A cross-modal space-time rhythm coordination scheduling method and system

Abstract

The present application belongs to the technical field of logistics scheduling and supply chain management, and relates to a cross-modal transportation space-time rhythm collaborative scheduling method and system, comprising: receiving a behavior model of a to-be-scheduled transportation unit, which contains an encrypted probability state transition graph and a plaintext behavior list file; analyzing the list file for environment verification, generating a data access credential in a sandbox isolation area; decrypting the encrypted transition graph using a hub node private key, and combining virtual deduction parameters with the credential to build a deduction parameter set; inputting the parameter set into the transition graph, and obtaining the arrival time probability distribution matrix of the unit through graph traversal algorithm deduction; obtaining the probability distribution matrix of each unit in the region, and calculating a global scheduling scheme based on a multi-objective evaluation function, and generating a scheduling instruction for issuing and allocating resources accordingly. The present application solves the problem of how to ensure the data security of each transportation subject while realizing the deep collaboration and dynamic scheduling of cross-modal transportation resources.

Description

Technical Field

[0001] This invention belongs to the technical field of logistics scheduling and supply chain management, and relates to a cross-modal transportation spatiotemporal rhythm collaborative scheduling method and system. Background Technology

[0002] In modern logistics systems, the collaborative operation of large transportation hubs is an important research direction. Hub nodes aggregate numerous independent transportation units from different modes of transport, such as road and waterway, each following the scheduling logic of its respective operator. To improve the overall throughput capacity of the hub, spatiotemporal collaborative scheduling of heterogeneous transportation units is necessary. In multi-party collaborative scenarios, the real-time dynamics and decision-making logic of transportation units constitute core data for the operator. Directly sharing this information poses a risk of data leakage and increases the technical difficulty of achieving deep collaboration.

[0003] Currently, the industry primarily uses centralized information sharing platforms or standardized message exchange mechanisms to achieve scheduling collaboration. In centralized platform solutions, each transport provider uploads raw data such as the real-time location and estimated arrival time of its transport units to a unified database. The dispatch center then uses this aggregated data and a unified algorithm model to formulate a global scheduling plan. In information exchange mechanism solutions, parties collaborate through predefined interfaces and data formats, enabling point-to-point reservations and status updates.

[0004] The aforementioned traditional methods have certain limitations when dealing with complex transportation networks. On the one hand, centralized platform solutions require all parties to report raw data on a large scale, which easily increases the risk of exposing data privacy, such as operational strategies. Furthermore, the unified algorithm model is not easy to finely characterize the operational characteristics of each independent transporter, thus limiting its applicability to non-standard situations. On the other hand, the information exchange dimensions based on standardized messages are relatively limited, usually only providing static reservation information or coarse-grained status updates. The means for joint extrapolation and risk assessment of future uncertainties are relatively simple, which limits the system's dynamic adjustment space in response to emergencies. Summary of the Invention

[0005] In a first aspect, the present invention provides a cross-modal transportation spatiotemporal rhythm cooperative scheduling method, comprising the following steps: S1. Receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probabilistic state transition diagram and a plaintext behavior list file; the probabilistic state transition diagram is used to represent the discrete state and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior list file contains the scene input mode, runtime environment signature and resource request token parameters. S2. Parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. S3. Use the private key of the target hub node to decrypt the encrypted probability state transition diagram to obtain the decrypted probability state transition diagram. Combine the preset virtual inference parameters with the data access credentials to construct an inference parameter set. S4. Input the deduction parameter set into the probability state transition diagram, and perform deduction calculations through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; S5. Obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area. Based on the preset multi-objective evaluation function, aggregate and calculate each arrival time probability distribution matrix to obtain a global scheduling scheme. Generate scheduling instructions based on the global scheduling scheme to allocate hub resources and issue them to the corresponding transportation units.

[0006] A further aspect of the present invention involves constructing a probabilistic state transition diagram, comprising the following steps: Acquire multi-source operational data of the target transportation network and transform positioning data and status parameters of different formats into a standardized operational dataset with a unified spatiotemporal reference; Based on the standardized operational dataset, the stationing status of transportation units during historical operations is extracted as the reference node of the directed graph. The conditional transition probability logic between baseline nodes is encoded into executable function code with input parameters; The executable function code is mapped to the edges between the corresponding baseline nodes to generate a probabilistic state transition graph.

[0007] A further aspect of this invention involves extracting the stationary position status of transportation units during historical operations as the baseline node of a directed graph, based on a standardized operational dataset, and includes the following steps: Load the coordinates of historical trajectory points from the standardized runtime dataset; Perform a density-based spatial clustering algorithm to calculate the spatial distance between the coordinates of historical trajectory points; Core objects whose spatial distance is less than the preset neighborhood radius and whose number of data points in the neighborhood exceeds the preset minimum sample size are aggregated into geospatial clusters. Extract the centroid of each geospatial cluster and define the centroid as a reference node with a unique identifier.

[0008] A further aspect of the present invention generates a transport unit behavior model, comprising the following steps: Based on the input requirements of the executable function code, extract and create a behavior manifest file; The transporter's private key on the local side of the corresponding transport unit is invoked to digitally sign the constructed probabilistic state transition diagram; The public key of the target hub node is used to perform asymmetric encryption on the probabilistic state transition graph with digital signature to obtain the encrypted probabilistic state transition graph. The behavior list file and the encrypted probabilistic state transition diagram are linked and concatenated to generate a transportation unit behavior model.

[0009] A further aspect of the present invention generates data access credentials, comprising the following steps: Extract the runtime environment signature from the behavior manifest file and perform environment compatibility verification; After the compatibility check is passed, the resource request token parameter in the behavior list file is read and parsed to extract the list of target fields required for the current simulation. Within the sandbox isolation operation zone, data access credentials that are only available for the current simulation round are generated based on the target field list; Allocate the corresponding target computing resources to the parsing process in the sandbox isolated operation area.

[0010] A further aspect of this invention involves constructing a set of projection parameters, including the following steps: The encrypted probabilistic state transition diagram is decrypted using the private key of the target hub node to obtain the decrypted probabilistic state transition diagram and digital signature. The public key of the corresponding transporter is retrieved to compare and verify the digital signature. If the verification is successful, the decrypted probability state transition diagram is confirmed to be valid. Generate virtual simulation data containing expected berth allocation information and on-site facility status parameters; The virtual simulation data and data access credentials are merged by field selection to form a simulation parameter set.

[0011] A further aspect of the present invention generates an arrival time probability distribution matrix, comprising the following steps: Determine the starting node of the probabilistic state transition diagram; Load the derivation parameter set into the execution memory area of ​​the executable function code corresponding to the edge connected to the starting node; After the data access credentials are verified, the executable function code is triggered to read external real-time dynamic status data and time series prediction data, perform logical operations, and output the target conversion probability value to be transferred to the downstream node. The concurrent traversal engine is started, and random parallel pathfinding traversal is performed in the probabilistic state transition graph structure in combination with the destination-oriented path constraints. The arrival time probability distribution matrix is ​​statistically calculated and generated.

[0012] A further aspect of this invention involves aggregating and calculating the probability distribution matrices of each arrival time based on a preset multi-objective evaluation function to obtain a global scheduling scheme, including the following steps: Collect and align the arrival time probability distribution matrices received within the target scheduling area to form a set of arrival time probability distribution matrices; The arrival time probability distribution matrix is ​​input in parallel into a preset multi-objective evaluation function to calculate the resource idle loss value and connection penalty risk value corresponding to different scheduling assumption schemes. The solution that minimizes overall loss and risk is calculated based on the set optimization strategy and used as the global scheduling scheme.

[0013] A further aspect of this invention involves calculating the resource idle loss value and the connection penalty risk value corresponding to different scheduling assumptions, including the following steps: The expected time and resource idle loss value are calculated by performing a weighted summation operation on the time window for allocating specific tasks in the scheduling hypothesis scheme and the corresponding arrival time probability distribution matrix. Identify the time requirement threshold for downstream connection nodes in the current transportation chain; Based on the arrival time probability distribution matrix, the penalty probability for arriving later than the time requirement threshold is extracted, and the penalty probability is multiplied by a preset penalty coefficient to obtain the connection penalty risk value.

[0014] Secondly, this invention provides a cross-modal transportation spatiotemporal rhythm collaborative scheduling system, comprising the following modules: The model receiving module is used to receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probabilistic state transition diagram and a plaintext behavior list file; the probabilistic state transition diagram is used to represent the discrete state and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior list file contains the scene input mode, runtime environment signature and resource request token parameters. The environment verification and authorization module is used to parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. The parameter set construction module is used to decrypt the encrypted probability state transition diagram using the private key of the target hub node to obtain the decrypted probability state transition diagram, and to combine the preset virtual inference parameters with the data access credentials to construct the inference parameter set. The deduction and calculation module is used to input the deduction parameter set into the probability state transition diagram, and perform deduction and calculation through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; The scheduling instruction generation module is used to obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area, aggregate and calculate each arrival time probability distribution matrix based on a preset multi-objective evaluation function to obtain a global scheduling scheme; generate scheduling instructions based on the global scheduling scheme to allocate hub resources and send them to the corresponding transportation units.

[0015] In summary, the present invention has the following beneficial technical effects: 1. This invention abstracts the operational rules of a transportation unit into a probabilistic state transition diagram containing conditional transition probability logic, and attaches the compiled executable function code to the edges of the diagram, thus modularly encapsulating the dynamic behavior pattern of the transportation unit. This feature enables the system to deduce the response logic of the transportation unit based on changes in external conditions. Compared to simply transmitting static estimated arrival time data, it increases the dimension of characterizing dynamic behavioral features and realizes the transformation from state result output to model logic sharing.

[0016] 2. This invention constructs a two-layer transportation unit behavior model comprising plaintext lists and encrypted graph data, combining asymmetric encryption and digital signature technologies to achieve the separation of environmental verification and core logic. Coupled with a sandbox isolation environment and a temporary data access credential mechanism, a controlled data interaction channel can be established between hub nodes and transport parties. This mechanism ensures the integrity and legitimacy of model data without directly exposing underlying raw operational data and core algorithms, helping to solve data privacy protection issues in cross-trust domain collaboration.

[0017] 3. This invention combines virtual simulation data with temporary data access credentials to form a simulation parameter set, which is then input into a probability state transition diagram to drive graph traversal calculations. This enables the generation of an arrival time probability distribution matrix for transportation units by combining real-time dynamic data from the hub side. This calculation method outputs the arrival probability distribution under different time windows through simulation, replacing the prediction mode of a single time node. It provides quantitative probabilistic reference data for subsequent scheduling decisions, helping the system to more objectively assess the spatiotemporal uncertainties in future transportation processes.

[0018] 4. This invention aggregates the arrival time probability distribution matrix of each transportation unit within the target scheduling area and introduces a multi-objective evaluation function that includes resource idle loss and connection penalty risk for optimization calculation, generating a global scheduling scheme. This scheduling mechanism quantifies different operational influencing factors into a comprehensive risk assessment index, calculates resource allocation solutions from the perspective of overall system scheduling, and generates specific entry windows and other scheduling instructions for execution, thereby improving the foresight and multi-party collaboration capabilities of cross-modal hub resource scheduling. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0020] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0021] Figure 2 Structural schematic diagrams of embodiments of this application are disclosed. Detailed Implementation

[0022] The following is in conjunction with the appendix Figures 1-2 A preferred description of the present invention is provided below.

[0023] See attached document Figure 1 This invention proposes a cross-modal transportation spatiotemporal rhythm cooperative scheduling method, comprising the following steps: S1. Receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probabilistic state transition diagram and a plaintext behavior list file; the probabilistic state transition diagram is used to represent the discrete state and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior list file contains the scene input mode, runtime environment signature and resource request token parameters. S2. Parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. S3. Use the private key of the target hub node to decrypt the encrypted probability state transition diagram to obtain the decrypted probability state transition diagram. Combine the preset virtual inference parameters with the data access credentials to construct an inference parameter set. S4. Input the deduction parameter set into the probability state transition diagram, and perform deduction calculations through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; S5. Obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area. Based on the preset multi-objective evaluation function, aggregate and calculate each arrival time probability distribution matrix to obtain a global scheduling scheme. Generate scheduling instructions based on the global scheduling scheme to allocate hub resources and issue them to the corresponding transportation units.

[0024] In one embodiment of the present invention, constructing a probabilistic state transition graph includes the following steps: Acquire multi-source operational data of the target transportation network, and transform positioning data and status parameters of different formats into a standardized operational dataset with a unified spatiotemporal reference; based on the standardized operational dataset, extract the stationary position status of transportation units during historical operations as the reference node of the directed graph; The conditional transition probability logic between baseline nodes is encoded into executable function code with input parameters; The executable function code is mapped to the edges between the corresponding baseline nodes to generate a probabilistic state transition graph.

[0025] Based on a standardized operational dataset, the stationary position status of transportation units during historical operations is extracted as the baseline node of a directed graph, including the following steps: Load the coordinates of historical trajectory points in the standardized running dataset; execute a density-based spatial clustering algorithm to calculate the spatial distance between the coordinates of historical trajectory points; aggregate core objects whose spatial distance is less than the preset neighborhood radius and whose number of data points in the neighborhood exceeds the preset minimum sample size into geospatial clusters; extract the centroid of each geospatial cluster and define the centroid as a reference node with a unique identifier.

[0026] Specifically, in the process of generating the transportation unit behavior model, the raw, multi-source, heterogeneous transportation data is processed and modeled as a probabilistic state transition graph containing dynamic logic. This step is performed by a model building processor deployed on a cloud server or edge computing node.

[0027] The model building processor accesses and collects heterogeneous signal streams from various transportation units within the target transportation network through its built-in data acquisition interfaces. These interfaces follow preset communication protocols, such as calling the application network interface to obtain structured location stream data containing longitude, latitude, timestamps, and speed information provided by the fleet management system, or parsing network transmission protocols to obtain standardized message statements issued by the Automatic Identification System (AIS). Simultaneously, the processor queries status parameters associated with the location data from business systems such as Transportation Management System (TMS) or Warehouse Management System (WMS) via a database connection component, including cargo loading status or waybill numbers. Since the collected data sources have varying formats and inconsistent spatiotemporal references, the processor then calls the spatiotemporal calibration module to uniformly convert all data timestamps to Coordinated Universal Time (UTC) and geographic coordinates to the WGS-84 coordinate system. After data cleaning and formatting, a structured, standardized operational dataset is generated.

[0028] The model building processor uses the standardized runtime dataset generated in the preceding steps to train the state transition logic. The processor loads the coordinates of all historical trajectory points of a specific transport unit or group of units within the preset time period from this dataset.

[0029] To extract typical discrete location states with significant clustering characteristics in geospatial space, the processor executes the density-based spatial clustering algorithm DBSCAN. This algorithm calculates the Euclidean distance between trajectory points, categorizing them into neighborhoods with spatial distances less than a preset radius. And the number of data points in the neighborhood exceeds the minimum sample size. The core objects and their density-reachable points are aggregated into clusters. The centroid of each successfully formed cluster, i.e., the average geographic coordinates of all points within the cluster, is defined as a baseline node and assigned a unique node identifier, such as a warehouse. ,port All these identified baseline nodes together form the node set of the subsequent directed graph.

[0030] The model building processor logically encodes the transfer relationships between baseline nodes. For any pair of baseline nodes with historical direct paths, the processor analyzes the conditional variables affecting the transfer time and success rate between them. These variables originate from business status parameters in the standardized operational dataset or external data sources such as real-time weather and traffic congestion indices. The processor uses these conditional variables as input parameters to construct a pure function for calculating conditional transfer probabilities. This pure function encapsulates specific business logic; for example, a function with traffic congestion indices as input preferably uses a piecewise function or a small decision tree model internally, outputting a probability value between 0 and 1.

[0031] To achieve cross-platform compatibility and execution security, the pure function is compiled by a bytecode compiler into platform-independent bytecode, which becomes the aforementioned executable function code. The processor then appends this code to the directed edges connecting the two corresponding base nodes. By traversing all connected node pairs and repeating this process, a complete probabilistic state transition graph is finally generated, consisting of the base nodes and the edges with the attached executable function code. This graph is stored in a serialized structured graph format for later use.

[0032] It should be noted that the standardized operational dataset is a time-ordered collection of records. Each record contains at least a transport unit ID, a timestamp calibrated to Coordinated Universal Time (UTC), geographic location coordinates based on the WGS-84 coordinate system, and relevant operational status data. Typical discrete location states are key locations in physical space where transport units frequently stop or intersect, such as warehouses, ports, terminals, or transportation hubs. The neighborhood radius used in the DBSCAN algorithm for their identification process... and minimum sample size These are key parameters.

[0033] Neighborhood radius The value is set based on the physical dimensions of a typical logistics terminal, for example, within a range of 50m to 200m, to ensure complete coverage of an independent functional area. Minimum sample size The number of units is determined based on the sampling frequency of the trajectory data and the typical dwell time of the units. For example, for GPS data with a sampling frequency of once per minute, if the goal is to identify areas with an average dwell time exceeding 15 minutes, then... It can be set to greater than 15. Executable function code is an executable code fragment that does not depend on external state. For the same input parameters, it always returns the same output result, ensuring the determinism and reproducibility of the deduction process. Its input parameters come from the virtual deduction scenario injected in subsequent steps, ensuring the model's responsiveness to future scenarios.

[0034] For example, the transportation unit is a truck, and some of its historical trajectory data has been stored in a standardized operational dataset. The model building processor extracts the set of coordinates of the truck's trajectory points within the target operating area. To extract its typical stopping points, the processor sets the neighborhood radius for the DBSCAN algorithm. Minimum sample size: 100m There are 20 data points. After the algorithm runs, at the first coordinate point... A dense cluster of points was detected nearby. The processor calculated the centroid coordinates of this cluster as follows: And mark it as a baseline node. This corresponds to the first logistics node, such as the originating warehouse. Similarly, at the second coordinate point... Another reference node was identified nearby. This corresponds to the second logistics node, such as the destination hub terminal. Historical data shows that this truck has a large number of origins from... arrive The processor analysis revealed that congestion on the main road connecting the two baseline nodes was a key factor affecting transportation time. Therefore, in order to construct... arrive The processor defines a pure function that takes a traffic congestion index `traffic_index` as input, with a value ranging from 0 to 1. The function's logic is as follows: if `traffic_index` is less than 0.5, it returns a probability of 0.95; if `traffic_index` is between 0.5 and 0.8, it returns a probability of 0.7; and if `traffic_index` is greater than 0.8, it returns a probability of 0.4. This function is then compiled into executable code. The resulting probabilistic state transition graph is represented by a specific data structure. Its node set contains the coordinates of the starting and destination nodes, its edge set records the relationships between the starting and ending points, and the compiled executable function code is attached to the edges as logical attributes. This structure fully represents the discrete states of the truck from the warehouse to the dock and the dynamic transition logic between them.

[0035] In one embodiment of the present invention, generating a transport unit behavior model includes the following steps: Based on the input requirements of the executable function code, extract and create a behavior manifest file; The transporter's private key on the local side of the corresponding transport unit is invoked to digitally sign the constructed probabilistic state transition diagram; The public key of the target hub node is used to perform asymmetric encryption on the probabilistic state transition graph with digital signature to obtain the encrypted probabilistic state transition graph. The behavior list file and the encrypted probabilistic state transition diagram are linked and concatenated to generate a transportation unit behavior model.

[0036] After completing the construction of the probabilistic state transition diagram, the model encapsulation process begins. This process aims to encapsulate the probabilistic state transition diagram generated in the previous step, which contains the core operational logic of the transporter, into an independent transport unit behavior model that can securely interact with external systems while protecting its own trade secrets through a rigorous cryptographic process.

[0037] First, the encapsulation processor receives and parses the probabilistic state transition graph. The processor traverses each edge of the graph, performing static analysis on the executable function code attached to each edge. By analyzing the function signature and metadata of the bytecode, the processor extracts the types and structures of the external input parameters required for its execution, forming a scene input pattern definition.

[0038] The processor records the compiler version, dependent libraries, and target runtime environment specifications used to compile the bytecode, and calculates the cryptographic hash value of this specification information, such as SHA-256, to generate a unique runtime environment signature.

[0039] In addition, the processor identifies data field identifiers that it needs to query from the target hub node in real time during execution from bytecode annotations or associated configuration files, and summarizes these identifiers into a resource request token parameter list.

[0040] The processor extracts and integrates the scene input patterns, runtime environment signatures, and resource request token parameters from all executable function code in the diagram, organizes and serializes them into a structured plaintext behavior list file, which is usually in JSON or XML format.

[0041] The encapsulation processor performs encryption and signing operations on the core graph data. The processor first retrieves the transporter's private key corresponding to the current transport unit from the local security keystore; this private key is the unique digital identity credential of the transport unit owner. The processor serializes the complete probabilistic state transition graph into binary format, and then calls an asymmetric encryption algorithm library such as OpenSSL to perform a digital signature operation on the binary data block using the transporter's private key. For example, it uses a preset asymmetric digital signature algorithm to generate a digital signature byte string separate from the graph data. This signature ensures the integrity of the graph data, i.e., tamper-proof, and the non-repudiation of its origin.

[0042] The processor obtains the target hub node's scheduling public key from a public directory service or through a pre-exchange mechanism. This public key is publicly disclosed by the target hub node for receiving encrypted information. The processor concatenates the serialized graph data with the generated digital signature, and then uses the scheduling public key to encrypt the entire concatenated data block using a preset asymmetric encryption algorithm. After encryption, the original graph structure and logic become an unreadable binary data block to an external observer; that is, a probabilistic state transition graph that appears as a black box after encryption.

[0043] Finally, the encapsulation processor completes the data encapsulation, generating the final transport unit behavior model. The processor creates a new data container with a two-layer data structure. The first layer of the container, or public layer, stores the behavior manifest file generated in the preceding steps. The second layer of the container, or private layer, stores the encrypted probabilistic state transition diagram data.

[0044] By constructing specific structured data objects, the plaintext behavior list file content is used as the first field, and the encrypted graph data converted using a preset encoding format, such as Base64, is used as the second field. This concatenation generates a structurally independent transportation unit behavior model with clearly defined rights and responsibilities. This model, as a whole, can be securely transmitted between different trust domains. Its plaintext portion is used for interactive negotiation and environment verification, while its encryption core protects the transportation party's core intellectual property.

[0045] The behavior manifest file is a data contract defining the interface for calling the cryptographic model. The scenario input pattern outlines the data interface for the external dynamic parameters required for the model's runtime. The runtime environment signature ensures the cryptographic model executes in a verified interpreter. The resource request token parameter specifies the range of data the model needs to access in a single simulation. The transport unit's private key and the target hub node's public key constitute an asymmetric encryption key pair system.

[0046] In a specific application scenario, the packaging processor receives data containing nodes. , The processor analyzes and detects that the congestion index floating-point number is required as input, and generates a scenario input pattern accordingly. The processor calculates the hash value of the current interpreter environment specification as the runtime environment signature and identifies the corresponding data extraction fields to generate resource request token parameters. These three items are written into a structured behavior manifest file. Next, the processor retrieves the transporter's private key to sign the serialized graph; then, it obtains the target hub node's public key to encrypt the concatenated data. Finally, a transport unit behavior model containing the aforementioned manifest file and encrypted data is generated and sent to the target hub node's scheduling system.

[0047] In one embodiment of the present invention, generating data access credentials includes the following steps: Extract the runtime environment signature from the behavior manifest file and perform environment compatibility verification; after the compatibility verification is passed, read the resource request token parameter in the behavior manifest file and parse and extract the target field list required for the current simulation; within the sandbox isolated runtime area, generate data access credentials that are only available for the current simulation round based on the target field list; allocate the corresponding target computing resources to the parsing process in the sandbox isolated runtime area.

[0048] Specifically, after the target hub node receives the transport unit behavior model, the authorized interpreter configured in the target hub node begins to execute this step. This interpreter is a software application deployed on the hub node server, whose task is to securely parse and prepare for execution of the externally received transport unit behavior model.

[0049] The interpreter first receives the transport unit behavior model, which contains a two-layer data format, through its network interface. Upon receiving the data, the interpreter does not directly process the encrypted private layer data; instead, it prioritizes parsing the behavior manifest file containing the plaintext state of the public layer. It then invokes the built-in JSON or XML parser to read the key-value pairs from the manifest file. The primary verification action is environment verification; the interpreter extracts the runtime environment signature from the behavior manifest file, which is a hash value representing the expected runtime environment of the model, such as the interpreter version, operating system, and set of dependent libraries.

[0050] Meanwhile, the interpreter calculates the real-time hash value of its operating environment based on various specifications. Through rigorous binary comparison, the interpreter verifies whether these two hash values ​​are completely identical. If they are inconsistent, the compatibility check fails, the interpreter refuses to execute the model, records a security alert, and the process aborts.

[0051] After the compatibility check passes, indicating that the current interpreter environment meets the model's execution requirements, the process enters the resource authorization phase. The authorized interpreter continues to parse the behavior manifest file, reading the resource request token parameter. This parameter is a list of explicit strings, each item precisely corresponding to a unique identifier of a data field that can be provided externally in the hub-side database or real-time data stream. By parsing this list, the interpreter obtains a list of explicit fields of the hub-side real-time dynamic state data and time-series prediction data required for the current inference flow, thus knowing all the data items that the model intends to access in subsequent executions.

[0052] The authorized interpreter executes subsequent operations within its sandboxed execution zone. This sandboxed execution zone is a restricted execution environment created using operating system-level virtualization technologies such as containers or process-level security policies. It effectively isolates the execution of model code, preventing it from making unauthorized access to or affecting the host system.

[0053] Within the sandbox isolation zone, the interpreter dynamically generates structured data access credentials that are only available for the current simulation round, based on the target field list obtained in the previous step. The data access credentials are configured with the session ID corresponding to this session, the details of the fields authorized for access, and the expiration timestamp.

[0054] Simultaneously, the interpreter communicates with the resource scheduling module of the hub node to pre-lock computational cycle resources for the current sandbox process based on model complexity and historical execution data. This includes allocating fixed CPU time slices and memory quotas to ensure the model's performance and real-time capabilities during subsequent decryption and deduction processes. At this point, the environment verification and pre-authorization process is complete, and the generated data access credentials are stored in the sandbox's memory, awaiting invocation in the next step.

[0055] The authorized interpreter configured on the target hub node constitutes the isolated call component between the external model and the internal system. Verification of the runtime environment signature in the behavior manifest file and the interpreter's own environment ensures that the model logic executes within a matching environment. The resource request token parameter is designed according to the principle of least privilege in information security, requiring the model to declare its data requirements in advance. The sandboxed isolated runtime area of ​​the interpreter provides an isolated runtime environment for model execution, with strict restrictions on its memory, computation, network, and file system access. Data access credentials are temporary tokens generated and used internally within the sandbox. They serve as credentials for calling the hub's internal data interfaces in subsequent steps, and their lifespan is bound to a single simulation task, becoming invalid immediately upon task completion, thus preventing permission overreach.

[0056] In a specific execution, the authorized interpreter running on the scheduling system of the target hub node receives the transport unit behavior model. The interpreter first parses the plaintext manifest field to compare it with the hash value of the environment signature. If the two match perfectly, the compatibility check passes. Next, the interpreter reads the congestion data resource request parameters that need to be invoked. Subsequently, within the sandbox isolation zone, the interpreter generates a temporary data access credential for this simulation session. This credential object encapsulates the session identifier, the specific set of congestion data fields authorized for access, and dynamically calculates and sets a lifecycle expiration timestamp. Simultaneously, the interpreter requests and locks CPU and memory quotas exclusively for this session from the host resource manager, and the generated credential object is stored in the restricted memory address of the sandbox for subsequent invocation.

[0057] In one embodiment of the present invention, constructing the inference parameter set includes the following steps: The encrypted probabilistic state transition diagram is decrypted using the private key of the target hub node to obtain the decrypted probabilistic state transition diagram and digital signature; the public key of the corresponding transporter is retrieved to compare and verify the digital signature. If the verification is successful, the decrypted probabilistic state transition diagram is confirmed to be valid; virtual simulation data containing expected berth allocation information and on-site facility status parameters is generated; the virtual simulation data and data access credentials are merged by field and assembled into a simulation parameter set.

[0058] The inference controller in the authorized interpreter securely reconstructs the model's internal logic and constructs a complete data environment for simulation. First, the inference controller accesses the transport unit behavior model, which has passed compatibility checks, and extracts the encrypted private layer data, namely the probabilistic state transition graph in an encrypted state. The controller then invokes the hub private key deployed within the target hub node's own secure container, such as a Hardware Security Module (HSM) or Trusted Execution Environment (TEE). Using this private key, the controller initiates an asymmetric decryption operation, such as employing the RSA-OAEP algorithm, to decrypt the encrypted probabilistic state transition graph. The output of this operation is a plaintext binary data block containing the original serialized graph data and an accompanying digital signature.

[0059] After successful decryption, to ensure the data's source is trustworthy and it has not been tampered with during transmission, the inference controller immediately performs a signature comparison verification. The controller retrieves the public key of the corresponding transporter obtained from the behavior manifest file in the aforementioned steps.

[0060] The decrypted binary data block is separated into two parts: the original probabilistic state transition graph data and a digital signature generated by the transporter's private key. The controller uses a digital signature algorithm matching the one used during signature generation, such as ECDSA, to calculate a verification digest of the probabilistic state transition graph data using the transporter's public key, and compares it with the decrypted digital signature. If the comparison matches, the integrity and legitimacy of the graph data are confirmed, the decrypted probabilistic state transition graph is marked as trustworthy, and loaded into memory for execution. If the comparison fails, the process immediately terminates and a security alert is generated.

[0061] After the probabilistic state transition diagram is validated, the simulation controller begins assembling the data environment required for the simulation. Internally, the controller constructs virtual simulation data for this simulation iteration. This data, serving as simulation parameters, includes the expected berth allocation scheme {berth_id:“P07”} and related on-site facility status parameters {crane_status:“AVAILABLE”,yard_density:0.75}. This data structure strictly adheres to the scenario input pattern defined in the behavior manifest file. The controller merges this dynamically generated virtual simulation data with data access credentials, forming a complete simulation parameter set, which will serve as the input parameters driving the model in the next step.

[0062] The private key within the target hub node's own secure container serves as the hub node's security credential, paired with the publicly released public key for decrypting information. The corresponding transporter's public key is provided by the transport unit and declared in the behavior manifest file, used to verify the digital signature generated by the transporter's private key. The virtual simulation data represents one or more scheduling simulation parameters for the future. The simulation parameter set is an instantaneous data aggregate that binds the scheduling simulation conditions reflected in the virtual simulation data and the execution data access permissions reflected in the data access credentials, forming the information package that drives the probabilistic state transition diagram for simulation.

[0063] In one specific embodiment, the inference controller of the target hub node decodes the received encrypted model data block, using the hub's private key in the hardware security module to decrypt it into a plaintext data block containing a serialized graph and a signature. Next, it retrieves the public key of the corresponding trucking company to perform signature verification. Successful verification indicates the probabilistic state transition graph is trustworthy. Subsequently, the controller decides to simulate a set high-congestion scenario, thus constructing an object containing congestion index parameters as virtual inference data. Finally, the controller merges and combines the fields of this inference data with the structure containing session authorization credentials generated in the previous step, forming a final inference parameter set containing multi-dimensional feature descriptions. This inference parameter set is then passed to subsequent steps to drive graph traversal calculations.

[0064] In one embodiment of the present invention, generating an arrival time probability distribution matrix includes the following steps: Determine the starting node of the probabilistic state transition graph; load the inference parameter set into the execution memory area of ​​the executable function code corresponding to the edge connected to the starting node; after verifying the data access credentials, trigger the executable function code to read external real-time dynamic state data and time-series prediction data to perform logical operations, and output the target transition probability value to the downstream node; start the concurrent traversal engine, combine the endpoint-oriented path constraints in the probabilistic state transition graph structure to perform random parallel pathfinding traversal, and statistically calculate the arrival time probability distribution matrix.

[0065] The multi-threaded concurrent traversal engine is launched to perform a random parallel pathfinding traversal covering the entire probabilistic state transition graph structure, and to aggregate and calculate the arrival time probability distribution matrix, including: Initialize multiple virtual iterator instances within the multi-threaded concurrent traversal engine, starting from the initial node; At each node, a weighted random selection is made based on the exact conversion probability value calculated by the executable function code to construct an uninterrupted traversal path; Aggregate statistics on the arrival frequency of all uninterrupted traversal paths to the terminal node; The arrival time frequency is divided by the total number of valid virtual traversal instances that successfully reach the terminal node to convert it into a standardized probability value, which is then fitted into an arrival time probability distribution matrix.

[0066] Within a secure and controlled simulation loop, the decrypted model logic is executed to generate predictions about the future. First, the graph traversal controller determines the starting node of the probabilistic state transition graph. Subsequently, the controller transfers the generated set of inference parameters to the execution memory area of ​​the executable function code corresponding to the edge connected to that starting node for loading and execution.

[0067] Once the inference parameter set is loaded, the execution of the executable function code is triggered. As the first security check during its execution, the bytecode internal logic accesses and parses the data access credentials contained in the passed inference parameter set. When the bytecode logic needs to obtain dynamic status data and time-series prediction data from the hub side, such as a real-time congestion index or future trend prediction value declared by a resource request token parameter, it uses these credentials to initiate a request to the interpreter's internal data interface.

[0068] Upon receiving a request, the interpreter's data interface module rigorously authenticates and verifies the validity of the data access credentials, including their session ID, expiration date, and whether the requested field is in the authorization list. Only after successful authentication will the interface return the requested real-time and predicted data values.

[0069] After acquiring all necessary input parameters—virtual simulation data and authorized real-time and prediction data—the core internal logic operations of the bytecode, such as a pre-trained decision tree or mathematical function, are executed. The result is a precise numerical value representing the transition probability from the current node to the downstream node connected by that edge.

[0070] To explore the complex spatiotemporal path possibilities inherent in the entire graph model, the graph traversal controller initiates a multi-threaded concurrent traversal engine. This engine employs the Monte Carlo method, simulating the behavior of transport units by performing large-scale random sampling.

[0071] Specifically, the engine initializes and concurrently executes a large number of, for example, tens of thousands to millions of, independent virtual traversal instances. Each traversal starts from the starting node, and at each node, it first eliminates invalid reverse edges according to the global optimization heuristic rules for reaching the target node. Then, it performs a weighted random selection based on the transformation probability values ​​calculated by the executable function code on the remaining outgoing edges of the current node to determine the next step.

[0072] Additionally, each edge is accompanied by a travel time calculated based on the input parameters. As the traverser moves through the graph, it accumulates the travel time of each path traversed. This process continues until the traverser reaches a predefined terminal node or terminates due to the lack of a valid path. The entire engine performs one random parallel pathfinding traversal covering the entire graph structure. After all traversals have finished running, the engine statistically analyzes all uninterrupted traversal paths that successfully reached the terminal node, aggregates their respective final arrival times, and calculates the arrival frequency within each preset time window.

[0073] By dividing the frequency of each time window by the number of valid traversers that eventually successfully reach the termination node, the frequency distribution is fitted and transformed into a standardized probability distribution, forming the arrival time probability distribution matrix of the transport unit under the current simulation conditions.

[0074]

[0075] In the formula, The elements in the arrival time probability distribution matrix represent the arrival times of the transport units under the current simulation conditions. The termination node, and the arrival time falls on the . Conditional probability within a time window. It is a traversal that starts from the initial node and eventually successfully reaches the th node. The termination node is located at the th termination node and the total time falls within the th th . The total number of uninterrupted traversal paths within a time window. This represents the total number of valid virtual traversal instances that successfully reach any terminal node in the Monte Carlo simulation. This formula normalizes the discrete random traversal results, ensuring the sum of the probabilities of the valid distributions equals 1, thus yielding a quantized probability distribution.

[0076] The starting node is the entry point for the traversal algorithm in graph theory, uniquely determining the starting point of the deduction. The execution memory area is an isolated memory space dynamically allocated by the sandbox for each bytecode call, ensuring that the logical calculations of different edges do not interfere with each other.

[0077] The multi-threaded concurrent traversal engine is a key component that leverages the parallel computing capabilities of modern multi-core processors to significantly reduce the time required for large-scale Monte Carlo simulations.

[0078] The Monte Carlo method, when applied here, approximates the true distribution of a complex stochastic process through a large number of random path choices; its accuracy increases with the number of valid instance samples. Increased by the increase, usually requiring The value of is in arrive The magnitude is used to balance computational efficiency and result accuracy. An uninterrupted traversal path refers to a complete and uninterrupted traversal trajectory from the starting node to any ending node in the simulation.

[0079] The arrival time probability distribution matrix is ​​the final structured output of this step. It quantifies in matrix form the probability that a transport unit will arrive at different destinations and at different times under a given virtual scenario assumption, providing direct data input for subsequent multi-objective risk assessment.

[0080] For example, after receiving a set of inference parameters containing environmental feature parameters and a verified probability state transition graph, the graph traversal controller locates the starting node and loads the inference parameter set into the corresponding execution memory area. In a set multi-level node network, such as from the starting node through intermediate nodes to the destination node, the controller starts a multi-threaded concurrent traversal engine containing a preset number of traversers. At each node branch, the traverser triggers the attached executable function code, reads the parameters in the inference parameter set to dynamically calculate the transition probability of the current branch. Based on this transition probability, the engine determines the direction and travel time of each traverser in multiple random parallel pathfinding operations. Finally, the engine calculates the total time of all traversal paths that have not been interrupted and have successfully reached the destination node, distributes it to various time windows of preset width, and divides the arrival frequency in each window by the total number of traversers that have successfully reached the destination node, thereby generating a quantified arrival time probability distribution matrix, realizing an objective prediction of the future arrival time distribution of the target transportation unit.

[0081] In one embodiment of the present invention, a global scheduling scheme is obtained by aggregating and calculating the probability distribution matrices of various arrival times based on a preset multi-objective evaluation function, including the following steps: Collect and align the arrival time probability distribution matrices received within the target scheduling area to form an arrival time probability distribution matrix set; input the arrival time probability distribution matrix set into a preset multi-objective evaluation function in parallel to calculate the resource idle loss value and connection penalty risk value corresponding to different scheduling assumption schemes; calculate the solution that minimizes the overall loss and risk based on the set optimization strategy, and use it as the global scheduling scheme.

[0082] Calculate the resource idle loss value and the connection penalty risk value for different scheduling assumptions, including the following steps: The time window for allocating specific tasks in the scheduling hypothesis scheme is weighted and summed with the corresponding arrival time probability distribution matrix to calculate the expected time and resource idle loss value; the time requirement threshold of the downstream connecting node in the current transportation chain is identified; based on the arrival time probability distribution matrix, the penalty probability of arriving later than the time requirement threshold is extracted, and the penalty probability is multiplied by a preset penalty coefficient to obtain the connection penalty risk value.

[0083] Specifically, after generating a forecast for a single transport unit, the process proceeds to the collaborative scheduling decision-making step. This step is executed by the global scheduling optimizer of the target hub node, aiming to aggregate and analyze the future behavior probabilities of all units to be scheduled within the region. First, the global scheduling optimizer continuously aggregates the arrival time probability distribution matrices generated by extrapolation for each transport unit within the region. The optimizer performs time-series alignment and standardization on these matrix data, which may have discrepancies in their time window divisions, forming a set of aggregated arrival time probability distribution matrices under a common time reference.

[0084] The global scheduling optimizer inputs this matrix set into a pre-defined multi-objective evaluation function in parallel, quantifying the comprehensive risks the system may face under different scheduling strategies. The optimizer generates a series of different virtual allocation schemes, which are assumptions about the future allocation of hub resources, such as allocating berth 1 to a ship. For each virtual dispatch scheme, the function performs two types of computations in parallel: The first type is to calculate the resource idle loss value, which involves performing a weighted summation operation on the resource time window allocated to a task in the virtual dispatch scheme and the arrival time probability distribution matrix of the corresponding transportation unit to calculate the expected time of resource idleness or waiting, and converting it into cost loss.

[0085] The second category is calculating the connection penalty risk value, which involves identifying the downstream connection relationship of the transportation chain, calculating the cumulative probability that the preceding transportation unit departs later than the downstream unit, and multiplying this probability by a preset penalty coefficient.

[0086] The global scheduling optimizer, based on a pre-defined objective function optimization strategy, typically seeks a solution that minimizes the total weighted risk to converge and obtain the global target scheduling scheme. The optimization strategy used can be based on heuristic algorithms, such as simulated annealing or genetic algorithms, which can efficiently search for near-optimal solutions in a vast solution space. After several iterations, the optimizer converges to a virtual assignment scheme that minimizes the total weighted risk; this scheme is then determined as the final global target scheduling scheme.

[0087] Based on this scheme, the optimizer further generates specific business instructions for each specific physical transportation unit. For example, it generates a business entry window authorization instruction for a truck that includes precise information such as "go to Gate 7 to complete the entry procedures between 14:30 and 14:45".

[0088] These instructions are then sent through the hub's business execution system to the dispatch systems of various transport parties or directly pushed to the drivers' terminal devices via electronic messages or API calls, thereby completing a full dispatch operation from data perception to collaborative decision-making to closed-loop execution.

[0089] It should be noted that the aggregated arrival time probability distribution matrix is ​​the data foundation for the entire collaborative decision-making process, encompassing the future dynamics of all relevant parties. The multi-objective evaluation function is the core decision-making model, unifying multiple, even conflicting, objectives in system operation within a quantifiable risk framework for assessment. Resource idle time depletion directly translates temporal uncertainty into economic costs, making optimization objectives more specific. The connection penalty risk value quantifies the risk of the most critical and unacceptable failure events in the system.

[0090] The global target scheduling scheme is the output of the entire complex optimization problem, representing a resource allocation strategy that improves the overall efficiency of the system and reduces risks under the current information. The business entry window authorization instruction is the final, executable scheduling result, which transforms the abstract optimization scheme into specific action guidance for specific units in the physical world.

[0091] For example, suppose the global scheduling optimizer of the target hub node receives the arrival time probability distribution matrices of the first and second trucks at the same time. Both trucks need to load cargo onto the same ship. The optimizer generates multiple virtual dispatch schemes with different entry orders and inputs them into the evaluation function. During the evaluation process, the risk function calculates that executing a certain order may cause some vehicles to miss the ship's departure time, thus triggering a high connection penalty risk value; while executing another optimized order can ensure that both trucks complete their work before departure, with a lower overall penalty risk value. The optimizer ultimately selects the global target scheduling scheme with the lowest total risk value based on the optimization strategy, and generates instructions based on this scheme: an authorization instruction to assign a specific berth to the first truck and allow immediate entry, and an instruction to assign a specific berth to the second truck and require entry after a preset waiting time. These instructions are then issued and executed, completing the scheduling loop.

[0092] See appendix Figure 2 This invention also proposes a cross-modal transportation spatiotemporal rhythm collaborative scheduling system, comprising the following modules: The model receiving module is used to receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probabilistic state transition diagram and a plaintext behavior list file; the probabilistic state transition diagram is used to represent the discrete state and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior list file contains the scene input mode, runtime environment signature and resource request token parameters. The environment verification and authorization module is used to parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. The parameter set construction module is used to decrypt the encrypted probability state transition diagram using the private key of the target hub node to obtain the decrypted probability state transition diagram, and to combine the preset virtual inference parameters with the data access credentials to construct the inference parameter set. The deduction and calculation module is used to input the deduction parameter set into the probability state transition diagram, and perform deduction and calculation through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; The scheduling instruction generation module is used to obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area, aggregate and calculate each arrival time probability distribution matrix based on a preset multi-objective evaluation function to obtain a global scheduling scheme; generate scheduling instructions based on the global scheduling scheme to allocate hub resources and send them to the corresponding transportation units.

[0093] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0094] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method, characterized in that, Includes the following steps: S1. Receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probability state transition diagram and a plaintext behavior list file; The probabilistic state transition diagram is used to represent the discrete states and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior manifest file contains the scene input mode, runtime environment signature and resource request token parameters; S2. Parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. S3. Use the private key of the target hub node to decrypt the encrypted probability state transition diagram to obtain the decrypted probability state transition diagram. Combine the preset virtual inference parameters with the data access credentials to construct an inference parameter set. S4. Input the deduction parameter set into the probability state transition diagram, and perform deduction calculations through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; S5. Obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area, and perform aggregation calculation on each arrival time probability distribution matrix based on the preset multi-objective evaluation function to obtain the global scheduling scheme. Based on the global scheduling scheme, scheduling instructions are generated to allocate hub resources and then sent to the corresponding transportation units.

2. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 1, characterized in that, Constructing a probabilistic state transition diagram includes the following steps: Acquire multi-source operational data of the target transportation network and transform positioning data and status parameters of different formats into a standardized operational dataset with a unified spatiotemporal reference; Based on the standardized operational dataset, the stationing status of transportation units during historical operations is extracted as the reference node of the directed graph. The conditional transition probability logic between baseline nodes is encoded into executable function code with input parameters; The executable function code is mapped to the edges between the corresponding baseline nodes to generate a probabilistic state transition graph.

3. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 2, characterized in that, Based on a standardized operational dataset, the stationary position status of transportation units during historical operations is extracted as the baseline node of a directed graph, including the following steps: Load the coordinates of historical trajectory points from the standardized runtime dataset; Perform a density-based spatial clustering algorithm to calculate the spatial distance between the coordinates of historical trajectory points; Core objects whose spatial distance is less than the preset neighborhood radius and whose number of data points in the neighborhood exceeds the preset minimum sample size are aggregated into geospatial clusters. Extract the centroid of each geospatial cluster and define the centroid as a reference node with a unique identifier.

4. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 3, characterized in that, Generating a transportation unit behavior model includes the following steps: Based on the input requirements of the executable function code, extract and create a behavior manifest file; The transporter's private key on the local side of the corresponding transport unit is invoked to digitally sign the constructed probabilistic state transition diagram; The public key of the target hub node is used to perform asymmetric encryption on the probabilistic state transition graph with digital signature to obtain the encrypted probabilistic state transition graph. The behavior list file and the encrypted probabilistic state transition diagram are linked and concatenated to generate a transportation unit behavior model.

5. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 4, characterized in that, Generating data access credentials includes the following steps: Extract the runtime environment signature from the behavior manifest file and perform environment compatibility verification; After the compatibility check is passed, the resource request token parameter in the behavior list file is read and parsed to extract the list of target fields required for the current simulation. Within the sandbox isolation operation zone, data access credentials that are only available for the current simulation round are generated based on the target field list; Allocate the corresponding target computing resources to the parsing process in the sandbox isolated operation area.

6. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 5, characterized in that, Constructing the inference parameter set includes the following steps: The encrypted probabilistic state transition diagram is decrypted using the private key of the target hub node to obtain the decrypted probabilistic state transition diagram and digital signature. The public key of the corresponding transporter is retrieved to compare and verify the digital signature. If the verification is successful, the decrypted probability state transition diagram is confirmed to be valid. Generate virtual simulation data containing expected berth allocation information and on-site facility status parameters; The virtual simulation data and data access credentials are merged by field selection to form a simulation parameter set.

7. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 6, characterized in that, Generating the arrival time probability distribution matrix includes the following steps: Determine the starting node of the probabilistic state transition diagram; Load the derivation parameter set into the execution memory area of ​​the executable function code corresponding to the edge connected to the starting node; After the data access credentials are verified, the executable function code is triggered to read external real-time dynamic status data and time series prediction data, perform logical operations, and output the target conversion probability value to be transferred to the downstream node. The concurrent traversal engine is started, and random parallel pathfinding traversal is performed in the probabilistic state transition graph structure in combination with the destination-oriented path constraints. The arrival time probability distribution matrix is ​​statistically calculated and generated.

8. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 1, characterized in that, Based on a preset multi-objective evaluation function, the probability distribution matrices of each arrival time are aggregated and calculated to obtain a global scheduling scheme, including the following steps: Collect and align the arrival time probability distribution matrices received within the target scheduling area to form a set of arrival time probability distribution matrices; The arrival time probability distribution matrix is ​​input in parallel into a preset multi-objective evaluation function to calculate the resource idle loss value and connection penalty risk value corresponding to different scheduling assumption schemes. The solution that minimizes overall loss and risk is calculated based on the set optimization strategy and used as the global scheduling scheme.

9. A cross-modal transportation spatiotemporal rhythm collaborative scheduling method according to claim 8, characterized in that, Calculate the resource idle loss value and the connection penalty risk value for different scheduling assumptions, including the following steps: The expected time and resource idle loss value are calculated by performing a weighted summation operation on the time window for allocating specific tasks in the scheduling hypothesis scheme and the corresponding arrival time probability distribution matrix. Identify the time requirement threshold for downstream connection nodes in the current transportation chain; Based on the arrival time probability distribution matrix, the penalty probability for arriving later than the time requirement threshold is extracted, and the penalty probability is multiplied by a preset penalty coefficient to obtain the connection penalty risk value.

10. A cross-modal transportation spatiotemporal rhythm collaborative scheduling system, characterized in that, Includes the following modules: The model receiving module is used to receive the transportation unit behavior model sent by the transportation unit to be scheduled; the transportation unit behavior model includes an encrypted probability state transition diagram and a plaintext behavior list file; The probabilistic state transition diagram is used to represent the discrete states and state transition logic of the transportation unit, and executable function code is attached to the edges; the behavior manifest file contains the scene input mode, runtime environment signature and resource request token parameters; The environment verification and authorization module is used to parse the behavior list file to perform environment verification, and generate the corresponding data access credentials in the sandbox isolation operation area after the verification is successful. The parameter set construction module is used to decrypt the encrypted probability state transition diagram using the private key of the target hub node to obtain the decrypted probability state transition diagram, and to combine the preset virtual inference parameters with the data access credentials to construct the inference parameter set. The deduction and calculation module is used to input the deduction parameter set into the probability state transition diagram, and perform deduction and calculation through the graph traversal algorithm to obtain the arrival time probability distribution matrix of the transportation unit; The scheduling instruction generation module is used to obtain the arrival time probability distribution matrix of each transportation unit to be scheduled within the target scheduling area, and to perform aggregation calculation on each arrival time probability distribution matrix based on a preset multi-objective evaluation function to obtain a global scheduling scheme. Based on the global scheduling scheme, scheduling instructions are generated to allocate hub resources and then sent to the corresponding transportation units.

Images