An integrated logistics management system for cargo transportation

Abstract

This invention provides an integrated logistics management system for cargo transportation, comprising a multimodal collaborative perception module, a semantic reasoning module, a scenario adaptation decision-making module, an intelligent execution module, and a logistics trust chain module. The multimodal collaborative perception module collects status data across the entire logistics scenario through a sensor network; the semantic reasoning module constructs a knowledge graph based on a graph attention network to achieve unified modeling of multi-source information; the scenario adaptation decision-making module combines a pulsed coupled neural network (PCNN) and a swarm intelligence collaborative mechanism to optimize routes; the intelligent execution module employs a hierarchical task management and dynamic adjustment mechanism to adjust task plans based on execution anomalies and credit assessments; and the logistics trust chain module constructs a logistics trust system through a hierarchical consensus mechanism, a credit scoring model, and a trust propagation and early warning mechanism to achieve risk prevention and control throughout the entire process. This system solves the technical problems of traditional logistics systems in areas such as data integration, dynamic monitoring, decision support, collaborative scheduling, and security traceability.

Description

Technical Field

[0001] This invention relates to the field of logistics information technology, and specifically to an integrated logistics management system for cargo transportation. Background Technology

[0002] With the deepening of globalization and the rapid rise of new retail models, modern logistics and transportation are facing unprecedented challenges. Current logistics management systems mainly suffer from the following problems:

[0003] Insufficient system integration. Functional modules such as order management, warehouse scheduling, and transportation tracking operate independently, resulting in severe data silos; seamless integration between systems of different participants is difficult; in multimodal transport scenarios, information exchange between various modes of transport is hindered; and customs clearance, commodity inspection, and other aspects of cross-border logistics are poorly connected with the transportation system.

[0004] Dynamic monitoring capabilities are limited. Relying solely on GPS positioning and RFID identification cannot achieve comprehensive status monitoring; specialized monitoring methods for special goods are insufficient; the real-time nature and accuracy of monitoring data are difficult to guarantee; and monitoring blind spots are widespread.

[0005] Insufficient decision support. Historical data analysis remains superficial, with insufficient in-depth value extraction; there is a lack of intelligent decision-making capabilities under multiple scenarios and constraints; and decision recommendations lack interpretability.

[0006] The efficiency of collaborative dispatching is low. Fleet management lacks overall planning and is mainly based on single-vehicle dispatching; cross-regional collaboration capabilities are weak, and resources are difficult to share between regions; multi-vehicle collaborative operations lack a unified command and dispatch mechanism.

[0007] Safety traceability is challenging. There is a lack of mechanisms for full-process tracking and verification of cargo integrity; insufficient monitoring of the physical condition during transportation; and difficulty in timely detection of safety hazards caused by human factors.

[0008] In conclusion, the integration, intelligence, and collaboration of traditional logistics management systems are insufficient to meet the demands of the complex modern logistics environment, and they urgently need to be optimized and improved through innovative technologies. Summary of the Invention

[0009] This invention provides an integrated logistics management system for cargo transportation. Through the systematic integration of multiple functional modules, it achieves intelligent management of the entire cargo transportation process, addressing technical challenges in existing logistics systems related to data integration, dynamic monitoring, decision support, collaborative scheduling, and safety traceability. The system constructs a complete technical framework from cargo transportation perception and analysis to execution, with each module working collaboratively to support integrated management of the entire cargo transportation process. Specifically, it includes the following technical modules:

[0010] 1. Multimodal cooperative sensing module:

[0011] As the data foundation layer of the system, a network of temperature, humidity, vibration, and tilt sensors installed in vehicles, warehouses, and loading / unloading areas enables status monitoring across the entire logistics scenario. A deep belief network with a specific network structure is used, in which each type of sensor data is configured with an independent feature extraction layer to ensure the feature representation of different types of data. Edge computing nodes process data locally through a pre-set federated learning model and transmit feature information to the central node, reducing data transmission load. This module provides real-time, multi-dimensional perception data support for the semantic reasoning module and the scene adaptation decision module.

[0012] 2. Semantic Reasoning Module:

[0013] As the knowledge processing layer of the system, the knowledge graph constructed based on the graph attention network will uniformly model the multi-dimensional state data obtained from the multimodal collaborative perception module, along with order information, warehouse status, transportation routes, vehicle information, and cargo attributes. Attention scores will be assigned to different types of information through a node weight matrix to highlight key information. The set knowledge graph completion algorithm will be used to infer node relationships and fill in information gaps. An ontology mapping method will be used to realize semantic connections between road, rail, and air transportation. This module provides structured knowledge support for the scenario adaptation decision module and provides the processed semantic information to the logistics trust chain module for credibility assessment.

[0014] 3. Scene Adaptation Decision Module:

[0015] As the intelligent decision-making layer of the system, it includes a Pulse Coupled Neural Network (PCNN) layer and a swarm intelligence collaboration layer. The PCNN layer constructs the logistics network as a neuron matrix structure, with each neuron corresponding to a logistics node. The connection weights between neurons are determined by a normalized weighted calculation of four indicators: path distance, travel time, transportation cost, and path reliability. The maximum processing capacity of each node is the core basis for the activation threshold, and is dynamically adjusted in conjunction with processing time and caching capacity. The swarm intelligence collaboration layer sets up multiple decision-making agents, each responsible for a region within a specified radius. Its status information includes road condition index, vehicle availability, warehouse capacity, and the proportion of urgent orders. Agents in adjacent regions exchange status information within a fixed period through a preset hierarchical communication protocol. The initial routing scheme output by the PCNN layer serves as a soft constraint for the agent's decision-making. The agent optimizes the route based on the regional status information it possesses, and the optimization result dynamically updates the connection weights of the PCNN layer according to a preset learning rate parameter. Finally, the module transmits the decision results to the intelligent execution module and provides structured decision data support for the logistics trust chain module.

[0016] 4. Intelligent Execution Module:

[0017] As the task execution layer of the system, a hierarchical task management mechanism is adopted to decompose the transportation tasks output by the scenario adaptation decision module into three sub-tasks: trunk transportation, branch line scheduling, and last-mile delivery. Based on the constraint model of time window, route traffic status, and node processing capacity, the rationality and feasibility of the execution plan are ensured. Through the anomaly detection results of the multimodal collaborative perception module and the credit assessment of the logistics trust chain module, the task execution plan is dynamically adjusted. At the same time, the real-time status of task execution is fed back to the multimodal collaborative perception module to form a closed-loop control, thereby enhancing the adaptability and stability of the system.

[0018] 5. Logistics Trust Chain Module:

[0019] As the system's trust assurance layer, a logistics trust system is constructed through a layered consensus mechanism, a credit scoring model, and a trust propagation and early warning mechanism. Consensus weights are allocated based on the roles and attributes of carriers, warehousing providers, and consignees / consignees, combined with accumulated credit scores and transaction data. The credit scoring model evaluates on-time delivery rate, transaction volume ratio, positive review rate, and complaint rate, inheriting the initial trust assessment results from the intelligent execution module. The trust propagation model calculates the trust relationships between nodes, implementing a tiered early warning mechanism. Data collection strategies, route planning decisions, resource allocation methods, and transaction consensus requirements are adjusted based on trust levels and risk assessment results. Simultaneously, the module continuously receives real-time data from the multimodal collaborative perception module and task feedback from the intelligent execution module, dynamically updating the trust assessment results to form an efficient closed-loop trust management mechanism.

[0020] The beneficial effects of this invention are:

[0021] 1. Comprehensive Perception and Monitoring: Employing multimodal sensor networks and federated learning models improves the coverage and data processing efficiency of logistics scenario status monitoring, and reduces system monitoring blind spots.

[0022] 2. Semantic information fusion: Knowledge graphs are constructed through graph attention networks to achieve interconnection and interoperability of multimodal transport information, reduce information silos, and improve information utilization efficiency.

[0023] 3. Intelligent decision optimization: Combining PCNN with swarm intelligence collaborative mechanism, route planning is completed under multiple constraints, improving the accuracy and timeliness of system decision-making.

[0024] 4. Task execution control: Based on hierarchical task management and a multi-dimensional constraint model, the system enables coordinated scheduling of trunk and branch lines, ensuring the feasibility and efficiency of the execution plan.

[0025] 5. End-to-end trust verification: By using a layered consensus mechanism and smart contracts, combined with dynamic credit scoring, a trustworthiness assessment system for the logistics process is established to enhance security traceability capabilities. Attached Figure Description

[0026] Figure 1 The system architecture diagram of the present invention is shown;

[0027] Figure 2 A system workflow diagram of the present invention is shown. Detailed Implementation

[0028] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] Combination Figure 1 This invention provides an integrated logistics management system for cargo transportation, comprising the following components:

[0030] 1. Multimodal collaborative sensing module:

[0031] As the system's data foundation layer, efficient status monitoring across the entire logistics scenario is achieved through scenario-optimized sensor network deployment, deep belief network feature extraction, and a distributed processing mechanism combining edge computing and federated learning. Specific implementation includes:

[0032] A1) Sensor Network Deployment

[0033] Temperature, humidity, vibration, and tilt sensors are deployed in transport vehicles, storage facilities, and loading / unloading areas, and optimized using a scenario-adaptive layered coverage mechanism.

[0034] a) Dynamic Calculation of Coverage Range: The effective coverage range of the sensor is dynamically adjusted through the sensing quality function Q, as shown in the following formula:

[0035]

[0036] in: For initial perceived quality; is the distance attenuation coefficient; e is the natural constant; :sensor to monitoring point The distance; Environmental impact factors.

[0037] b) Layered Deployment Approach: Based on the characteristics of the logistics scenario, the sensors are deployed in two layers: Core Monitoring Layer: Deployed at key nodes (loading and unloading points, warehouse entrances), using high-precision sensors to ensure the accuracy of data in key areas. Collaborative Sensing Layer: Deployed in regular areas, using low-cost sensors to provide broad coverage support.

[0038] A2). Deep Belief Network (DBN) architecture

[0039] A deep belief network with a specific network structure is used to extract features from multimodal data. For temperature, humidity, vibration, and tilt angle data in logistics scenarios, a constrained hierarchical feature extraction mechanism is designed.

[0040] a). DBN Energy Function: The energy function for DBN feature extraction is:

[0041]

[0042] in: The visible layer neuron state vector; Let be the state vector of the i-th neuron in the visible layer;

[0043] h is the state vector of the hidden layer neurons; Let be the state vector of the j-th neuron in the hidden layer;

[0044] m and n are the number of neurons in the hidden layer and the visible layer, respectively; This represents the connection weights between the visible and hidden layers. For bias terms; The constraints for the logistics scenario include temperature constraints, vibration amplitude constraints, and tilt angle range constraints, defined as follows:

[0045]

[0046] In the formula:

[0047] These are the weighting coefficients; ,satisfy This is used to balance the various constraint terms;

[0048] Temperature change rate constraint:

[0049]

[0050] To limit the continuity of temperature data and suppress abnormal fluctuations, the formula is as follows: The smoothing coefficient is t, where t is the current time step and T is the total number of time steps.

[0051] Vibration amplitude constraints:

[0052]

[0053] Limit vibration amplitude to exceed a set threshold In the formula, This is the vibration suppression coefficient; The second norm of the vibration amplitude;

[0054] Tilt angle range constraints:

[0055]

[0056] Limit tilt angle beyond safe range In the formula, This is the tilt angle constraint coefficient; This is an indicator function; it returns 1 if the condition is true, and 0 otherwise. b) Feature extraction layer design: Considering the characteristics of multimodal data, the following independent feature extraction layers are designed:

[0057] Temperature and humidity data: Input dimension 1, number of hidden layer nodes is [64, 32, 16];

[0058] Vibration data: Input dimension 3, number of hidden layer nodes [128, 64, 32];

[0059] Tilt angle data: Input dimension 2, number of hidden layer nodes is [64, 32, 16].

[0060] A3). Edge Computing and Federated Learning

[0061] By combining edge computing and federated learning, distributed data processing and model training were achieved, effectively reducing data transmission load and improving computational efficiency.

[0062] a) Local training of edge nodes: Each edge node is trained based on a local dataset. Train the model and update the parameters. :

[0063]

[0064] in: , These represent the model parameters of the edge nodes at time steps t+1 and t, respectively. The learning rate is initially set to 0.01. This is the gradient operator, used to differentiate a function; The loss function for a node-based dataset is defined as follows:

[0065]

[0066] in: For edge nodes In the Model parameters in the next iteration; For edge nodes The local dataset used; For dataset The size of the dataset is the number of samples in the dataset. For dataset One of the input samples; To obtain input data through the model The predicted output obtained from reconstruction; The regularization coefficient is used. For model parameters The square of the second norm; For input data Rather than reconstructing data The L2 norm squared error between them.

[0067] b) Model aggregation and optimization: The center node aggregates the edge node models according to the following weighted average formula:

[0068]

[0069] Weight The calculation method is as follows:

[0070]

[0071] in: The reliability coefficient of the node; The data quality score for a node is defined as follows:

[0072]

[0073] in: Weighting coefficients are used to balance the impact of different data quality indicators, satisfying the following:

[0074] The completeness score measures the proportion of missing values ​​in the data; a higher score indicates more complete data. It is defined as follows:

[0075] in: This represents the number of missing data points. This represents the total number of data points in the dataset.

[0076] A freshness score measures the timeliness of the data; a higher score indicates "fresher" data, ensuring that the model uses timely data. Low-freshness data may affect the model's real-time performance and accuracy. The score is defined as follows:

[0077] in: This is the current timestamp; The timestamp of the most recent data update; This is the time decay coefficient, which controls the rate of time decay.

[0078] An accuracy score measures the accuracy of the data, based on outlier detection. A higher score indicates more accurate data, defined as:

[0079] in: This is the average of historical data; Standard deviation of external historical data; For indicator functions, if data Values ​​that deviate from the historical mean by more than three times the standard deviation are considered outliers. For dataset The number of samples in the middle. This scoring mechanism comprehensively considers the completeness, timeliness, and accuracy of the data, and is used to assign weights to the data quality of different nodes during model aggregation. High-quality data can be assigned higher weights, prompting the model to rely more on the data from these nodes during aggregation, ultimately improving the overall model performance.

[0080] A4) Multi-dimensional perception data support

[0081] By jointly processing sensor networks and deep belief networks, high-quality data support is provided for semantic reasoning and decision-making modules.

[0082] Real-time data on temperature, humidity, vibration, and tilt angle are transformed into structured perception information through feature extraction; the real-time performance and integrity of the data are ensured through efficient collaboration between edge computing and federated learning.

[0083] 2. Semantic Reasoning Module

[0084] As the system's knowledge processing layer, a knowledge graph is constructed based on a Graph Attention Network (GAT) to unify the modeling of multi-dimensional data collected by the multimodal collaborative perception module with order information, warehouse status, transportation routes, vehicle information, and cargo attributes. A dynamic attention mechanism is used to allocate semantic association weights between nodes, highlighting key information; a knowledge graph completion algorithm is employed to predict and improve node relationships, filling information gaps; and an ontology mapping method is combined to achieve semantic connections in multimodal transport scenarios. Specific implementation includes:

[0085] B1) Knowledge Graph Construction

[0086] Building knowledge graphs Contains a set of nodes Edge set and relation set It is used to unify the modeling of heterogeneous information in logistics systems.

[0087] a) Node set definition Node set Including the following types: order nodes This includes core information describing logistics needs, such as order ID, goods type, quantity, and delivery time requirements; warehousing nodes. This includes warehouse ID, storage location, capacity, and inventory status, reflecting goods storage information; path nodes. Records road segment ID, origin and destination, distance, and traffic status to describe the transportation route; vehicle node. This includes vehicle ID, type, load capacity, and current location, reflecting the real-time status of the logistics transportation vehicle; cargo nodes. This includes cargo ID, attributes, status, and location information, used to monitor cargo movement.

[0088] b) Define the edge set Defined as a concrete instance of a relationship between entity nodes, including:

[0089]

[0090] in: The starting node; It is a relation type; Define the same for the target node. ; This refers to edge attributes; A set of nodes; A set of relation types; The set of attributes for the edges includes the following categories:

[0091] Time attributes: Relationship establishment time, relationship termination time (or estimated termination time if in progress), and most recent status update time;

[0092] State attributes: current state identifier, previous state record, expected state transition;

[0093] Weighted attributes: Basic relationship weight, timeliness weight, service quality weight, cost-related weight

[0094] Business attributes: capacity / load-related parameters, efficiency-related parameters, resource usage parameters, and constraint parameters.

[0095] c) Definition of relation sets include:

[0096]

[0097] Wherein: stored in: the relationship between goods and warehousing, indicating that the goods are stored in a certain warehouse;

[0098] Transportation: The relationship between goods and vehicles, indicating that the goods are transported by a particular vehicle;

[0099] Belonging to: The relationship between a vehicle and a company or between a warehouse and a logistics center, indicating that a vehicle belongs to a certain company or a warehouse belongs to a certain logistics center;

[0100] Connection: The relationship between path nodes, representing the connection relationship of the transportation path;

[0101] Carrying: The relationship between a vehicle and an order, indicating that a vehicle carries a certain order; Including: The relationship between an order and goods or between a warehouse and storage locations, indicating that an order includes multiple goods or a warehouse includes multiple storage locations.

[0102] B2). Graph Attention Network Implementation

[0103] This module uses graph attention networks to model knowledge graphs, completing the dynamic weight allocation and information aggregation of node features.

[0104] a) Node feature encoding: Encode node features based on node type.

[0105]

[0106] in: Represents a node eigenvectors; For node type; These are the original features of the node; For encoders specific to node types.

[0107] The encoder parameters are configured as follows:

[0108] Order node: Input dimension 12, output dimension 64; Warehouse node: Input dimension 8, output dimension 64; Path node: Input dimension 10, output dimension 64; Vehicle node: Input dimension 6, output dimension 64; Goods node: Input dimension 8, output dimension 64.

[0109] b) The attention mechanism assigns semantic association weights between nodes through a graph attention network, calculated using the following formula:

[0110]

[0111] in: Represents a node with neighboring nodes Attention weights between them; For nodes , The eigenvectors of k; The feature transformation weight matrix; This is the attention weight vector; For nodes The set of neighbors; exp represents the natural exponential function; is the linear rectified activation function with leakage correction; || denotes the vector concatenation operation.

[0112] c) Information aggregation updates node features through a multi-head attention mechanism:

[0113]

[0114] in:

[0115] To increase the number of attention heads (set to 8), enhance the diversity of information aggregation between nodes; is the activation function used to introduce non-linear features; || represents the concatenation operation of the outputs of K attention heads.

[0116] B3). Knowledge Graph Completion

[0117] To fill the information gaps in the knowledge graph, the TransE model is used for relation prediction and completion:

[0118] a) Relation prediction: Predict missing relations based on the embedding vectors of the head entity, tail entity, and relation.

[0119]

[0120] in: These are vectors representing the head entity, relation, and tail entity, respectively; the embedding dimension is set to 100.

[0121] b) Confidence assessment: A confidence assessment is performed on the predicted relationship.

[0122]

[0123] in: The function is sigmoid; the confidence threshold is set to 0.8 to filter reliable relationships.

[0124] B4). Multimodal transport semantic mapping

[0125] To achieve semantic connectivity among highway, railway, and air transport, ontology mapping technology is used for standardization.

[0126] a) Ontology Construction: Constructing the highway transportation ontology Railway transportation itself and air transport body This provides a unified description of the core concepts of each mode of transportation.

[0127] b. Ontology mapping calculates semantic similarity between concepts by measuring information content:

[0128]

[0129] in: There are two concepts to be mapped; The most recent common ancestor; The information content reflects the importance of the concepts.

[0130] c. Rule generation: Rules are generated across transportation modes through mapping relationships.

[0131]

[0132] in: Represents the rules from state i to j; This represents a certain attribute or condition in state i; This is the logical AND operator, indicating that two conditions must be met simultaneously. This represents the mapping relationship from state i to j; It represents a certain attribute or condition under state j.

[0133] The processed semantic information is passed to the decision-making module for path planning and scheduling optimization; node relationships and confidence levels are given to the trust chain module for dynamically adjusting the credibility of the logistics network.

[0134] 3. Scenario Adaptation Decision Module

[0135] As the intelligent decision-making layer of the system, this module constructs an adaptive route optimization framework under multiple constraints through an innovative combination of pulsed coupled neural networks (PCNN) and swarm intelligence collaborative mechanisms. This framework fully considers the node processing capacity, path characteristics, and regional state information in the logistics scenario, enabling efficient collaboration between global planning and local optimization, and providing reliable support for intelligent decision-making in complex logistics networks. Specific implementation includes:

[0136] 3.1 PCNN Network Construction and Optimization

[0137] C1) Logistics network modeling maps each logistics node in the logistics network to a neuron in a PCNN. The activation threshold of the neuron... The formula reflecting the node's maximum processing capacity is as follows:

[0138]

[0139] in: For nodes The maximum processing capacity (pieces / hour) is obtained from historical data statistics; For nodes The average processing time (in hours) represents the processing efficiency. For nodes The caching capacity (in units) is used to reflect temporary storage capacity; The weighting coefficients are obtained through regression optimization using historical operational data.

[0140] C2). Optimize the path connection weights between nodes. Taking into account route distance, travel time, transportation cost, and route reliability, the calculation formula is as follows:

[0141]

[0142] in: This is the path distance, reflecting the physical distance between nodes; The travel time represents the impact of traffic conditions on path efficiency. For transportation costs, it represents route economics; The path reliability index is calculated based on the historical on-time rate of the path. Weighting coefficients, satisfying It was obtained through experimental optimization.

[0143] The dynamic behavior of neurons in PCNN is described by the following equation:

[0144]

[0145] in: , : These represent feed input and link input, respectively; Internal activity of neurons; The output status is 1, indicating that the node is active, and 0, indicating that it is not active. Input attenuation coefficient; Input amplitude coefficient; This refers to the link strength.

[0146] C3) Swarm intelligence collaborative optimization

[0147] Each decision-making agent Maintain a state vector to describe the logistics status of its responsible area:

[0148]

[0149] in:

[0150] The road condition index, ranging from [0,1], reflects traffic conditions and is calculated as the ratio of the current vehicle speed to the ideal free-flow speed.

[0151] Vehicle availability, ranging from [0,1], represents the proportion of vehicles that can be dispatched in the current area; The storage capacity is in the range [0,1] and reflects the storage utilization rate, which is calculated as the proportion of available space to total space. The percentage of urgent orders indicates the proportion of urgent orders within a region. Orders with less than 20% of the remaining time are considered urgent.

[0152] The agents exchange state information using a hierarchical gossip protocol, and the state update formula is as follows:

[0153]

[0154] in:

[0155] Representing intelligent agents respectively Directly adjacent regions and second-order adjacent regions; These are the direct neighborhood and far neighborhood influence factors, respectively. This represents the information fusion coefficient.

[0156] C4) Two-level collaborative decision-making mechanism

[0157] The activation states output by PCNN serve as soft constraints for the agent, as shown in the following formula:

[0158]

[0159] in:

[0160] These are weight parameters;

[0161] : Measures the deviation of the agent's decision from the PCNN routing scheme, calculated as the ratio of the deviation between the actual path and the planned path to the planned path;

[0162] Local feasibility assessment, used to limit infeasible paths, is calculated using the following formula:

[0163]

[0164] in:

[0165] For time feasibility, it indicates whether there is sufficient time. ; Resource feasibility indicates whether the supply and demand of resources are matched. ; Cost feasibility indicates whether the budget is sufficient; min(·) represents the minimum value function;

[0166] Taking into account timeliness, economy, path reliability, and load balancing, the reward function is defined as follows:

[0167]

[0168] in:

[0169] These are weight parameters;

[0170] , These represent the actual time required to complete the task and the time limit for completing the task, respectively.

[0171] , These represent the actual costs incurred and the budgeted costs, respectively.

[0172] The path reliability score is calculated using the following formula:

[0173]

[0174] in:

[0175] For time reliability, it represents the proportion of tasks completed on time. ;

[0176] To ensure cost reliability, it reflects the deviation between actual and planned costs. ; For safety and reliability, it is an inverse indicator of the accident rate. .

[0177] C5) Decision-making process and convergence judgment

[0178] Initialize the PCNN network and agent state, setting learning parameters and path weights; generate an initial routing scheme from the PCNN, then optimize the path scheme through the agent, update the PCNN weights, and check the convergence condition. The convergence condition is determined when the rate of change of the objective function is less than a threshold.

[0179]

[0180] in: , , Let represent the objective function values ​​at time t and time t-1, respectively.

[0181] 4. The intelligent execution module, as the system's task execution layer, achieves precise execution and real-time adjustment of transportation tasks through a hierarchical task management mechanism, multi-dimensional constraints, dynamic feedback and anomaly response mechanisms, and resource optimization strategies, ensuring the system's efficient and reliable operation in complex logistics scenarios. Specific implementation includes:

[0182] D1). Task Decomposition Mechanism: Task decomposition adopts a three-layer structure, namely: : Collection of trunk line transportation tasks; : A set of branch scheduling tasks; : A collection of last-mile delivery tasks.

[0183] General Task The definition of is:

[0184]

[0185] The set of tasks at each level satisfies the following condition:

[0186]

[0187] in: Indicates the first The first layer of tasks Sub-tasks; k represents the level; For the first Number of tasks per layer.

[0188] Task decomposition is achieved by optimizing the model, with the optimization objective being to minimize the execution cost and interaction cost of the tasks.

[0189]

[0190]

[0191] in: : No. The first layer Sub-tasks; The execution cost of a single task; Task interaction cost; : Weighting factor for task interaction cost.

[0192] D2) Multidimensional Constraint Model

[0193] a) Time window constraints: Tasks must be completed within a specified time frame.

[0194]

[0195] in: The earliest and latest arrival times for the task; This refers to the actual arrival time. This refers to the departure time; This represents the maximum processing time.

[0196] b) Path feasibility is evaluated using the following model, which is constrained by path traffic conditions:

[0197]

[0198] in: This represents the internal activity of PCNN neurons, dynamically reflecting real-time changes in path state;

[0199] For activation threshold and proportionality coefficient This is jointly determined, representing the dynamic lower bound of path reliability; This is the path reliability index, dynamically calculated during the PCNN network training process.

[0200] The congestion level is calculated using the following formula:

[0201]

[0202] in: To measure actual travel time and transportation costs; The preset passage time and cost; Let be the weighting coefficient, satisfying .

[0203] c) Node processing capacity constraints: The processing capacity of a node is defined by the following model:

[0204]

[0205] in: For the task At the node Use resources Variables; For resources Unit demand; For nodes Maximum processing capacity.

[0206] D3). Dynamic adjustment mechanism a). Abnormal event assessment

[0207] The risk value of abnormal events is assessed using the following model:

[0208]

[0209] in:

[0210] The probability of an abnormal event is calculated as follows:

[0211]

[0212] in: Rate the data quality; The influence coefficient represents the attenuation effect of control path reliability on probability, and its value is [value missing]. ; It is a path reliability index that dynamically reflects the reliability of the path.

[0213] The higher the data quality score, the lower the probability of anomalies; the higher the path reliability index, the more exponentially the probability of anomalies decreases.

[0214] To determine the influence of intensity, the calculation formula is as follows:

[0215]

[0216] in: The delay time represents the difference between the actual travel time and the preset travel time. Preset passage time; Additional costs, including costs incurred due to current road conditions or other unforeseen events; To pre-set toll costs; The weighting coefficients and .

[0217] b) Dynamic adjustment of tasks

[0218] In all available options In the middle, choose utility and initial trust level The solution with the largest weighted product value :

[0219]

[0220] in:

[0221] Let the utility function be defined as:

[0222]

[0223] in: This describes the internal activity of neurons in PCNN. is the neuron activation threshold in PCNN, representing the baseline for evaluating path performance; This refers to the actual travel time. Preset passage time; Let be the weighting coefficient, satisfying .

[0224] The initial level of trust is defined as:

[0225]

[0226] in: To account for the cumulative effect of execution deviations, In the formula, For the first The deviation for each execution is calculated by comparing the actual execution result with the preset value. The maximum allowable deviation is used for normalization; The attenuation coefficient controls the impact of deviation on trust level.

[0227] D4). Real-time feedback mechanism

[0228] a) Task status monitoring

[0229] Task state vector definition:

[0230]

[0231] Where: Location is the task execution location, based on the node index of PCNN; Progress is the task completion progress, normalized to... Resource indicates resource usage; Exception indicates exception flags, with 1 for exceptions and 0 for normal operations.

[0232] Overall condition rating:

[0233]

[0234] For the scoring function:

[0235] in: These are planning reference values ​​from the scenario adaptation decision module; This is the state deviation coefficient, set according to the importance of the indicator; This is an indicator function used for discrete state determination.

[0236] b) Feedback cycle Adjustment

[0237]

[0238] in:

[0239] Basic feedback cycle; This is the high state threshold. This is the low state threshold.

[0240] D5) Implement optimization strategies

[0241] a) Resource scheduling optimization

[0242] Dynamic programming optimization objective:

[0243] in: This represents the instant reward function. In the formula, The weighting coefficients and To perform performance evaluation, take The standardized value; Resource consumption assessment, In the formula, This refers to the actual cost of resource consumption. Preset resource costs.

[0244] b) Dynamic adjustment of task priorities

[0245]

[0246] in:

[0247] Weighting coefficient ,satisfy ;

[0248] urgency :

[0249]

[0250] In the formula, The deadline for the task; The current time; Total allowed execution time; This represents the time urgency coefficient.

[0251] Dependency :

[0252]

[0253] In the formula, The number of subsequent tasks that depend on this task; This represents the total number of tasks in the current system.

[0254] Completeness :

[0255]

[0256] 5. Logistics Trust Chain Module

[0257] As the system's trust guarantee layer, a highly efficient and reliable logistics trust system is constructed through a layered consensus mechanism, smart contracts, and dynamic credit assessment. Specific implementation includes:

[0258] E1). Layered consensus mechanism

[0259] The hierarchical consensus mechanism supports fair participation and decision-making among multiple parties in the logistics system through role attributes, credit scores, dynamic weight allocation, and flexible consensus rules.

[0260] a) Character Attribute Definition

[0261] Basic role: Carrier Responsible for cargo transportation; warehousing party Responsible for the storage and allocation of goods; consignor and consignee They are responsible for sending and receiving goods.

[0262] b) Cumulative credit score

[0263] Credit scores are calculated by defining a set of quantitative indicators based on the key business points of the role type and then comprehensively calculating them by weight.

[0264] Scoring formula:

[0265]

[0266] in:

[0267] Carrier rating index ( ):

[0268] On-time rate The percentage of tasks completed on time; path matching degree The degree to which the actual route matches the planned route; the rate of cargo integrity. Supplementing the proportion of damaged goods; abnormal response rate The proportion of cases that are handled promptly; weighting .

[0269] Warehouse rating indicators ( ):

[0270] Storage accuracy The ratio of stored records to actual goods; loading and unloading efficiency. The loading and unloading volume completed per unit time (unit: pieces / hour) is then standardized; inventory accuracy. The ratio of inventory records to actual goods; error rate supplementation. The proportional complement of storage or operational errors; weights are assigned values ​​sequentially. ;

[0271] Rating indicators for consignors and consignees ( ):

[0272] Timeliness of settlement On-time payment ratio; information accuracy The ratio of information conveyed to reality; the degree of cooperation. Success rate is measured and recorded based on task completion status; document validity. The completeness of the transmitted and confirmed documents; the weights are assigned sequentially. .

[0273] c) Consensus weight allocation

[0274] Consensus weights are dynamically calculated by combining role attributes and historical behavior.

[0275] Weighting formula:

[0276]

[0277] Among them: Role type weight ( Carrier weighting Warehouse weight ,

[0278] Weight of consignor and consignee .

[0279] Credit score ( ):

[0280] in: ;

[0281] Transaction frequency , The number of transactions for a specific role;

[0282] Transaction amount Total transaction amount for a specific role;

[0283] Weighting coefficient , , .

[0284] d) Consensus Reaching Mechanism

[0285] like If so, then the consensus variable Agreement = 1 (a consensus has been reached);

[0286] like If no consensus is reached, then the consensus variable Agreement = 0 (no consensus has been reached).

[0287] in, The weights of each node and corresponding value The sum of the products of, The consensus threshold is determined based on the decision type, and the calculation method is as follows:

[0288]

[0289] in, It is the maximum number of fault-tolerant nodes. (n-1) / 3 and round down; This represents the total number of nodes. It is an adjustment factor, the value of which is determined by the decision type. For carrier-related decisions, Regarding warehousing-related decisions, Regarding the decisions made by the shipper and consignee, .

[0290] Furthermore, the participation requirements differ for different roles: carriers require at least 40% of carrier nodes to participate in the decision-making vote, and only when this participation requirement is met will the consensus formula calculation result be meaningful; warehousing requires at least 30% of warehousing nodes to participate in the decision-making vote; and the shipper and consignee require explicit confirmation from both parties for the decision to pass, and are not subject to the unilateral constraints of the consensus formula calculation result.

[0291] E2). Smart Contracts

[0292] a) Status verification and transition verification of the current operation Does it satisfy all rules? If the condition is met, a state transition is allowed. The state transition depends on the current state. ,operate and condition set .

[0293] b) Positioning verification checks whether the deviation between the current position and the planned position is within the allowable range.

[0294] c) Timeliness verification: Check whether the deviation between the operation completion time and the planned time meets expectations.

[0295] E3). Credit scoring model

[0296] a) Credit score calculation model: The credit score is calculated by weighting multiple indicators.

[0297]

[0298] in: As the indicator weight, In the formula, This is the weight update coefficient, with a value of 0.3; The magnitude of the indicator change; As the benchmark indicator weight; These are indicator values, including: on-time rate. Normal transaction volume percentage Positive feedback rate Complaint rate supplement The initial trust level obtained by the intelligent execution module .

[0299] b) Dynamic update mechanism: Credit scores are dynamically updated using the following formula:

[0300]

[0301] in: To update the coefficients, the fusion ratio of the old and new scores is controlled; , which is the inheritance coefficient, balancing the impact of initial trust level and new ratings; Changes in the scores of the newly added indicators.

[0302] E4). Dynamic Trust Assessment and Risk Early Warning Mechanism

[0303] a) Trust propagation model: The formula for trust propagation between nodes is:

[0304]

[0305] in: :node For nodes Trust level; Trust chains that propagate through intermediate nodes.

[0306] b) System monitoring and early warning: The early warning level is determined by the comprehensive risk value.

[0307]

[0308] in: Risk values ​​for abnormal events obtained by the intelligent execution module; These are the high and low risk thresholds.

[0309] E5). Early Warning Response and Coordinated Regulation

[0310] Based on the aforementioned risk warning mechanism, the system adopts corresponding coordinated control measures for different warning levels:

[0311] a) High-risk status

[0312] Transportation control: Restrict carriers from adding new waybills; increase the frequency of GPS positioning and sensor data collection to once every 5 minutes; require vehicles to travel on the main route planned by PCNN and prohibit alternative routes; enforce cargo status checks at each handover point.

[0313] Warehouse management: Increase the processing capacity threshold of warehouse nodes to 120%; increase the inventory counting frequency to once a day; require warehouses to provide additional video surveillance records.

[0314] Information control: All transaction information requires additional confirmation nodes to participate in consensus; increase the node association threshold of the knowledge graph; shorten the execution timeout of smart contracts to 50% of the standard value.

[0315] b) Medium-risk status

[0316] Transportation control: New waybills are allowed, but additional node confirmation is required; sensor data collection frequency is adjusted to once every 15 minutes; alternative routes planned by PCNN are allowed, but route changes must be reported in real time; cargo status checks are conducted at key handover points.

[0317] Warehouse management: Maintain standard processing capacity thresholds for warehouse nodes; maintain regular inventory counting frequency, but with detailed records; strengthen monitoring of key areas.

[0318] Information control: Transaction information requires a benchmark number of nodes to participate in consensus; maintain the standard association threshold of the knowledge graph; and implement a standard timeout mechanism for executing smart contracts.

[0319] c) Low-risk status

[0320] Transportation control: Normal access to new waybills; sensor data collection frequency maintained at 30-minute intervals; flexible use of all optional paths planned by PCNN allowed; cargo status checks performed according to standard procedures.

[0321] Warehouse management: Allow warehouse nodes to flexibly adjust their processing capacity; implement routine inventory management mechanisms; maintain standard monitoring frequency.

[0322] Information control: Transaction information is processed in accordance with standard consensus mechanisms; dynamic correlation thresholds of the knowledge graph are allowed to be adjusted; smart contracts are executed according to standard procedures.

[0323] Combination Figure 2 As shown, the integrated logistics management system for cargo transportation of the present invention operates in the following five stages:

[0324] Phase 1: Multimodal Data Perception and Edge Processing

[0325] The system utilizes a multimodal collaborative sensing module deployed in a sensor network across transport vehicles, warehousing facilities, and loading / unloading areas to collect logistics status data such as temperature, humidity, vibration, and tilt angle. The sensor network employs a layered deployment approach: Core Monitoring Layer: High-precision sensors are deployed at key nodes such as loading / unloading points and warehouse entrances to ensure accurate status monitoring in critical areas. Collaborative Sensing Layer: Low-cost sensors are deployed in regular areas to extend the monitoring coverage.

[0326] Data is transmitted to edge computing nodes, where deep belief networks (DBNs) are used to extract hierarchical features from the collected multimodal data and generate hidden layer feature representations.

[0327] Federated learning model processing: Edge computing nodes train models based on local datasets and upload the model parameters to the central node. The central node integrates the parameters from each edge node using a weighted aggregation algorithm to generate compressed global feature data, providing real-time, multi-dimensional data support for the semantic reasoning module.

[0328] Phase Two: Semantic Reasoning and Knowledge Graph Construction

[0329] The system receives multi-dimensional state data from the multimodal collaborative perception module through the semantic reasoning module, and constructs a knowledge graph by combining order information, warehouse status, transportation routes, vehicle information, and cargo attributes. Node set Including order nodes Warehouse nodes Path nodes Vehicle nodes and cargo nodes Edge set and relation set Define the semantic relationships and association types between nodes respectively.

[0330] Graph Attention Network (GAT) is used to dynamically assign semantic weights between nodes and aggregate key information.

[0331] To address the information gaps in knowledge graphs, the TransE model is used for relation prediction and completion, and reliable completion results are selected through confidence assessment.

[0332] To support multimodal transport scenarios, the semantic reasoning module utilizes ontology mapping technology to construct a unified semantic model among road, rail, and air transport, providing structured knowledge support for the scenario adaptation decision-making module.

[0333] Phase 3: PCNN and Swarm Intelligence Routing Decision

[0334] Pulse Coupled Neural Network (PCNN) Modeling: The scene adaptation decision module maps the logistics network into a PCNN neuron matrix structure and dynamically calculates the activation threshold of neurons according to the following rules: node processing capacity and path characteristics (distance, travel time, transportation cost and reliability) are used as key decision indicators; the connection weights between neurons are generated by normalized weighted calculation based on the above indicators.

[0335] Swarm intelligence collaborative optimization mechanism: The scene adaptation decision module sets up multiple decision agents, each maintaining a state vector of its responsible area, including road condition index, vehicle availability, warehouse capacity, and emergency order ratio; neighboring agents periodically exchange state data through a hierarchical communication protocol; the initial routing scheme output by PCNN serves as a soft constraint for agent decision-making, and agents optimize routes based on the state information of their own area, with the optimization results dynamically updating the connection weights of PCNN.

[0336] The optimized routing scheme is then passed to the intelligent execution module, while simultaneously providing structured decision data to the logistics trust chain module.

[0337] Phase Four: Hierarchical Logistics Task Execution

[0338] Task decomposition: The intelligent execution module receives the routing scheme output by the scenario adaptation decision module and decomposes it into three levels of tasks: trunk transportation, branch line scheduling, and last-mile delivery. Each task must meet the constraints of time window, path access status, and node processing capacity.

[0339] Dynamic adjustment mechanism: When the perception module detects abnormal events (such as route interruption, vehicle failure, etc.), the intelligent execution module dynamically adjusts the task execution plan based on the abnormality detection results and the credit assessment provided by the logistics trust chain module to ensure the smooth completion of the task.

[0340] Real-time feedback mechanism: The task execution status is transmitted to the multimodal collaborative perception module in real time through the feedback module, forming a closed-loop control to support subsequent task optimization.

[0341] Phase 5: Building a Trust Chain and Preventing Risks in Logistics

[0342] Layered consensus mechanism: The logistics trust chain module is based on the role attributes of carriers, warehouses and consignees. It calculates cumulative credit scores by combining role-specific indicators such as on-time delivery rate, storage accuracy rate and settlement timeliness. Consensus weights are allocated through factors such as transaction frequency and transaction amount to realize trust assessment of all parties in the logistics network.

[0343] Credit Scoring and Propagation: The module evaluates credit scores based on indicators such as on-time delivery rate, transaction volume ratio, positive review rate, and complaint rate using a scoring model, and updates credit scores using a time decay mechanism; it also uses a trust propagation model to calculate trust relationships between nodes, thereby building a trust network.

[0344] Risk warning and collaborative control: Based on trust level and risk assessment results, a tiered early warning system is implemented, and differentiated control measures are adopted for different risk levels, including adjusting data collection frequency, restricting path selection, changing resource allocation methods, and adjusting consensus mechanism parameters, and collaborating with other modules to complete risk prevention and control.

[0345] Through the collaborative work of the above five stages, the integrated logistics management system for cargo transportation of this invention achieves closed-loop operation from data collection and processing, knowledge reasoning, intelligent decision-making to task execution and risk prevention. The coordinated cooperation between the various modules of the system ensures the controllability of the logistics process.

[0346] In summary, this invention provides an integrated logistics management system for cargo transportation. It constructs a complete technical framework from logistics scene perception and knowledge processing to task execution through a multimodal collaborative perception module's hierarchical perception network and edge computing processing, a semantic reasoning module's knowledge graph construction and ontology mapping, a scene adaptation decision module's PCNN and swarm intelligence collaboration, a hierarchical task management module for intelligent execution, and a hierarchical consensus mechanism for the logistics trust chain module. Through data interaction and feedback mechanisms between modules, the system achieves intelligent management of the entire logistics process, effectively solving the technical problems of existing logistics systems in areas such as data integration, dynamic monitoring, decision support, collaborative scheduling, and security traceability.

[0347] While this document illustrates specific applications of the invention through particular embodiments, these embodiments are for illustrative purposes only and do not imply limitation of the scope of protection of the invention. The scope of protection of the invention is defined by the claims, and any appropriate modifications, equivalent substitutions, or improvements based on the principles of the invention should be considered to fall within the scope of protection of the invention. Therefore, the scope of protection of the invention should be given the broadest interpretation to cover all such modifications, equivalent structures, and functions.

Claims

1. A cargo transportation integrated logistics management system, characterized by, Comprise: A multi-modal collaborative perception module for collecting multi-dimensional state data of a logistics scene through a deployed temperature, humidity, vibration, and inclination sensor network, implementing local processing of the data based on edge computing nodes and a federated learning model, and aggregating feature information at a central node; A semantic reasoning module for constructing a knowledge graph based on a graph attention network, uniformly modeling the multi-dimensional state data with order information, warehouse status, transportation path, vehicle information, and cargo attributes, and implementing semantic connection in a multi-modal transport scene through an ontology mapping method; A scene-adaptive decision-making module for optimizing adaptive routing of a logistics network under multiple constraints of path distance, transportation cost, and node processing capacity based on a pulse-coupled neural network (PCNN) and a swarm intelligence collaborative mechanism; the node processing capacity includes maximum processing capacity, average processing time, and cache capacity; the PCNN constructs the logistics network as a neuron matrix structure, with each neuron corresponding to a logistics node, and the connection weights between neurons being determined by normalized weighting calculation of path distance, transportation cost, and node processing capacity; the swarm intelligence collaborative mechanism sets multiple decision-making agents, each responsible for a specified area, exchanges state information through a hierarchical communication protocol, and optimizes the initial routing scheme output by the PCNN; An intelligent execution module for decomposing the routing scheme output by the scene-adaptive decision-making module into mainline transportation, branch scheduling, and end-of-line distribution tasks, and adjusting the task execution scheme based on real-time feedback mechanism; A logistics trust chain module for evaluating the trust relationship between carriers, warehouse operators, and consignees based on a hierarchical consensus mechanism and a dynamic credit scoring model, and generating graded early warning and dynamic control measures based on risk assessment.

2. The freight transportation integrated logistics management system of claim 1, wherein, The multi-modal collaborative perception module uses a deep belief network (DBN) to extract features from temperature, humidity, vibration, and inclination data, and optimizes the data coverage range through the core monitoring layer and collaborative perception layer of the hierarchical perception network.

3. The freight transportation integrated logistics management system of claim 1, wherein, The multi-modal collaborative perception module optimizes the aggregation model of the central node through the distributed training mechanism of federated learning, and the weight distribution is based on data integrity, freshness, and accuracy scores.

4. The freight transportation integrated logistics management system of claim 1, wherein, The semantic reasoning module predicts missing relationships between nodes through a knowledge graph completion algorithm, and filters reliable relationships through confidence evaluation.

5. The freight transportation integrated logistics management system of claim 1, wherein, The semantic reasoning module uses an ontology mapping method to realize semantic connection between road, rail, and air transportation, assigns semantic weights between nodes using a graph attention network, and highlights key information.

6. The freight transportation integrated logistics management system of claim 1, wherein, The scene-adaptive decision-making module models logistics nodes through a pulse-coupled neural network (PCNN), and the activation threshold of the node is dynamically calculated based on maximum processing capacity, average processing time, and cache capacity.

7. The freight transportation integrated logistics management system of claim 1, wherein, The scene-adaptive decision-making module optimizes the routing scheme through a swarm intelligence collaborative mechanism, and the decision-making agents adjust the path based on road condition index, vehicle availability, warehouse capacity, and emergency order proportion in the region.

8. The freight transportation integrated logistics management system of claim 1, wherein, The intelligent execution module dynamically decomposes and adjusts mainline transportation, branch scheduling, and end-of-line distribution tasks based on constraints of time window, path traffic state, and node processing capacity.

9. The freight transportation integrated logistics management system of claim 1, wherein, The logistics trust chain module calculates the trust relationship between nodes through a trust propagation model, and dynamically adjusts the data collection frequency, resource allocation mode and path planning strategy based on the abnormal event risk value.

Images