Transportation risk early warning method and system for smart logistics

Abstract

This invention provides a transportation risk early warning method and system for smart logistics, relating to the field of smart logistics technology. It acquires cargo status perception sequences and transportation environment perception sequences associated with the transportation task to be monitored. The method performs heterogeneous disturbance decoupling processing on these sequences, separating the sudden change base sequence in the transportation environment perception sequence and the lag offset sequence in the cargo status perception sequence, thus constructing a change propagation mapping. It then invokes a logistics risk recursive discovery network to detect the recursive propagation of change events along the cargo transportation path based on the change propagation mapping, obtaining a multi-level change propagation path set containing node sequences and accumulated propagated changes. Based on this multi-level change propagation path set, it performs risk localization, determining the risk initiation node and affected area, and generating transportation risk early warning information. This invention effectively separates heterogeneous disturbances and characterizes risk propagation paths, achieving accurate multi-level risk localization.

Description

Technical Field

[0001] This invention relates to the field of smart logistics technology, and more specifically, to a transportation risk early warning method and system for smart logistics. Background Technology

[0002] Current logistics and transportation risk monitoring typically involves installing cameras or vibration sensors on transport vehicles to periodically collect and monitor the condition of goods during transportation, or identifying potential risks by setting thresholds for environmental parameters such as temperature, humidity, and acceleration. However, existing technologies have significant limitations when facing complex and ever-changing real-world transportation scenarios.

[0003] On the one hand, traditional methods mostly assess risk based on a single data source, processing cargo status perception data and transportation environment perception data independently. This fails to fully consider the inherent correlation between the two data sources, making it difficult to accurately distinguish the root cause and transmission path of risk when faced with abnormal cargo status caused by sudden environmental changes. On the other hand, existing risk warning methods often employ static rule matching or simple threshold triggering mechanisms, which cannot effectively depict the dynamic transmission process of risk events along the transportation path between multiple cargo placement units. Especially when risk events have lag and cumulative characteristics, existing technologies often can only identify the abnormal state at the current moment, but cannot trace the starting point of the risk or estimate its scope. Furthermore, in actual transportation, disturbances to cargo placement units may originate from sudden changes in the external environment or from inherent factors of the cargo itself. These two types of disturbances are often coupled and difficult to separate in the perception sequence. Existing technologies lack effective decoupling methods for heterogeneous disturbances, significantly limiting the accuracy and timeliness of risk assessment. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a transportation risk early warning method for smart logistics, the method comprising: Acquire cargo status perception sequence and transportation environment perception sequence associated with the transportation task to be monitored. The cargo status perception sequence is collected by the sensors built into the cargo placement unit, and the transportation environment perception sequence is collected by the sensors external to the vehicle. The cargo status perception sequence and the transportation environment perception sequence are subjected to heterogeneous disturbance decoupling processing to separate the sudden change base sequence in the transportation environment perception sequence and the hysteresis offset sequence generated by the cargo status perception sequence following the sudden change base sequence, and a change transmission mapping between the sudden change base sequence and the hysteresis offset sequence is constructed. The existing logistics risk recursive discovery network is invoked, and the recursive propagation of change events along the cargo transportation path is detected and processed according to the change propagation mapping to obtain a multi-level change propagation path set. The multi-level change propagation path set includes the sequence of cargo placement unit nodes through each change propagation path and the cumulative amount of propagated change for each cargo placement unit node. Based on the multi-level change propagation path set, risk localization processing is performed on the transportation task to be monitored to determine the risk initiation cargo placement unit node and risk-affected cargo placement unit area in the transportation task to be monitored, and transportation risk early warning information for the transportation task to be monitored is generated.

[0005] Furthermore, embodiments of the present invention also provide a transportation risk early warning system for smart logistics, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described transportation risk warning method for smart logistics by executing the machine-executable instructions.

[0006] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions stored in a computer-readable storage medium, a processor of a transportation risk warning system for smart logistics reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the transportation risk warning system for smart logistics to execute the aforementioned transportation risk warning method for smart logistics.

[0007] Based on the above, by acquiring the cargo status perception sequence collected by the built-in sensors of the cargo placement unit and the transportation environment perception sequence collected by the external sensors of the transport vehicle, and by performing heterogeneous disturbance decoupling processing on the two types of perception sequences, it is possible to effectively separate the sudden change base sequence in the transportation environment perception sequence and the lag offset sequence generated by the sudden change base sequence in the cargo status perception sequence, and construct the change transmission mapping between the two. This solves the problem in the prior art that it is difficult to distinguish the causal relationship and temporal difference between environmental disturbance and cargo response. Furthermore, by calling the constructed logistics risk recursive discovery network, the recursive propagation of change events along the cargo transportation path is detected according to the change transmission mapping. This can identify a multi-level change propagation path set, which not only includes the cargo placement unit node sequence passed through each change propagation path, but also the cumulative amount of propagated change at each node. This overcomes the deficiency of the prior art that can only perform single-point static risk judgment and cannot perform path-level dynamic risk tracking. Ultimately, risk localization processing based on a multi-level change propagation path set can accurately determine the initial cargo placement unit node and the affected cargo placement unit area, generating transportation risk early warning information for the transportation tasks to be monitored. It has the dual capabilities of locating the source of risk and estimating the scope of impact, significantly improving the accuracy and operability of transportation risk early warning, and realizing the transformation of transportation risk early warning from passive response to proactive tracing, and from local judgment to global transmission and analysis. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the execution flow of the transportation risk early warning method for smart logistics provided in an embodiment of the present invention.

[0009] Figure 2 This is a logical management diagram of the transportation risk early warning method for smart logistics provided in this embodiment of the invention.

[0010] Figure 3 This is a schematic diagram of exemplary hardware and software components of a transportation risk early warning system for smart logistics provided in an embodiment of the present invention. Detailed Implementation

[0011] Figure 1 This is a flowchart illustrating a transportation risk early warning method for smart logistics provided in one embodiment of the present invention, which will be described in detail below.

[0012] The transportation risk early warning method for smart logistics provided in this application can be applied to scenarios involving real-time risk monitoring and early warning of transported goods during road freight transportation. In this scenario, the transport vehicle is equipped with cargo status sensing devices and transportation environment sensing devices. The cargo status sensing devices include vibration sensors and tilt sensors, while the transportation environment sensing devices include road elevation sensors and weather sensors. The transport vehicle continuously collects cargo status data and transportation environment data during its operation. The method analyzes the disturbance transmission relationship between these two types of heterogeneous data to discover the recursive propagation pattern of changes along the cargo transport path, identify the initial location and scope of risk impact, and generate transportation risk early warning information for the transport vehicle control unit and remote logistics scheduling unit to respond to the risk. It should be noted that the sensing data collection involved in this embodiment only targets the physical state of the cargo and transportation environment parameters, and does not involve the collection of personal information.

[0013] Step S110: Obtain the cargo status perception sequence and the transportation environment perception sequence associated with the transportation task to be monitored. The cargo status perception sequence is collected by the sensors built into the cargo placement unit, and the transportation environment perception sequence is collected by the sensors external to the vehicle.

[0014] The cargo status sensing sequence is collected at a fixed sampling frequency by vibration sensors and tilt sensors located at the bottom of each cargo placement unit within the cargo compartment of the vehicle. The vibration sensors employ microelectromechanical triaxial accelerometers, outputting vibration acceleration components in three orthogonal directions. These components are converted from analog to digital and output as digital values, with each sample yielding a three-dimensional vibration acceleration vector. The tilt sensors utilize a tilt measurement module combining a microelectromechanical gyroscope and an accelerometer, outputting the tilt angle of the cargo placement unit relative to the direction of gravity. The vibration and tilt sensor data are aggregated to the cargo compartment data acquisition terminal via a controller area network bus. The data acquisition terminal adds a timestamp to each sample and stores it in chronological order within the cargo status sensing sequence. The transportation environment sensing sequence is collected at the same fixed sampling frequency by a road surface elevation sensor mounted at the front of the vehicle chassis and a weather sensor mounted on the roof. The road surface elevation sensor uses a laser ranging module to emit pulsed laser beams towards the road surface and receive the reflected echoes. It calculates the vertical distance from the sensor to the road surface using the time-of-flight method, and after vehicle attitude compensation, obtains the height deviation value of the road surface relative to a reference surface. The meteorological sensors include an anemometer and a rain gauge. The anemometer uses ultrasonic wind speed measurement to output a scalar wind speed, and the rain gauge uses a tipping bucket rain gauge to output a scalar rainfall intensity per unit time. Each sampling point in the transportation environment perception sequence is also accompanied by a timestamp.

[0015] For example, combining Figure 2As shown, the sequence of cargo placement unit nodes, cargo status perception sequence, and transportation environment perception sequence of logistics vehicles together constitute the input data basis for heterogeneous disturbance decoupling processing.

[0016] In detail, the cargo placement unit node sequence is obtained. The attribute information of all cargo placement units loaded with cargo in this transportation mission is read from the cargo compartment layout configuration file of the transport vehicle. This includes the node identifier of each cargo placement unit, its row and column level three-dimensional spatial coordinates in the cargo compartment coordinate system, and a loading status marker indicating whether the cargo placement unit actually carries cargo. All cargo placement units with loaded cargo are organized into a cargo placement unit node sequence according to their spatial arrangement. This sequence prioritizes the longitudinal arrangement of the cargo compartment from front to back, followed by the transverse arrangement from left to right, and then the vertical arrangement from bottom to top. Spatial correspondence is established. Based on the three-dimensional spatial coordinates of each node in the cargo placement unit node sequence, spatial registration is performed with the installation position coordinates of the external sensors on the transport vehicle. The three-dimensional Euclidean distance between each cargo placement unit node and each external sensor is calculated, constructing a node-sensor spatial distance matrix. Simultaneously, the cargo status perception sequence is assigned according to the cargo placement unit identifier from which the data was collected, so that each node in the cargo placement unit node sequence is associated with the cargo status perception sub-sequence collected by its built-in sensors. The transportation environment perception sequence is labeled with the identifiers of the external sensors from which it was collected, binding the perception data of each external sensor to its installation location coordinates. The aforementioned cargo placement unit node sequence, node-sensor spatial distance matrix, cargo status perception sequence after attribution, and location-labeled transportation environment perception sequence together serve as the complete input data structure for the heterogeneous disturbance decoupling process in step S120. In step S120, after separating the sudden change base sequence from the transportation environment perception sequence, the sudden change is weighted according to the node-sensor spatial distance matrix and allocated to the cargo placement unit nodes within the affected range, providing a spatial constraint basis for the subsequent extraction of the lag offset sequence. In step S130, the cargo placement unit node sequence directly serves as the spatial carrier for the recursive propagation of changes in the logistics risk recursive discovery network, and the adjacency relationship between nodes is determined by the spatial arrangement order of the nodes in the sequence.

[0017] Step S120: Perform heterogeneous disturbance decoupling processing on the cargo state perception sequence and the transportation environment perception sequence, separate the sudden change base sequence in the transportation environment perception sequence and the hysteresis offset sequence generated by the cargo state perception sequence following the sudden change base sequence, and construct the change transmission mapping between the sudden change base sequence and the hysteresis offset sequence.

[0018] Step S121: Extract time-series continuous cargo status sampling points from the cargo status perception sequence to form a cargo status sampling point sequence; extract time-series continuous transportation environment sampling points from the transportation environment perception sequence to form a transportation environment sampling point sequence.

[0019] Following the ascending order of sampling timestamps in the cargo state perception sequence, cargo state sampling points are extracted sequentially from front to back at each sampling moment. Each cargo state sampling point contains a sampling timestamp, a three-dimensional vibration acceleration vector, and a tilt angle scalar. All extracted cargo state sampling points are then stored in a linear list of length M in their original chronological order; this linear list constitutes the cargo state sampling point sequence. Similarly, following the ascending order of sampling timestamps in the transportation environment perception sequence, transportation environment sampling points are extracted sequentially from front to back at each sampling moment. Each transportation environment sampling point contains a sampling timestamp, a road surface elevation difference scalar, and a meteorological parameter vector. All extracted transportation environment sampling points are then stored in a linear list of the same length M in their original chronological order; this linear list constitutes the transportation environment sampling point sequence. Sampling timestamps at the same list index position correspond to each other in the two sequences.

[0020] Step S122: Perform stationary trend stripping processing on the transportation environment sampling point sequence to remove the progressive trend component from the transportation environment sampling point sequence, and obtain the trend-stripped transportation environment sampling point sequence. The progressive trend component is determined based on the long-term moving average curve of the transportation environment sampling point sequence.

[0021] A moving average filter is used to smooth the transportation environment sampling point sequence, with the half-width of the sliding window denoted as W. For the road surface elevation difference scalar of the i-th sampling point in the transportation environment sampling point sequence, the road surface elevation difference scalar values ​​of the W sampling points before and after this sampling point are taken. These 2W+1 road surface elevation difference scalar values ​​are summed and divided by 2W+1; the quotient is the long-term moving average of the road surface elevation difference at the i-th sampling point. The long-term moving averages of the road surface elevation difference of all sampling points in the transportation environment sampling point sequence are connected to form a curve, resulting in the long-term moving average curve of the road surface elevation difference. The long-term moving average curves of the wind speed scalar and rainfall intensity scalar in the meteorological parameter vector are also calculated. The gradual trend component is the value of each of the above long-term moving average curves at each sampling point. The trend-free transportation environment sampling point sequence is obtained by subtracting the long-term moving average of the road surface elevation difference at each sampling point from the road surface elevation difference scalar, subtracting the long-term moving average of the wind speed at each sampling point from the wind speed scalar, and subtracting the long-term moving average of the rainfall intensity at each sampling point from the rainfall intensity scalar.

[0022] Step S123: Perform transient jump identification processing on the transportation environment sampling point sequence after trend stripping. Mark the sampling points whose amplitude changes more than the transient jump threshold within a unit time interval as sudden change points. Combine the sudden change points in time order to form a sudden change base sequence.

[0023] For the transportation environment sampling point sequence after trend stripping, the amplitude change between adjacent sampling points is compared pairwise from front to back. For the road surface elevation scalar, the amplitude change is equal to the absolute value of the difference between the road surface elevation scalar of the next sampling point and the road surface elevation scalar of the previous sampling point. For the wind speed scalar, the amplitude change is equal to the absolute value of the difference between the wind speed scalar of the next sampling point and the wind speed scalar of the previous sampling point. For the rainfall intensity scalar, the amplitude change is equal to the absolute value of the difference between the rainfall intensity scalar of the next sampling point and the rainfall intensity scalar of the previous sampling point. The transient jump thresholds for road surface elevation, wind speed, and rainfall intensity are all preset constant threshold values. If the amplitude change of the road surface elevation scalar exceeds the transient jump threshold for road surface elevation, or the amplitude change of the wind speed scalar exceeds the transient jump threshold for wind speed, or the amplitude change of the rainfall intensity scalar exceeds the transient jump threshold for rainfall intensity, then the next sampling point is marked as a sudden change point. All the marked sudden change points are arranged in ascending order of sampling timestamps to form a sudden change base sequence.

[0024] Step S124: Perform long-term drift compensation processing on the cargo state sampling point sequence to eliminate the slow offset components related to the cargo's own attributes in the cargo state sampling point sequence, and obtain the drift-compensated cargo state sampling point sequence.

[0025] During long-distance transportation, the baseline values ​​of vibration acceleration and tilt angle of goods will slowly shift over time due to factors such as gravity settling, creep of packaging materials, and gradual loosening of strapping. This slow shift is independent of external road surface disturbances. An exponentially weighted moving average method is used to estimate the long-term drift of the vibration acceleration amplitude and tilt angle values ​​at each sampling point in the cargo state sampling point sequence. The long-term drift estimate of the vibration acceleration amplitude at the i-th sampling point is calculated by multiplying the long-term drift estimate at the i-th (-1)-th sampling point by 1 minus the attenuation coefficient, and then adding the product of the vibration acceleration amplitude at the i-th sampling point multiplied by the attenuation coefficient, where the attenuation coefficient is a constant between 0 and 1. Subtracting the long-term drift estimate from the vibration acceleration amplitude at the i-th sampling point, and subtracting the long-term drift estimate from the tilt angle at the i-th sampling point, yields the drift-compensated cargo state sampling point sequence.

[0026] Step S125: Take each sudden change point in the sudden change base sequence as a reference time point, and extract the cargo state sampling point segment within the time delay search window corresponding to the reference time point from the cargo state sampling point sequence after drift compensation. The start time of the time delay search window is later than the reference time point.

[0027] For the k-th sudden change point in the sudden change base sequence, its sampling timestamp is read. The starting offset of the delay search window is denoted as τ1, and the ending offset as τ2, where τ1 and τ2 are both positive numbers and τ1 is less than τ2. In the drift-compensated cargo state sampling point sequence, all cargo state sampling points between the sampling timestamp of the sudden change point plus τ1 and the sampling timestamp of the sudden change point plus τ2 are extracted. This extracted segment of cargo state sampling points is the cargo state sampling point fragment within the delay search window corresponding to the sudden change point. The values ​​of τ1 and τ2 are set according to the vibration transmission delay characteristics from the transport vehicle chassis to the cargo compartment.

[0028] Step S126: Perform offset extraction processing on the cargo status sampling point segment within each time delay search window, obtain the time history curve of the deviation of the cargo status sampling point segment relative to the unperturbed state reference value, and use the time history curve of the deviation as the hysteresis offset sequence corresponding to the sudden change point.

[0029] The undisturbed state baseline value is calculated from a segment of cargo state sampling points collected before the transportation task begins, taken from a stationary, docked state. The arithmetic mean of the vibration acceleration amplitude in the stationary state is taken as the vibration acceleration baseline value, and the arithmetic mean of the tilt angle values ​​in the stationary state is taken as the tilt angle baseline value. For each sampling point in the cargo state sampling point segment, the absolute value of the difference between the vibration acceleration amplitude and the vibration acceleration baseline value at that sampling point is calculated as the vibration offset, and the absolute value of the difference between the tilt angle value and the tilt angle baseline value at that sampling point is calculated as the tilt offset. The vibration offset and tilt offset values ​​of each sampling point are arranged in chronological order according to the sampling timestamps to form a deviation time history curve. This deviation time history curve is the hysteresis offset sequence corresponding to the sudden change point.

[0030] Step S127: Associate the time delay, amplitude response ratio and recovery morphology parameters between each sudden change point and its corresponding hysteresis offset sequence to form a change transmission correlation pair. Summarize all change transmission correlation pairs to obtain the change transmission mapping.

[0031] The time delay is equal to the difference between the sampling timestamp of the sampling point in the hysteresis migration sequence where the first deviation exceeds the preset response threshold and the sampling timestamp of the sudden change point. The amplitude response ratio is equal to the maximum vibration deviation in the hysteresis migration sequence divided by the absolute value of the scalar change in road surface elevation difference corresponding to the sudden change point. The recovery morphology parameter describes the time it takes for the deviation to decay from its peak value to the point where the peak value is divided by the natural constant e. This parameter is determined by searching the deviation time history curve from the peak value to the point where the deviation first drops to the point where the peak value is divided by the natural constant e. Each sudden change point and its corresponding hysteresis migration sequence are associated using the three parameters: time delay, amplitude response ratio, and recovery morphology parameter, forming a change transmission association pair. The sum of all change transmission association pairs corresponding to sudden change points constitutes a change transmission mapping.

[0032] Step S130: Invoke the constructed logistics risk recursive discovery network, and perform discovery processing on the recursive propagation of change events along the cargo transportation path according to the change propagation mapping to obtain a multi-level change propagation path set. The multi-level change propagation path set includes the sequence of cargo placement unit nodes passed through each change propagation path and the cumulative amount of propagated change of each cargo placement unit node.

[0033] Step S131: Each pair of change transmission associations in the change transmission mapping is taken as an initial propagation unit. The occurrence time, occurrence location, and change intensity index of the sudden change point are extracted from the initial propagation unit as the three-element attributes of the initial propagation source. The occurrence location corresponds to the installation location of the external sensing device of the vehicle that generated the sudden change point.

[0034] For the k-th variation transmission correlation pair in the variation transmission mapping, the sampling timestamp of the sudden variation point is extracted from the sudden variation basis sequence as the occurrence time. The specific installation coordinates of the road elevation sensor corresponding to the sudden variation point on the vehicle chassis are retrieved from the vehicle's external sensor installation location record file. These installation coordinates are three-dimensional spatial coordinates, with three components representing the distance along the vehicle's longitudinal direction, the distance along the vehicle's lateral direction, and the height relative to the road surface. The amplitude response ratio in the variation transmission correlation pair is taken as the variation intensity index. The occurrence time, installation coordinates, and variation intensity index are combined into a ternary attribute of the initial propagation source.

[0035] Step S132: Based on the location of the initial propagation source, determine the cargo placement unit node that is spatially adjacent to the location of the initial propagation source, and use the ternary attribute of the initial propagation source and the hysteresis offset sequence corresponding to the initial propagation source as the input variation characteristics of the adjacent cargo placement unit node.

[0036] The cargo compartment of the transport vehicle is equipped with multiple cargo placement units for securing goods. Each cargo placement unit occupies a rectangular space, and all cargo placement units are fixed in the cargo compartment in a three-dimensional array arrangement of rows, columns, and layers. The spatial position of each cargo placement unit node is taken as the three-dimensional coordinates of the geometric center point of its rectangular space in the cargo compartment coordinate system. Taking the coordinates of the initial propagation source's occurrence position as the center, the three-dimensional Euclidean distance from this occurrence position coordinate to the spatial positions of each cargo placement unit node in the cargo compartment is calculated, and the Q cargo placement unit nodes with the smallest distances are selected as spatially adjacent cargo placement unit nodes. The input variation characteristics are a data structure containing three attribute fields: the first field is the ternary attribute of the initial propagation source, the second field is the complete deviation time history curve data of the hysteresis offset sequence corresponding to the initial propagation source, and the third field is the node identifier of the adjacent cargo placement unit node.

[0037] Step S133: Based on the magnitude response ratio of the change intensity index and the hysteresis offset sequence in the incoming change characteristics, and combined with the historical response characteristic records of adjacent cargo placement unit nodes, the derived change characteristics of adjacent cargo placement unit nodes after being affected by the incoming change characteristics are calculated. The derived change characteristics include the derived change intensity index and the derived hysteresis offset sequence.

[0038] Each cargo placement unit node maintains a historical response characteristic record, which includes a response gain coefficient and a response delay coefficient. The response gain coefficient is the average ratio of the output variation intensity to the input variation intensity of the cargo placement unit node in historical transportation tasks, and the response delay coefficient is the average time lag of the output variation relative to the input variation of the cargo placement unit node. The derived variation intensity index is calculated by multiplying the variation intensity index in the input variation characteristic by the response gain coefficient. The magnitude response ratio of the derived lag offset sequence is calculated by multiplying the magnitude response ratio of the lag offset sequence in the input variation characteristic by the response gain coefficient. The time base of the derived lag offset sequence is shifted backward by the response delay coefficient by the time base of the lag offset sequence in the input variation characteristic. The derived variation characteristic is also a data structure containing two attribute fields: the derived variation intensity index and the derived lag offset sequence.

[0039] Step S134: Take the adjacent cargo placement unit node as the new starting point for change propagation, take the derived change feature as the new input change feature, and perform change recursion propagation processing along the cargo transportation path to the next level of adjacent cargo placement unit nodes to generate the derived change feature of the next level cargo placement unit node.

[0040] The propagation direction is determined according to the spatial arrangement of cargo placement units along the longitudinal direction of the vehicle body from front to back within the cargo compartment. The spatially adjacent cargo placement unit nodes obtained in step S133 are used as the starting point for the propagation of changes at the current propagation level, and their derived change characteristics are used as new incoming change characteristics for propagation to the next level. In the cargo compartment layout model, the cargo placement unit node located directly behind the current propagation level cargo placement unit node along the longitudinal direction of the vehicle body is the next level spatially adjacent cargo placement unit node. For the next level spatially adjacent cargo placement unit nodes, their respective historical response characteristic records are read, and the calculation process of step S133 is repeated to generate the derived change characteristics of the next level cargo placement unit nodes.

[0041] Step S135: Repeat the step of using the outgoing change feature of the current cargo placement unit node as the incoming change feature of the next level cargo placement unit node and propagating it step by step to subsequent cargo placement unit nodes until the change propagation stopping condition is met, and obtain a change propagation path. The change propagation path includes all cargo placement unit nodes that are sequentially traversed from the initial propagation source.

[0042] The propagation of a change stops when the intensity of the induced change in the induced change characteristics of the current cargo placement unit node is lower than a preset propagation termination threshold, or when the current cargo placement unit node is the last cargo placement unit node in the longitudinal direction of the cargo compartment. All cargo placement unit nodes experienced by all propagation levels before the stopping condition is met are arranged in the order of propagation, forming a change propagation path. The first node in the propagation path is the adjacent cargo placement unit node directly affected by the initial propagation source, followed by the cargo placement unit nodes of each subsequent propagation level.

[0043] Step S136: Compare the differences between the incoming change characteristics and the outgoing change characteristics of each cargo placement unit node in the change propagation path to obtain the cumulative amount of transmitted change corresponding to the cargo placement unit node. The cumulative amount of transmitted change represents the net increase or decrease in the intensity of the change after passing through the cargo placement unit node.

[0044] For the j-th cargo placement unit node on the change propagation path, read the change intensity index from its incoming change characteristics and the induced change intensity index from its induced change characteristics. The cumulative amount of propagated change for this cargo placement unit node is equal to the difference between the induced change intensity index and the change intensity index. If the cumulative amount of propagated change is positive, it indicates that the cargo placement unit node has an amplification effect on the change, and the change intensity increases after passing through this node; if the cumulative amount of propagated change is negative, it indicates that the cargo placement unit node has an attenuation effect on the change, and the change intensity weakens after passing through this node.

[0045] Step S137: Collect all the variation propagation paths corresponding to all initial propagation units and the cumulative amount of transmission variation of each cargo placement unit node on each variation propagation path to form a multi-level variation propagation path set.

[0046] Steps S131 to S136 are performed on all change propagation pairs in the change propagation mapping, generating a total number of change propagation paths equal to the total number of change propagation pairs. Each change propagation path records the sequence of cargo placement unit nodes it passes through and the cumulative amount of propagated change at each cargo placement unit node. All change propagation paths and the cumulative amount of propagated change on each path are summarized into a set, which is the multi-level change propagation path set.

[0047] Step S140: Based on the multi-level change propagation path set, perform risk localization processing on the transportation task to be monitored, determine the risk originating cargo placement unit node and the risk-affected cargo placement unit area in the transportation task to be monitored, and generate transportation risk warning information for the transportation task to be monitored.

[0048] Step S141: Extract the ternary attributes of the initial propagation source of each multi-level change propagation path from the set of change propagation paths, compare the change intensity index in the ternary attributes with the preset change intensity classification table, and screen out the initial propagation sources whose change intensity index belongs to the risk level as risk starting candidate points.

[0049] The variation intensity grading table divides the variation intensity index into three continuous and non-overlapping numerical intervals, each corresponding to a risk level. The three risk levels, from lowest to highest, are: safe level, attention level, and risk level. The safe level corresponds to a variation intensity index below a first threshold; the attention level corresponds to a variation intensity index greater than or equal to the first threshold and less than a second threshold; and the risk level corresponds to a variation intensity index greater than or equal to the second threshold. The variation intensity index of the initial propagation source for each variation propagation path is compared with the numerical intervals in the variation intensity grading table. If the variation intensity index falls within the risk level's numerical interval, the initial propagation source is marked as a candidate risk initiation point.

[0050] Step S142: Match the cargo placement unit node corresponding to the risk initiation candidate point with the cargo placement unit node layout map of the transportation task to be monitored, and determine the risk initiation candidate point located within the coverage area of ​​the transportation task to be monitored as the risk initiation cargo placement unit node.

[0051] The cargo placement unit node layout diagram for the monitored transportation task records the spatial coordinates and node identifiers of each cargo placement unit node that actually carries cargo in this transportation task. The spatial coordinates of the cargo placement unit node corresponding to the risk initiation candidate point are compared one by one with the spatial coordinates of each node in the cargo placement unit node layout diagram. If the three-dimensional Euclidean distance between the two spatial coordinates is less than a preset overlap threshold, the risk initiation candidate point is determined to be within the coverage area of ​​this transportation task, and the risk initiation candidate point is marked as the risk initiation cargo placement unit node.

[0052] Step S143: For each risk-initiating cargo placement unit node, retrieve the target change propagation path starting from the risk-initiating cargo placement unit node in the multi-level change propagation path set, and extract the cumulative amount of propagated changes of all subsequent cargo placement unit nodes on the target change propagation path.

[0053] In the set of multi-level change propagation paths, the record of each change propagation path is read one by one. If the node identifier of the starting cargo placement unit node of a certain change propagation path is consistent with the node identifier of the current risk starting cargo placement unit node, then this change propagation path is the target change propagation path. For each target change propagation path, according to the arrangement order of the cargo placement unit node sequence on the change propagation path, starting from the first subsequent cargo placement unit node after the starting point, the cumulative amount of transmitted change corresponding to each cargo placement unit node is extracted sequentially.

[0054] Step S144: Compare the cumulative amount of transmitted changes of each subsequent cargo placement unit node on the target change propagation path with the preset cumulative amount risk threshold, and judge node by node along the recursive direction of the target change propagation path. When a cargo placement unit node appears where the cumulative amount of transmitted changes is lower than the cumulative amount risk threshold for the first time, all cargo placement unit nodes before that cargo placement unit node are included in the risk impact node set.

[0055] The cumulative risk threshold is a constant value determined based on statistical analysis of the risk impact range in historical transportation accidents. Starting from the first cargo placement unit node after the starting point of the target change propagation path, the process proceeds sequentially along the propagation recursion direction. For the q-th cargo placement unit node reached, its cumulative propagated change is compared with the cumulative risk threshold. If the cumulative propagated change is greater than or equal to the cumulative risk threshold, the process continues to the next cargo placement unit node. If the cumulative propagated change is less than the cumulative risk threshold for the first time, the process stops, and all cargo placement unit nodes from the first cargo placement unit node after the starting point to the stopping position are added to the risk impact node set.

[0056] Step S145: Perform geometric envelope generation on the spatial regions corresponding to all cargo placement unit nodes in the risk-affected node set on the cargo placement unit node layout diagram to obtain the risk-affected cargo placement unit region defined by the envelope boundary coordinate sequence.

[0057] On the layout diagram of cargo placement unit nodes, the coordinates of all eight vertices of the cuboid space occupied by each cargo placement unit node in the risk-affected node set are extracted to form a spatial point set. The Graham scan algorithm is used to calculate a two-dimensional convex hull on the projection points of the spatial point set onto the horizontal plane. Then, the two-dimensional convex hull is stretched vertically to cover the height range of all risk-affected nodes, resulting in a three-dimensional envelope geometry. The three-dimensional spatial coordinates of all boundary vertices on the surface of this three-dimensional envelope geometry are extracted and arranged in adjacency order to form an envelope boundary coordinate sequence. The spatial region defined by this envelope boundary coordinate sequence is the risk-affected cargo placement unit region.

[0058] Step S146: Combine the node identifier and spatial coordinates of the risk-initiating cargo placement unit node, the envelope boundary coordinate sequence of the risk-affected cargo placement unit area, and the maximum cumulative amount of transmitted change on the target change propagation path to form transportation risk early warning information.

[0059] The node identifier and three-dimensional spatial coordinates of the cargo placement unit node at the origin of the risk are obtained. The envelope boundary coordinate sequence of the cargo placement unit area affected by the risk is obtained. The maximum value among the cumulative propagated changes of all cargo placement unit nodes along multiple target change propagation paths is taken as the maximum cumulative propagated change. The node identifier, spatial coordinates, envelope boundary coordinate sequence, and maximum cumulative propagated change are combined into a transportation risk warning information data packet according to a predefined data message format. The header of this transportation risk warning information data packet contains an information type identifier and a message length field, and the message body stores the above four data items sequentially.

[0060] Step S210: Extract the meteorological element perception subsequence and the road condition perception subsequence from the transportation environment perception sequence. The meteorological element perception subsequence is collected by the meteorological sensor and the road condition perception subsequence is collected by the road surface adhesion sensor.

[0061] The meteorological element sensing subsequence is extracted from the meteorological parameter vector of the transportation environment sensing sequence. The meteorological sensing devices include ultrasonic anemometers and tipping bucket rain gauges. The meteorological element sensing subsequence contains a scalar time series of wind speed and a scalar time series of rainfall intensity. Each sampling point includes a sampling timestamp, wind speed value, and rainfall intensity value. The road surface condition sensing subsequence is extracted from the transportation environment sensing sequence. The road surface adhesion sensing device is an optical road surface condition sensor installed at the front of the vehicle chassis. This optical road surface condition sensor emits an infrared beam towards the road surface and receives the reflection spectrum. By analyzing the characteristics of the reflection spectrum, it determines the dryness and wetness of the road surface and the state of ice and snow coverage, and outputs an estimated value of the road surface adhesion coefficient. Each sampling point in the road surface condition sensing subsequence includes a sampling timestamp and a road surface adhesion coefficient value.

[0062] Step S220: Perform meteorological change segment detection processing on the meteorological element sensing subsequence to obtain the time of occurrence of meteorological change and the intensity level of meteorological change; perform road condition jump detection processing on the road condition sensing subsequence to obtain the time of occurrence of road condition jump and the level of road condition deterioration.

[0063] For the wind speed scalar time series and rainfall intensity scalar time series in the meteorological element sensing subsequence, abrupt change detection algorithms using a sliding window variance mutation detection algorithm are employed to detect abrupt changes. For each sampling point, the variance of wind speed values ​​within a window length before and after that sampling point is calculated. The ratio of the variance of the later window to the variance of the earlier window is used as the wind speed mutation index. When the wind speed mutation index exceeds a preset wind speed mutation judgment threshold, the sampling point is marked as the time of wind speed mutation occurrence, and the mutation intensity is divided into three levels based on the magnitude of the wind speed mutation index. Similarly, the rainfall intensity mutation index is calculated, and the time of rainfall mutation occurrence is marked. For the pavement adhesion coefficient values ​​in the pavement condition sensing subsequence, the change in pavement adhesion coefficient between adjacent sampling points is calculated. When the absolute value of the change exceeds a preset pavement jump judgment threshold, the sampling point is marked as the time of pavement condition jump. The pavement condition deterioration level is divided into three levels—slight deterioration, moderate deterioration, and severe deterioration—based on the numerical range of the pavement adhesion coefficient value after the jump.

[0064] Step S230: Perform time-series correlation analysis on the time of meteorological change and the time of road surface condition change to determine the causal relationship between meteorological events and road surface condition events, and calculate the induction time lead and induction intensity transmission coefficient from meteorological events to road surface condition events.

[0065] For each meteorological abrupt change, a search is performed within a preset association search time window to determine if a road condition abrupt change occurs. If a road condition abrupt change is found within the preset association search time window, this pair of meteorological abrupt change and road condition abrupt change occurrences is marked as a candidate causal relationship pair. For this candidate causal relationship pair, the induced time lead is equal to the difference between the road condition abrupt change occurrence and the meteorological abrupt change occurrence. The induced intensity transmission coefficient is equal to the quotient of the road condition deterioration level divided by the meteorological abrupt change intensity level. A statistical significance test is performed on this candidate causal relationship pair. The statistical significance test is conducted by calculating the conditional probability ratio of the meteorological abrupt change and the road condition abrupt change across all sampling times. If the conditional probability ratio exceeds a preset significance threshold, the candidate causal relationship pair is confirmed as a causal relationship pair.

[0066] Step S240: Add the causal relationship pair, the induced time advance, and the induced intensity transmission coefficient as new variable transmission pairs to the variable transmission mapping, and update the variable transmission mapping.

[0067] The occurrence time of meteorological abrupt changes in the causal relationship pair is mapped to the occurrence time of newly added sudden change points. The wind speed change index or rainfall change index corresponding to the occurrence time of the meteorological abrupt change is used as the change intensity index, and the installation location of the meteorological sensor is used as the occurrence location, forming a ternary attribute of the newly added sudden change point. The change sequence of the road adhesion coefficient within a window after the occurrence time of the road surface condition jump in the causal relationship pair is used as the corresponding hysteresis offset sequence. The induced time advance is used as the time delay of the change transmission pair, and the induced intensity transmission coefficient is used as the amplitude response ratio of the change transmission pair. The above newly added change transmission pair is added to the original change transmission mapping, and the change transmission mapping is updated.

[0068] Step S250: Re-execute the change propagation path discovery process of the logistics risk recursive discovery network using the updated change propagation mapping to obtain the updated multi-level change propagation path set.

[0069] Using the updated change propagation mapping as input, the complete logistics risk recursive discovery network processing flow from steps S131 to S137 is re-executed to generate an updated set of multi-level change propagation paths. The updated set of multi-level change propagation paths includes new change propagation paths originating from meteorological events.

[0070] Step S260: Based on the updated multi-level change propagation path set, perform secondary risk localization processing on the transportation task to be monitored, generate supplementary transportation risk warning information, merge the transportation risk warning information and the supplementary transportation risk warning information to form comprehensive transportation risk warning information for the transportation task to be monitored. The supplementary transportation risk warning information includes the starting node of secondary risks caused by the joint changes in meteorology and road surface and the area affected by secondary risks.

[0071] Using the updated set of multi-level change propagation paths as input, the risk location processing flow from steps S141 to S146 is re-executed. The risk originating cargo placement unit node generated in this round of processing is marked as the secondary risk originating node, and the risk-affected cargo placement unit area is marked as the secondary risk affected area, generating supplementary transportation risk warning information. The transportation risk warning information generated in step S146 is merged with the supplementary transportation risk warning information. During the merging, the respective risk originating node identifier and risk affected area boundary coordinate sequence are retained, and a source marker field to distinguish between the original risk and the secondary risk is added to form comprehensive transportation risk warning information.

[0072] Step S310: Obtain cargo attribute registration information corresponding to the transportation task to be monitored. The cargo attribute registration information includes cargo vulnerability level identifier and cargo tolerance environmental boundary parameters.

[0073] Cargo attribute registration information is retrieved from the cargo information database of the warehouse management system, with the query criteria being the cargo number associated with the transportation task to be monitored. The cargo vulnerability level is a comprehensive rating of the cargo's resistance to vibration, tilting, and impact during transportation, with values ​​ranging from 1 to 5, where a higher value indicates greater vulnerability. The cargo's environmental tolerance boundary parameters include three boundary values: maximum withstand vibration amplitude, maximum withstand tilt angle, and maximum withstand impact acceleration.

[0074] Step S320: Extract the change intensity index of each sudden change point from the change propagation mapping, and extract the cumulative amount of propagation change of each level of cargo placement unit node on each change propagation path from the multi-level change propagation path set.

[0075] Traverse all change propagation pairs in the change propagation map and extract the change intensity index corresponding to each change propagation pair. Traverse the cumulative amount of propagated change for each cargo placement unit node on each change propagation path in the multi-level change propagation path set.

[0076] Step S330: Using the cargo tolerance environment boundary parameters as tolerance benchmarks, the change intensity index of each sudden change point is mapped to the standardized risk value of the corresponding change type, and the boundary threshold of the same type in the cargo tolerance environment boundary parameters is also mapped to the standardized tolerance threshold. By comparing the standardized risk value with the corresponding standardized tolerance threshold, it is determined whether the sudden change point triggers a tolerance warning.

[0077] The volatility index is standardized by dividing it by the corresponding cargo tolerance boundary parameter. The standardized risk value is equal to the volatility index divided by the corresponding type boundary threshold. The standardized tolerance threshold is always 1. If the standardized risk value is greater than 1, the sudden change is considered to trigger a tolerance warning.

[0078] Step S340: Map the cumulative amount of transmission changes of each level of cargo placement unit node on each change propagation path to a standardized cumulative risk value. By comparing the standardized cumulative risk value with the corresponding standardized tolerance threshold, determine the first breakthrough node and the amount of the standardized cumulative risk value exceeding the standardized tolerance threshold.

[0079] The standardization method for the cumulative amount of transmission change is the same as that for the change intensity index. The standardized cumulative risk value is equal to the cumulative amount of transmission change divided by the corresponding type boundary threshold. The standardized cumulative risk value is compared with the standardized tolerance threshold node by node along the change propagation path from the starting point to the end point. The cargo placement unit node corresponding to the first time the standardized cumulative risk value exceeds the standardized tolerance threshold is recorded as the first breakthrough node. The difference between the standardized cumulative risk value and the standardized tolerance threshold is the exceedance amount.

[0080] Step S350: Based on the cargo vulnerability level label, determine the sensitivity response coefficient of the cargo to the changing factors, and combine the overrun amount of the first breakthrough node to calculate the estimated damage level classification of the cargo in the affected cargo placement unit.

[0081] For each increase in the vulnerability level rating of cargo, the sensitivity response coefficient is multiplied by a preset multiplier factor. The estimated damage level equals the exceedance multiplied by the sensitivity response coefficient. Based on the numerical range of the estimated damage level, cargo is classified into three levels: minor damage, moderate damage, and severe damage.

[0082] Step S360: Based on the estimated damage level classification, mark the cargo placement unit sub-regions corresponding to different damage levels within the risk-affected cargo placement unit area to form a risk-affected cargo placement unit area distribution map with damage level classification labels, and attach it to the transportation risk warning information to generate differentiated transportation risk warning information adapted to cargo attributes.

[0083] Within the risk-affected cargo placement unit area, consecutive cargo placement unit nodes with the same estimated damage level are aggregated into a sub-region, and this sub-region is filled and marked with the corresponding damage level color. The boundary coordinates and damage level codes of all sub-regions are written as additional data segments into the transportation risk warning information.

[0084] Step S410: Extract the cargo vibration sensing subsequence and cargo tilt sensing subsequence from the cargo status sensing sequence. The cargo vibration sensing subsequence is collected by a vibration sensing device, and the cargo tilt sensing subsequence is collected by a tilt angle sensing device.

[0085] The cargo vibration sensing subsequence is extracted from the three-dimensional vibration acceleration vector of the cargo state sensing sequence, and the composite amplitude of the three-axis vibration acceleration is taken as the vibration intensity value of each sampling point in the cargo vibration sensing subsequence. The cargo tilt sensing subsequence is extracted from the tilt angle scalar of the cargo state sensing sequence.

[0086] Step S420: Perform vibration event segmentation processing on the cargo vibration sensing subsequence to obtain the vibration event start time, vibration event end time and vibration event duration; perform tilt event segmentation processing on the cargo tilt sensing subsequence to obtain the tilt event start time, tilt event end time and tilt event duration.

[0087] For the cargo vibration sensing subsequence, vibration events are detected using a vibration intensity threshold triggering method. When the duration for which the vibration intensity value continuously exceeds the preset vibration event triggering threshold exceeds the preset minimum event duration, the moment when the vibration event triggering threshold is first exceeded is recorded as the vibration event start time, and the moment when the vibration intensity value falls back below the vibration event triggering threshold is recorded as the vibration event end time. The duration of the vibration event is equal to the vibration event end time minus the vibration event start time. The cargo tilt sensing subsequence detects tilt events in the same way.

[0088] Step S430: Perform time axis overlap analysis on the start time of the vibration event, the end time of the vibration event, the start time of the tilt event, and the end time of the tilt event to determine the concurrent event segments and independent event segments of vibration and tilt, and extract the vibration peak intensity feature and tilt peak angle feature for each concurrent event segment.

[0089] The intersection of the time intervals of vibration events and tilt events on the time axis is calculated. The time periods where the two time intervals overlap are called concurrent event segments, and the time periods that occur only in one of the time intervals and do not overlap are called independent event segments. Within the time range of concurrent event segments, the maximum value of vibration intensity in the cargo vibration sensing subsequence is taken as the vibration peak intensity feature, and the maximum value of tilt angle in the cargo tilt sensing subsequence is taken as the tilt peak angle feature.

[0090] Step S440: The vibration peak intensity characteristics and tilt peak angle characteristics of the concurrent event segments are compared with the preset vibration intensity risk level mapping table and tilt angle risk level mapping table, respectively, to obtain the vibration risk level and tilt risk level.

[0091] The vibration intensity risk level mapping table divides the range of vibration intensity values ​​into five consecutive sub-intervals, each corresponding to a vibration risk level of 1 to 5. The tilt angle risk level mapping table divides the range of tilt angle values ​​into five consecutive sub-intervals, each corresponding to a tilt risk level of 1 to 5. The vibration risk level corresponding to the sub-interval in which the peak vibration intensity characteristic falls is used as the vibration risk level, and the tilt risk level corresponding to the sub-interval in which the peak tilt angle characteristic falls is used as the tilt risk level.

[0092] Step S450: Based on the preset composite risk rules, the vibration risk level and tilt risk level are combined to generate a vibration and tilt composite risk level, which serves as a vibration and tilt composite risk factor representing the degree of threat to the stability of the cargo under the combined effect of vibration and tilt.

[0093] The composite risk rule uses a two-dimensional risk level matrix lookup table. The row index of the two-dimensional risk level matrix represents the vibration risk level, the column index represents the tilt risk level, and the matrix elements represent the combined vibration and tilt risk levels, with values ​​ranging from 1 to 5. The numerical value of the combined vibration and tilt risk factor is equal to the combined vibration and tilt risk level value obtained by looking up the two-dimensional risk level matrix.

[0094] Step S460: The vibration tilt composite risk factor, single vibration risk factor, and single tilt risk factor are added as new cargo status risk sources. Internal risk change transmission correlation pairs starting from cargo status are added to the change transmission mapping, and the change transmission mapping is supplemented and updated.

[0095] The vibration and tilt composite risk factor, the single vibration risk factor, and the single tilt risk factor are each considered as new cargo status risk sources. For each new cargo status risk source, the starting time of the concurrent event segment is taken as the change occurrence time, the spatial location of the cargo placement unit node that generated the concurrent event segment is taken as the occurrence location, and the risk factor value is taken as the change intensity index, forming a ternary attribute for the new internal risk change transmission correlation pair. The cargo status perception sequence segment following the concurrent event segment is taken as the corresponding hysteresis offset sequence. The new internal risk change transmission correlation pair is added to the change transmission mapping.

[0096] Step S470: Regenerate the multi-level change propagation path set using the supplemented and updated change propagation mapping, and adjust the cumulative amount of propagation changes of cargo placement unit nodes on the corresponding propagation path based on the vibration tilt composite risk factor, single vibration risk factor and single tilt risk factor. Re-execute risk positioning processing based on the adjusted multi-level change propagation path set to generate composite transportation risk warning information containing internal risk sources.

[0097] The processing steps S131 to S137 are re-executed to generate a new set of multi-level variation propagation paths. When calculating the cumulative amount of transmitted variation, the vibration tilt composite risk factor, the single vibration risk factor, and the single tilt risk factor are weighted according to preset weighting coefficients and then superimposed onto the original cumulative amount of transmitted variation to obtain an adjusted cumulative amount of transmitted variation. Based on the adjusted set of multi-level variation propagation paths, the risk location processing steps S141 to S146 are re-executed to generate composite transportation risk early warning information.

[0098] Step S510: Obtain the spatial topology of the predetermined transportation path of the transportation task to be monitored. The spatial topology includes the path node sequence and the path segment connection relationship.

[0099] The planned transportation routes are obtained from the transportation task planning system. The spatial topology is stored in the form of a directed graph. The nodes of the directed graph represent key locations on the route, such as loading points, transfer points, and unloading points. The edges of the directed graph represent path segments connecting two key locations. Each path node contains a node identifier and geographic coordinates, and each path segment contains a segment identifier, a start node identifier, an end node identifier, and a segment length.

[0100] Step S520: Map each change propagation path in the multi-level change propagation path set onto the spatial topology, and determine the set of path segments and the set of path nodes traversed by each change propagation path.

[0101] For each change propagation path, based on the actual geographical coordinates of the transport vehicle corresponding to each cargo placement unit node on the change propagation path at the time of the change, these actual geographical coordinates are overlaid and matched with the geographical coverage of each path segment in the spatial topology to determine which path segments the change propagation path traverses. The actual geographical coordinates of the cargo placement unit node at the starting point of the change propagation path are then matched with the nearest path node in the spatial topology to determine the set of path nodes along the change propagation path.

[0102] Step S530: Count the frequency of each path segment being traversed by the change propagation path in the multi-level change propagation path set, and calculate the statistical characteristic value of the maximum cumulative amount of the maximum propagation change of the path segment based on the maximum cumulative amount of the maximum propagation change corresponding to all the change propagation paths that traverse the path segment.

[0103] For the j-th path segment in the path segment set, the total number of change propagation paths traversing this path segment in the multi-level change propagation path set is counted as the traversal frequency. The maximum cumulative amount of transmitted change corresponding to each change propagation path traversing this path segment is extracted, and the arithmetic mean of these maximum cumulative amounts of transmitted change is calculated as a statistical feature value.

[0104] Step S540: Normalize the statistical characteristic values ​​of the crossing frequency and the maximum cumulative amount of transmission change for each path segment to obtain normalized crossing frequency and normalized cumulative amount characteristic values; and calculate the segment risk exposure index for each path segment by weighted fusion based on preset weights.

[0105] The crossing frequencies of all path segments are normalized from minimum to maximum. The normalized crossing frequency equals the original crossing frequency minus the minimum crossing frequency among all path segments, divided by the maximum crossing frequency minus the minimum crossing frequency. The statistical characteristic values ​​of all path segments are normalized in the same way. The segment risk exposure index equals the normalized crossing frequency multiplied by the first weight, plus the normalized cumulative characteristic value multiplied by the second weight. The sum of the first and second weights equals 1.

[0106] Step S550: Count the frequency of change initiation for each path node as the starting point of change in the multi-level change propagation path set, and calculate the statistical characteristic value of the initial change intensity index of the path node based on the initial change intensity index of all change propagation paths starting from the path node.

[0107] For the i-th path node, the total number of change propagation paths corresponding to the path nodes of the cargo placement unit node that is the starting point of the change propagation path is counted as the change initiation frequency. The initial change intensity index corresponding to each of the change propagation paths starting from the i-th path node is extracted, and the arithmetic mean of these initial change intensity indices is calculated as the statistical feature value.

[0108] Step S560: Normalize the statistical characteristic values ​​of the change initiation frequency and initial change intensity index of each path node to obtain normalized initiation frequency and normalized intensity characteristic value; and calculate the node risk source index of each path node by weighted fusion based on preset weights.

[0109] The normalization process is the same as in step S540. The node risk source index is equal to the normalization starting frequency multiplied by the third weight plus the normalization intensity characteristic value multiplied by the fourth weight, and the sum of the third weight and the fourth weight is equal to 1.

[0110] Step S570: Mark the segment risk exposure index of each path segment onto the path segment connection relationship corresponding to the spatial topology, and mark the node risk source index of each path node onto the path node sequence corresponding to the spatial topology to form a transportation path risk distribution topology map, and attach it to the transportation risk warning information to generate transportation risk warning information containing path topology risk distribution.

[0111] In a directed graph data structure with spatial topology, a risk exposure index attribute field is added to each edge, and the segment risk exposure index value of that path segment is written in it. Similarly, a risk source index attribute field is added to each node, and the node risk source index value of that path node is written in it. The directed graph labeled with risk indices is serialized into a transportation path risk distribution topology graph data segment, which is then appended to the transportation risk early warning information.

[0112] Step S610: Obtain historical transportation task records related to the transportation task to be monitored, and extract historical cargo status perception sequences and historical transportation environment perception sequences from the historical transportation task records.

[0113] Retrieve records of completed historical transportation tasks from the historical task database of the transportation management system that have the same transportation route, the same vehicle type, and the same cargo type as the current transportation task to be monitored, and extract the historical cargo status perception sequence and the historical transportation environment perception sequence stored in these historical transportation task records.

[0114] Step S620: Perform heterogeneous disturbance decoupling processing on the historical transportation environment perception sequence and the historical cargo status perception sequence to obtain the historical change propagation mapping. Call the logistics risk recursive discovery network on the historical change propagation mapping to obtain the set of historical multi-level change propagation paths.

[0115] For each extracted set of historical cargo status perception sequences and historical transportation environment perception sequences, heterogeneous disturbance decoupling processing is performed according to the processing flow of steps S121 to S127 to obtain the historical change propagation mapping corresponding to each set of historical data. For each set of historical change propagation mapping, the logistics risk recursive discovery network is invoked according to the processing flow of steps S131 to S137 to obtain the set of historical multi-level change propagation paths.

[0116] Step S630: Compare and label the starting cargo placement unit node of each historical change propagation path in the historical multi-level change propagation path set with the accident origin node in the actual transportation accident record to establish a mapping relationship between the propagation path and the accident result.

[0117] Retrieve the associated actual transportation accident records from historical transportation task records. These records contain the identifiers of the cargo placement unit nodes that confirmed the origin of the accident through accident investigation. Compare the node identifier of the starting cargo placement unit node for each historical change propagation path with the accident origin node identifier in the actual transportation accident records. If they match, the historical change propagation path is marked as a positive sample mapping; otherwise, it is marked as a negative sample mapping. All marking results constitute the mapping relationship from propagation path to accident outcome.

[0118] Step S640: Based on the mapping relationship, extract the transmission characteristics of typical change propagation paths that lead to actual transportation accidents from the historical multi-level change propagation path set. The transmission characteristics include the change curve shape of the cumulative amount of transmission changes at each level of cargo placement unit node and the critical amount of the final cumulative amount.

[0119] From the historical change propagation paths corresponding to the positive sample mapping, the change sequence of the cumulative amount of transmitted change along the propagation direction of each cargo placement unit node is extracted as the cumulative amount of transmitted change change curve. A point-by-point arithmetic average is performed on multiple positive sample change curves to obtain the typical shape of the cumulative amount of transmitted change curve. The cumulative amount of transmitted change of the last cargo placement unit node on the change propagation path leading to the accident in the positive sample is taken as the critical value of the final cumulative amount.

[0120] Step S650: The transmission characteristics of typical change propagation paths are used as risk propagation reference templates. Template matching is performed between these templates and the transmission characteristics of each change propagation path in the currently generated multi-level change propagation path set. The matching degree between each current change propagation path and the risk propagation reference template is calculated.

[0121] Template matching employs a dynamic time warping algorithm, which dynamically warps and aligns the cumulative change curve of the current propagation path with the typical cumulative change curve of the risk propagation reference template, calculating the warped path distance. The matching degree is equal to 1 divided by 1 plus the warped path distance, with the matching degree ranging from 0 to 1; a larger value indicates a more similarity between the two curves.

[0122] Step S660: Mark the current change propagation path with a matching degree exceeding the matching degree threshold as a risk change propagation path, and prioritize the inclusion of the risk originating cargo placement unit node and the risk affected cargo placement unit area into the transportation risk early warning information to generate transportation risk early warning information enhanced by historical experience.

[0123] The preset matching threshold is a constant value between 0 and 1. Current change propagation paths with a matching degree greater than the matching threshold are marked as risk change propagation paths. A historical experience enhancement marker field is added to the transportation risk warning information. The risk originating cargo placement unit node and the risk-affected cargo placement unit area corresponding to the risk change propagation path are written into this field and set to the highest priority.

[0124] For example, the method may further include: step S710, distributing transportation risk warning information to the vehicle control unit and the remote logistics scheduling unit, wherein the transportation risk warning information includes the node identifier of the cargo placement unit where the risk originates, the boundary coordinate sequence of the cargo placement unit area affected by the risk, and the maximum cumulative amount of transmission change.

[0125] The transportation risk warning information is distributed to the vehicle control unit via the vehicle-mounted communication terminal through the cellular mobile communication network or satellite communication link, and simultaneously distributed to the remote logistics dispatch unit through the logistics dispatch data dedicated line.

[0126] In step S720, after receiving the transportation risk warning information, the vehicle control unit extracts the physical location of the cargo placement unit corresponding to the risk initiation cargo placement unit node identifier on the vehicle, generates a local stabilization adjustment instruction for the cargo placement unit, and extracts the boundary coordinate sequence of the risk-affected cargo placement unit area to generate an overall tightening and reinforcement instruction for all cargo placement units in the affected area.

[0127] The vehicle control unit maintains a table mapping each cargo placement unit node identifier to its physical location within the cargo compartment. Based on the initial risk cargo placement unit node identifier, its physical location is obtained from the table. A local stabilization adjustment command, containing the target physical location and reinforcement force parameters, is generated and sent to the automatic fastening device for execution. All cargo placement units within the boundary coordinate sequence of the risk-affected cargo placement unit area are traversed, generating a batch overall fastening reinforcement command containing multiple target physical locations.

[0128] In step S730, after receiving the transportation risk warning information, the remote logistics scheduling unit evaluates the future change risk prediction value of each alternative path among the alternative paths between the current location of the vehicle and the destination based on the node identifier of the risk originating cargo placement unit and the maximum cumulative amount of transmission change, and selects the alternative path with the smallest future change risk prediction value as the corrected transportation path.

[0129] The remote logistics scheduling unit generates multiple alternative routes in the route planning module based on the current location and destination of the transport vehicle. For each route segment on each alternative route, the corresponding segment risk exposure index is read from the transport route risk distribution topology map generated in step S570. The segment risk exposure indices of all route segments on the alternative routes are weighted and accumulated to obtain the predicted future change risk value of the alternative route. The alternative route with the smallest predicted future change risk value is selected as the corrected transport route.

[0130] In step S740, the remote logistics scheduling unit generates an advance unloading priority ranking of the affected cargo placement units and an emergency reception preparation instruction for the destination based on the boundary coordinate sequence of the risk-affected cargo placement unit area. The unit then sends the corrected transportation route and the emergency reception preparation instruction for the destination to the vehicle control unit and the destination logistics receiving terminal, thus jointly forming a closed-loop response information for transportation risks.

[0131] For cargo placement units within the boundary coordinate sequence of the risk-affected cargo placement unit area, an early unloading priority ranking is generated by sorting them in descending order of cumulative transmission changes. Based on the early unloading priority ranking, a destination emergency reception preparation instruction containing cargo number and priority is generated, and the corrected transportation route and destination emergency reception preparation instruction are sent to the vehicle control unit and the destination logistics receiving terminal.

[0132] Step S810: Acquire the time-series path environment image stream collected by the external image acquisition device of the vehicle along the travel path of the transport task to be monitored, perform dynamic obstacle trajectory analysis processing on the time-series path environment image stream, and extract the movement trajectory point sequence and obstacle shape contour change sequence of dynamic obstacles entering the preset warning range of the travel path.

[0133] The external image acquisition device for the vehicle is a binocular stereo camera mounted at the front of the vehicle. The binocular stereo camera acquires images of the road ahead at a fixed frame rate. Each frame of image is processed using a deep learning-based obstacle detection and tracking algorithm. The detection algorithm outputs the bounding box coordinates and obstacle category of the obstacle, while the tracking algorithm assigns a unique obstacle identifier to each detected obstacle and associates its bounding box position across frames. For obstacles entering a preset radial distance warning range centered on the vehicle, the sequence of its bounding box center point positions in consecutive frames is recorded as a movement trajectory point sequence, and the sequence of changes in its bounding box width and height is recorded as an obstacle shape contour change sequence.

[0134] Step S820: The sequence of moving trajectory points and the sequence of transportation environment perception are spatiotemporally fused to establish a spatiotemporally synchronous correlation between the moving state of dynamic obstacles and changes in the transportation environment. The moving state includes instantaneous moving speed and moving direction angle.

[0135] The timestamp of each trajectory point in the movement trajectory point sequence is registered with the transportation environment sampling points of the same timestamp in the transportation environment perception sequence. The instantaneous movement rate is obtained by calculating the spatial displacement between adjacent trajectory points and the time interval, and the movement direction angle is obtained by calculating the angle between the line connecting the trajectory points and the direction of travel of the vehicle. The instantaneous movement rate and movement direction angle are associated and stored with the road surface elevation difference scalar at the same timestamp to establish a spatiotemporal synchronization relationship.

[0136] Step S830: Calculate the relative approach rate change rate between the dynamic obstacle and the vehicle based on the sequence of moving trajectory points, and identify the obstacle type classification of the dynamic obstacle based on the obstacle outline change sequence. The obstacle type classification includes fixed-shape obstacle category and variable-shape obstacle category.

[0137] The relative approach rate is equal to the change in relative distance between two adjacent trajectory points divided by the time interval. The relative approach rate change rate is equal to the difference between the current relative approach rate and the previous relative approach rate divided by the time interval. The changes in the bounding box width and height in the obstacle's shape contour change sequence are statistically analyzed. If the changes in both the bounding box width and height throughout the entire tracking period are less than a preset shape change threshold, the obstacle is identified as a fixed-shape obstacle; otherwise, it is identified as a variable-shape obstacle.

[0138] Step S840: Input the relatively close rate of change and obstacle type into the constructed obstacle threat evolution prediction network to generate a threat evolution sequence of dynamic obstacles to the vehicle's driving stability. The threat evolution sequence contains multiple step threat prediction values ​​within a preset future time period.

[0139] The obstacle threat evolution prediction network employs a sequence prediction model based on a Long Short-Term Memory (LSTM) network. This LSM network consists of two LSM unit layers, each containing 64 LSM units, followed by a fully connected output layer. The input is a concatenated sequence of the relative approach rate change sequence over a past period and a one-hot encoded vector indicating the obstacle type. The output is the predicted threat level for each time step within a preset future period, with the predicted threat level being a continuous value between 0 and 1. This obstacle threat evolution prediction network is trained using historical obstacle approach event data.

[0140] Step S850: Extract the first predicted moment from the threat evolution sequence when the predicted threat level exceeds the threat level warning limit as the threat trigger moment, and take the spatial location of the vehicle corresponding to the threat trigger moment as the potential location of the road impact.

[0141] The threat level warning limit is a preset constant value. The predicted threat level is compared with the warning limit step-by-step along the threat evolution sequence from near to far. The moment when the predicted threat level first exceeds the warning limit is marked as the threat trigger moment. The spatial location of the vehicle along its current path at the threat trigger moment is calculated by multiplying the vehicle's current speed by the predicted time step corresponding to the threat trigger moment. This spatial location is then used as the potential location of the road impact.

[0142] Step S860: The potential road impact location and the corresponding threat level prediction value are injected into the change transmission mapping as a new sudden change point, and the cargo placement unit node corresponding to the sudden change point is determined according to the spatial correspondence between the potential road impact location and the cargo placement unit node.

[0143] The time of threat triggering is used as the occurrence time of the new sudden change point, the predicted threat level is used as the change intensity index, and the location of potential road impact is used as the occurrence location, forming a three-element attribute for the new sudden change point. A new change transmission correlation pair is added to the change transmission mapping. The Euclidean distance between the potential road impact location and the spatial locations of each cargo placement unit node within the cargo compartment is calculated, and the cargo placement unit node with the smallest distance is taken as the cargo placement unit node corresponding to the sudden change point.

[0144] Step S870: Re-execute the change propagation path discovery process of the logistics risk recursive discovery network on the change propagation mapping after injecting the newly added sudden change points, and obtain a set of multi-level change propagation paths for the supplementary obstacle impact risk.

[0145] Using the change propagation mapping after injecting new sudden change points as input, the processing flow from steps S131 to S137 is re-executed to obtain a set of multi-level change propagation paths for supplementing obstacle impact risk.

[0146] Step S880: Based on the set of multi-level change propagation paths for supplementary obstacle impact risk, supplementary risk localization processing is performed on the transportation task to be monitored to generate transportation risk warning information that includes the risk induced by obstacle impact.

[0147] Using the set of multi-level change propagation paths for the risk of obstacle impact as input, the risk location processing flow from step S141 to step S146 is re-executed, and the newly added risk starting cargo placement unit node is marked as the obstacle impact induced risk node, generating transportation risk warning information containing the obstacle impact induced risk identifier.

[0148] Step S910: Obtain the attitude perception sequence of the attachment unit towed by the vehicle in the transportation task to be monitored. The attitude perception sequence is collected by the angular displacement sensor installed at the joint of the attachment unit and the sway sensor installed at the main body of the attachment unit.

[0149] The angular displacement sensor is a wire-type displacement sensor installed at the joint connecting the coupling unit and the vehicle. This wire-type displacement sensor measures the opening and closing of the gap at the joint during transportation and outputs a gap displacement signal. The sway sensor is a triaxial accelerometer installed at the center of gravity of the coupling unit. This triaxial accelerometer measures the sway acceleration of the coupling unit in the lateral and longitudinal directions, and converts it into a sway displacement signal after double integration.

[0150] Step S920: Perform joint loosening feature analysis on the attitude perception sequence to separate the gap variation signal representing the change in the gap between the connecting joints of the hook unit and the abnormal swaying signal representing the irregular swing of the main body of the hook unit. The gap variation signal includes the gap opening displacement and the gap closing impact.

[0151] The gap displacement signal undergoes high-frequency filtering. A high-pass filter with a preset cutoff frequency is used to filter out the slow gap changes of the connecting joint during normal operation, retaining the transient signal components representing sudden gap opening and closing, thus obtaining the gap variation signal. The swaying displacement signal undergoes regular swaying removal processing. Correlation analysis is performed between the swaying displacement signal and the vehicle's speed. Regular swaying components with a linear correlation to speed are removed, leaving irregular swaying components as abnormal swaying signals.

[0152] Step S930: Construct a dynamic change cycle diagram of the gap of the connecting joint using the gap variation signal. The dynamic change cycle diagram of the gap records the maximum opening amplitude and the maximum impact force of the gap closing of the connecting joint of the hook unit within each transportation environment variation cycle.

[0153] Using the timestamp of each sudden change point in the transportation environment sensing sequence as the period boundary, the gap variation signal is divided into multiple period segments. For each period segment, the maximum value of the gap opening displacement within that period segment is taken as the maximum gap opening amplitude, and the peak value of the displacement change rate when the gap closes within that period segment is multiplied by the mass of the hook unit as the maximum impact force when the gap closes. A dynamic change cycle diagram of the gap is plotted using the maximum gap opening amplitude of each period as the horizontal axis coordinate value and the maximum impact force when the gap closes as the vertical axis coordinate value.

[0154] Step S940: Construct a body sway phase trajectory diagram of the main body of the hanging unit using abnormal sway signals. The body sway phase trajectory diagram records the two-dimensional phase space evolution path of the sway displacement and sway rate of the main body of the hanging unit in the lateral and longitudinal directions.

[0155] Using the yaw displacement in the abnormal sway signal as the first dimension coordinate value, the yaw rate as the second dimension coordinate value, the pitch displacement as the third dimension coordinate value, and the pitch rate as the fourth dimension coordinate value, phase trajectory diagrams in the yaw and pitch directions are plotted respectively. The yaw phase trajectory diagram uses yaw displacement as the horizontal axis and yaw rate as the vertical axis, while the pitch phase trajectory diagram uses pitch displacement as the horizontal axis and pitch rate as the vertical axis.

[0156] Step S950: Extract loosening event segments from the gap dynamic change cycle diagram where the maximum gap opening exceeds the loosening threshold, and extract instability event segments from the main body swaying phase trajectory diagram where the phase space evolution path undergoes unsteady divergence.

[0157] The loosening threshold is the upper limit of the allowable gap determined based on the design gap and safety factor of the connecting joint. Continuous periodic segments in the gap dynamic change cycle diagram where the maximum gap opening exceeds the loosening threshold are marked as loosening event segments. Unsteady divergence is defined as the phase trajectory in the main body swaying phase trajectory diagram spiraling outward from the central stable region with a continuously increasing divergence radius. Continuous time periods satisfying the characteristics of unsteady divergence are marked as instability event segments.

[0158] Step S960: Overlap and compare the occurrence time intervals of the loosening event segment and the instability event segment to identify the joint dangerous period when the loosening event segment and the instability event segment overlap on the time axis, and use the peak value of the gap opening displacement and the peak value of the swaying rate within the joint dangerous period as the risk feature pair of the hanging unit.

[0159] The intersection of the time intervals of the loosening event segment and the instability event segment on the time axis is calculated. The time interval with a non-empty intersection is the joint danger period. Within the joint danger period, the maximum value of the gap opening displacement is taken as the peak gap opening displacement, and the maximum value of the yaw rate and the pitch rate is taken as the peak sway rate. The peak gap opening displacement and the peak sway rate constitute the risk characteristic pair of the coupling unit.

[0160] In step S970, the risk characteristics of the hook unit are converted into the equivalent road impact disturbance intensity, and the equivalent road impact disturbance intensity is added to the change transmission mapping as a new sudden change point. The equivalent road impact disturbance intensity is positively correlated with the peak value of the gap opening displacement and the peak value of the swaying rate.

[0161] The formula for calculating the equivalent road impact disturbance intensity is the peak value of the gap opening displacement multiplied by the first conversion factor plus the peak value of the sway rate multiplied by the second conversion factor. The time of occurrence of the newly added sudden change point is determined by the center moment of the joint hazardous period, the equivalent road impact disturbance intensity is used as the change intensity index, and the installation position of the connecting joint on the vehicle is used as the occurrence position. These three attributes constitute the newly added sudden change point and are added to the change transmission mapping.

[0162] Step S980 involves calling the logistics risk recursive discovery network on the supplemented change propagation mapping to perform change propagation path discovery processing, resulting in a multi-level change propagation path set considering the instability risk of the attached unit. Based on this set, risk localization processing is performed on the transportation task to be monitored, generating transportation risk early warning information that includes the risk induced by the instability of the attached unit.

[0163] The processing flow of steps S1010 to S1060 and steps S1110 to S1170 is similar to that of steps S910 to S980 above. They respectively perform corresponding feature extraction, risk modeling, change transmission mapping supplementation, change propagation path discovery and risk location processing on the transport formation collaborative perception data and the cargo stacking configuration sensing data, and finally generate formation collaborative transportation risk warning information and transportation risk warning information containing stacking instability risk classification identifier.

[0164] Step S1010: Obtain the formation collaborative perception data of the transportation formation consisting of multiple vehicles in the transportation task to be monitored. The formation collaborative perception data includes the transportation environment perception sequence and cargo status perception sequence of each vehicle, as well as the vehicle spacing perception sequence between adjacent vehicles.

[0165] The platoon's collaborative sensing data is shared and aggregated in real time through the inter-vehicle communication links established between the various vehicles within the platoon. Each vehicle aligns its collected transportation environment sensing sequence and cargo status sensing sequence with a unified time base, and then broadcasts them to other vehicles in the platoon via the inter-vehicle communication links. The vehicle-to-vehicle distance sensing sequence is collected by a millimeter-wave ranging radar installed at the front of each vehicle. The millimeter-wave ranging radar transmits a frequency-modulated continuous wave signal forward and receives the echo signal reflected from the vehicle ahead. The inter-vehicle distance value is obtained by calculating the frequency difference between the transmitted and received signals, and the sampling frequency is consistent with the transportation environment sensing sequence. Each sampling point in the vehicle-to-vehicle distance sensing sequence includes a sampling timestamp and the inter-vehicle distance value. The vehicles in the platoon are numbered sequentially from the front to the back, and the platoon's collaborative sensing data summarizes the sensing sequences of all vehicles.

[0166] Step S1020: Extract sudden change basis sequences from the transportation environment perception sequence of each vehicle to obtain independent change basis sequences corresponding to each vehicle. Align the independent change basis sequences of each vehicle according to the time axis to form a platoon change basis sequence set.

[0167] For the transportation environment perception sequence of the m-th vehicle in the convoy, the independent variable basis sequence corresponding to that vehicle is extracted according to the processing flow of steps S122 and S123. The extraction process includes stationary trend stripping and transient jump identification processing to obtain the occurrence time and intensity index of all sudden change points on that vehicle. The independent variable basis sequences of all vehicles are aligned and arranged according to a unified time axis to form a convoy variable basis sequence set. The convoy variable basis sequence set records the temporal information of the external environmental abrupt changes experienced by each vehicle in the convoy.

[0168] Step S1030: Perform change propagation direction identification processing on the set of change base sequences of the formation. Determine the propagation direction of the change in the formation based on the chronological order of the occurrence of sudden change points on each vehicle, and obtain the formation-level change propagation direction vector.

[0169] For sudden change points of the same type in the base sequence set of convoy changes, the occurrence time of each sudden change point of this type on each vehicle is extracted according to the arrangement order of the vehicles from the head to the tail of the convoy. If the occurrence time of the vehicles from the head to the tail increases sequentially, that is, the occurrence time of the vehicle at the head of the convoy is the earliest and the occurrence time of the vehicle at the tail of the convoy is the latest, then the change propagation direction is determined to be from the head to the tail of the convoy, and the convoy-level change propagation direction vector is positive. If the occurrence time of the vehicles from the tail to the head of the convoy increases sequentially, then the change propagation direction is determined to be from the tail to the head of the convoy, and the convoy-level change propagation direction vector is negative. If the occurrence times of each vehicle are similar and there is no obvious increasing trend, then the change is determined to be simultaneous, and the convoy-level change propagation direction vector is zero.

[0170] Step S1040: Perform vehicle spacing response characteristic analysis on the vehicle spacing sensing sequence, obtain the response contraction amount and response contraction rate of vehicle spacing at each sudden change point in the sudden change base sequence, and establish the vehicle spacing change response function.

[0171] For each sudden change point, the inter-vehicle distance value is extracted from the inter-vehicle spacing sensing sequence for a window length before and after the occurrence of the sudden change point. The response contraction amount is equal to the difference between the minimum inter-vehicle distance value after the occurrence of the sudden change point and the baseline inter-vehicle distance value before the occurrence of the sudden change point. The baseline value is the arithmetic mean of the inter-vehicle distance values ​​during a period of stable driving before the occurrence of the sudden change point. The response contraction rate is equal to the response contraction amount divided by the time it takes for the inter-vehicle distance value to decrease from the baseline value to the minimum value. The inter-vehicle spacing change response function is expressed as a piecewise function. When the input change intensity index is less than the trigger threshold, the output is zero. When the input change intensity index is greater than or equal to the trigger threshold, the output is the product of the response contraction amount and the response contraction rate, which is the inter-vehicle spacing change response function value.

[0172] Step S1050: Using the formation-level change propagation direction vector as the propagation guidance constraint and the vehicle spacing change response function as the propagation intensity modulation factor, a cross-vehicle formation-level change propagation correlation pair is established in the change propagation mapping. The formation-level change propagation correlation pair performs cross-vehicle correlation between the hysteresis offset sequence of the cargo placement unit node on the previous vehicle and the incoming change characteristics of the cargo placement unit node on the next vehicle.

[0173] The source and target vehicles for change propagation are determined based on the sign of the convoy-level change propagation direction vector. If the convoy-level change propagation direction vector is positive, the source vehicle is the vehicle positioned at the front of the convoy, and the target vehicle is the vehicle immediately following behind it. The hysteresis offset sequence of the cargo placement unit node on the source vehicle is used as the propagation source, and the incoming change characteristics of the cargo placement unit node at the corresponding spatial position on the target vehicle are used as the propagation target. The amplitude response ratio of the convoy-level change propagation correlation pair is equal to the amplitude response ratio of the hysteresis offset sequence of the source vehicle multiplied by the vehicle spacing change response function value. The time delay of the convoy-level change propagation correlation pair is equal to the time delay on the source vehicle plus the vehicle-to-vehicle distance value divided by the convoy's travel speed, resulting in the propagation time.

[0174] Step S1060: Add the formation-level change propagation association pairs to the change propagation mapping to form a formation-coordinated change propagation mapping. Use the formation-coordinated change propagation mapping to call the logistics risk recursive discovery network to perform change propagation path discovery processing, and obtain a set of formation-level multi-level change propagation paths that cross the vehicle boundary.

[0175] All formation-level change propagation pairs are added to the change propagation mapping, which originally only contained change propagation pairs within a single vehicle, to form a formation-coordinated change propagation mapping. This mapping includes both change propagation relationships within a vehicle and formation-level change propagation relationships across vehicles. Using the formation-coordinated change propagation mapping as input, the logistics risk recursive discovery network processing flow from steps S131 to S137 is re-executed. In step S134, the search scope for spatially adjacent cargo placement unit nodes at the next level is expanded from within a single vehicle to include all cargo placement unit nodes within the cargo compartment of the following vehicle. The resulting set of formation-level multi-level change propagation paths can span from cargo placement unit nodes on the preceding vehicle to cargo placement unit nodes on the following vehicle.

[0176] Step S1070: Based on the multi-level change propagation path set at the formation level, perform risk location processing on each vehicle in the monitored transportation task, identify the cross-vehicle transmission starting node where the change first propagates across vehicles within the formation and the set of vehicles affected by the cross-vehicle risk, and generate formation collaborative transportation risk warning information that includes the cross-vehicle transmission starting node identifier, the cross-vehicle risk affected vehicle identifier, and the formation level maximum cumulative amount of transmitted change.

[0177] For each change propagation path in the platoon-level multi-level change propagation path set, the node where the vehicle identifier of the cargo placement unit node first changes is detected and marked as the cross-vehicle transmission initiation node. The cross-vehicle transmission initiation node represents the first affected cargo placement unit where the change propagates from one vehicle to another. The vehicle identifiers of all cargo placement unit nodes along the change propagation path after the cross-vehicle transmission initiation node are collected after deduplication to form the cross-vehicle risk-affected vehicle set. The maximum value of the cumulative amount of transmitted changes across all cargo placement unit nodes in the platoon-level multi-level change propagation path set is taken as the platoon-level maximum cumulative amount of transmitted changes. The cross-vehicle transmission initiation node identifier, the cross-vehicle risk-affected vehicle identifier, and the platoon-level maximum cumulative amount of transmitted changes are combined to form platoon-coordinated transportation risk warning information.

[0178] Step S1110: Obtain cargo stacking configuration sensing data inside the cargo placement unit in the transportation task to be monitored. The cargo stacking configuration sensing data is collected by the distributed pressure sensing array and internal structured light sensing device inside the cargo placement unit.

[0179] The distributed pressure sensing array consists of multiple arrayed thin-film pressure sensors embedded in the base plate of the cargo placement unit. Each thin-film pressure sensor measures the pressure value per unit area above it, and the pressure values ​​of all the thin-film pressure sensors constitute a pressure distribution matrix. The pressure distribution matrix is ​​organized in a row and column format, with the row and column numbers corresponding to the physical positions of the thin-film pressure sensors on the base plate plane, and the values ​​of the matrix elements representing the pressure values. The internal structured light sensing device is a small structured light depth camera mounted on the top of the cargo placement unit. This small structured light depth camera projects an coded structured light pattern onto the cargo stack surface and captures reflected images. Three-dimensional point cloud data of the cargo stack surface is calculated using the triangulation principle. Each point in the three-dimensional point cloud data contains the x-coordinate, y-coordinate, and depth coordinates relative to the depth camera's coordinate system.

[0180] Step S1120: Perform stacking center of gravity offset identification processing on the cargo stacking configuration sensing data, calculate the real-time center of gravity projection coordinates of the entire cargo stack based on the pressure distribution matrix collected by the distributed pressure sensing array, and calculate the three-dimensional morphological change of the cargo stack surface based on the cargo surface point cloud data collected by the internal structured light sensing device.

[0181] The real-time center-of-gravity projection coordinates of the entire cargo stack are calculated using the moment balance equation of the pressure distribution matrix. The horizontal coordinate of the center-of-gravity projection is equal to the sum of the products of each row's pressure value and its row coordinate, divided by the sum of all pressure values. The vertical coordinate of the center-of-gravity projection is equal to the sum of the products of each column's pressure value and its column coordinate, divided by the sum of all pressure values. The three-dimensional shape changes on the cargo stack surface are obtained by registering and differencing two adjacent sets of three-dimensional point cloud data. First, the iterative nearest-point algorithm is used to accurately register two adjacent sets of three-dimensional point cloud data. After registration, the Euclidean distance between corresponding points in the two frames of point cloud is calculated as the shape change of that point. The shape changes of all points are then organized spatially into a three-dimensional shape change distribution map.

[0182] Step S1130: Generate a centroid offset trajectory line based on the real-time centroid projection coordinates at consecutive time intervals, and generate a stacked morphology evolution sequence based on the three-dimensional morphology changes at consecutive time intervals. The centroid offset trajectory line includes the centroid offset direction and the cumulative centroid offset distance, and the stacked morphology evolution sequence includes the growth rate of the local convex region and the expansion rate of the concave region on the stacked surface.

[0183] The real-time centroid projection coordinates of multiple consecutive sampling times are connected sequentially in chronological order to form a trajectory line in the planar coordinate system on the base plate. This trajectory line is the centroid offset trajectory line. The centroid offset direction is the angle of the line connecting two adjacent points on the trajectory line, and the cumulative centroid offset distance is the total arc length of the trajectory line from the starting point to the current point. The stacked morphology evolution sequence is generated by identifying continuous regions with morphology changes exceeding a preset threshold in the 3D morphology change distribution map. Continuous regions with positive morphology changes are marked as protruding regions, and continuous regions with negative morphology changes are marked as concave regions. The growth rate of a protruding region is equal to the change in the area of ​​the protruding region between two adjacent time points divided by the time interval, and the expansion rate of a concave region is equal to the change in the area of ​​the concave region between two adjacent time points divided by the time interval. The growth rates of the protruding regions and the expansion rates of the concave regions at each time point are arranged in chronological order to form the stacked morphology evolution sequence.

[0184] Step S1140: Perform joint analysis and processing of the center of gravity offset trajectory line and the cargo status perception sequence to establish a response causal chain between the center of gravity offset event and the sudden change point in the sudden change base sequence, and determine the corresponding sudden change point that causes the center of gravity offset to exceed the center of gravity offset threshold as the stacking instability trigger source.

[0185] The center of gravity offset is defined as the Euclidean distance between the projected coordinates of the center of gravity at the current moment and the projected coordinates of the center of gravity at the initial moment. The center of gravity offset threshold is the safe upper limit of the offset determined based on cargo stacking stability analysis. When the center of gravity offset first exceeds the center of gravity offset threshold, this moment is recorded as the moment of the center of gravity offset event. The earliest sudden change point within a preset retrospective time window before the moment of the center of gravity offset event is retrieved from the sudden change base sequence and identified as the stacking instability trigger source. The response causal chain includes the timestamp of the sudden change point and the change intensity index of the stacking instability trigger source, the moment of the center of gravity offset event, and the magnitude of the center of gravity offset exceeding the limit.

[0186] Step S1150: Perform time-series correlation processing on the stack morphology evolution sequence and the hysteresis offset sequence corresponding to the stack instability trigger source to obtain the synchronous growth relationship between the growth rate of the protruding region in the stack morphology evolution sequence and the amplitude response ratio in the hysteresis offset sequence.

[0187] Using the occurrence time of the stack instability trigger as the origin of the time axis, the amplitude response ratio of the hysteresis migration sequence and the growth rate of the protrusion region in the stack morphology evolution sequence are aligned to the same time axis. The Pearson correlation coefficient between the two is calculated at each time point. If the Pearson correlation coefficient exceeds a preset correlation coefficient threshold, a synchronous growth relationship is determined. Specifically, for every unit increase in the amplitude response ratio, the growth rate of the protrusion region increases proportionally; this proportional increase is the synchronous growth coefficient.

[0188] Step S1160: Construct a stack instability risk prediction function based on the synchronous growth relationship. The input of the stack instability risk prediction function is the amplitude response ratio of the hysteresis offset sequence, and the output is the predicted value of the growth rate of the convex region in the stack morphology evolution sequence in the future period.

[0189] The stack instability risk prediction function is a linear regression function. The slope of this linear regression function is equal to the synchronous growth coefficient obtained in step S1150, and the intercept is equal to the baseline value of the bulge region growth rate under undisturbed conditions. During prediction, the amplitude response ratio of the hysteresis offset sequence of a certain cargo placement unit node at the current moment is input into this linear regression function, and the output is the predicted value of the bulge region growth rate within the preset prediction period.

[0190] Step S1170: Use the stacking instability risk prediction function to assess the stacking instability risk of each cargo placement unit node on each change propagation path in the multi-level change propagation path set, and calculate the probability of stacking overturning of each cargo placement unit node in the future period.

[0191] For the j-th cargo placement unit node on each propagation path in the multi-level change propagation path set, the amplitude response ratio of its hysteresis offset sequence is read and substituted into the stack instability risk prediction function to obtain the predicted value of the bulge region growth rate. The predicted growth rate of the bulge region is multiplied by the prediction time length to obtain the predicted growth area of ​​the bulge region. The ratio of the predicted growth area of ​​the bulge region to the bottom area of ​​the cargo placement unit node is used as the stack overturning risk probability. The stack overturning risk probability ranges from 0 to 1, with a higher value indicating a higher probability of future stack overturning. A constant stack overturning risk threshold is preset, and cargo placement unit nodes whose calculated stack overturning risk probability exceeds the threshold are marked as stack instability risk nodes. Within the risk-affected cargo placement unit area, the spatial location corresponding to the stack instability risk node is highlighted, and a stack instability risk classification identifier field is added to the transportation risk warning information. This field records the node identifier and stack overturning risk probability value of each stack instability risk node, generating transportation risk warning information containing the stack instability risk classification identifier.

[0192] In one exemplary embodiment, a transportation risk early warning system for smart logistics is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 3 As shown, this transportation risk early warning system for smart logistics includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a transportation risk early warning method for smart logistics. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of a transportation risk warning system for smart logistics, or an external keyboard, touchpad, or mouse, etc.

[0193] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A transportation risk early warning method for smart logistics, characterized in that, The method includes: Acquire cargo status perception sequence and transportation environment perception sequence associated with the transportation task to be monitored. The cargo status perception sequence is collected by the sensors built into the cargo placement unit, and the transportation environment perception sequence is collected by the sensors external to the vehicle. The cargo status perception sequence and the transportation environment perception sequence are subjected to heterogeneous disturbance decoupling processing to separate the sudden change base sequence in the transportation environment perception sequence and the hysteresis offset sequence generated by the cargo status perception sequence following the sudden change base sequence, and a change transmission mapping between the sudden change base sequence and the hysteresis offset sequence is constructed. The existing logistics risk recursive discovery network is invoked, and the recursive propagation of change events along the cargo transportation path is detected and processed according to the change propagation mapping to obtain a multi-level change propagation path set. The multi-level change propagation path set includes the sequence of cargo placement unit nodes through each change propagation path and the cumulative amount of propagated change for each cargo placement unit node. Based on the multi-level change propagation path set, risk localization processing is performed on the transportation task to be monitored to determine the risk initiation cargo placement unit node and risk-affected cargo placement unit area in the transportation task to be monitored, and transportation risk early warning information for the transportation task to be monitored is generated.

2. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The process involves acquiring the cargo status perception sequence and the transportation environment perception sequence associated with the transportation task to be monitored, performing heterogeneous disturbance decoupling processing on the cargo status perception sequence and the transportation environment perception sequence, separating the sudden change base sequence in the transportation environment perception sequence and the hysteresis offset sequence generated by the cargo status perception sequence following the sudden change base sequence, and constructing a change transmission mapping between the sudden change base sequence and the hysteresis offset sequence, including: Extract time-series continuous cargo status sampling points from the cargo status perception sequence to form a cargo status sampling point sequence; extract time-series continuous transportation environment sampling points from the transportation environment perception sequence to form a transportation environment sampling point sequence. The transportation environment sampling point sequence is subjected to stationary trend stripping processing to remove the progressive trend component from the transportation environment sampling point sequence, resulting in a trend-stripped transportation environment sampling point sequence. The progressive trend component is determined based on the long-term moving average curve of the transportation environment sampling point sequence. The sampling point sequence of the transportation environment after trend stripping is subjected to transient jump identification processing. Sampling points whose amplitude changes more than the transient jump threshold within a unit time interval are marked as sudden change points. The sudden change points are combined in chronological order to form the sudden change base sequence. Long-term drift compensation processing is performed on the cargo status sampling point sequence to eliminate the slow offset components related to the cargo's own attributes in the cargo status sampling point sequence, resulting in a drift-compensated cargo status sampling point sequence. Each sudden change point in the sudden change base sequence is used as a reference time point. A cargo state sampling point segment corresponding to the reference time point is extracted from the cargo state sampling point sequence after drift compensation. The start time of the time delay search window is later than the reference time point. Offset extraction processing is performed on each cargo status sampling point segment within the time delay search window to obtain the time history curve of the deviation of the cargo status sampling point segment relative to the undisturbed state reference value. The time history curve of the deviation is used as the hysteresis offset sequence corresponding to the sudden change point. The time delay, amplitude response ratio, and recovery morphology parameters between each sudden change point and its corresponding hysteresis offset sequence are associated to form a change transmission correlation pair. All change transmission correlation pairs are summarized to obtain the change transmission mapping.

3. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The process involves acquiring the cargo status perception sequence and transportation environment perception sequence associated with the transportation task to be monitored, and then invoking the constructed logistics risk recursive discovery network. Based on the change propagation mapping, the recursive propagation of change events along the cargo transportation path is processed to obtain a multi-level change propagation path set, including: Each pair of change transmission associations in the change transmission mapping is taken as an initial propagation unit. The occurrence time, occurrence location, and change intensity index of the sudden change point are extracted from the initial propagation unit as the three-element attributes of the initial propagation source. The occurrence location corresponds to the installation location of the external sensing device of the vehicle that generated the sudden change point. Based on the location of the initial propagation source, determine the cargo placement unit node that is spatially adjacent to the location of the initial propagation source, and use the ternary attribute of the initial propagation source and the hysteresis offset sequence corresponding to the initial propagation source as the input variation characteristics of the adjacent cargo placement unit node. Based on the amplitude response ratio of the change intensity index and the hysteresis offset sequence in the incoming change characteristics, and combined with the historical response characteristic records of the adjacent cargo placement unit nodes, the induced change characteristics of the adjacent cargo placement unit nodes after being affected by the incoming change characteristics are calculated. The induced change characteristics include the induced change intensity index and the induced hysteresis offset sequence. The adjacent cargo placement unit node is used as the new starting point for change propagation, and the derived change feature is used as the new input change feature. The change propagation process is carried out along the cargo transportation path to the next level spatially adjacent cargo placement unit node to generate the derived change feature of the next level cargo placement unit node. Repeat the step of using the outgoing change feature of the current cargo placement unit node as the incoming change feature of the next level cargo placement unit node and then propagating it to subsequent cargo placement unit nodes step by step until the change propagation stopping condition is met, thus obtaining a change propagation path. The change propagation path includes all cargo placement unit nodes that are sequentially traversed from the initial propagation source. The differences between the incoming change characteristics and outgoing change characteristics of each cargo placement unit node in the change propagation path are compared to obtain the cumulative amount of transmitted change corresponding to the cargo placement unit node. The cumulative amount of transmitted change represents the net increase or decrease in the intensity of the change after passing through the cargo placement unit node. The cumulative amount of transmission changes of all the change propagation paths corresponding to all the initial propagation units and each cargo placement unit node on each change propagation path is collected to form the multi-level change propagation path set.

4. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The process involves acquiring the cargo status perception sequence and transportation environment perception sequence associated with the transportation task to be monitored, and performing risk localization processing on the transportation task to be monitored based on the multi-level change propagation path set. This process determines the risk initiation cargo placement unit node and the risk-affected cargo placement unit area in the transportation task to be monitored, and generates transportation risk early warning information for the transportation task to be monitored, including: Extract the ternary attribute of the initial propagation source of each multi-level change propagation path from the set of multi-level change propagation paths, compare the change intensity index in the ternary attribute with the preset change intensity classification table, and screen out the initial propagation sources whose change intensity index belongs to the risk level as risk starting candidate points. The cargo placement unit node corresponding to the risk initiation candidate point is matched with the cargo placement unit node layout diagram of the transportation task to be monitored, and the risk initiation candidate point located within the coverage area of ​​the transportation task to be monitored is determined as the risk initiation cargo placement unit node. For each of the risk-initiating cargo placement unit nodes, retrieve the target change propagation path starting from the risk-initiating cargo placement unit node in the multi-level change propagation path set, and extract the cumulative amount of propagated changes of all subsequent cargo placement unit nodes on the target change propagation path. The cumulative amount of transmitted changes of each subsequent cargo placement unit node on the target change propagation path is compared with the preset cumulative amount risk threshold. The process is performed node by node along the recursive direction of the target change propagation path. When a cargo placement unit node appears where the cumulative amount of transmitted changes is lower than the cumulative amount risk threshold for the first time, all cargo placement unit nodes before that cargo placement unit node are included in the risk impact node set. Geometric envelope generation is performed on the spatial regions corresponding to all cargo placement unit nodes in the risk-affected node set on the cargo placement unit node layout diagram to obtain the risk-affected cargo placement unit region defined by the envelope boundary coordinate sequence. The transportation risk early warning information is formed by combining the node identifier and spatial coordinates of the risk-initiating cargo placement unit node, the envelope boundary coordinate sequence of the risk-affected cargo placement unit area, and the maximum cumulative amount of transmitted change on the target change propagation path.

5. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The method further includes: Meteorological element perception subsequence and road condition perception subsequence are extracted from the transportation environment perception sequence. The meteorological element perception subsequence is collected by a meteorological sensor, and the road condition perception subsequence is collected by a road surface adhesion sensor. The meteorological element sensing subsequence is subjected to meteorological sudden change segment detection processing to obtain the time of occurrence of meteorological sudden change and the intensity level of meteorological sudden change; the road condition sensing subsequence is subjected to road condition jump detection processing to obtain the time of occurrence of road condition jump and the level of road condition deterioration. The timing of the meteorological change and the timing of the road surface condition jump are analyzed by time series correlation analysis to determine the causal relationship between the meteorological event and the road surface condition event, and to calculate the induction time advance and induction intensity transfer coefficient from the meteorological event to the road surface condition event. The causal relationship pair, the induced time advance, and the induced intensity transmission coefficient are added as new variable transmission pair to the variable transmission mapping, and the variable transmission mapping is updated. The change propagation path discovery process of the logistics risk recursive discovery network is re-executed using the updated change propagation mapping to obtain an updated set of multi-level change propagation paths. Based on the updated multi-level change propagation path set, the transportation task to be monitored is subjected to secondary risk localization processing to generate supplementary transportation risk warning information. The transportation risk warning information and the supplementary transportation risk warning information are merged to form comprehensive transportation risk warning information for the transportation task to be monitored. The supplementary transportation risk warning information includes the starting node of secondary risks caused by the combined changes in meteorology and road surface and the area affected by secondary risks.

6. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The method further includes: Obtain cargo attribute registration information corresponding to the transportation task to be monitored. The cargo attribute registration information includes cargo vulnerability level identifier and cargo tolerance environment boundary parameters. Extract the change intensity index of each sudden change point from the change transmission mapping, and extract the cumulative amount of transmission change of each level of cargo placement unit node on each change propagation path from the multi-level change propagation path set. Using the cargo tolerance environment boundary parameters as tolerance benchmarks, the change intensity index of each sudden change point is mapped to the standardized risk value of the corresponding change type, and the boundary threshold of the same type in the cargo tolerance environment boundary parameters is also mapped to the standardized tolerance threshold. By comparing the standardized risk value with the corresponding standardized tolerance threshold, it is determined whether the sudden change point triggers a tolerance warning. The cumulative amount of transmission changes at each level of cargo placement unit node on each change propagation path is mapped to a standardized cumulative risk value. By comparing the standardized cumulative risk value with the corresponding standardized tolerance threshold, the first breakthrough node and the amount of exceeding the standardized tolerance threshold are determined. Based on the cargo vulnerability level identifier, the sensitivity response coefficient of the cargo to changing factors is determined, and combined with the overshoot amount of the first breach node, the estimated damage level classification of the cargo in the affected cargo placement unit is calculated. Based on the estimated damage level classification, different damage levels of cargo placement unit sub-regions are marked within the risk-affected cargo placement unit area to form a risk-affected cargo placement unit area distribution map with damage level classification labels, and this map is added to the transportation risk warning information to generate differentiated transportation risk warning information adapted to cargo attributes.

7. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The method further includes: The cargo vibration sensing subsequence and the cargo tilt sensing subsequence are extracted from the cargo status sensing sequence. The cargo vibration sensing subsequence is collected by a vibration sensing device, and the cargo tilt sensing subsequence is collected by a tilt angle sensing device. The vibration event segmentation process is performed on the cargo vibration sensing subsequence to obtain the vibration event start time, vibration event end time and vibration event duration. The tilt event segmentation process is performed on the cargo tilt sensing subsequence to obtain the tilt event start time, tilt event end time and tilt event duration. The start time and end time of the vibration event are subjected to time axis overlap analysis with the start time and end time of the tilting event to determine the concurrent event segments and independent event segments of vibration and tilting, and the vibration peak intensity feature and tilt peak angle feature are extracted for each concurrent event segment. The vibration peak intensity feature and the tilt peak angle feature of the concurrent event segment are compared with the preset vibration intensity risk level mapping table and tilt angle risk level mapping table, respectively, to obtain the vibration risk level and tilt risk level. Based on the preset composite risk rules, the vibration risk level and the tilt risk level are combined to generate a vibration-tilt composite risk level, which serves as a vibration-tilt composite risk factor representing the degree of threat to cargo stability under the combined effect of vibration and tilt. The vibration tilt composite risk factor, single vibration risk factor, and single tilt risk factor are added as new cargo status risk sources. An internal risk change transmission association pair starting from cargo status is added to the change transmission mapping to supplement and update the change transmission mapping. The multi-level change propagation path set is regenerated using the supplemented and updated change propagation mapping. Based on the vibration tilt composite risk factor, the single vibration risk factor, and the single tilt risk factor, the cumulative amount of the propagation change of the cargo placement unit node on the corresponding propagation path is adjusted. Based on the adjusted multi-level change propagation path set, risk positioning processing is re-executed to generate composite transportation risk warning information containing internal risk sources.

8. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The method further includes: Obtain the spatial topology of the predetermined transportation path of the transportation task to be monitored, wherein the spatial topology includes the path node sequence and the path segment connection relationship; Map each change propagation path in the multi-level change propagation path set onto the spatial topology, and determine the set of path segments and the set of path nodes traversed by each change propagation path; The frequency of each path segment being traversed by the change propagation path in the multi-level change propagation path set is counted, and the statistical characteristic value of the maximum cumulative amount of conducted change for the path segment is calculated based on the maximum cumulative amount of conducted change corresponding to all change propagation paths traversing the path segment. Based on the statistical characteristic values ​​of the crossing frequency and the maximum cumulative amount of transmission variation for each path segment, normalization processing is performed to obtain normalized crossing frequency and normalized cumulative amount characteristic values; based on preset weights, the normalized crossing frequency and the normalized cumulative amount characteristic values ​​are weighted and fused to calculate the segment risk exposure index for each path segment. The frequency of change initiation for each path node as the starting point of change in the multi-level change propagation path set is counted, and the statistical characteristic value of the initial change intensity index of the path node is calculated based on the initial change intensity index of all change propagation paths starting from the path node. Based on the statistical characteristic values ​​of the change initiation frequency and the initial change intensity index of each path node, normalization processing is performed to obtain the normalized initiation frequency and normalized intensity characteristic value; based on the preset weight, the normalized initiation frequency and the normalized intensity characteristic value are weighted and fused to calculate the node risk source index of each path node. The risk exposure index of each path segment is marked on the path segment connection relationship corresponding to the spatial topology, and the node risk source index of each path node is marked on the path node sequence corresponding to the spatial topology to form a transportation path risk distribution topology map, which is then attached to the transportation risk warning information to generate transportation risk warning information containing path topology risk distribution.

9. The transportation risk early warning method for smart logistics according to claim 1, characterized in that, The method further includes: Obtain historical transportation task records related to the transportation task to be monitored, and extract historical cargo status perception sequences and historical transportation environment perception sequences from the historical transportation task records; The heterogeneous disturbance decoupling process is performed on the historical transportation environment perception sequence and the historical cargo status perception sequence to obtain the historical change propagation mapping. The logistics risk recursive discovery network is called on the historical change propagation mapping to obtain the set of historical multi-level change propagation paths. The starting cargo placement unit node of each historical change propagation path in the set of historical multi-level change propagation paths is compared and labeled with the accident origin node in the actual transportation accident record to establish a mapping relationship from propagation path to accident result. Based on the mapping relationship, the transmission characteristics of typical change propagation paths that lead to actual transportation accidents are extracted from the set of historical multi-level change propagation paths. The transmission characteristics include the change curve shape of the cumulative amount of transmission changes at each level of cargo placement unit node and the critical amount of the final cumulative amount. The transmission characteristics of the typical change propagation path are used as a risk propagation reference template. Template matching is performed on the transmission characteristics of each change propagation path in the currently generated multi-level change propagation path set to calculate the matching degree between each current change propagation path and the risk propagation reference template. The current change propagation path with a matching degree exceeding the matching degree threshold is marked as a risk change propagation path, and the risk originating cargo placement unit node and the risk affected cargo placement unit area corresponding to the risk change propagation path are preferentially incorporated into the transportation risk warning information to generate transportation risk warning information enhanced by historical experience.

10. A transportation risk early warning system for smart logistics, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the transportation risk warning method for smart logistics as described in any one of claims 1 to 9 by executing the machine-executable instructions.

Images