times can be understood. For example, on weekday mornings and evenings, residents travel more frequently, forming obvious morning and evening peaks; while on weekends or holidays, residents' travel times may be more dispersed. By combining time, mode of transportation, and location information from travel log datasets, we can identify residents' activity types, such as commuting, shopping, and entertainment. For example, if a resident chooses to drive to a shopping mall on a weekend afternoon and stays there for a considerable time, we can infer that they are shopping. By analyzing the time intervals and dwell time information in the travel log datasets, we can understand the frequency and duration of residents' different activity types. For example, if a resident chooses to go to the gym every weekday evening and stays for about an hour, we can infer that they exercise once a week for one hour each time.

[0048] In one feasible embodiment, the method for obtaining a travel log dataset may include at least: Public datasets: finding and downloading publicly available datasets related to travel. User-generated data: collecting user-generated travel log data through social media, mobile applications, or online surveys. Synthetic data: generating travel log data using data generation tools (such as GPT-4) or simulation models.

[0049] In one feasible embodiment, after obtaining the travel log dataset, the data can be further filtered, for example, by removing duplicate, erroneous, or incomplete data, to ensure data diversity and structure. The filtered data is called the standard log dataset, which will be used for subsequent model training.

[0050] In a feasible embodiment, step 220 involves using a pre-trained large model (such as a Qwen model with 7B parameters) as the initial trajectory generation model. A standard log dataset is input into this model for iterative training. By continuously adjusting the model's parameters (i.e., the first parameter), the model learns the characteristics and patterns of the travel log data. After multiple iterations of training, an intermediate trajectory generation model is obtained, which already possesses a certain trajectory generation capability.

[0051] In one feasible embodiment, in step 230, historical disaster-affected area information and historical shelter information within the region are crucial for the model to understand travel patterns under disaster scenarios. Disaster-affected area information may include the time, location, and type of the disaster; shelter information may include the location and capacity of shelters.

[0052] In a feasible embodiment, in step 240, historical disaster site information and historical shelter information are input as additional contextual information into the intermediate trajectory generation model. By adjusting the model's parameters (i.e., the second parameter), the model can better understand the impact of travel under disaster scenarios and stably infer possible travel patterns of individuals under unconventional circumstances. After this training step, a pre-trained trajectory generation model can be obtained. This model can not only generate reasonable trajectories under normal circumstances but also provide reliable trajectory generation services when dealing with complex disaster scenarios.

[0053] In one feasible embodiment, the pre-trained trajectory generation model can utilize POI data from the Chroma vector database. For example, model training: Trajectory generation models typically require a large amount of historical trajectory data for training. This historical trajectory data can contain information about POIs visited by the user at different times. During training, the model can learn the correlations between POIs and the user's movement patterns. POI data query: When a new trajectory needs to be generated, the trajectory generation model can query POI data from the Chroma vector database. Through similarity search or other query methods, the model can find POIs related to given conditions (such as user location, time, interests, etc.). Based on the retrieved POI data and other relevant information (such as road networks, traffic conditions, etc.), the trajectory generation model can generate a trajectory that matches the user's needs and movement patterns.

[0054] In one feasible embodiment, the core objective of travel log generation is to accurately simulate and record the daily activity logs of urban residents, particularly their movement paths and activity patterns within the urban space, using a pre-trained trajectory generation model. In generating residents' travel logs based on their identity types, resident identity types can first be constructed based on the residents' travel patterns within the region; then, residents' travel intentions can be generated based on these identity types, including travel time, current location, destination location, and purpose; finally, the travel intentions are serialized to obtain the residents' travel logs.

[0055] In one feasible embodiment, in the trajectory generation model, given an input x (such as time, weather, etc.), the model can generate a corresponding output y, i.e., a simulated trajectory or activity log, based on its internal algorithm and training data. During this process, the probability p1 of the output y (e.g., ...) The probability of the model generating y without additional information or constraints is reflected. However, to more accurately simulate the behavior of urban residents, resident identity types (personas) can be introduced as additional input information. These persons represent different resident types, such as students, office workers, and IT workers, each with its own unique travel patterns. By introducing differentiated persona distributions, the distribution of the model's generated output y can be influenced. The probability of outputting y can then be represented by P², where... This optimizes the diversity of trajectory patterns. Specifically, when the model receives a specific persona input, it can adjust its generation strategy based on the characteristics of that type of resident. For example, for students, the model might increase trajectories between school and home; while for office workers, it might increase trajectories between residence and workplace. This adjustment not only improves the realism of the simulation but also makes the generated trajectories more consistent with the behavioral characteristics of specific user types. Furthermore, differentiated persona distributions help the model generate more diverse trajectory patterns. Even for the same type of resident identity, the model can generate differentiated trajectories due to individual differences (such as job nature, lifestyle habits, etc.).

[0056] In a feasible implementation, based on resident identity types, the trajectory generation model can further generate residents' travel intentions. These travel intentions not only include basic information such as travel time, current location, and planned destination, but also explicitly indicate the purpose of the trip. This process is the result of the model inferring from multiple factors, including time factors, user type (i.e., resident identity type), and typical activity patterns. Specifically, the model can pre-determine some typical activity patterns based on resident identity types (such as students, office workers, IT workers, etc.). For example, students typically go to school in the morning, may stay at school or go home for lunch, and continue studying or participate in extracurricular activities in the afternoon; while office workers typically go to the office from home in the morning, may go out for lunch or stay at the office for lunch, continue working in the afternoon, and return home in the evening. When generating travel intentions, the model can first narrow down the possible travel purposes based on the current time period. For example, around 8:00 AM, a reasonable travel intention for an office worker might be "8:00 - home - office - go to work." The model will infer this intention based on this time point and user type. Similarly, around noon, for the same user type, the model might infer a travel intention of "12:00 - company - nearby restaurant - to eat at a restaurant near the company".

[0057] It's important to note that trajectory generation models may face limitations in their knowledge of specific city geography when generating travel logs. To address this and maintain the model's general adaptability, fuzzy Points of Interest (POIs) can be used. Fuzzy POIs are location descriptions that don't point to specific names or locations, but rather use more abstract and general terms to describe a location or area of ​​activity. For example, using "the library in the city center" instead of a specific library name, or "a nearby café" instead of a particular café. This descriptive approach helps the model maintain consistency when processing geographical data from different cities, avoiding generation biases caused by regional specificity. The advantage of using fuzzy POIs is that it allows the model to be more flexible and general in generating travel logs. The model doesn't need in-depth knowledge of every specific location in every city; instead, it can generate reasonable travel logs based on residents' travel intentions and activity patterns. For example, if the model knows that a resident is a law student and frequently visits the library to read law books at a certain time, then the model can generate a travel intention like "going to the library in the city center to read law books," without needing to know which specific library.

[0058] In one feasible implementation, a resident's movement and activity patterns throughout the day can be serialized into a series of travel intentions, which together constitute a complete travel log. These logs can record the user's activities at different time points in a structured manner, providing robust framework support for subsequent POI (Point of Interest) mapping and trajectory generation. A travel intention can be represented as (t... i ,pos i ,e i The travel log is defined as travel time - destination POI type - travel purpose. A resident's travel log can be an ordered pair of travel intentions, for example, a resident's travel log can be represented as <(t1,pos1,e1),(t2,pos2,e2),...,(t n ,pos n ,e n )>.

[0059] In one feasible embodiment, the process of mapping travel logs to a set of geographic entities essentially transforms the travel intentions in the travel logs into specific geographic locations, thereby obtaining an initial set of target locations. This process can be broken down into the following steps: First, extract the target location from the travel logs. Travel logs typically contain at least one travel intention, each detailing the travel time, current location, planned destination (i.e., target location), and travel purpose. This information allows for the identification of at least one specific target location. Second, determine the location type corresponding to the target location. Each target location has a specific location type, i.e., POI type, such as a library, internet company, or restaurant. Detailed analysis of the target location accurately identifies its corresponding location type. Finally, based on the location type and the set of geographic entities, obtain the initial set of target locations. The set of geographic entities is a database or collection containing various geographic entities (such as POIs) associated with specific location types. After determining the location type of the target location, the corresponding locations can be found in the set of geographic entities, and the initial set of target locations is formed based on these locations.

[0060] In one feasible embodiment, such as Figure 3 As shown, the specific process of filtering the initial set of target locations to obtain the set of travel target locations may include, but is not limited to, steps 310 to 350.

[0061] Step 310: Vectorize the travel intention to obtain the intention embedding vector;

[0062] Step 320: Vectorize the location information to obtain the location embedding vector;

[0063] Step 330: Calculate the cosine similarity between the intent embedding vector and the location embedding vector to obtain the similarity result;

[0064] Step 340: Based on the similarity results, select a set of candidate target locations from the initial target location set;

[0065] Step 350: Based on the current location and the preset balance factor, filter the set of candidate destinations to obtain the set of travel destinations.

[0066] In a feasible implementation, typically, natural language processing (NLP) techniques, such as word embeddings, sentence embeddings, or more complex deep learning models, can be used to transform the various elements of travel intent (travel time, current location, planned destination, and purpose) into vectors. Correspondingly, location information can also be transformed into structured vectors for comparison and analysis with the intent embedding vectors.

[0067] In one feasible embodiment, the degree of matching (i.e., similarity result) between travel intentions and various locations can be quantified by calculating the cosine similarity between the intention embedding vector and the location embedding vector (e.g., the embedding vector of each POI in a city POI dataset). Based on these similarity results, a candidate target location set can be selected from the initial target location set. Specifically, the top k POIs with the highest similarity to the travel intention embedding vector can be selected as the candidate set, where k is a positive integer. The calculation of cosine similarity is shown in Equation (1), which helps to quickly and effectively measure the similarity between two vectors:

[0068]

[0069] Here, v1 represents the intent embedding vector, and v2 represents the location embedding vector. This method allows for rapid narrowing of the search scope and ensures that the selected POIs not only meet the geographical requirements but also highly align with the travel intent semantically.

[0070] In one feasible embodiment, such as Figure 4 As shown, the specific process of step 350 may include, but is not limited to, steps 410 and 430.

[0071] Step 410: Calculate the spatial distance between the current location and multiple target travel points;

[0072] Step 420: Based on the preset balance factor and spatial distance, score multiple target travel destinations to obtain a travel score sequence;

[0073] Step 430: Filter the candidate destination set according to the travel score sequence to obtain the travel destination set.

[0074] In one feasible embodiment, step 410 aims to quantify the actual physical distance between the current location and each potential travel destination. Typically, the straight-line distance or driving distance from the current location to each target travel destination can be accurately calculated using map APIs or GPS data.

[0075] In one feasible embodiment, after calculating the spatial distances between the current location and multiple target travel destinations, these spatial distances can be filtered. During the filtering process, a reasonable distance threshold can be set according to actual needs. Only target travel destinations whose distance from the current location is within the threshold range will be included in subsequent consideration. This effectively excludes target travel destinations that are geographically too remote or inconvenient to reach, thereby improving the practicality and accuracy of the filtering results.

[0076] In a feasible embodiment, after obtaining spatial distance information, a comprehensive scoring mechanism can be used to evaluate the suitability of each point of interest (POI). Specifically, a preset balancing factor can be introduced, which can be used to balance semantic matching degree and spatial distance. By combining spatial distance with the balancing factor, a comprehensive travel score can be calculated for each target point of interest. In this way, a sequence sorted by travel score can be obtained, providing a basis for subsequent screening steps. As shown in formula (2), this formula can be used to calculate the total score of a point of interest (POI), which comprehensively considers the semantic score and the distance from the current location.

[0077]

[0078] Here, β is a balancing factor used to balance the contributions of semantic score and distance score. The value of β typically ranges from 0 to 1. If β is close to 1, it indicates that the semantic score contributes more to the total score; if β is close to 0, it indicates that the distance score contributes more to the total score. Semant icscore is the semantic score of the point of interest, which is usually calculated using semantic analysis algorithms based on textual information such as the name and category of the point of interest. (1-β) is a balancing factor for the distance score, complementary to β, ensuring that the total weight of the two factors is 1. 1 / distance(current_location, POI_location): This is the calculation method for the distance score. Here, the reciprocal of the distance is used to represent the score, meaning that the closer the distance, the higher the score. Distance(current_location, POI_location) represents the actual distance between the current location (current_location) and the location of the point of interest (POI_location).

[0079] In one feasible embodiment, after calculating the travel score for each target destination and obtaining a travel score sequence, a score threshold can be set, and locations exceeding this threshold can be included in the final set of travel target destinations. Alternatively, if the number of travel targets is limited, the top-ranked locations can be selected as the final travel targets.

[0080] In one feasible embodiment, based on the detailed situation of the current disaster site, information on existing shelters, the current location of residents, and a pre-selected set of travel destinations, the movement trajectory of residents during a disaster can be determined. This process can include the following two key steps: Step 1: Combining the severity of the current disaster site, the capacity and location distribution of shelters, and the current location information of residents, the most suitable shelter or safe location is selected as the target location from the set of travel destinations. Step 2: Once the target location is determined, one or more feasible movement routes for residents during a disaster can be planned using the current location information and the location of the target location. These routes are designed to ensure that residents can reach the target location safely and efficiently.

[0081] In one feasible embodiment, when determining the movement trajectory of residents during a disaster based on the target location and current location information, firstly, advanced path planning algorithms are used to calculate a series of possible movement paths, forming a path set, by combining the geographical information of the target location and the residents' current location. These path planning algorithms can comprehensively consider road conditions, obstacle distribution, traffic rules, and the potential impact of the disaster to ensure the feasibility and safety of the selected paths. Next, the path most suitable for the current disaster situation and the needs of the residents is selected from the calculated path set as the target path. Finally, the determined target path will be regarded as the movement trajectory of residents during the disaster, providing residents with clear evacuation or refuge guidance to help them safely and quickly reach the designated target location.

[0082] Reference Figure 5 This application also discloses an electronic device 500, which includes: at least one processor 510; at least one memory 520 for storing at least one program; when the at least one program is executed by the at least one processor 510, it can implement the disaster-time population movement trajectory generation method described above.

[0083] This application also discloses a computer-readable storage medium storing a processor-executable computer program. When the processor executes the processor-executable computer program, it can implement the disaster-time population movement trajectory generation method described above.

[0084] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating population movement trajectories during a disaster, characterized in that, include: The system acquires a set of geographic entities within the region, information on current disaster-stricken areas, information on current shelters, and information on the current locations of residents. It then inputs the information on current disaster-stricken areas, information on current shelters, information on current locations, and the set of geographic entities into a pre-trained trajectory generation model. The set of geographic entities includes information on multiple locations. Based on the resident's identity type, a travel log for the resident is generated. The travel log includes at least one travel intention, which includes the current location. Extract the target location of the trip from the travel log and determine the location type corresponding to the target location; Based on the location type and the set of geographic entities, an initial set of target locations is obtained; The travel intention in the travel log is vectorized to obtain an intention embedding vector, and the location information is vectorized to obtain a location embedding vector. Calculate the cosine similarity between the intent embedding vector and the location embedding vector to obtain a similarity result, and select a candidate target location set from the initial target location set based on the similarity result. The candidate target location set includes multiple target travel locations. Calculate the spatial distance between the current location and multiple target travel destinations, and score the multiple target travel destinations according to a preset balance factor and the spatial distance to obtain a travel score sequence; The candidate destination set is filtered based on the travel score sequence to obtain the travel destination set; Based on the current disaster-affected area information, the current shelter information, the current location information, and the set of travel destinations, the movement trajectory of residents during the disaster is obtained.

2. The method for generating population movement trajectories during a disaster according to claim 1, characterized in that, The training steps for the pre-trained trajectory generation model include: Obtain a set of travel log data and filter the set of travel log data to obtain a standard log data set; An initial trajectory generation model is constructed, and the standard log data set is input into the initial trajectory generation model for iterative training. The first parameter of the initial trajectory generation model is adjusted to obtain an intermediate trajectory generation model. Obtain information on historical disaster sites and historical shelters within the area; The intermediate trajectory generation model is trained based on the historical disaster site information and the historical shelter information, and the second parameter of the intermediate trajectory generation model is adjusted to obtain a pre-trained trajectory generation model.

3. The method for generating population movement trajectories during a disaster according to claim 1, characterized in that, The step of generating the resident's travel log based on the resident's identity type includes: Based on the travel patterns of residents within the area, construct resident identity types; The resident's travel intention is generated based on the identity type, and the travel intention includes travel time, current location, destination location, and purpose; The travel intention is serialized to obtain the resident's travel log.

4. The method for generating population movement trajectories during a disaster according to claim 1, characterized in that, The step of obtaining the resident movement trajectory during the disaster based on the current disaster-affected area information, the current shelter information, the current location information, and the set of travel destinations includes: Based on the current disaster site information, the current shelter information, and the current location information, determine the target location from the set of travel target locations; Based on the target location and the current location information, the movement trajectory of residents during the disaster is obtained.

5. The method for generating population movement trajectories during a disaster according to claim 4, characterized in that, The step of obtaining the resident movement trajectory during the disaster based on the target location and the current location information includes: The path set for the target location and the current location information is calculated based on the path planning algorithm; Select a target path from the set of paths to serve as the movement trajectory of residents during a disaster.

6. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When at least one of the programs is executed by at least one of the processors, the disaster-time population movement trajectory generation method as described in any one of claims 1 to 5 is implemented.

7. A computer-readable storage medium, characterized in that, It stores a processor-executable computer program, which, when executed by the processor, is used to implement the disaster-time population movement trajectory generation method as described in any one of claims 1 to 5.

Images