Cold-chain logistics information evidence method and system based on multi-dimensional perception and blockchain

Abstract

The present application relates to the technical field of Internet of Things data processing and information security, and discloses a cold-chain logistics information evidence storage method and system based on multi-dimensional sensing and blockchains.The method comprises the following steps: collecting continuous sensing data streams of a mobile monitoring unit and mapping them to a multi-dimensional state space, constructing a covariance matrix based on steady-state data to generate an adaptive hyper-ellipsoid manifold; constructing a collaborative response cone model, generating an event descriptor to construct a data anchor point, and synchronizing to a distributed ledger.The present application effectively solves the semantic analysis and trust anchoring problems of cold-chain full-link data, and realizes the integrity verification and tamper-proof evidence storage of information.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) data processing and information security technology, and more specifically, to a method and system for storing cold chain logistics information based on multi-dimensional perception and blockchain. Background Technology

[0002] Existing cold chain transportation status monitoring technologies primarily rely on temperature sensors, humidity sensors, and positioning terminals deployed on mobile monitoring units to collect environmental parameters. These terminal devices convert the collected environmental physical quantities into discrete digital signals and transmit them to a server for centralized storage and processing via wireless communication networks. Current information processing systems typically use preset fixed threshold values ​​to linearly compare uploaded time-series data. When monitored values ​​exceed a set range, an alarm record is generated, thus achieving preliminary digital monitoring of the environmental status during transportation and historical data retrieval. However, in actual multi-stage cold chain handover engineering scenarios, quality control personnel often encounter a phenomenon of data logic verification failure: when the controlled target object arrives at the receiving node for physical inspection, it already exhibits thermal damage characteristics, but historical environmental monitoring data sequences show that the values ​​throughout the process are within the preset compliance range. The root cause of this failure lies in the fact that existing data processing methods only perform instantaneous threshold determination on single-dimensional scalar data, ignoring the spatiotemporal correlation between multi-source heterogeneous sensor data streams. This results in the data model being unable to characterize the nonlinear relationship between transient disturbances in non-stationary environments and the cumulative heat exposure of the controlled object. Furthermore, due to the lack of cryptographically based metadata anchoring mechanisms in existing data storage architectures, time-series data only has a weak logical connection with the identity and geographical location of the responsible entity that generated the data during the generation, transmission, and storage stages. This results in a lack of inherent integrity self-verification capabilities in the data flow process. This deficiency at the data processing level leads to a logical disconnect between digital monitoring records and the physical transportation process. Consequently, the stored evidence loses its technical uniqueness and immutability when faced with anomaly tracing, failing to meet the verification requirements for the validity of electronic data evidence in the flow of high-value controlled goods.

[0003] In the prior art, Chinese patent application CN120258667A discloses a blockchain-based cold chain warehousing data security traceability system. This system includes a data acquisition module, a data analysis module, a risk management module, and an execution monitoring module. The data acquisition module collects historical logs, parameter information, and current logistics details of refrigerated trucks; the data analysis module generates temperature field evolution videos through physical simulation models, and divides the affected areas of the goods after frame decomposition; the risk management module calculates the average temperature of the affected areas to analyze the risk index of the goods; and the execution monitoring module marks the goods based on the risk index and formulates priority execution plans. It achieves full-chain control of the cold chain through multi-source data perception and dynamic risk modeling via the Internet of Things, improving the quality of fresh produce and transportation efficiency. Chinese patent CN119398642B discloses a multimodal AIGC cold chain logistics information processing method and system. This system employs a multimodal AIGC model to extract and enhance features from video, temperature, and location data, generating an enhanced feature matrix to support digital twin modeling. It combines deep learning technology with information latency index calculation to generate a real-time data synchronization scheme. Through deep graph network encoding, it obtains the distribution network representation vector, enabling multi-agent collaborative decision-making and dynamic updates of scheduling schemes. The aim is to solve the difficulties in cold chain scheduling and control, and improve the intelligence of scheduling and the accuracy of control in complex scenarios.

[0004] However, while the two existing technologies mentioned above have certain application value in calculating risk index and optimizing scheduling decisions for cold chain logistics, they have failed to solve the technical challenges of deep semantic parsing and underlying trust anchoring of multi-source heterogeneous data streams. Among them, the Chinese patent application with publication number CN120258667A focuses on risk index calculation and scheme execution, but does not analyze the physical correlation of multi-dimensional sensor data, cannot map sensor changes to specific state mutations, and lacks causal inference capabilities; the Chinese patent with authorization announcement number CN119398642B relies on multimodal data and digital twins to achieve scheduling optimization, but does not solve the sensor drift problem, nor does it establish an active risk boundary adjustment mechanism based on event prediction. Moreover, neither of them focuses on numerical statistics or feature extraction under discrete time slices at the data processing level, and lacks the ability to model the geometric manifold characteristics of sensor data streams in continuous multi-dimensional state space as a whole. This makes it difficult for the system to accurately characterize the nonlinear causal relationship between small state mutations and cumulative heat exposure in non-stationary environments through mathematical models. At the same time, at the data storage level, the above technologies do not establish a metadata anchoring mechanism based on cryptography, so that there is only a weak logical association between the key state verification data packet and the identity of the responsible node that generated the data. This makes it impossible for the system to provide electronic evidence data with intrinsic integrity verification capabilities and immutability when facing cross-domain traceability. Summary of the Invention

[0005] This invention is applicable to cold chain logistics information storage scenarios with high requirements for data integrity and traceability verifiability, and can meet the deep semantic analysis needs of multi-source heterogeneous data streams. By acquiring continuous data streams from heterogeneous sensors of mobile monitoring units and mapping them to a multi-dimensional state space, a covariance matrix is ​​constructed based on the data set under steady-state operation mode to generate a hyperellipsoidal manifold that adapts to the environment. Combined with a cooperative response cone model that characterizes the temporal features of state abrupt changes, a structured state abrupt change event descriptor is generated, enabling accurate semantic differentiation between disturbance signals and valid events under non-stationary environments, thus solving the problem of difficult anomaly location and traceability. Relying on an anomaly assessment model that includes cumulative heat exposure and instantaneous thermal shock rate, the steady-state convergence domain of the controlled target object is delineated, and a dynamic safety boundary is formulated in combination with the state abrupt change event descriptor, transforming static threshold judgment into dynamic anomaly control based on manifold geometric features. A data anchoring mechanism is constructed based on state mutation event descriptors and dynamic security boundaries. The Merkle tree algorithm is used to generate a root hash value containing the digital identity of the node and synchronize it to the distributed ledger. This generates an immutable evidence with cryptographic trust properties, preventing data tampering and post-entry, meeting the requirements of high compliance scenarios for data continuity and verifiability, and improving the security and credibility of cold chain logistics information evidence.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Methods for storing cold chain logistics information based on multi-dimensional perception and blockchain include: The system acquires continuous data streams from heterogeneous sensors during the operation of the mobile monitoring unit, extracts the core data from the continuous data streams to obtain the original feature vectors, maps the original feature vectors to a pre-constructed multidimensional state space to obtain a set of state points, constructs a hyperellipsoid and a standard response axis in the multidimensional state space based on the set of state points, and constructs a cooperative response cone representing the event response based on the standard response axis of historical real events; acquires the perturbation trajectory of the hyperellipsoid, performs geometric matching between the perturbation trajectory and the cooperative response cone, and generates a structured event descriptor. An anomaly assessment model is constructed for specific goods, and a steady-state convergence region representing the goods satisfying the constraints is defined in the anomaly assessment model. Dynamic safety boundaries are formulated in the steady-state convergence region based on event descriptors. A data storage mechanism is built based on event descriptors and dynamic security boundaries, a distributed ledger with reliable and traceable records is generated, and responsibilities are divided according to the distributed ledger to obtain a handover agreement with clear responsibility division.

[0007] Furthermore, the method for obtaining the ellipsoid includes: The original feature vectors are preprocessed to obtain a set of feature vectors; Any point in the multidimensional state space is a state point, which corresponds to a set of feature vectors; When the mobile monitoring unit enters the steady-state operation mode, it collects the feature vector set corresponding to the state point according to the preset data collection window, and integrates the obtained feature vector set according to the time sequence of the data collection window to obtain the state point set. The steady-state operation mode refers to the transportation stage on the highway where the mobile monitoring unit continuously travels at a speed greater than a preset speed threshold for a time greater than a preset time threshold. A covariance matrix is ​​constructed by using a set of state points. The covariance matrix is ​​then solved to obtain multiple eigenvalues ​​and the orthogonal eigenvectors corresponding to each eigenvalue. The mean vector is calculated from the set of state points and used as the geometric center of the hyperellipsoid. Using the directions of the orthogonal eigenvectors themselves as the directions of the principal axes of the hyperellipsoid, the semi-axis lengths of each principal axis are obtained by calculating the eigenvalues, and finally the hyperellipsoid is obtained.

[0008] Furthermore, the method for obtaining the event descriptor includes: For real-time state points collected in real time, it is determined whether they are located inside the hyperellipsoid. If it is detected that the real-time state point has crossed the boundary of the hyperellipsoid, the motion trajectory of the state point is recorded to form a disturbance trajectory. The perturbation trajectory is geometrically matched with the cooperative response cone. When the geometric matching verification condition is met, a structured event descriptor is generated. The event descriptor is a data structure containing multiple key fields, including event type, start timestamp, duration, and confidence score.

[0009] Furthermore, the geometric matching includes: Geometric matching includes path length verification and point count percentage verification; The criteria for path length verification are as follows: if the path segment length of the disturbance trajectory inside the cooperative response cone is greater than the preset minimum length threshold, then the path length verification is satisfied. Here, the path segment length inside the cooperative response cone refers to the sum of the path lengths corresponding to all consecutive state points located inside the cooperative response cone among the state points constituting the disturbance trajectory. The criteria for the point percentage verification are as follows: if the proportion of the number of state points falling within the cooperative response cone of the disturbance trajectory to the total number of state points of the disturbance trajectory is greater than the preset minimum proportion threshold, then the point percentage verification is satisfied. When both path length verification and point percentage verification are satisfied, it is determined that a valid state mutation corresponding to the cooperative response cone has been identified.

[0010] Furthermore, the anomaly assessment model includes: The anomaly assessment model is a two-dimensional Cartesian coordinate system. The X-axis of the anomaly assessment model is defined as the cumulative heat exposure, and the Y-axis is defined as the instantaneous thermal shock rate. The physical meaning of cumulative heat exposure is the product of the degree to which the core temperature of the cargo deviates from the preset ideal temperature range and time. It is used to measure cumulative damage and is obtained by integrating the difference between the core temperature of the cargo and a preset reference temperature over time. The core temperature of the cargo is obtained in real time by temperature sensors installed inside the cargo box or at representative locations. The physical meaning of instantaneous thermal shock rate is the rate of change of the core temperature of the cargo. It is used to measure instantaneous impact and is calculated as the derivative of the core temperature of the cargo with respect to time.

[0011] Furthermore, the method for obtaining the dynamic security boundary includes: Extract the confidence score of the event descriptor. When the confidence score is greater than the preset trigger threshold, trigger the dynamic boundary adjustment mechanism. The dynamic boundary adjustment mechanism searches for the impact direction of the state change on the current cargo state point from a predefined event impact mapping table based on the event type in the event descriptor. The impact direction is represented by a two-dimensional vector in the anomaly assessment model, and the two-dimensional vector is defined as the risk impact direction vector. The steady-state convergence region is a closed polygonal region consisting of four vertices. To preventively shrink the steady-state convergence region, each vertex of the steady-state convergence region is translated inward along the direction opposite to the direction vector of the risk impact by a certain distance, which is the translation distance. The new polygonal region formed after the shrinkage is the dynamic safety boundary.

[0012] Furthermore, the cooperative response cone includes: The co-response cone is a cone-shaped deviation region centered on the standard response axis, and the radius of the co-response cone is a function that varies along the standard response axis. The function obtains multiple confirmed real state point trajectories belonging to the same state change from historical real events, performs non-linear alignment on the real state point trajectories, and obtains an aligned trajectory set. Starting from the surface of the ellipsoid, a series of ordered axis points are obtained by stepping at a preset step size; For any axis point, select all corresponding real state points from the alignment trajectory set, calculate the distance set by combining all real state points with the axis point, and calculate the arithmetic mean and standard deviation of the distance set. Then, formulate the function based on the arithmetic mean and standard deviation.

[0013] Furthermore, the data storage mechanism includes: The data evidence storage mechanism is triggered when a state change ends or when the mobile monitoring unit reaches a predetermined handover point. The data storage mechanism aggregates all key information related to the time period, generates a causal evidence package, and calculates the first hash value from the causal evidence package. From the event descriptor, extract the start timestamp and duration fields to define a state change time window. From the historical continuous data stream, filter out all data whose timestamps are within the state change time window. Divide the filtered data into a series of independent data blocks according to the timestamps. Calculate the root hash value for each data block. The root hash value, the first hash value, and the preset digital identity of the responsible party are concatenated and calculated to obtain the responsibility anchor hash value. The responsibility anchor hash value is digitally signed to form a signed transaction, which is then recorded on the distributed ledger.

[0014] Furthermore, the method for obtaining the handover protocol includes: When the mobile monitoring unit arrives at the responsibility handover point, the handover agreement is executed between the transferring and receiving parties. When the mobile monitoring unit arrives at the responsibility handover point, the handover agreement is executed between the transferring and receiving parties. The transferring party refers to the party responsible for supervising the goods before the predetermined handover point, and the receiving party refers to the next party responsible for taking over the supervision of the goods after the handover point. Generate a corresponding responsibility anchor hash value for the transferor and record it in the distributed ledger. Before officially taking over the goods, the receiving party reads the responsibility anchor hash value submitted by the transferor and the signed transaction record to verify the authenticity of the transferor's signature. Once the signature is confirmed to be genuine and valid, the receiving party obtains the first hash value from the transaction record, recalculates the first hash value for the causal evidence package of the transferring party, records it as the verification hash value, and compares the verification hash value with the first hash value in the transaction record. Once the verification hash value matches the first hash value, the receiving party digitally countersigns the transaction record on the distributed ledger. The digital countersignature and the transaction record together constitute a handover agreement with a clear division of responsibilities.

[0015] A cold chain logistics information storage system based on multi-dimensional sensing and blockchain is used to implement the aforementioned cold chain logistics information storage method based on multi-dimensional sensing and blockchain. The system includes: Event parsing module: Used to acquire continuous data streams representing the states of multiple physical quantities within the mobile monitoring unit during its operation; maps the continuous data streams to a multi-dimensional state space; constructs a hyperellipsoid that adaptively adjusts its position according to environmental changes and a standard response axis in the multi-dimensional state space; defines a cooperative response cone representing the event response in terms of geometric shape based on the standard response axis; acquires real-time state points and detects the disturbance trajectories of the real-time state points and the hyperellipsoid; geometrically matches the disturbance trajectories with the cooperative response cone to generate structured event descriptors; Risk management module: used to build an anomaly assessment model for specific goods, and to define a steady-state convergence region in the anomaly assessment model that represents the goods satisfying the constraints, and to formulate dynamic safety boundaries in the steady-state convergence region based on event descriptors; Evidence storage and traceability module: It is used to build a data evidence storage mechanism based on event descriptors and dynamic security boundaries, generate a distributed ledger with reliable traceability of records, and generate a handover agreement with clear division of responsibilities based on the distributed ledger.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires continuous data streams from heterogeneous sensors of a mobile monitoring unit and maps them to a multidimensional state space. Based on the covariance matrix, it constructs a hyperellipsoidal manifold and cooperative response cone model that adaptively adjusts with environmental parameters, generating a structured state change event descriptor. This enables accurate identification of state changes and effectively distinguishes between environmental disturbance signals and sensor noise, addressing the technical pain point of traditional solutions that lack multidimensional semantic expression and causal inference, leading to difficulties in anomaly localization and tracing. Furthermore, relying on an anomaly assessment model that includes cumulative heat exposure and instantaneous thermal shock rate, it delineates the steady-state convergence region of the controlled target object and establishes... By defining dynamic security boundaries, the traditional passive threshold alarm is transformed into proactive anomaly control based on manifold geometry characteristics, breaking through the technical bottleneck of "monitoring after the fact and loss of control first," and avoiding mismatch between data representation models and actual environmental thermodynamic evolution processes. A trusted distributed ledger is generated through a data anchoring mechanism, combined with tamper-proof evidence with clearly defined responsible nodes, eliminating data tampering and post-event recording, clearly defining the data verification boundaries of multi-node collaboration, meeting the requirements of data continuity and integrity verification in high-compliance scenarios such as pharmaceuticals and active foods, and improving the security and credibility of cold chain logistics information evidence storage. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the method for storing cold chain logistics information based on multi-dimensional perception and blockchain, as provided in an embodiment of the present invention. Figure 2 A logical judgment diagram for geometric matching provided in an embodiment of the present invention; Figure 3 A flowchart illustrating the process of obtaining the hash value of the responsibility anchor point using the data storage mechanism provided in this embodiment of the invention. Figure 4 This is a functional module diagram of a cold chain logistics information storage system based on multi-dimensional perception and blockchain provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 Please see Figure 1 As shown, this embodiment provides a method for storing cold chain logistics information based on multi-dimensional perception and blockchain, including: Step S10: Acquire the continuous data stream from the heterogeneous sensors during the operation of the mobile monitoring unit, extract the core data from the continuous data stream to obtain the original feature vector, map the original feature vector to the pre-constructed multi-dimensional state space to obtain the set of state points, construct a hyperellipsoid and a standard response axis in the multi-dimensional state space based on the set of state points, and construct a cooperative response cone representing the event response based on the standard response axis of historical real events; acquire the perturbation trajectory of the hyperellipsoid, perform geometric matching between the perturbation trajectory and the cooperative response cone, and generate a structured event descriptor.

[0021] Further, step S10 includes: Step S11: Obtain a continuous data stream characterizing the state of multiple physical quantities within the mobile monitoring unit during its operation; extract the core data from the continuous data stream; map the core data into a multidimensional state space; and construct a hyperellipsoid in the multidimensional state space that adaptively adjusts its position according to environmental changes.

[0022] In cold chain transportation practice, when the external environment of the mobile monitoring unit changes, the sensor data representing its stable operating state will shift accordingly. If the system cannot adaptively follow this shift, sensor fluctuations within the normal range, caused by physical laws, will be incorrectly identified as abnormal, leading to numerous missed or false alarms. The mobile monitoring unit refers to the medium used in cold chain transportation, such as refrigerated trucks. Therefore, constructing a hyperellipsoid that can accurately depict the inherent physical correlation between sensor readings in a stable state and adaptively adjust its position according to environmental changes provides a precise and dynamic reference system for subsequent operations.

[0023] Specifically, a continuous data stream is acquired from heterogeneous sensors deployed within the mobile monitoring unit. These heterogeneous sensors refer to a set of devices installed at different locations within the mobile monitoring unit, used to simultaneously collect signals of various physical quantities, such as temperature sensors, cabin-internal pressure differential sensors, reed door magnetic sensors, acoustic array sensors, and heat flow meters. The continuous data stream refers to a time-series dataset formed by the heterogeneous sensors aligned with pre-set frequencies and time-stamped. For example, the temperature sensor measures a specific real-time floating-point value of temperature at a pre-set frequency, and the cabin-internal pressure differential sensor measures a floating-point value representing the air pressure difference between the inside and outside of the cabin at a pre-set frequency.

[0024] The core data is extracted from the continuous data stream of heterogeneous sensors, and the core data features are processed to eliminate the influence of different physical units and numerical ranges. The processed core data features are then mapped to a unified dimensionless multidimensional state space. The core data refers to a single quantifiable value that represents the core physical state of the heterogeneous sensors. For example, the temperature sensor extracts the arithmetic mean of temperature, representing the macroscopic thermodynamic state within the entire mobile monitoring unit. This arithmetic mean is calculated using the arithmetic mean formula from all temperature sensor measurements within the mobile monitoring unit. The cabin-internal pressure difference sensor directly extracts the instantaneous pressure difference value measured in real time, reflecting the airtightness of the cabin. The reed door magnetic sensor processes the raw binary output of the reed door magnetic sensor and extracts the percentage of time the door is open using a continuous spatial analysis method. The acoustic array sensor performs a fast Fourier transform on the collected complex sound waveforms to obtain the sound spectrum, and then performs acoustic energy integration on the spectrum. The resulting acoustic integral value is the core data, quantifying the compressor's operating intensity. The heat flux meter extracts the instantaneous heat flux density value measured in real time, representing the heat passing through a unit area per unit time. The core data obtained from each sensor in the heterogeneous sensors are integrated to obtain the original feature vector.

[0025] Data preprocessing methods, such as Z-score transformation, are used to process the original feature vectors, transforming them into dimensionless standardized values ​​with an approximate mean of 0 and a standard deviation of 1, resulting in a feature vector set. This feature vector set is then mapped to a multidimensional state space. The spatial dimensions and coordinate axes of this multidimensional state space are determined based on heterogeneous sensors and core data. For example, if there are 5 types of heterogeneous sensors, the multidimensional state space has 5 dimensions. Each coordinate axis of the multidimensional state space corresponds to a specific core data value in the feature vector set. Any state point in the multidimensional state space corresponds to a feature vector set, and this state point represents any point in the multidimensional state space, functioning similarly to a coordinate point in a planar coordinate system. In cold chain transportation, when the mobile monitoring unit enters a steady-state operation mode, it collects the feature vector set corresponding to the state point according to a preset data collection window. The obtained feature vector set is then integrated according to the time sequence of the data collection window to obtain a state point set. The steady-state operation mode represents a specific transportation stage where the mobile monitoring unit continuously travels on a highway at a speed exceeding a preset threshold for a time exceeding a preset threshold. The speed threshold is set based on general regulations for minimum highway speeds and the practical need to isolate the impact of complex urban conditions. It filters out low-speed or stop-and-go non-steady-state conditions such as congested urban roads, maneuvering within stations, and temporary stops, and excludes severe disturbances introduced by frequent starts and stops, door openings and closings, etc., in urban driving. For example, it is set to 60 kilometers per hour. The time threshold represents the shortest time required for the mobile monitoring unit's speed to remain above the speed threshold. It is set based on the time period required for the mobile monitoring unit's thermodynamic system to reach dynamic thermal equilibrium after entering a stable operating condition, and also considers the experience of avoiding the transient adjustment impact when the vehicle first enters a highway section. The purpose is to ensure that data acquisition enters a truly periodic and regular stable operating mode, ensuring that the collected state points accurately reflect the situation rather than providing transient responses. For example, it is set to 15 minutes. The data collection window refers to a preset time length for continuously collecting state points during the steady-state operating mode. It is set based on the minimum observation time required in statistics to ensure that the sample can fully cover multiple complete cycles of a random process and obtain robust mean and variance. For the mobile monitoring unit, the time span covers the complete start-stop cycles of several refrigeration compressors. The aim is to ensure that the collected set of status points has a sufficient time span to capture all typical fluctuation patterns occurring in steady-state operation, such as the periodic coordinated changes in temperature, acoustics, and power consumption caused by compressor start-stop; for example, this is set to 30 minutes.

[0026] A covariance matrix is ​​constructed using a set of state points. This covariance matrix is ​​a symmetric square matrix, and the value of any element C(i,j) in the matrix is ​​the covariance calculated based on the i-th and j-th components of all state points in the set. Each component refers to any one of the components that constitutes the feature vector set. For example, if the multidimensional state space is a five-dimensional space, and the data in the feature vector set corresponding to any state point are, in order, the arithmetic mean of temperature, the instantaneous pressure difference, the percentage of time the door magnet is open, the acoustic integral value, and the instantaneous heat flux density value, then the first component is the arithmetic mean of temperature. i and j represent indices in the eigenvector set. The range of i and j is determined by the number of dimensions in the multidimensional state space. For example, if the multidimensional state space is 5-dimensional, then i and j are less than or equal to 5 and greater than or equal to 1. When i and j are equal, they are diagonal elements of the covariance matrix; otherwise, they are off-diagonal elements. Diagonal elements measure the intensity of the component's own fluctuations under steady-state operating conditions. Off-diagonal elements quantify the linear relationship between the i-th and j-th components under steady-state operating conditions. Solving the covariance matrix yields multiple eigenvalues ​​and orthogonal eigenvectors corresponding to each eigenvalue. These orthogonal eigenvectors constitute a new coordinate system in the multidimensional state space. The direction of the orthogonal eigenvectors represents the main direction of data change in the set of state points, while the eigenvalues ​​corresponding to each orthogonal eigenvector measure the degree of dispersion of the data in the corresponding direction. The number of eigenvalues ​​is consistent with the number of dimensions in the multidimensional state space; for example, a 5-dimensional space has 5 eigenvalues. A dynamically stable hyperellipsoid is constructed in a multidimensional state space. The mean vector of the set of state points is used as the geometric center of the hyperellipsoid, which is obtained by averaging each component of the set of eigenvectors corresponding to the set of state points. The directions of the orthogonal eigenvectors are used as the principal axes of the hyperellipsoid. The semi-axis length of each principal axis is obtained by multiplying the square root of the eigenvalue corresponding to each principal axis direction by a confidence coefficient, which is a constant used to control the volume and confidence level of the hyperellipsoid, set according to statistics. For example, it is set to 3. The tilt shape of the hyperellipsoid can accurately depict the inherent physical relationship between the core data in the steady-state operation mode. When the mobile monitoring unit naturally shifts due to changes in the external environment, by continuously updating the set of state points used to calculate the covariance matrix, that is, re-executing the data collection window in the new steady-state operation mode, the geometric center, principal axis directions, and semi-axis lengths of the hyperellipsoid can be slowly adjusted accordingly, thereby dynamically adapting to different environmental benchmarks and providing an accurate and adaptive reference system for subsequent operations.

[0027] Step S12: Define a standard response axis with temporal characteristics in the multidimensional state space, and construct a cooperative response cone with geometric shape to characterize the event response mode based on the standard response axis.

[0028] In order to further interpret the disturbance represented by a state point that deviates from the hyperellipsoid as a state mutation with clear business semantics, such as opening a door for loading and unloading or temporarily merging cabinets, a collaborative response cone is constructed to transform the abstract event logic into a concrete spatial geometry.

[0029] To address state abrupt changes, a standard response axis with temporal characteristics is defined in the multidimensional state space. This standard response axis is an idealized, ordered curve representing the movement of state points in the multidimensional state space at the time of the state abrupt change. The state abrupt change refers to a physical process occurring during cold chain transportation that causes identifiable coordinated changes in the core data of heterogeneous sensors and possesses clear business or operational semantics. State abrupt changes differ from random noise or irregular environmental fluctuations, exhibiting a reproducible response pattern that follows physical causality. For example, taking the temporary opening of the door for loading and unloading in cold chain transportation practice as a state change for analysis, the starting point of the standard response axis of the state change is located on the surface of the hyperellipsoid. When the door of the motion monitoring unit is opened, the core data of the state change will change significantly in a specific order. For example, the proportion of time that represents the door magnetic state as open changes, causing the third component of the state point to shift; at the same time, due to the connection between the inside and outside air, the instantaneous pressure difference value changes rapidly, causing the second component of the state point to shift; then, due to the convection of hot and cold air, the arithmetic mean of the temperature rises, causing the first component of the state point to shift; simultaneously, the instantaneous heat flux density value spikes, causing the fifth component of the state point to shift. Connecting these significantly changed state points in sequence in the multidimensional state space forms a curve starting from the surface of the hyperellipsoid and sequentially passing through the specific component change region; this curve is the standard response axis.

[0030] To accommodate reasonable variations in actual operation, such as the speed of door opening and the magnitude of temperature differences in the external environment, which affect the amplitude and rate of the response, a cooperative response cone is constructed based on the standard response axis to geometrically characterize the event response. This cooperative response cone is a conical deviation region in a multidimensional state space centered on the standard response axis. The radius of the cooperative response cone is a function that varies along the standard response axis, determined through point-by-point statistical calculations based on historical data of similar state mutations. Specifically, a set of real state point trajectories, including multiple confirmed trajectories belonging to the same state mutation, is obtained. These trajectories are nonlinearly aligned using a time series alignment algorithm to obtain an aligned trajectory set. Starting from the surface of an ellipsoid, the standard response axis is discretized into a series of ordered axis points with a preset step size. The step size is set based on the arc length of the standard response axis itself. For any axis point, find all true state points corresponding to the axis point from the aligned trajectory set, and denote all true state points as the cross-sectional state point set. Calculate the Euclidean distance values ​​from all true state points to the axis point, and combine these Euclidean distance values ​​into a distance set. Calculate the arithmetic mean D1 and standard deviation D2 of all Euclidean distance values ​​using the distance set. Construct the variation function R corresponding to the radius of the cooperative response cone using D1 and D2. Where k represents the cone confidence coefficient, used to ensure that the cooperative response cone can contain the vast majority of historical real event trajectories; for example, it is set to 2 or 3. The reason for using the arithmetic mean and standard deviation to construct the variation function corresponding to the radius of the cooperative response cone is that the arithmetic mean geometrically defines the average deviation of the motion trajectory of all historical real state points at a specific position on the standard response axis, constituting the base value of the cooperative response cone at that specific position. The standard deviation quantifies the uncertainty range of the degree of deviation of the motion trajectory of all historical real state points. By multiplying the standard deviation of the distance by the cone confidence coefficient, a statistically reliable safety margin is added to the base value of the radius, thereby ensuring that the cooperative response cone can accommodate the vast majority of reasonable operational variations.

[0031] The robustness of the cooperative response cone far exceeds that of simple multi-threshold judgment. For example, in a scenario where a magnetic sensor on a reed door is falsely triggered due to severe vehicle jolting, although the third component of the state point will change, causing the state point to move out of the hyperellipsoid, the state point's trajectory cannot enter the cooperative response cone that requires multi-dimensional coordination because core data such as instantaneous pressure difference and temperature arithmetic mean do not generate a cooperative response. Thus, it is effectively filtered out, avoiding false alarms.

[0032] Step S13: Obtain the real-time state point, detect the perturbation trajectory between the real-time state point and the hyperellipsoid, perform geometric matching between the perturbation trajectory and the cooperative response cone, and generate a structured event descriptor.

[0033] To accurately and reliably determine the perturbation trajectory formed by a state point deviating from the hyperellipsoid as an effective state mutation corresponding to the cooperative response cone, and to quantify the reliability of the determination, a dual judgment mechanism is constructed, ultimately generating a structured event descriptor containing a confidence score.

[0034] For real-time state points acquired in real time, it is determined whether they are located inside the hyperellipsoid. If a real-time state point is detected to have crossed the boundary of the hyperellipsoid, a disturbance is identified, and the movement trajectory of the state point is recorded, forming a disturbance trajectory. The disturbance trajectory is an ordered path composed of continuous state points that deviate from the hyperellipsoid. Specifically, the criterion for determining whether a real-time state point has crossed the boundary of the hyperellipsoid is as follows: according to the method described in step S11, the real-time state point is transformed from the multidimensional state space to a new coordinate system defined in the hyperellipsoid. In the new coordinate system, the coordinate value of the real-time state point in the new coordinate system is divided by the semi-axis length, and the result is squared. The squared results of each component of the feature vector of the state point are summed to obtain a normalized distance squared value. When the normalized distance squared value is greater than 1, it is determined that the real-time state point is located outside the hyperellipsoid; otherwise, it is determined that the real-time state point is located inside the hyperellipsoid or on its boundary. The determination that a real-time state point has crossed the boundary of the hyperellipsoid is as follows: if the squared normalized distance of the real-time state point at the current sampling time is greater than 1, and the squared normalized distance of the real-time state point at the previous sampling time is less than or equal to 1, then the real-time state point has crossed the boundary of the hyperellipsoid.

[0035] Geometric matching of the perturbation trajectory with the cooperative response cone, such as... Figure 2As shown, geometric matching includes path length verification and point proportion verification. The path length verification is based on the following criteria: if the length of the path segment within the cooperative response cone of the perturbation trajectory is greater than a preset minimum length threshold, then the path length verification is satisfied. The path segment length within the cooperative response cone refers to the sum of the path lengths corresponding to all consecutive state points within the cooperative response cone that constitute the perturbation trajectory. The minimum length threshold is determined based on the total length of the standard response axis corresponding to the cooperative response cone; for example, it can be set to 20% of the total length of the standard response axis. This aims to exclude invalid noise trajectories that briefly and accidentally cross the edge of the cooperative response cone. The point proportion verification is based on the following criteria: if the proportion of the number of state points within the cooperative response cone of the perturbation trajectory to the total number of state points from the beginning to the end of the perturbation trajectory is greater than a preset minimum proportion threshold, then the point proportion verification is satisfied. For example, the minimum proportion threshold is set to 70% to ensure that the main process of the perturbation trajectory conforms to a predetermined pattern. When both path length verification and point percentage verification are satisfied, a valid state mutation corresponding to the cooperative response cone is identified. Simultaneously, a structured event descriptor is generated. This event descriptor is a data structure containing multiple key fields, including event type, start timestamp, duration, and confidence score. The event type is a string uniquely identifying the category of the identified state mutation; the start timestamp precisely records the start time of the identified state mutation; the duration represents the time elapsed from the start to the end of the identified state mutation; and the confidence score is a floating-point value between 0 and 1, used to quantify the geometric similarity between the perturbation trajectory and the standard pattern of the state mutation. Specifically… The formula is as follows: The arc length of the perturbation trajectory refers to the actual arc length of the perturbation trajectory within the cooperative response cone, obtained by accumulating the Euclidean distances between consecutive state points constituting the path segment within the perturbation trajectory. The projection length refers to the length of the projected path obtained by orthogonally projecting the path segment of the perturbation trajectory within the cooperative response cone onto the standard response axis of the cooperative response cone. This is obtained by orthogonally projecting the start and end points of the path segment within the perturbation trajectory onto the standard response axis and calculating the difference in arc lengths between the two projected points on the standard response axis. The average orthogonal distance refers to the average orthogonal distance from all state points within the cooperative response cone to the standard response axis. This is obtained by calculating the orthogonal distances from all state points within the perturbation trajectory to the standard response axis and taking the arithmetic mean of all orthogonal distance values. The average cone radius refers to the average radius of the cooperative response cone in the region where the perturbation trajectory is located. This is obtained by obtaining the radius value of the cooperative response cone at the position of the standard response axis corresponding to each state point within the perturbation trajectory and taking the arithmetic mean of all radius values. The first ratio term in the formula is... This reflects the consistency between the disturbance trajectory and the standard response axis in terms of direction and timing. The second ratio term is... This reflects the atypicality of the disturbance trajectory; the smaller the ratio term, the closer the trajectory is to the standard response axis.

[0036] Step S10 addresses the technical challenges in existing cold chain logistics information management by constructing a hyperellipsoid, defining standard response axes, building collaborative response cones, and establishing structured event descriptors. These challenges include the sparse dimensions of heterogeneous IoT data, the lack of online calibration for sensor drift, the failure to promptly identify scene changes, and the lack of structured representation of environmental semantic states and verifiable causal inferences, leading to difficulties in locating and tracing anomalies. Step S10 achieves accurate identification of sudden state changes in cold chain transportation, accurate differentiation between disturbance signals and valid events, and standardized output of structured event information. Specifically, the hyperellipsoid provides a precise and adaptive reference system for steady-state operation, avoiding misjudgments of sensor collaborative fluctuations due to environmental changes; the collaborative response cone transforms abstract event logic into a concrete spatial geometry, accommodating reasonable changes in actual operation while effectively filtering invalid noise trajectories when sensors are falsely triggered; and the structured events provide semantically clear and formatted data support for subsequent risk assessment and accountability.

[0037] Step S20: Construct an anomaly assessment model for specific goods, and define a steady-state convergence region in the anomaly assessment model that represents the goods satisfying the constraints. Formulate a dynamic boundary adjustment mechanism based on the event descriptor, and formulate a dynamic safety boundary in the steady-state convergence region.

[0038] Further, step S20 includes: Step S21: Construct an anomaly assessment model for specific goods.

[0039] In order to upgrade single temperature monitoring into a comprehensive quantitative assessment of the cumulative damage and instantaneous impact risk to goods, an anomaly assessment model is constructed to unify the two key modes of goods quality deterioration in the same view for quantification, providing a basic quantifiable status input for subsequent processing.

[0040] The anomaly assessment model is a two-dimensional Cartesian coordinate system. The X-axis of the anomaly assessment model is defined as the cumulative heat exposure, and the Y-axis is defined as the instantaneous thermal shock rate. The cumulative heat exposure t1 is physically represented as the product of the degree to which the cargo core temperature t2 deviates from the ideal temperature range and time, used to measure cumulative damage. It is obtained by integrating the difference between the cargo core temperature and a preset reference temperature t3 over time; that is, the cumulative heat exposure is calculated as an integral over time T. The cargo core temperature t2 is obtained in real time by temperature sensors installed inside the cargo container or at representative locations. The reference temperature t3 is a preset boundary value of an ideal temperature range, set according to cargo requirements. For example, for cargo requiring a temperature between 2 and 8 degrees Celsius, the reference temperature is equal to 8 degrees Celsius when the cargo core temperature is greater than 8 degrees Celsius; and equal to 2 degrees Celsius when the cargo core temperature is less than 2 degrees Celsius. The instantaneous thermal shock rate w1 is the rate of change of the cargo core temperature, used to measure instantaneous impact. It is calculated as the derivative of the cargo core temperature with respect to time. ,in, This represents a preset short time window. Indicates the temperature before the core temperature of the cargo. The core temperature of the cargo at that time; because the formula uses an approximation of the temperature change difference within a short time window, it is set according to the engineering practice requirement of ensuring the real-time calculation of the temperature change rate while effectively filtering out the instantaneous noise of the sensor. For example, it is set to 1 minute.

[0041] By continuously acquiring readings from the core temperature sensors of the cargo, the current cumulative heat exposure and instantaneous thermal shock rate are calculated in real time and continuously. These two values ​​(cumulative heat exposure and instantaneous thermal shock rate) together constitute a coordinate point of the cargo on the anomaly assessment model, i.e., the current cargo status point.

[0042] Step S22: Define a steady-state convergence region in the anomaly assessment model that represents the goods satisfying the constraints.

[0043] To transform abstract, non-uniform cargo quality requirements into a concrete, computable geometric boundary, a parameter-to-vertex mapping method was constructed. A steady-state convergence region was customized for each type of cargo, resolving the mismatch between a one-size-fits-all temperature control range and the actual tolerance of the cargo, and providing personalized evaluation criteria for subsequent processing.

[0044] The cargo risk constraint parameters are obtained and mapped onto a series of constraint vertices on the anomaly assessment model. These cargo risk constraint parameters are a set of specific values ​​derived from cargo specifications, safety regulations, or transportation contracts, used to define the red lines for cargo quality and safety. For example, for a certain vaccine, the maximum allowable cumulative heat exposure is 4℃·h; the maximum absolute value of the instantaneous thermal shock rate is 0.5℃ per minute. The constraint vertices represent the extreme coordinate points of the specific constraint conditions on the anomaly assessment model. The constraint vertices include the cumulative exposure limit H, the shock rate limit Q, and the origin. The cumulative exposure limit refers to the maximum sum of the product of temperature deviation and time that the cargo can withstand after deviating from the ideal temperature range; it is the limit value for measuring the cargo's ability to withstand chronic damage. The coordinate point mapped to the cumulative exposure limit on the anomaly assessment model, i.e., vertex one, is (H, ε), where ε is a very small positive number representing the ideal low shock rate. The value of ε must be set to avoid computational singularity caused by vertex one falling on the coordinate axis, and the value must be small enough to meet computational stability requirements. The upper limit of the impact rate refers to the maximum rate of temperature change that cargo can withstand per unit time; it is the limit value for measuring the cargo's ability to withstand instantaneous impacts. The upper limit of the impact rate is mapped to the coordinate points (vertex 2 and vertex 3) on the anomaly assessment model, which are (δ, Q) and (δ, -Q) respectively. Here, δ is a very small positive cumulative heat exposure value. The value of δ is set based on the physical requirement that the cumulative exposure is close to zero when an instantaneous impact occurs, while avoiding the values ​​of vertex 2 and vertex 3 falling on the coordinate axes, thus ensuring computational stability. The origin represents the coordinate origin of the anomaly assessment model, where both cumulative heat exposure and instantaneous thermal shock rate are zero, representing the most ideal risk-free state. This origin is also denoted as vertex 4.

[0045] Connecting vertices one, two, three, and four sequentially forms a closed polygonal region in the anomaly assessment model. This polygonal region represents the steady-state convergence region of the cargo. Each vertex of the polygonal region precisely represents the limit of the cargo's risk tolerance in one dimension. Each edge of the polygonal region materializes a specific commercial or technical constraint. Therefore, the interior of the polygonal region geometrically constitutes an intersection that satisfies all constraints, representing the absolute safety zone that the cargo should maintain under all circumstances.

[0046] Step S23: Based on the event descriptor, a dynamic boundary adjustment mechanism is established to define a dynamic safety boundary in the steady-state convergence domain.

[0047] To transform perceived state mutations into direct impacts on control strategies and thus achieve proactive risk avoidance, a dynamic boundary adjustment mechanism based on state mutations is constructed using event descriptors. By proactively shrinking the steady-state convergence domain before the actual risk materializes, sufficient buffer space is reserved for impending risk shocks, thereby overcoming the technical bottleneck of prioritizing loss of control over monitoring.

[0048] Upon receiving an event descriptor, its confidence score is extracted. When the confidence score exceeds a preset trigger threshold, a dynamic boundary adjustment mechanism is triggered. This threshold is a value between 0 and 1, set based on the confidence statistical analysis of the historical real-state point trajectory. For example, it is set to 0.75. Specifically, the dynamic boundary adjustment mechanism, based on the event type in the event descriptor, searches a predefined event impact mapping table for the typical risk impact direction corresponding to that event type in the anomaly assessment model. This impact direction is represented in the anomaly assessment model by a two-dimensional vector, defined as the risk impact direction vector. For example, a door opening and unloading event typically leads to a simultaneous positive increase in both cumulative heat exposure and instantaneous heat shock rate; therefore, the risk impact direction vector corresponding to the state change is located in the first quadrant of the anomaly assessment model. The event impact mapping table establishes a unique mapping relationship between various state mutations represented by event types of event descriptors and a specific two-dimensional vector in the anomaly assessment model. Specifically, for each state mutation to be identified, the physical process of the state mutation is qualitatively analyzed to determine the impact of the state mutation on the cumulative heat exposure and the instantaneous thermal shock rate. For example, in the state mutation of door opening and loading / unloading, for the instantaneous thermal shock rate, when the door is opened, the rate of increase of the cargo core temperature increases, so the instantaneous thermal shock rate shifts towards the Y-axis square of the anomaly assessment model, that is, the cumulative heat exposure increases; for the cumulative heat exposure, since the cargo core temperature is higher than the reference temperature, the cumulative heat exposure shifts towards the positive X-axis direction of the anomaly assessment model, that is, the cumulative heat exposure increases. Based on qualitative analysis, the risk impact direction vector is quantitatively calibrated through historical data analysis or physical simulation. By acquiring the actual trajectory of the current cargo state point in the anomaly assessment model from a large number of similar state mutations, the displacement vector formed by the deviation of the actual state point trajectory from the origin to the point of most significant event impact is calculated for each actual state point movement. The point of most significant event impact refers to the coordinate point corresponding to the moment when the instantaneous thermal shock rate reaches its local maximum during the movement of the current cargo state point in the anomaly assessment model within the duration of a specific, already occurred state mutation. Geometrically, this represents the instantaneous moment when the state mutation causes the most severe instantaneous impact on the cargo state. A representative average impact vector is obtained by averaging the displacement vectors calculated from all similar events. After normalizing the average impact vector, the risk impact direction vector is obtained. Various state mutations of all event types are paired with the risk impact direction vector to construct and store the event impact mapping table.

[0049] Preventative contraction is performed on the steady-state convergence region. Specifically, each vertex of the steady-state convergence region is shifted inward by a distance A, in the opposite direction to the risk impact direction vector. The formula for the shift distance A is: A = Confidence Score × Base Contraction Amount × Event Severity Gain. The base contraction amount is a preset distance value representing the minimum safety margin, set according to the engineering requirements of the minimum safe buffer for the subsequent control system response delay when a state mutation occurs. For example, it can be set to 5% of the length of the shortest side formed by the four vertices in the steady-state convergence region. The event severity gain is a dimensionless multiplier factor used to amplify the severity of the state mutation, and is set in stages based on the duration field in the structured event descriptor. For example, if the duration is less than 1 minute, the event severity gain is set to 1.0. If the duration is between 1 and 5 minutes (greater than or equal to 1 minute and less than or equal to 5 minutes), the event severity gain is set to 1.5. If the duration is greater than 5 minutes, the event severity gain is set to 2.0. The selection of confidence score, base shrinkage amount, and event severity gain as parameters comprehensively considers three core dimensions assessed when responding to a state mutation: basic safety, event certainty, and event severity. This ensures that the magnitude and intensity of boundary shrinkage have a baseline guarantee while being able to intelligently and reasonably adjust according to the certainty and severity of the state mutation. The new polygonal region formed after shrinkage is defined as the dynamic safety boundary, which is a more strictly conservative safety region compared to the steady-state convergence region.

[0050] Step S20 addresses the technical challenges of traditional cold chain logistics, which relies solely on single-temperature monitoring, failing to quantify the dual risks of cumulative thermal damage and instantaneous thermal shock, and exhibiting a mismatch between fixed temperature control ranges and the actual tolerance of goods, as well as control strategies lagging behind the occurrence of risks. This is achieved through the construction of an anomaly assessment model, the delineation of steady-state convergence domains, and the generation of dynamic safety boundaries. Specifically, the anomaly assessment model unifies cumulative heat exposure and instantaneous thermal shock rate into a two-dimensional geometric space, quantifying two key quality degradation modes. The steady-state convergence domain maps cargo risk constraint parameters to vertices, forming closed polygons, resolving the mismatch between a one-size-fits-all temperature control range and the actual tolerance of goods, and customizing exclusive safety standards for different goods. The dynamic safety boundary contracts the vertices of the steady-state convergence domain inwards along the opposite direction of risk impact, reserving buffer space for upcoming risk impacts and overcoming the technical bottleneck of monitoring after the fact and loss of control before it occurs.

[0051] Step S30: Based on event descriptors and dynamic security boundaries, construct a data storage mechanism, generate a distributed ledger with reliable and traceable records, and generate a handover agreement with clear division of responsibilities based on the distributed ledger.

[0052] Further, step S30 includes: Step S31: Based on the event descriptor and dynamic security boundary, construct a data storage mechanism and generate a distributed ledger with reliable and traceable records.

[0053] In order to solidify event descriptors and dynamic security boundaries into an immutable and traceable evidence in a multi-party collaborative environment, a data storage mechanism is constructed to generate a distributed ledger that is recognized by multiple parties and has a trusted traceability capability.

[0054] like Figure 3 As shown, the data evidence storage mechanism is triggered when a state change ends or the mobile monitoring unit reaches a predetermined responsibility handover point. The responsibility handover point refers to a specific geographical location pre-defined in the cold chain transportation plan where the supervision or physical control of goods is formally transferred from one responsible party to another. This point is set according to the responsibility division nodes clearly stipulated in the transportation contract for different carriers or logistics links. For example, in a cross-border transportation of pharmaceuticals from a factory to an airport to a destination warehouse, responsibility handover point one could be set as the airport, and responsibility handover point two could be set as a port or transit warehouse in the destination area; the purpose is to ensure that the responsible party in the next transportation segment takes over the goods with full knowledge and consent. This fundamentally eliminates the phenomenon of responsibility division arising from unclear responsibility boundaries.

[0055] The data storage mechanism aggregates all key information related to the time period to generate a causal evidence package. This package includes an event descriptor and the coordinates of the four vertices of a dynamic safety boundary generated at the time of the state change or the handover of responsibility. From the event descriptor generated by the completed state change, the start timestamp and duration fields are extracted. The start timestamp and duration together define a state change time window. From the historical continuous data stream, all data whose timestamps fall within the state change time window are selected, and the selected data is divided into a series of independent data blocks according to the timestamp. The hash value of each data block is calculated using a hash algorithm. A Merkle tree construction algorithm is used, with the hash value of each data block as a leaf node. Hash operations are performed from bottom to top to ultimately construct a complete Merkle tree. The Merkle tree is a multi-level hash tree data structure built from the bottom up. Its purpose is to condense a large dataset into a single, fixed-length root hash value through layer-by-layer hash digests. This root hash value uniquely represents the content and order of the continuous data stream from heterogeneous sensors corresponding to the state change time window. Any small change to the selected data will lead to a significant change in the root hash value. After converting the causal evidence package into a string format, a hash algorithm is used to calculate the hash value representing the content of the causal evidence package. This hash value representing the content of the causal evidence package is recorded as the first hash value. The root hash value, the first hash value, and a preset digital identity identifier of the responsible party are concatenated to form a unique string. This string is then hashed using a hash algorithm to obtain the second hash value, which is recorded as the responsibility anchor hash value. The responsible party's digital identity is a digital credential used to uniquely and verifiably identify the current operating entity on the distributed ledger. The current operating entity may be a specific carrier or driver. This responsible party's digital identity is obtained using a standard public-private key generation algorithm. Specifically, the algorithm generates a key pair for the responsible party: a private key and a public key. The private key is kept by the responsible party and used to digitally sign the responsibility anchor hash value to prove the authorization of the operation. The public key is the responsible party's digital identity and is publicly disclosed. The carrier uses its private key to digitally sign the responsibility anchor hash value, forming a signed transaction. The signed transaction is recorded on the distributed ledger. The distributed ledger is a digital record system shared, synchronized, and replicated among multiple participants. Its purpose is to maintain immutable transaction records, establishing a common and verifiable single source of truth for participants who previously lacked mutual trust. Participants include cargo owners, carriers at various levels, and regulatory agencies. The signed transaction record recorded on the distributed ledger also includes the responsibility anchor hash value, a first hash value, and the current responsible party's digital identity.

[0056] Step S32: Based on the distributed ledger, generate a handover agreement with clearly defined responsibilities.

[0057] To ensure that the hash values ​​of responsibility anchors recorded in the distributed ledger serve as verification tools in actual logistics handover processes, a handover protocol is designed to upgrade the traditional, subjectively trust-based signature-based handover to an automated algorithmic handover. Through a dual verification process, the ambiguous issue of responsibility allocation is transformed into a computable mathematical problem, resolving commercial disputes arising from unclear responsibility division in multi-carrier collaborations.

[0058] When the mobile monitoring unit arrives at the responsibility handover point, the transferor and receiver of responsibility execute the aforementioned handover agreement. The transferor of responsibility refers to the party responsible for supervising the goods before the predetermined handover point. The receiver of responsibility refers to the next party responsible for taking over the supervision of the goods after the handover point. The specific process of the handover agreement is as follows: A corresponding responsibility anchor hash value is generated for the transferor of responsibility using the data storage mechanism described above, and recorded in the distributed ledger. Before formally taking over the goods, the receiving party reads the responsibility anchor hash value and the signed transaction record submitted by the transferring party, and verifies the authenticity of the signature using the transferring party's digital identity. Once the signature is confirmed to be authentic and valid, the receiving party obtains the first hash value from the transaction record and simultaneously sends a request to the transferring party to obtain the causal evidence package corresponding to the responsibility anchor hash value. After successfully obtaining the causal evidence package from the transferring party, the receiving party recalculates the first hash value using the same hash algorithm as in step S31, and records it as the verification hash value. The verification hash value is compared with the first hash value in the transaction record. If they are inconsistent, it indicates that the causal evidence package data provided by the transferring party has been tampered with.

[0059] Once the verification hash value matches the first hash value, the receiving party digitally countersigns the transaction record on the distributed ledger. This digital countersignature is a digital signature generated by the receiving party using its own private key on the transaction information of the transferring party. The digital countersignature and the transaction record together constitute a clearly defined transfer agreement, marking the formal transfer of cargo supervision responsibility from the transferring party to the receiving party. Both the receiving party and the transferring party are responsible parties and each has its own private and public keys.

[0060] Step S30 addresses the technical challenges of ambiguous responsibility boundaries, low reliability of single-source data, and gray areas of data tampering and post-event recording in complex multi-carrier collaborations, which make it difficult to meet the requirements of continuity, traceability, and verifiability in high-compliance scenarios, through causal evidence packages, distributed ledgers, and handover protocols. It achieves tamper-proof storage of cold chain data, clear division of transportation responsibilities, and reliable traceability of multi-party collaborations. Specifically, the distributed ledger ensures data authenticity and immutability; the responsibility handover protocol is executed when the mobile monitoring unit arrives at the responsibility handover point, upgrading traditional subjective signature handover to automated algorithmic handover, eliminating commercial disputes caused by unclear responsibility division, and establishing a jointly verifiable source of facts.

[0061] Example 2 This embodiment, based on Embodiment 1, provides a cold chain logistics information storage system based on multi-dimensional perception and blockchain, such as... Figure 4 As shown, it includes: Event parsing module: Used to acquire continuous data streams from heterogeneous sensors during the operation of the mobile monitoring unit, extract core data from the continuous data streams to obtain raw feature vectors, map the raw feature vectors to a pre-constructed multi-dimensional state space to obtain a set of state points, construct a hyperellipsoid and a standard response axis in the multi-dimensional state space based on the set of state points, and construct a cooperative response cone representing the event response based on the standard response axis of historical real events; acquire the perturbation trajectory of the hyperellipsoid, perform geometric matching between the perturbation trajectory and the cooperative response cone, and generate a structured event descriptor; Risk management module: used to build an anomaly assessment model for specific goods, and to define a steady-state convergence region in the anomaly assessment model that represents the goods satisfying the constraints, and to formulate dynamic safety boundaries in the steady-state convergence region based on event descriptors; Evidence storage and traceability module: It is used to build a data evidence storage mechanism based on event descriptors and dynamic security boundaries, generate a distributed ledger with reliable traceability of records, divide responsibilities based on the distributed ledger, and obtain a handover agreement with clear division of responsibilities.

[0062] In the event analysis module, the continuous data stream from heterogeneous sensors during the operation of the mobile monitoring unit is acquired, the core data of the continuous data stream is extracted to obtain the original feature vector, the original feature vector is mapped to a pre-constructed multi-dimensional state space to obtain a set of state points, a hyperellipsoid and a standard response axis are constructed in the multi-dimensional state space based on the set of state points, and a cooperative response cone representing the event response is constructed based on the standard response axis of historical real events; the perturbation trajectory of the hyperellipsoid is acquired, and the perturbation trajectory is geometrically matched with the cooperative response cone to generate a structured event descriptor, including: Step S11: Obtain a continuous data stream characterizing the state of multiple physical quantities within the mobile monitoring unit during its operation; extract the core data from the continuous data stream; map the core data to a multidimensional state space; and construct a hyperellipsoid in the multidimensional state space that adaptively adjusts its position according to environmental changes. Step S12: Define a standard response axis with temporal characteristics in the multidimensional state space, and construct a cooperative response cone with geometric shape to characterize the event response mode based on the standard response axis; Step S13: Obtain the real-time state point, detect the perturbation trajectory between the real-time state point and the hyperellipsoid, perform geometric matching between the perturbation trajectory and the cooperative response cone, and generate a structured event descriptor.

[0063] In the risk management module, the anomaly assessment model is constructed for specific goods, and a steady-state convergence region representing the goods satisfying the constraints is defined in the anomaly assessment model. Dynamic safety boundaries are then established within the steady-state convergence region based on event descriptors, including: Step S21: Construct an anomaly assessment model for specific goods; Step S22: Define a steady-state convergence region in the anomaly assessment model to represent the goods satisfying the constraints. Step S23: Based on the event descriptor, a dynamic boundary adjustment mechanism is established to define a dynamic safety boundary in the steady-state convergence domain.

[0064] In the evidence storage and traceability module, a data storage mechanism is constructed based on event descriptors and dynamic security boundaries to generate a distributed ledger for reliable traceability. Responsibilities are then assigned based on the distributed ledger, resulting in a clearly defined handover agreement, including: Step S31: Based on event descriptors and dynamic security boundaries, construct a data storage mechanism and generate a distributed ledger with reliable and traceable records; Step S32: Based on the distributed ledger, generate a handover agreement with clearly defined responsibilities.

[0065] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0066] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0067] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for storing cold chain logistics information based on multi-dimensional perception and blockchain, characterized in that, The method includes: The system acquires continuous data streams from heterogeneous sensors during the operation of the mobile monitoring unit, extracts the core data from the continuous data streams to obtain the original feature vectors, maps the original feature vectors to a pre-constructed multidimensional state space to obtain a set of state points, constructs a hyperellipsoid and a standard response axis in the multidimensional state space based on the set of state points, and constructs a cooperative response cone representing the event response based on the standard response axis of historical real events; acquires the perturbation trajectory of the hyperellipsoid, performs geometric matching between the perturbation trajectory and the cooperative response cone, and generates a structured event descriptor. An anomaly assessment model is constructed for specific goods, and a steady-state convergence region representing the goods satisfying the constraints is defined in the anomaly assessment model. Dynamic safety boundaries are formulated in the steady-state convergence region based on event descriptors. A data storage mechanism is built based on event descriptors and dynamic security boundaries, a distributed ledger with reliable and traceable records is generated, and responsibilities are divided according to the distributed ledger to obtain a handover agreement with clear responsibility division.

2. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 1, characterized in that, The method for obtaining the hyperellipsoid includes: The original feature vectors are preprocessed to obtain a set of feature vectors; Any point in the multidimensional state space is a state point, which corresponds to a set of feature vectors; When the mobile monitoring unit enters the steady-state operation mode, it collects the feature vector set corresponding to the state point according to the preset data collection window, and integrates the obtained feature vector set according to the time sequence of the data collection window to obtain the state point set. The steady-state operation mode refers to the transportation stage on the highway where the mobile monitoring unit continuously travels at a speed greater than a preset speed threshold for a time greater than a preset time threshold. A covariance matrix is ​​constructed by using a set of state points. The covariance matrix is ​​then solved to obtain multiple eigenvalues ​​and the orthogonal eigenvectors corresponding to each eigenvalue. The mean vector is calculated from the set of state points and used as the geometric center of the hyperellipsoid. Using the directions of the orthogonal eigenvectors themselves as the directions of the principal axes of the hyperellipsoid, the semi-axis lengths of each principal axis are obtained by calculating the eigenvalues, and finally the hyperellipsoid is obtained.

3. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 2, characterized in that, The methods for obtaining the event descriptor include: For real-time state points collected in real time, it is determined whether they are located inside the hyperellipsoid. If it is detected that the real-time state point has crossed the boundary of the hyperellipsoid, the motion trajectory of the state point is recorded to form a disturbance trajectory. The perturbation trajectory is geometrically matched with the cooperative response cone. When the geometric matching verification condition is met, a structured event descriptor is generated. The event descriptor is a data structure containing multiple key fields, including event type, start timestamp, duration, and confidence score.

4. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 3, characterized in that, The geometric matching includes: Geometric matching includes path length verification and point count percentage verification; The criteria for path length verification are as follows: if the path segment length of the disturbance trajectory inside the cooperative response cone is greater than the preset minimum length threshold, then the path length verification is satisfied. Here, the path segment length inside the cooperative response cone refers to the sum of the path lengths corresponding to all consecutive state points located inside the cooperative response cone among the state points constituting the disturbance trajectory. The criteria for the point percentage verification are as follows: if the proportion of the number of state points falling within the cooperative response cone of the disturbance trajectory to the total number of state points of the disturbance trajectory is greater than the preset minimum proportion threshold, then the point percentage verification is satisfied. When both path length verification and point percentage verification are satisfied, it is determined that a valid state mutation corresponding to the cooperative response cone has been identified.

5. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 4, characterized in that, The anomaly assessment model includes: The anomaly assessment model is a two-dimensional Cartesian coordinate system. The X-axis of the anomaly assessment model is defined as the cumulative heat exposure, and the Y-axis is defined as the instantaneous thermal shock rate. The physical meaning of cumulative heat exposure is the product of the degree to which the core temperature of the cargo deviates from the preset ideal temperature range and time. It is used to measure cumulative damage and is obtained by integrating the difference between the core temperature of the cargo and a preset reference temperature over time. The core temperature of the cargo is obtained in real time by temperature sensors installed inside the cargo box or at representative locations. The physical meaning of instantaneous thermal shock rate is the rate of change of the core temperature of the cargo. It is used to measure instantaneous impact and is calculated as the derivative of the core temperature of the cargo with respect to time.

6. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 5, characterized in that, The method for obtaining the dynamic security boundary includes: Extract the confidence score of the event descriptor. When the confidence score is greater than the preset trigger threshold, trigger the dynamic boundary adjustment mechanism. The dynamic boundary adjustment mechanism searches for the impact direction of the state change on the current cargo state point from a predefined event impact mapping table based on the event type in the event descriptor. The impact direction is represented by a two-dimensional vector in the anomaly assessment model, and the two-dimensional vector is defined as the risk impact direction vector. The steady-state convergence region is a closed polygonal region consisting of four vertices. To preventively shrink the steady-state convergence region, each vertex of the steady-state convergence region is translated inward along the direction opposite to the direction vector of the risk impact by a certain distance, which is the translation distance. The new polygonal region formed after the shrinkage is the dynamic safety boundary.

7. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 6, characterized in that, The cooperative response cone includes: The co-response cone is a cone-shaped deviation region centered on the standard response axis, and the radius of the co-response cone is a function that varies along the standard response axis. The function obtains multiple confirmed real state point trajectories belonging to the same state change from historical real events, performs non-linear alignment on the real state point trajectories, and obtains an aligned trajectory set. Starting from the surface of the ellipsoid, a series of ordered axis points are obtained by stepping at a preset step size; For any axis point, select all corresponding real state points from the alignment trajectory set, calculate the distance set by combining all real state points with the axis point, and calculate the arithmetic mean and standard deviation of the distance set. Then, formulate the function based on the arithmetic mean and standard deviation.

8. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 7, characterized in that, The data storage mechanism includes: The data evidence storage mechanism is triggered when a state change ends or when the mobile monitoring unit reaches a predetermined handover point. The data storage mechanism aggregates all key information related to the time period, generates a causal evidence package, and calculates the first hash value from the causal evidence package. From the event descriptor, extract the start timestamp and duration fields to define a state change time window. From the historical continuous data stream, filter out all data whose timestamps are within the state change time window. Divide the filtered data into a series of independent data blocks according to the timestamps. Calculate the root hash value for each data block. The root hash value, the first hash value, and the preset digital identity of the responsible party are concatenated and calculated to obtain the responsibility anchor hash value. The responsibility anchor hash value is digitally signed to form a signed transaction, which is then recorded on the distributed ledger.

9. The cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in claim 8, characterized in that, The method for obtaining the handover protocol includes: When the mobile monitoring unit arrives at the responsibility handover point, the handover agreement is executed between the transferring and receiving parties. The transferring party refers to the party responsible for supervising the goods before the predetermined handover point, and the receiving party refers to the next party responsible for taking over the supervision of the goods after the handover point. Generate a corresponding responsibility anchor hash value for the transferor and record it in the distributed ledger. Before officially taking over the goods, the receiving party reads the responsibility anchor hash value submitted by the transferor and the signed transaction record to verify the authenticity of the transferor's signature. Once the signature is confirmed to be genuine and valid, the receiving party obtains the first hash value from the transaction record, recalculates the first hash value for the causal evidence package of the transferring party, records it as the verification hash value, and compares the verification hash value with the first hash value in the transaction record. Once the verification hash value matches the first hash value, the receiving party digitally countersigns the transaction record on the distributed ledger. The digital countersignature and the transaction record together constitute a handover agreement with a clear division of responsibilities.

10. A cold chain logistics information storage system based on multi-dimensional perception and blockchain, used to implement the cold chain logistics information storage method based on multi-dimensional perception and blockchain as described in any one of claims 1-9, characterized in that, The system includes: Event parsing module: Used to acquire continuous data streams from heterogeneous sensors during the operation of the mobile monitoring unit, extract core data from the continuous data streams to obtain raw feature vectors, map the raw feature vectors to a pre-constructed multi-dimensional state space to obtain a set of state points, construct a hyperellipsoid and a standard response axis in the multi-dimensional state space based on the set of state points, and construct a cooperative response cone representing the event response based on the standard response axis of historical real events; acquire the perturbation trajectory of the hyperellipsoid, perform geometric matching between the perturbation trajectory and the cooperative response cone, and generate a structured event descriptor; Risk management module: used to build an anomaly assessment model for specific goods, and to define a steady-state convergence region in the anomaly assessment model that represents the goods satisfying the constraints, and to formulate dynamic safety boundaries in the steady-state convergence region based on event descriptors; Evidence storage and traceability module: It is used to build a data evidence storage mechanism based on event descriptors and dynamic security boundaries, generate a distributed ledger with reliable traceability of records, divide responsibilities based on the distributed ledger, and obtain a handover agreement with clear division of responsibilities.

Images