A remote logistics risk early warning method, device and medium

Abstract

This application discloses a remote logistics risk early warning method, device, and medium, relating to the field of intelligent logistics technology. The method includes: collecting multi-source data from various data sources during the logistics transportation process and preprocessing the multi-source data to generate a standardized multi-source dataset; analyzing the standardized multi-source dataset using a pre-trained large language model to identify predefined risk types present in the logistics transportation process; if present, generating risk identification data containing risk descriptions and risk locations, and calculating the probability of occurrence and impact range of the identified risk based on the risk identification data using a preset risk assessment model; weighting the probability of occurrence and impact range according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data; and automatically determining the early warning level and the early warning recipient based on the risk assessment result data.

Description

Technical Field

[0001] This application relates to the field of intelligent logistics technology, and in particular to a remote logistics risk early warning method, device and medium. Background Technology

[0002] In modern logistics and transportation systems, timely identification and early warning of various risks during transportation are crucial to ensuring logistics efficiency and safety. Existing logistics monitoring systems typically possess basic data collection functions, such as GPS location tracking and using sensors to acquire cargo status information.

[0003] However, these systems have significant shortcomings in risk warning: on the one hand, their risk judgments mostly rely on preset simple rules or thresholds, such as triggering an alarm when the temperature exceeds a certain fixed value. This approach has a low level of intelligence and is difficult to cope with complex, variable, or hidden risk scenarios. On the other hand, the various links from data collection, risk identification, assessment to final warning are often isolated from each other, failing to form an organic and automated intelligent analysis closed loop.

[0004] Furthermore, the automation and intelligence levels of the entire risk warning process need improvement. The decision-making process from identification to warning still requires significant human intervention. For example, experienced professionals are needed to conduct secondary analysis and judgment of risk information to determine the risk level and assign it to the appropriate personnel. This reliance on manual decision-making not only results in a slow response but also makes it prone to missing the optimal response time when dealing with sudden and concurrent risk events due to limitations in human resources and processing efficiency.

[0005] Therefore, overcoming the shortcomings of existing technologies, such as reliance on manual labor, slow response, and inaccurate judgment in logistics risk assessment, has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] This application provides a remote logistics risk early warning method, device, and medium to solve the following technical problem: how to overcome the shortcomings of existing technologies in logistics risk assessment, such as reliance on manual labor, slow response, and insufficient accuracy.

[0007] In a first aspect, embodiments of this application provide a remote logistics risk early warning method, the method comprising: collecting multi-source data during the logistics transportation process through multiple data sources, and preprocessing the multi-source data to generate a standardized multi-source dataset; analyzing the standardized multi-source dataset using a pre-trained large language model to identify predefined risk types that exist in the logistics transportation; if they exist, generating risk identification data containing risk descriptions and risk locations, and calculating the probability of occurrence and the scope of impact of the identified risk based on the risk identification data using a preset risk assessment model; weighting the probability of occurrence and the scope of impact according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data; and automatically determining the early warning level and the early warning recipient based on the risk assessment result data.

[0008] In one embodiment of this application, the standardized multi-source dataset includes: standardized video data, environmental monitoring data, positioning trajectory data, and external environment data. Multi-source data is collected from various data sources during the logistics and transportation process, and preprocessed to generate the standardized multi-source dataset. Specifically, this includes: acquiring image data of the driver and cargo through a video surveillance system on the vehicle, and compressing and converting the image data to generate standardized video data; acquiring sensor data to be applied through sensors on the vehicle, and calibrating and filtering the sensor data to generate environmental monitoring data; wherein the environmental monitoring data includes: temperature data, humidity data, and vibration data; acquiring real-time location information of the vehicle through a positioning device, and parsing the real-time location information in conjunction with map data to generate positioning trajectory data; and acquiring traffic flow data and weather forecast data on the transportation route from a public data source through a network interface, and parsing the traffic flow data and weather forecast data to generate external environment data.

[0009] In one embodiment of this application, a pre-trained large language model is used to analyze a standardized multi-source dataset to identify predefined risk types that exist in logistics transportation. Specifically, this includes: performing multimodal data fusion on the standardized multi-source dataset; converting video data, environmental data, location data, and external environment data into a unified high-dimensional feature vector through the encoder layer of the large language model to generate a multimodal feature vector set; using the attention mechanism of the large language model to weight the multimodal feature vector set to extract key features related to risk and generate risk key feature data; inputting the risk key feature data into the classification output layer of the large language model, calculating similarity with the predefined risk type embedding vector, outputting the probability score of each risk type, and determining the existence of predefined risk types through threshold comparison.

[0010] In one embodiment of this application, generating risk identification data containing risk descriptions and risk locations specifically includes: retrieving a corresponding risk description template from a pre-built risk knowledge base based on the risk type; wherein the risk description template includes a risk type code corresponding to the risk type; obtaining the latitude and longitude coordinates of the current vehicle from the positioning trajectory data in a standardized multi-source dataset, mapping the latitude and longitude coordinates to geographic area information, generating risk location data, and adding timestamp information; and filling the risk location data and timestamp information into the risk description template to generate risk identification data.

[0011] In one embodiment of this application, based on risk identification data, a preset risk assessment model is used to calculate the probability of occurrence and the scope of impact of the identified risks. Specifically, this includes: parsing the risk identification data, extracting risk type codes, risk location data, and timestamp information to generate a risk feature parameter set; inputting the risk feature parameter set into the risk assessment model, and calculating the probability of occurrence data by combining historical risk event statistics and current environmental parameters through a probability calculation sub-model; and using the scope of impact calculation sub-model, simulating the risk diffusion path and scope of impact based on the risk location coordinates and logistics network topology data to generate scope of impact data.

[0012] In one embodiment of this application, the probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data. Specifically, this includes: retrieving the probability of occurrence weight value and the scope of impact weight value corresponding to the risk type from a weight configuration database to generate a weighting factor pair; normalizing the probability of occurrence data and the scope of impact data to convert the probability of occurrence into a standard probability value and the scope of impact data into a standard range index to generate a standardized probability and a standardized range; multiplying the probability of occurrence weight value in the weighting factor pair by the standardized probability and the scope of impact weight value by the standardized range to obtain a weighted probability value and a weighted range value; adding the weighted probability value and the weighted range value to obtain a comprehensive risk score, and converting the comprehensive risk score into a risk level code according to a predefined risk level threshold mapping table to generate risk assessment result data.

[0013] In one embodiment of this application, the warning level and warning recipient are automatically determined based on risk assessment result data. Specifically, this includes: parsing the risk assessment result data, extracting risk level codes and risk type information, and generating risk decision input data; querying the warning rule database, matching the warning level rules with the risk decision input data, and determining the warning level data; and retrieving the corresponding warning recipient list from the contact management database based on the warning level data, and generating warning recipient data by combining the recipient preference settings.

[0014] In one embodiment of this application, the method further includes: automatically sending warning information according to the warning level and the warning recipient, specifically including: selecting a corresponding warning message template from a message template library based on the warning level data to generate original warning message data; filling in the variable portion of the original warning message data to generate personalized warning message content data; wherein the variable portion includes risk description, risk location, and risk level; determining the preferred communication channel for each recipient according to the warning recipient data to generate channel selection data; and sending the personalized warning message content data to the terminal or system specified by the warning recipient through the communication channel in the channel selection data.

[0015] Secondly, embodiments of this application also provide a remote logistics risk warning device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a remote logistics risk warning method as described above.

[0016] Thirdly, embodiments of this application also provide a computer storage medium storing computer-executable instructions, which, when executed, implement a remote logistics risk warning method as described above.

[0017] This application provides a remote logistics risk early warning method, device, and medium, which has the following beneficial effects: By introducing a pre-trained large language model to perform in-depth analysis of standardized multi-source datasets, it achieves accurate and intelligent identification of multimodal and hidden risks in the logistics transportation process; furthermore, by constructing a risk assessment model and integrating weighted calculations, it can dynamically and quantitatively comprehensively assess the probability of occurrence and scope of impact of identified risks, and finally automatically determine the early warning level and recipient based on the assessment results. This application's solution effectively overcomes the shortcomings of existing technologies that rely on human experience and have slow response times, forming a fully automated and intelligent closed loop from data fusion, risk identification, assessment to early warning decision-making, significantly improving the overall efficiency, accuracy, and reliability of logistics risk early warning, and providing strong protection for the safety and timeliness of logistics transportation. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a remote logistics risk early warning method provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a remote logistics risk early warning device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] This application provides a remote logistics risk early warning method, device, and medium to solve the following technical problem: how to overcome the shortcomings of existing technologies in logistics risk assessment, such as reliance on manual labor, slow response, and insufficient accuracy.

[0021] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 A flowchart illustrating a remote logistics risk early warning method provided in this application embodiment. Figure 1 As shown in the figure, the remote logistics risk early warning method provided in this application embodiment specifically includes the following steps: Step 101: Collect multi-source data from various data sources during the logistics and transportation process, and preprocess the multi-source data to generate a standardized multi-source dataset.

[0023] In this embodiment, multi-source data refers to heterogeneous information obtained from multiple independent sources in the logistics and transportation process. Its purpose is to comprehensively cover various dynamic factors that may affect safety and efficiency during transportation, and to provide a rich and reliable data foundation for subsequent risk analysis.

[0024] In one embodiment of this application, the standardized multi-source dataset includes: standardized video data, environmental monitoring data, positioning trajectory data, and external environment data. Multi-source data is collected from various data sources during the logistics and transportation process, and preprocessed to generate the standardized multi-source dataset. Specifically, this includes: acquiring image data of the driver and cargo through a video surveillance system on the vehicle, and compressing and converting the image data to generate standardized video data; acquiring sensor data to be applied through sensors on the vehicle, and calibrating and filtering the sensor data to generate environmental monitoring data; wherein the environmental monitoring data includes: temperature data, humidity data, and vibration data; acquiring real-time location information of the vehicle through a positioning device, and parsing the real-time location information in conjunction with map data to generate positioning trajectory data; and acquiring traffic flow data and weather forecast data on the transportation route from a public data source through a network interface, and parsing the traffic flow data and weather forecast data to generate external environment data.

[0025] It is understood that this embodiment involves the collection and processing of various types of data, including standardized video data, environmental monitoring data, positioning trajectory data, and external environment data.

[0026] For example, a video surveillance system on a vehicle acquires real-time images of the driver's behavior (such as facial expressions or operational postures) and the cargo's condition (such as packaging integrity or securing status). These raw image data are then compressed and converted to a new format. Compression aims to reduce data size to optimize storage and transmission efficiency, while format conversion standardizes the data to a standardized video format such as H.264-encoded MP4, ensuring compatibility with subsequent analysis modules. This process not only improves data availability but also prevents analysis interruptions due to format incompatibility.

[0027] It should be noted that the environmental monitoring data is acquired through sensors deployed on vehicles, including temperature, humidity, and vibration data. These sensor data are calibrated and filtered. Calibration is used to correct inherent equipment errors (such as adjusting temperature readings by comparing them with standard reference values), while filtering is used to eliminate random noise (such as smoothing vibration data sequences using a moving average algorithm), thereby generating accurate and reliable environmental monitoring data, which is crucial for monitoring the transportation status of perishable goods or precision instruments.

[0028] In addition, positioning trajectory data is obtained by using positioning devices such as GPS or Beidou to acquire the real-time location information of vehicles and then analyzing this information in conjunction with high-precision map data. The analysis process includes mapping latitude and longitude coordinates to specific road names, administrative divisions or mileage markers, thereby generating positioning trajectory data with semantic information, such as for subsequent analysis of path deviation or travel delays.

[0029] Meanwhile, external environmental data is obtained from public data sources (such as traffic management departments or meteorological service APIs) via network interfaces to obtain traffic flow data and weather forecast data on transportation routes, and these data are parsed. The parsing operation involves extracting key fields (such as congestion level or rainfall probability) from unstructured data (such as JSON format) and converting them into a unified data structure defined within the system.

[0030] Ultimately, all these preprocessed data are integrated into a time-synchronized and spatially aligned standardized multi-source dataset. This dataset eliminates the heterogeneity of the original data, providing consistent and high-quality input for the subsequent risk identification module, thereby ensuring a smooth transition of the entire method from the data layer to the decision layer.

[0031] Step 102: Analyze the standardized multi-source dataset using a pre-trained large language model to identify predefined risk types that exist in logistics transportation.

[0032] In this embodiment, this step is the core of intelligent risk identification, and its key lies in transforming preprocessed, standardized, multi-source heterogeneous data into an effective assessment of transportation risks. It should be noted that the "large language model" in this embodiment is not limited to processing text data, but refers to a deep learning model architecture trained on large-scale multimodal data, possessing powerful feature extraction and pattern recognition capabilities. Through specific technical adaptations, it can handle various data types in this scenario, such as video, sensor data, and location information.

[0033] In one embodiment of this application, a pre-trained large language model is used to analyze a standardized multi-source dataset to identify predefined risk types that exist in logistics transportation. Specifically, this includes: performing multimodal data fusion on the standardized multi-source dataset; converting video data, environmental data, location data, and external environment data into a unified high-dimensional feature vector through the encoder layer of the large language model to generate a multimodal feature vector set; using the attention mechanism of the large language model to weight the multimodal feature vector set to extract key features related to risk and generate risk key feature data; inputting the risk key feature data into the classification output layer of the large language model, calculating similarity with the predefined risk type embedding vector, outputting the probability score of each risk type, and determining the existence of predefined risk types through threshold comparison.

[0034] Understandably, the process in this embodiment first involves multimodal data fusion of a standardized multi-source dataset. For example, through the encoder layer of a large language model, different types of raw data are mapped to a unified semantic space; for instance, video frame sequences are encoded into vectors representing driving behavior features, sensor data sequences such as temperature and vibration are encoded into vectors representing environmental stability, and location trajectories and external road condition information are encoded into vectors representing path traffic conditions. All these vectors together constitute a comprehensive multimodal feature vector set describing the current transportation status. Understandably, this fusion process breaks down barriers between different data sources, enabling the model to examine the transportation process from a holistic perspective rather than isolated dimensions.

[0035] Subsequently, the multimodal feature vector set is deeply processed using the attention mechanism unique to large language models. It should be noted that the attention mechanism can dynamically evaluate the differences in importance of different features for risk identification tasks and perform weighted processing accordingly. For example, when analyzing a transportation task, the model might assign higher attention weights to the video feature of "frequent driver blinking" and the sensor feature of "slight, high-frequency steering wheel vibration," while assigning a lower weight to the feature of "normal ambient humidity," thereby automatically focusing on the key information most relevant to the risk and generating refined risk key feature data.

[0036] Finally, the weighted risk key feature data is input into the model's classification output layer. At this stage, the model calculates the similarity between these features and predefined risk type embedding vectors (such as "fatigue driving," "loose cargo," "road congestion," etc.) and outputs a probability score for each risk type. Specifically, by setting a probability threshold (e.g., 0.7), the system determines the existence of a predefined risk type when the probability score for that risk type exceeds this threshold.

[0037] Understandably, the entire analysis process simulates the thinking process of human experts who synthesize multiple pieces of information to make judgments. However, thanks to the powerful computing capabilities of the large language model, it achieves high-speed and accurate identification of complex and hidden risks, providing accurate judgment basis for subsequent risk assessment and early warning.

[0038] Step 103: If it exists, generate risk identification data containing risk description and risk location, and based on the risk identification data, use a preset risk assessment model to calculate the probability of occurrence and the scope of impact of the identified risk.

[0039] It should be noted that the steps in this embodiment undertake the crucial task of transforming abstract risk identification results into specific and actionable risk information and conducting quantitative assessments.

[0040] In one embodiment of this application, generating risk identification data containing risk descriptions and risk locations specifically includes: retrieving a corresponding risk description template from a pre-built risk knowledge base based on the risk type; wherein the risk description template includes a risk type code corresponding to the risk type; obtaining the latitude and longitude coordinates of the current vehicle from the positioning trajectory data in a standardized multi-source dataset, mapping the latitude and longitude coordinates to geographic area information, generating risk location data, and adding timestamp information; and filling the risk location data and timestamp information into the risk description template to generate risk identification data.

[0041] In one embodiment of this application, based on risk identification data, a preset risk assessment model is used to calculate the probability of occurrence and the scope of impact of the identified risks. Specifically, this includes: parsing the risk identification data, extracting risk type codes, risk location data, and timestamp information to generate a risk feature parameter set; inputting the risk feature parameter set into the risk assessment model, and calculating the probability of occurrence data by combining historical risk event statistics and current environmental parameters through a probability calculation sub-model; and using the scope of impact calculation sub-model, simulating the risk diffusion path and scope of impact based on the risk location coordinates and logistics network topology data to generate scope of impact data.

[0042] In this embodiment, it is understood that the process of generating risk identification data first involves retrieving corresponding risk description templates from a pre-built risk knowledge base based on the identified risk types. For example, for the identified "cargo preservation risk," the system will retrieve a structured template similar to "Cargo located at <location> faces <risk type> risk due to <reason>." It should be noted that these templates contain standardized codes for the corresponding risk types and reserve placeholders for key variables. Simultaneously, the system obtains the precise latitude and longitude coordinates of the current vehicle from the location trajectory data in the standardized multi-source dataset, maps them to easily understandable geographic area information (e.g., "G4 Expressway K1253+500") through a geographic information system service, and combines this with precise timestamp information to form complete risk location data. Finally, the system fills the corresponding variable placeholders in the description template with the specific risk location data and specific parameter values ​​extracted from the key risk features (e.g., the specific value of vibration amplitude exceeding the standard), generating highly readable structured risk identification data, such as "Cargo located at G4 Expressway K1253+500 faces damage risk due to excessive vibration amplitude."

[0043] Understandably, generating standardized risk identification data lays the foundation for subsequent accurate quantitative assessment. Based on this, the system further utilizes a pre-set risk assessment model to calculate the probability of occurrence and the scope of impact of the identified risks.

[0044] Specifically, the system first parses the generated risk identification data, extracting the risk type code, precise risk location coordinates, and risk occurrence timestamp information. These key parameters are then combined into a machine-readable risk feature parameter set. It should be noted that this parameter set provides a standardized input interface for the risk assessment model. Subsequently, this risk feature parameter set is input into a specially trained risk assessment model. For example, the probability calculation sub-model within the model combines a historical risk event database (recording the statistical frequency of similar risks occurring in similar locations, weather conditions, and time periods) with currently acquired environmental parameters (such as road surface temperature and visibility) to calculate the actual probability of the risk occurring in the current context using built-in statistical algorithms (such as logistic regression or survival analysis).

[0045] Meanwhile, the sub-model for calculating the impact range within the overall model uses the risk location coordinates and topological data of the logistics network (including transportation route maps, distribution of transit warehouses, customer nodes, and other important node information) to simulate the potential spread path and impact boundaries of risk events through simulation algorithms (such as graph-based propagation models). For example, for a sudden landslide risk in a mountainous area, this sub-model will simulate and calculate the affected route length, the expected delay duration, and the number of subsequent transport trips that may be affected, thereby generating quantitative impact range data. Understandably, this step, by transforming qualitative risk assessments into quantitative probability and range indicators, provides precise data support for subsequent comprehensive risk level assessments and early warning decisions, enabling the entire risk warning process to move from the perception stage to the analysis and decision-making stage.

[0046] Step 104: Calculate the probability of occurrence and the scope of impact based on preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data.

[0047] It should be noted that the core purpose of the steps in this embodiment is to integrate the two different dimensions of quantified probability of occurrence and scope of impact into a unified and comparable comprehensive risk assessment through a scientific weighting mechanism. It is understandable that the importance of the probability of occurrence and scope of impact varies depending on the type of risk.

[0048] In one embodiment of this application, the probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data. Specifically, this includes: retrieving the probability of occurrence weight value and the scope of impact weight value corresponding to the risk type from a weight configuration database to generate a weighting factor pair; normalizing the probability of occurrence data and the scope of impact data to convert the probability of occurrence into a standard probability value and the scope of impact data into a standard range index to generate a standardized probability and a standardized range; multiplying the probability of occurrence weight value in the weighting factor pair by the standardized probability and the scope of impact weight value by the standardized range to obtain a weighted probability value and a weighted range value; adding the weighted probability value and the weighted range value to obtain a comprehensive risk score, and converting the comprehensive risk score into a risk level code according to a predefined risk level threshold mapping table to generate risk assessment result data.

[0049] In this embodiment, the system first retrieves the occurrence probability weight value and impact range weight value corresponding to the current risk type code from the weight configuration database, forming a pair of dedicated weight factors. For example, for high-risk types such as "hazardous chemical leaks," even though the probability of occurrence is low, its potential impact is extremely serious; therefore, the system will assign it a high impact range weight value. Conversely, for conventional risks such as "ordinary package delivery delays," the system may focus more on the probability of occurrence, thus the occurrence probability weight value will be relatively high. It is understandable that this differentiated weight configuration reflects the essential characteristics of different risk types, ensuring the rationality and relevance of the assessment results.

[0050] After obtaining the weighting factors, the system normalizes the occurrence probability data and influence range data calculated in step 103. It should be noted that the normalization process is necessary to eliminate differences in the units and numerical ranges of different indicators, enabling them to be compared and calculated under the same standard.

[0051] Specifically, the system uses a pre-defined normalization algorithm to convert the occurrence probability data into a standard probability value between 0 and 1. Simultaneously, it converts the impact range data (which may be a composite indicator encompassing multiple dimensions such as the number of affected nodes and delay duration) into a standard range index between 0 and 1. For example, the impact range data may be mapped to a standardized index range using a linear or non-linear function based on parameters such as the length of the affected logistics routes and the number of transit warehouses impacted. Subsequently, the system uses retrieved weight factor pairs to multiply the occurrence probability weight value by the standardized probability to obtain a weighted probability value, and multiplies the impact range weight value by the standardized range index to obtain a weighted range value. This multiplication process essentially amplifies or reduces the importance of the two assessment dimensions. Finally, the weighted probability value and the weighted range value are summed to obtain a comprehensive risk score between 0 and 100. The system compares this comprehensive risk score with a predefined risk level threshold mapping table. For example, a score of 0-30 corresponds to "low risk," 31-70 to "medium risk," and 71-100 to "high risk," thus converting the score into a clear risk level code. This risk level code, together with the original risk characteristic parameters, constitutes the final risk assessment result data. This data not only contains the essential information of the risk but also provides a scientifically calculated comprehensive rating, offering a direct and clear basis for subsequent early warning decisions. Step 105: Based on the risk assessment results, automatically determine the warning level and the warning recipient.

[0052] In one embodiment of this application, the warning level and warning recipient are automatically determined based on risk assessment result data. Specifically, this includes: parsing the risk assessment result data, extracting risk level codes and risk type information, and generating risk decision input data; querying the warning rule database, matching the warning level rules with the risk decision input data, and determining the warning level data; and retrieving the corresponding warning recipient list from the contact management database based on the warning level data, and generating warning recipient data by combining the recipient preference settings.

[0053] In one embodiment of this application, the method further includes: automatically sending warning information according to the warning level and the warning recipient, specifically including: selecting a corresponding warning message template from a message template library based on the warning level data to generate original warning message data; filling in the variable portion of the original warning message data to generate personalized warning message content data; wherein the variable portion includes risk description, risk location, and risk level; determining the preferred communication channel for each recipient according to the warning recipient data to generate channel selection data; and sending the personalized warning message content data to the terminal or system specified by the warning recipient through the communication channel in the channel selection data.

[0054] In this embodiment, this step marks the system's transition from the analysis and evaluation phase to the decision-making and execution phase, a crucial link in achieving closed-loop risk management. Specifically, the system first parses the risk assessment results data generated in the preceding steps, extracting the most critical risk level codes and risk type information to form refined risk decision input data. This data refinement process ensures that subsequent decisions are based on the most core risk characteristics. Subsequently, the system queries the early warning rule database, matching it against the risk decision input data to determine the early warning level. For example, the early warning rule database may pre-set a "high-risk level + route safety risk" combination to trigger a "Level 1 (Red) Early Warning," while a "medium-risk level + cargo preservation risk" combination corresponds to a "Level 2 (Yellow) Early Warning." It should be noted that this multi-dimensional rule matching mechanism makes the determination of early warning levels more accurate and in line with actual management needs.

[0055] After determining the warning level, the system retrieves the corresponding list of warning recipients from the contact management database based on the warning level data. Specifically, the contact management database not only stores contact information for personnel in different positions (such as drivers, fleet dispatch managers, customer service specialists, etc.), but also establishes association rules between personnel and warning levels and risk types. For example, when the system determines that a "Level 1 (Red) Warning" needs to be issued, it will automatically include the fleet manager, safety administrator, and relevant customers in the warning recipient list; while for a "Level 2 (Yellow) Warning," it may only need to notify the driver and the on-duty dispatcher. Understandably, this tiered notification mechanism ensures timely reporting and coordinated handling of important risks while avoiding excessive interference with unnecessary personnel.

[0056] Furthermore, this step also fully encompasses the generation and transmission of early warning information. Based on the determined early warning level data, the system selects the corresponding early warning message template from the message template library to generate raw early warning message data. It should be noted that these templates have pre-defined professional expression frameworks for different risk scenarios. Subsequently, the system uses specific parameters from the risk assessment results, including risk description, precise risk location, and clear risk level, as variables to fill the corresponding placeholders in the template, generating personalized early warning message content data. For example, a complete early warning message might read: "Warning! A risk of loose cargo securing (high risk) has been detected at K1253+500 on the Beijing-Hong Kong-Macau Expressway. Please verify immediately!" Simultaneously, the system determines the preferred communication channel based on the preference settings of each recipient recorded in the early warning recipient data, generating channel selection data. Finally, the personalized early warning message content data is accurately sent to the terminal or system designated by the early warning recipient through various communication channels such as SMS, email, application push, or system interface integration. Understandably, this end-to-end automated early warning process greatly shortens the response time from risk identification to early warning issuance, ensuring that critical risk information can be delivered to the most needed personnel in the most appropriate form, through the most effective channels, and at the earliest possible time. This buys valuable time for rapid risk mitigation measures and fully achieves the ultimate goal of intelligent risk early warning.

[0057] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a remote logistics risk early warning device, the structure of which is as follows: Figure 2 As shown.

[0058] Figure 2 This is a schematic diagram of the internal structure of a remote logistics risk early warning device provided in an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions executable by at least one processor, which are executed by at least one processor 201 to enable at least one processor 201 to: Multi-source data is collected from various data sources during the logistics and transportation process, and the multi-source data is preprocessed to generate a standardized multi-source dataset. We use pre-trained large language models to analyze standardized multi-source datasets to identify predefined risk types that exist in logistics and transportation. If present, generate risk identification data containing risk description and risk location, and based on the risk identification data, use a preset risk assessment model to calculate the probability of occurrence and scope of impact of the identified risk; The probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risks and generate risk assessment result data. Based on the risk assessment results, the warning level and the recipient of the warning are automatically determined.

[0059] Some embodiments of this application provide corresponding to Figure 1 A computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows: Multi-source data is collected from various data sources during the logistics and transportation process, and the multi-source data is preprocessed to generate a standardized multi-source dataset. We use pre-trained large language models to analyze standardized multi-source datasets to identify predefined risk types that exist in logistics and transportation. If present, generate risk identification data containing risk description and risk location, and based on the risk identification data, use a preset risk assessment model to calculate the probability of occurrence and scope of impact of the identified risk; The probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risks and generate risk assessment result data. Based on the risk assessment results, the warning level and the recipient of the warning are automatically determined.

[0060] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0061] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0062] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0063] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0064] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0066] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0067] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0068] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0069] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0070] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for remote logistics risk early warning, characterized in that, The method includes: Multi-source data is collected from various data sources during the logistics and transportation process, and the multi-source data is preprocessed to generate a standardized multi-source dataset. The standardized multi-source dataset is analyzed using a pre-trained large language model to identify predefined risk types that exist in logistics and transportation. If present, generate risk identification data containing risk description and risk location, and based on the risk identification data, use a preset risk assessment model to calculate the probability of occurrence and the scope of impact of the identified risk; The probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data. Based on the risk assessment results, the warning level and the warning recipient are automatically determined.

2. The remote logistics risk early warning method according to claim 1, characterized in that, The standardized multi-source dataset includes: standardized video data, environmental monitoring data, location trajectory data, and external environment data; Multi-source data is collected from various data sources during the logistics and transportation process, and this multi-source data is preprocessed to generate a standardized multi-source dataset, specifically including: The video surveillance system on the vehicle acquires image data of the driver and cargo, and compresses and converts the image data to generate standardized video data. Sensor data to be applied is acquired through sensors on vehicles, and the sensor data is calibrated and filtered to generate environmental monitoring data; wherein, the environmental monitoring data includes: temperature data, humidity data, and vibration data; The real-time location information of the vehicle is obtained through a positioning device, and the real-time location information is analyzed in combination with map data to generate positioning trajectory data; Traffic flow data and weather forecast data on transportation routes are obtained from public data sources through network interfaces, and the traffic flow data and weather forecast data are parsed to generate external environment data.

3. The remote logistics risk early warning method according to claim 1, characterized in that, The standardized multi-source dataset is analyzed using a pre-trained large language model to identify predefined risk types present in logistics transportation, specifically including: Multimodal data fusion is performed on the standardized multi-source dataset. The encoder layer of the large language model converts video data, environmental data, positioning data and external environment data into a unified high-dimensional feature vector to generate a multimodal feature vector set. The attention mechanism of a large language model is used to weight the multimodal feature vector set to extract key features related to risk and generate key risk feature data. The key risk feature data is input into the classification output layer of the large language model, and similarity is calculated with the predefined risk type embedding vector. The probability score of each risk type is output, and the existence of the predefined risk type is determined by threshold comparison.

4. The remote logistics risk early warning method according to claim 1, characterized in that, Generate risk identification data that includes risk descriptions and risk locations, specifically including: Based on the risk type, a corresponding risk description template is retrieved from a pre-built risk knowledge base; wherein, the risk description template includes a risk type code corresponding to the risk type; The latitude and longitude coordinates of the current vehicle are obtained from the location trajectory data in the standardized multi-source dataset, and the latitude and longitude coordinates are mapped to geographic area information to generate risk location data and add timestamp information. The risk location data and the timestamp information are filled into the risk description template to generate the risk identification data.

5. The remote logistics risk early warning method according to claim 1, characterized in that, Based on the risk identification data, a preset risk assessment model is used to calculate the probability of occurrence and the scope of impact of the identified risks, specifically including: The risk identification data is analyzed to extract risk type codes, risk location data, and timestamp information, and a risk feature parameter set is generated. The risk characteristic parameter set is input into the risk assessment model, and the probability of occurrence data is calculated by combining historical risk event statistics and current environmental parameters through the probability calculation sub-model. Using the impact range calculation sub-model, based on the risk location coordinates and logistics network topology data, the risk diffusion path and impact range are simulated to generate impact range data.

6. The remote logistics risk early warning method according to claim 5, characterized in that, The probability of occurrence and the scope of impact are weighted according to preset weighting factors to comprehensively assess the risk level of the identified risk and generate risk assessment result data, specifically including: Retrieve the occurrence probability weight value and impact range weight value corresponding to the risk type from the weight configuration database, and generate weight factor pairs; The occurrence probability data and the influence range data are normalized to convert the occurrence probability into a standard probability value and the influence range data into a standard range index, thereby generating a standardized probability and a standardized range. The weighted probability value and the weighted range value are obtained by multiplying the occurrence probability weight value in the weighted factor pair by the standardized probability and the influence range weight value by the standardized range. The weighted probability value and the weighted range value are added together to obtain the comprehensive risk score. Then, according to the predefined risk level threshold mapping table, the comprehensive risk score is converted into a risk level code to generate risk assessment result data.

7. The remote logistics risk early warning method according to claim 6, characterized in that, Based on the risk assessment results, the warning level and the warning recipient are automatically determined, specifically including: Analyze the risk assessment results data, extract the risk level code and risk type information, and generate risk decision input data; Query the early warning rule database, match the early warning level rules with the risk decision input data, and determine the early warning level data; Based on the warning level data, the corresponding list of warning recipients is retrieved from the contact management database, and warning recipient data is generated by combining the recipient preference settings.

8. The remote logistics risk early warning method according to claim 7, characterized in that, The method further includes: Based on the warning level and the warning recipient, warning information is automatically sent, specifically including: Based on the aforementioned warning level data, a corresponding warning message template is selected from the message template library to generate the original warning message data; Fill in the variable portion of the original early warning message data to generate personalized early warning message content data; wherein, the variable portion includes risk description, risk location, and risk level; Based on the data from the early warning recipients, the preferred communication channel for each recipient is determined, and channel selection data is generated. By selecting the communication channel from the data, the personalized early warning message content data is sent to the terminal or system designated by the early warning recipient.

9. A remote logistics risk early warning device, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform a remote logistics risk warning method as described in any one of claims 1-8.

10. A computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, a remote logistics risk early warning method as described in any one of claims 1-8 is implemented.

Images