Freight logistics information intelligent tracking management method and system

Abstract

The application discloses a freight logistics information intelligent tracking management method and system, belongs to the technical field of freight management, obtains unique identification information of a freight unit, and is associated with order content, carrier information and a preset transportation path to build a freight entity relationship graph, preprocesses and modality fuses multi-source real-time data in the transportation process, and extracts a structured cargo state feature set; based on a dynamic space-time modeling mechanism, intelligently judges the current freight state, generates a risk prediction result set, and predictively reconstructs the transportation path in combination with the risk result; after identifying an abnormal event, automatically performs path reconstruction, vehicle scheduling or user notification; synchronizes the evolution state and risk data to a cloud platform for subsequent task optimization; the application realizes full-process intelligent perception, real-time scheduling and abnormal closed-loop response in the logistics process, and significantly improves the transparency, robustness and time control ability of the transportation system.

Description

Technical Field

[0001] The present invention relates to the technical field of freight management, and in particular to a method and system for intelligent tracking and management of freight logistics information. Background Art

[0002] Freight logistics, a critical link in supply chain management, faces challenges such as dynamic transportation routes, diverse cargo conditions, and complex and changing environments. Existing freight tracking technologies, which mostly rely on GPS or RFID sensing, are limited to static location tracking of cargo. They lack the ability to comprehensively analyze real-time status, abnormal events, and environmental conditions during transportation, making them difficult to adapt to the requirements of modern smart logistics for full-process transparency and intelligent response.

[0003] Furthermore, traditional logistics information management systems often utilize centralized deployments, limited in processing power and real-time performance. These systems are unable to effectively integrate massive amounts of data from multiple, heterogeneous devices, nor are they capable of predicting and adaptively responding to unexpected events. These systems typically rely on fixed rules or manual monitoring to identify anomalies and adjust routes, resulting in low efficiency and a lack of dynamic learning and optimization mechanisms.

[0004] Therefore, there is an urgent need for an intelligent tracking and management method for freight logistics information that integrates multi-source data, performs real-time status analysis, intelligent anomaly prediction and response, adaptive routing adjustments, and edge collaborative processing. This method should move beyond traditional tracking technologies in this field and fundamentally enhance the intelligence and robustness of freight information processing by incorporating artificial intelligence modeling, edge computing collaboration, dynamic rule-driven development, and closed-loop feedback mechanisms. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for intelligent tracking and management of freight logistics information to address the deficiencies in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solutions: a method for intelligent tracking and management of freight logistics information, comprising:

[0007] S100, obtaining unique identification information of the freight unit, associating it with the order content, carrier information, and preset transportation route, and generating an initial freight entity relationship graph;

[0008] S200 uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets;

[0009] S300, based on a dynamically constructed spatiotemporal feature model, intelligently judges the current status of the freight unit and forms a freight risk prediction result set;

[0010] S400 builds a learnable multi-dimensional route assessment engine to perform predictive assessments on existing routes and replan transportation routes based on predicted risk information;

[0011] S500, when an abnormal event is detected, automatically reconfigure the route, dispatch vehicles, or notify the sender and receiver;

[0012] S600 synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks;

[0013] S700 outputs final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

[0014] Preferably, the S100 specifically includes:

[0015] Utilize a graph neural network-based entity recognition model to extract unique identification information for freight units from multi-source logistics data, including shipping documents, electronic tag information, and vehicle binding records.

[0016] The extracted identification information is combined with order details, carrier's historical fulfillment records, and preset transportation route parameters to establish association edges between entity nodes, thus constructing a multi-dimensional relationship graph with semantic weights.

[0017] Based on the graph embedding mechanism, the constructed freight entity graph is vector encoded to generate a structured freight graph vector representation for state recognition and path reasoning.

[0018] Preferably, the S200 specifically includes:

[0019] Real-time acquisition of multi-source sensing data including GPS positioning, acceleration, temperature and humidity, light, and door opening and closing status;

[0020] Use lightweight feature transformation models to perform edge preprocessing on collected data, perform outlier removal, sample synchronization, and extract short-term trends based on sliding windows;

[0021] A spatial correlation enhancement mechanism is used to collaboratively encode heterogeneous sensor data and output a structured cargo status feature set, which covers the abnormal values ​​of cargo inertial impact intensity and the coupled fluctuation values ​​of environmental disturbances.

[0022] Preferably, the calculation steps of the cargo inertia impact strength abnormal value ISI include: setting a fixed time sliding window T on the edge node, collecting the three-axis acceleration data stream ; Calculate the derivative of the total acceleration change rate in each window, that is, the mutation intensity; take the absolute value of the change rate and integrate it within the window to obtain the abnormal value of the cargo inertia impact intensity , the expression is: ;

[0023] The calculation steps of the environmental perturbation coupling fluctuation value EPCI include: collecting temperature temp(t), humidity hum(t), light intensity light(t) and vibration frequency vib(t) data in the same period; calculating the change rate of each data within the current sliding window; calculating the covariance matrix based on the standardized change rate to obtain the environmental perturbation coupling fluctuation value EPCI, which is expressed as: ;in, ; It represents the rate of change of the sensor variable x per unit time within the window; Var represents the joint variance among the four variables, and δ represents the sampling interval.

[0024] Preferably, the S300 specifically includes:

[0025] A spatiotemporal feature model based on joint modeling of trajectory history and sensor state is constructed, and the geographical path nodes of the freight unit, the sensor state evolution sequence and the event annotation data are aligned in time series to generate a multi-dimensional state trajectory stream;

[0026] The state trajectory stream is mapped into a multi-channel state representation matrix using a tensor embedding structure. Context enhancement is performed by combining entity relationship edges in the graph structure to extract risk vectors with path deviation memory and abnormal precursor sensitivity.

[0027] An adaptive sliding window mechanism is used to evaluate the trend and mutation of risk vectors, generating a set of freight risk prediction results, including delay warning, physical impact prediction, and environmental instability risk level assessment information.

[0028] Preferably, the S400 specifically includes:

[0029] A multi-dimensional path evaluation vector set is constructed based on historical transportation data, real-time traffic status and weather information, and the dynamic traffic characteristics and risk weight distribution between each path node are learned through a deep graph neural network.

[0030] Combined with the freight risk prediction result set, the edge weights and path priorities in the route graph are dynamically modified to form a time-sensitive route risk map;

[0031] The application of the reachability and robustness coupling optimization model is used to globally evaluate the current set of optional paths and select the path plan with the lowest overall risk and the best scheduling cost.

[0032] Preferably, the S600 specifically includes:

[0033] Based on the traceability coding mechanism, the state evolution data of each freight unit is structured and encapsulated to generate an evolution path sequence including time series state vectors, path node labels and abnormal event identifiers;

[0034] Upload the evolutionary path sequence and its corresponding risk prediction label to the cloud graph storage node through the distributed data synchronization center;

[0035] Based on the multidimensional clustering results of historical evolution paths and the risk trend regression model, key path parameters and context features are automatically extracted for adaptive optimization of path recommendation weights, transportation fault tolerance thresholds, and scheduling strategies for subsequent tasks.

[0036] The present invention also provides a freight logistics information intelligent tracking and management system, comprising:

[0037] The data modeling and entity graph module obtains the unique identification information of the freight unit, associates it with the order content, carrier information and preset transportation route, and generates the initial freight entity relationship graph;

[0038] The edge perception and preprocessing module uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets;

[0039] The intelligent analysis and risk identification module, based on a dynamically constructed spatiotemporal feature model, makes intelligent judgments on the current status of the freight unit and forms a freight risk prediction result set;

[0040] The route assessment and intelligent scheduling module builds a learnable multi-dimensional route assessment engine, performs predictive assessments on existing routes, and replans transportation routes based on predicted risk information.

[0041] The abnormal response and linkage disposal module automatically reconfigures routes, dispatches vehicles, or notifies both the sender and receiver when abnormal events are detected;

[0042] The cloud-based learning and task optimization module synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks;

[0043] The user interaction and visualization presentation module outputs the final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

[0044] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0045] 1. This invention achieves a collaborative, integrated processing mechanism for global perception and local response to freight status by constructing a full-process logistics management system encompassing freight unit identification and modeling, edge-sensing fusion, intelligent risk prediction, dynamic path assessment and reconstruction, anomaly linkage response, cloud-based learning optimization, and terminal visualization interaction. Compared to traditional logistics management methods that rely on static paths and single tracking methods, this invention not only possesses the ability to deeply integrate multi-source data and proactively identify anomaly trends, but also dynamically adjusts transportation routes and scheduling strategies based on risk status, significantly improving the safety, reliability, and flexibility of the transportation system.

[0046] 2. This invention, for the first time, integrates graph modeling, graph neural learning, spatiotemporal tensor analysis, and edge computing into logistics tracking and management scenarios, achieving an intelligent closed-loop from the perception layer to the analysis layer and the decision-making layer. The system boasts high scalability and model self-evolution capabilities, continuously optimizing path evaluation rules and transportation parameter configuration based on actual operational data. This reduces manual intervention costs and enhances the business's ability to rapidly adapt to emergencies. It is suitable for intelligent logistics management and control across multiple transportation tasks and complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 This is a mind map of the method of the present invention.

[0049] Figure 2 This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0051] Example 1, please refer to Figure 1 As shown, the method for intelligent tracking and management of freight logistics information described in this embodiment includes:

[0052] S100, obtaining unique identification information of the freight unit, associating it with the order content, carrier information, and preset transportation route, and generating an initial freight entity relationship graph;

[0053] S200 uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets;

[0054] S300, based on a dynamically constructed spatiotemporal feature model, intelligently judges the current status of the freight unit and forms a freight risk prediction result set;

[0055] S400 builds a learnable multi-dimensional route assessment engine to perform predictive assessments on existing routes and replan transportation routes based on predicted risk information;

[0056] S500, when an abnormal event is detected, automatically reconfigure the route, dispatch vehicles, or notify the sender and receiver;

[0057] S600 synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks;

[0058] S700 outputs final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

[0059] Obtain the unique identification information of the freight unit, associate it with the order content, carrier information and preset transportation route, and generate the initial freight entity relationship map. In order to achieve accurate management and intelligent tracking of the freight unit throughout the logistics process, this step is carried out through the following specific technical means:

[0060] Unique identification information acquisition: The system extracts the unique identification code (ID) by reading the physical identification device (including RFID tags, QR codes, NFC chips, etc.) attached to the shipping unit (such as a container, pallet, or packaging box). This ID serves as the primary index key for the entire tracking system, used to uniquely identify the corresponding relationship with other data sources.

[0061] Multi-source information analysis and association extraction: The system obtains the order content information corresponding to the identification code from the order processing platform, including the category of goods, quantity, shipping and receiving addresses, shipping time, etc.; at the same time, it retrieves the carrier information on the transportation service platform, such as carrier ID, historical performance evaluation, transportation vehicle information and driver information, etc.

[0062] Pre-defined transport route analysis: Based on the geographic information of the shipping and receiving locations, combined with a pre-configured transportation route database or transport scheduling engine, the system calculates and selects several candidate transport routes. Each route includes information such as transit nodes, cities passed through, estimated travel time, and risk level, which serves as the basis for subsequent route adjustments and risk assessments.

[0063] Constructing a freight entity relationship graph: Based on the extracted data, the system uses the "freight unit" as the core node and creates semantic edges between entities such as "order information," "carrier," and "transportation route." Each edge carries a weight, reflecting its importance or historical behavior (such as carrier punctuality or route congestion probability). The node and edge information in the graph is modeled and stored in a graph database (such as Neo4j or a vectorization framework based on graph neural networks).

[0064] Data structure output: The constructed initial freight entity map is stored in the form of a structured graph structure, which can be called by the subsequent state recognition module, path adjustment module and anomaly detection module, supporting fast query, path reasoning and intelligent judgment.

[0065] Edge computing nodes are used to preprocess and fuse multi-source real-time data during the transportation process to extract the cargo status feature set.

[0066] To achieve refined perception and dynamic monitoring of the environment and behavior during freight transportation, this step deploys embedded processing nodes with edge intelligence capabilities to perform on-site processing and integration of multimodal real-time data during transportation, reducing data transmission pressure and improving status response speed. The specific implementation method is as follows:

[0067] Edge sensing terminal deployment and data collection configuration: Embedded edge nodes with local computing capabilities are integrated into transport vehicles or cargo containers. These nodes integrate multiple sensor modules, including a three-axis accelerometer, temperature and humidity sensor, light intensity meter, access control status detector, and GPS module. Each module independently collects data based on the configured sampling frequency and uploads it to the edge processing unit.

[0068] Edge-level data preprocessing mechanism: Preprocessing operations are performed in edge nodes immediately after raw data is collected, including but not limited to:

[0069] Synchronous timing and sampling rate normalization ensure alignment of multi-source data;

[0070] Sliding window denoising algorithms, such as median filtering and local extreme value removal, remove data glitches;

[0071] The abnormal initial screening module identifies abnormal data segments based on rules such as upper and lower limits, time mutation amplitude, etc., and makes annotations or warnings.

[0072] Feature-level modal fusion mechanism: For sensory data from different sources and physical dimensions, a fusion mechanism based on spatial and semantic correlation is introduced. By analyzing the coupling relationship between acceleration changes and temperature, humidity, and light data, potential state anomalies can be identified. For example, if a sudden acceleration change is accompanied by a sudden drop in temperature and a sudden change in light intensity, the system will fuse it into an "exposure-type impact event."

[0073] State feature set generation and encoding: Preprocessed data enters the edge intelligent encoding module, which extracts abnormal values, including cargo inertial impact strength, and coupled fluctuations due to environmental disturbances. The feature results are output in a structured format (e.g., JSON or vector encoding) for use in subsequent state assessment modules. Some data is also uploaded to the cloud for further modeling.

[0074] The calculation steps for the cargo inertia shock intensity (ISI) include:

[0075] Set a fixed time sliding window T (such as 1 second) on the edge node to collect the three-axis acceleration data stream ;

[0076] Calculate the total acceleration rate derivative within each window, that is, the mutation intensity;

[0077] Take the absolute value of the rate of change and integrate it within the window to obtain the abnormal value of the cargo inertia impact strength , the expression is: ; Compare the calculated result with the preset threshold θISI. If the abnormal value of the cargo inertia impact intensity is greater than the preset threshold θISI, it is considered as an inertia impact abnormality.

[0078] The calculation steps of the environmental perturbation coupled fluctuation value EPCI include:

[0079] Collect temperature temp(t), humidity hum(t), light intensity light(t) and vibration frequency vib(t) data in the same period;

[0080] Calculate the rate of change (differential or difference) of each data within the current sliding window;

[0081] The covariance matrix is ​​calculated based on the standardized rate of change to obtain the environmental perturbation coupling fluctuation value EPCI, which is expressed as: ;in, ; represents the rate of change of the sensor variable x per unit time within the window; Var represents the joint variance among the four variables, which is used to measure the perturbation coupling strength; δ represents the sampling interval; a higher EPCI value indicates a more drastic mutation and a stronger synergy.

[0082] The environmental disturbance coupling fluctuation value EPCI is compared with the empirical disturbance threshold θEPCI. If the environmental disturbance coupling fluctuation value EPCI is greater than the empirical disturbance threshold θEPCI, it is determined that the environmental disturbance is abnormal.

[0083] Based on the dynamically constructed spatiotemporal feature model, the current status of the freight unit is intelligently judged to form a freight risk prediction result set.

[0084] To achieve pre-identification and intelligent prediction of potential risks during transportation, this step relies on a dynamic spatiotemporal feature modeling mechanism, integrating the trajectory behavior, state perception, and historical event data of the freight unit to form a reasonable state evolution structure and ultimately output a risk prediction result set. The specific implementation is as follows:

[0085] Trajectory-state sequence construction: The system uses freight units as the object, synchronously collecting their GPS coordinates and state characteristics (such as inertial impact intensity (ISI), environmental perturbation (EPCI), and position integrity index) in a time series, and constructing a set of time-stamped state trajectory vector streams. This vector stream not only includes spatial path changes but also superimposes the dynamic changes in the state of the cargo itself.

[0086] Dynamic spatiotemporal feature model construction: Using a sliding window mechanism, the trajectory vector stream is segmented into segments by time period. This is then expanded into a three-dimensional tensor representation of "time × state feature × path segment." This tensor is fed into an embedded spatiotemporal feature extraction network (which can be a lightweight graph convolutional temporal network or a cross-attention mechanism model) to automatically extract potential patterns between current freight behavior and typical abnormal events.

[0087] Graph-structured context enhancement mechanism: The aforementioned tensor results are integrated with the initial freight graph (generated by S100). During feature encoding, the system considers contextual semantic edge weights, such as carrier performance characteristics, historical route accident rates, and weather conditions, to form a risk identification vector with more business context. The entity graph structure enhances context sensitivity to the current state and improves judgment accuracy.

[0088] Freight risk prediction output mechanism: After the spatiotemporal feature encoding is completed, the system evaluates the potential risk level of the current freight status based on trend change rate, pattern similarity, or disturbance cascade criteria, and outputs a freight risk prediction result set, including but not limited to:

[0089] Predictive delay risks (e.g., route congestion trends);

[0090] Shock / drop risk (e.g. sudden change in inertia);

[0091] Risk of temperature control failure (e.g., drastic fluctuations in temperature and humidity within a short period of time);

[0092] Abnormal behavior warning (such as abnormal parking, route deviation).

[0093] The prediction results are synchronously fed back to the user end and transmitted to the cloud model center for continuous model training and parameter updates.

[0094] Build a learnable multi-dimensional path assessment engine to perform predictive assessments on existing paths and replan transportation paths based on predicted risk information.

[0095] To enhance the intelligent selection of freight routes and enable dynamic adjustments driven by risk perception during transportation, the system builds a route assessment engine with machine learning capabilities. This engine integrates historical behavior, real-time status, and environmental variables to perform multi-dimensional evaluation and replanning of route plans. The specific implementation is as follows:

[0096] The system first obtains a set of candidate paths for the planned route, which is generated by combining the map engine, historical trajectory data, and business rules. Each candidate path is split into multiple path segment nodes, and each node is associated with the following dynamic and static feature dimensions:

[0097] Static structural characteristics: route segment length, road grade, historical transportation costs;

[0098] Dynamic state features: current traffic flow, weather conditions, and regional event information (such as traffic control and road closures);

[0099] Risk input features: The abnormal level in the freight risk prediction result set output by step S300 is mapped to the route segment (such as high impact, high temperature fluctuation or deviation trend in a certain section of the route).

[0100] Based on the above features, the system constructs a multi-dimensional path node vector flow as the input feature set.

[0101] The core of the engine is a set of path modeling structures based on graph neural networks (GNNs), which include:

[0102] Each path is represented as a directed graph G=(V,E), where the node V represents the path segment and the edge E represents the spatiotemporal transfer relationship between nodes;

[0103] Node vectors are propagated through multi-layer graph convolution calculations, integrating the risk, flow, and stability of upstream and downstream path segments;

[0104] The engine uses path accessibility, estimated time, comprehensive risk value and change scheduling cost as training objectives, and outputs a comprehensive score vector for each path.

[0105] The model can be continuously optimized based on actual execution results (such as whether delays or accidents occur) through online learning mechanisms.

[0106] Based on the multi-dimensional path scoring results and business priority strategies (such as "safety first" and "time first"), the system automatically selects the path plan with the best current comprehensive score as the recommended plan; if the score of the current execution path is lower than the set threshold (or is identified as a high-risk state), the re-planning mechanism is triggered, and the scheduling system adjusts the execution task according to the new scoring path and notifies the user end and the carrier.

[0107] Replanning can not only partially replace high-risk path segments, but also reconstruct the entire path when necessary, and adjust the plan based on factors such as the current position of the terminal vehicle and the remaining time.

[0108] When an abnormal event is detected, the route is automatically reconstructed, the vehicle is dispatched, or the sender and receiver are notified.

[0109] To ensure safety, timeliness, and information symmetry during freight transportation, in this step, the system, based on the established risk prediction and route assessment engine, automatically triggers a coordinated response process when a certain type of emergency or warning state is detected. This process includes anomaly identification, route / vehicle intelligent scheduling, and multi-party collaborative alerting. The specific implementation is as follows:

[0110] During the real-time analysis of freight status, the system continuously monitors status streams from edge nodes and the cloud, including but not limited to:

[0111] Abnormal inertial impact intensity (such as ISI value exceeding the threshold);

[0112] Environmental disturbance coupled fluctuations (such as EPCI surges);

[0113] GPS path deviation or stagnation abnormality;

[0114] The carrier's performance behavior is abnormal (e.g. the task does not proceed according to the preset trajectory).

[0115] Once an abnormal event is confirmed or the risk prediction confidence exceeds the threshold, the system triggers the response process.

[0116] The system calls the path evaluation engine built in S400, eliminates the current high-risk path segment, re-evaluates the remaining candidate path set, and re-plans a new path plan based on the current vehicle position, remaining task nodes and time requirements. This path reconstruction supports:

[0117] Local dynamic detour: implement on-site avoidance for abnormal road sections;

[0118] Global path replacement: replans the entire route in the event of severe path blockage or systemic anomalies.

[0119] The new route will be pushed to the vehicle terminal in real time and the task scheduling platform will be updated synchronously.

[0120] If the system determines that the current transport unit cannot complete its mission on time (e.g., due to vehicle failure, loss of connection, or unreachable route), it automatically matches a nearby unloaded vehicle and initiates a "temporary connection" or "transfer handover" process through the scheduling engine to achieve a seamless task transfer. This process is based on a comprehensive decision-making process based on parameters such as the carrier's responsiveness, vehicle location, and acceptable cargo types.

[0121] The system synchronizes exception information, route change details, and possible transportation impacts to:

[0122] Shippers and consignees provide cargo status updates, estimated delays, and action suggestions.

[0123] The carrier's task monitoring center supports remote intervention, scheduling coordination, and responsibility identification;

[0124] System log center, for event archiving and post-event model training.

[0125] The notification supports adaptive content classification (warning level, fault level, accident level) and channel diversification (SMS, APP push, API docking, etc.).

[0126] To improve the robustness and intelligent response capabilities of the transportation system under uncertain conditions, the system has built a cloud-based adaptive route recommendation and strategy optimization model driven by historical evolution data. Specifically, it includes:

[0127] Input data encapsulation:

[0128] Evolutionary path sequence: the historical transport trajectory of each freight unit (including path node number, timestamp, and state characteristics);

[0129] Abnormal label sequence: risk identification results during each transportation process, such as shock, temperature control instability, and path deviation;

[0130] External context features: including carrier identification, weather data, traffic index, etc.

[0131] Feature extraction: Perform sliding window encoding on the evolution path and extract high-dimensional features such as "path segment stability rate", "abnormal event density", and "transportation timeliness deviation rate" as clustering and regression inputs.

[0132] Using a density-based trajectory clustering algorithm (such as DBSCAN or OPTICS variants), historical routes are clustered based on similarity (based on state dynamics, anomaly distribution, and node structure), forming several "typical path behavior clusters." Each cluster represents a transportation characteristic pattern (such as "high stability and low efficiency" or "high efficiency and high volatility"). Trajectory clustering not only considers path similarity but also integrates the temporal distribution of state evolution and the density of risk events to construct a multidimensional similarity metric.

[0133] A regression model (which can be a regularized gradient boosted tree or graph attention regression network) is trained within each route behavior cluster. Using the route evolution sequence as input, it predicts the probability of risk abrupt changes and the degree of potential transport deviations in future route segments. The formula is: Rt = f(Xt, Ct, Et); where Xt represents the route state feature vector (speed variability, temperature and humidity fluctuations, etc.); Ct represents the contextual variables (weather, time of day, carrier ID); Et represents the embedding vector of historical abnormal events; and Rt represents the predicted intensity of future route risk.

[0134] Based on the regression results and cluster labels, the system constructs a path decision recommendation table and extracts the following optimization factors:

[0135] Path recommendation weight: assign a risk sensitivity adjustment coefficient to each candidate path;

[0136] Transport fault tolerance threshold: Set the maximum acceptable deviation based on the mission importance and historical risk trends;

[0137] Dynamic scheduling strategy: Propose scheduling suggestions such as whether to enable double-vehicle cross-delivery, node speed limit, etc.

[0138] After each task is completed, the system will compare the execution path with the prediction results, and adjust the risk prediction model and path deviation regression weight based on the error, forming a continuous iterative optimization closed loop.

[0139] Output the final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

[0140] To achieve full transparency, status visibility, and controllable responses during the freight process, this step builds a mechanism for integrating status information and linking it with the user interface. By uniformly packaging and graphically outputting the final freight status, end users can understand the freight status in real time and respond promptly to key abnormal events. This step includes the following key technical links:

[0141] Status field definition and aggregation: The system summarizes the core information fields output by edge nodes, risk models, and scheduling modules throughout the entire transportation cycle of the freight unit, including but not limited to:

[0142] Current location information (latitude and longitude, city name, site ID);

[0143] Status summary indicators (such as current location temperature, acceleration, and access control status);

[0144] Accumulated risk score (such as impact risk index, delay probability, and temperature control fluctuation rate during this transportation process);

[0145] Execution path (actual driving route and timestamp);

[0146] Abnormal event records (event type, occurrence time, and handling status);

[0147] Current task progress (such as the ratio of completed nodes to total nodes, estimated arrival time).

[0148] Unified encapsulation structure design: All information is encapsulated into a standardized data structure (such as JSON or Protobuf format), with hierarchical relationships, field self-description capabilities and scalability, and can be adapted to different terminal platforms (mobile, web, console, etc.).

[0149] Front-end visual layout design:

[0150] Main map view: The map component is the main interface, showing the location information of the freight unit, the path trajectory and the geographical event annotation in real time;

[0151] Status panel: displays key indicators such as temperature and humidity, vibration status, and access control status, using color coding / icons to indicate status health;

[0152] Abnormal event timeline: lists risk events in chronological order, and users can click to view details and system response process;

[0153] Task progress chart: Display the progress of transportation task completion in the form of Gantt chart or bar chart;

[0154] Risk scoring radar chart: Scores the five-dimensional safety indicators of the transportation mission (time, temperature control, path, impact, and communication) to form a visual "mission health chart."

[0155] View Update Mechanism: The system uses lightweight protocols like WebSocket or MQTT to enable real-time push and rendering of freight status information. Upon receiving updates, front-end components perform minimal incremental updates based on the changes, ensuring responsiveness and display efficiency.

[0156] Multi-level alarm strategy triggering: For medium- and high-risk events identified from S300-S500, the system notifies users in different ways based on the event level (general / serious / urgent): General level: interface reminder or email prompt; Serious level: APP pop-up window and confirmation request; Emergency level: SMS, phone call or API direct dispatch system automatic intervention.

[0157] User response operation interface: provides interactive operation buttons, such as: "Request reroute"; "Delay receipt"; "Report exceptions and feedback"; "Confirm receipt / terminate task", etc.; all operations will be transmitted back to the cloud scheduling system and enter the task operation link.

[0158] Users can choose to export the complete tracking data of the current freight, or view the transportation history track and event response log on the platform to meet business needs such as quality audit and proof of compensation.

[0159] Example 2, please refer to Figure 2 As shown, the freight logistics information intelligent tracking and management system described in this embodiment includes:

[0160] The data modeling and entity graph module obtains the unique identification information of the freight unit, associates it with the order content, carrier information and preset transportation route, and generates the initial freight entity relationship graph;

[0161] The edge perception and preprocessing module uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets;

[0162] The intelligent analysis and risk identification module, based on a dynamically constructed spatiotemporal feature model, makes intelligent judgments on the current status of the freight unit and forms a freight risk prediction result set;

[0163] The route assessment and intelligent scheduling module builds a learnable multi-dimensional route assessment engine, performs predictive assessments on existing routes, and replans transportation routes based on predicted risk information.

[0164] The abnormal response and linkage disposal module automatically reconfigures routes, dispatches vehicles, or notifies both the sender and receiver when abnormal events are detected;

[0165] The cloud-based learning and task optimization module synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks;

[0166] The user interaction and visualization presentation module outputs the final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

[0167] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for intelligent tracking and management of freight logistics information, characterized by: include: S100, obtaining unique identification information of the freight unit, associating it with the order content, carrier information, and preset transportation route, and generating an initial freight entity relationship graph; S200 uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets; S200 specifically includes: real-time acquisition of multi-source sensor data including GPS positioning, acceleration, temperature and humidity, lighting, and door opening and closing status; edge preprocessing of the collected data using a lightweight feature transformation model, performing outlier removal, sampling synchronization, and short-term trend extraction based on a sliding window; collaborative encoding of heterogeneous sensor data using a spatial correlation enhancement mechanism, and outputting a structured cargo status feature set that covers outliers in cargo inertial impact strength and coupled fluctuations in environmental disturbances; The calculation steps of the cargo inertia impact intensity abnormal value ISI include: setting a fixed time sliding window T on the edge node, collecting the three-axis acceleration data stream ; Calculate the derivative of the total acceleration change rate in each window, that is, the mutation intensity; take the absolute value of the change rate and integrate it within the window to obtain the abnormal value of the cargo inertia impact intensity , the expression is: ; The calculation steps of the environmental perturbation coupling fluctuation value EPCI include: collecting temperature temp(t), humidity hum(t), light intensity light(t) and vibration frequency vib(t) data in the same period; calculating the change rate of each data within the current sliding window; calculating the covariance matrix based on the standardized change rate to obtain the environmental perturbation coupling fluctuation value EPCI, which is expressed as: ;in, ; Δx represents the rate of change of the sensor variable x per unit time within the window; Var represents the joint variance among the four variables, and δ represents the sampling interval; S300, based on a dynamically constructed spatiotemporal feature model, intelligently judges the current status of the freight unit and forms a freight risk prediction result set; S400 builds a learnable multi-dimensional route assessment engine to perform predictive assessments on existing routes and replan transportation routes based on predicted risk information; S500, when an abnormal event is detected, automatically reconfigure the route, dispatch vehicles, or notify the sender and receiver; S600 synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks; S700 outputs final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

2. The method for intelligent tracking and management of freight logistics information according to claim 1, characterized in that: The S100 specifically includes: Utilize a graph neural network-based entity recognition model to extract unique identification information for freight units from multi-source logistics data, including shipping documents, electronic tag information, and vehicle binding records. The extracted identification information is combined with order details, carrier's historical fulfillment records, and preset transportation route parameters to establish association edges between entity nodes, thus constructing a multi-dimensional relationship graph with semantic weights. Based on the graph embedding mechanism, the constructed freight entity graph is vector encoded to generate a structured freight graph vector representation for state recognition and path reasoning.

3. The method for intelligent tracking and management of freight logistics information according to claim 1, characterized in that: The S300 specifically includes: A spatiotemporal feature model based on joint modeling of trajectory history and sensor state is constructed, and the geographical path nodes of the freight unit, the sensor state evolution sequence and the event annotation data are aligned in time series to generate a multi-dimensional state trajectory stream; The state trajectory stream is mapped into a multi-channel state representation matrix using a tensor embedding structure. Context enhancement is performed by combining entity relationship edges in the graph structure to extract risk vectors with path deviation memory and abnormal precursor sensitivity. An adaptive sliding window mechanism is used to evaluate the trend and mutation of risk vectors, generating a set of freight risk prediction results, including delay warning, physical impact prediction, and environmental instability risk level assessment information.

4. The method for intelligent tracking and management of freight logistics information according to claim 1, characterized in that: The S400 specifically includes: A multi-dimensional path evaluation vector set is constructed based on historical transportation data, real-time traffic status and weather information, and the dynamic traffic characteristics and risk weight distribution between each path node are learned through a deep graph neural network. Combined with the freight risk prediction result set, the edge weights and path priorities in the route graph are dynamically modified to form a time-sensitive route risk map; The application of the reachability and robustness coupling optimization model is used to globally evaluate the current set of optional paths and select the path plan with the lowest overall risk and the best scheduling cost.

5. The method for intelligent tracking and management of freight logistics information according to claim 1, characterized in that: The S600 specifically includes: Based on the traceability coding mechanism, the state evolution data of each freight unit is structured and encapsulated to generate an evolution path sequence including time series state vectors, path node labels and abnormal event identifiers; Upload the evolutionary path sequence and its corresponding risk prediction label to the cloud graph storage node through the distributed data synchronization center; Based on the multidimensional clustering results of historical evolution paths and the risk trend regression model, key path parameters and context features are automatically extracted for adaptive optimization of path recommendation weights, transportation fault tolerance thresholds, and scheduling strategies for subsequent tasks.

6. A freight logistics information intelligent tracking and management system, used to implement the freight logistics information intelligent tracking and management method according to any one of claims 1 to 5, characterized in that: include: The data modeling and entity graph module obtains the unique identification information of the freight unit, associates it with the order content, carrier information and preset transportation route, and generates the initial freight entity relationship graph; The edge perception and preprocessing module uses edge computing nodes to preprocess and fuse multi-source real-time data during transportation and extract cargo status feature sets; The intelligent analysis and risk identification module, based on a dynamically constructed spatiotemporal feature model, makes intelligent judgments on the current status of the freight unit and forms a freight risk prediction result set; The route assessment and intelligent scheduling module builds a learnable multi-dimensional route assessment engine, performs predictive assessments on existing routes, and replans transportation routes based on predicted risk information. The abnormal response and linkage disposal module automatically reconfigures routes, dispatches vehicles, or notifies both the sender and receiver when abnormal events are detected; The cloud-based learning and task optimization module synchronizes the state evolution path and risk prediction information of each freight unit to the distributed cloud platform for path selection and parameter optimization of subsequent transportation tasks; The user interaction and visualization presentation module outputs the final freight status information for users to perform visual tracking and early warning response operations through smart terminals.

Images