Logistics transportation personalized recommendation path planning method and system based on cloud platform

Abstract

The invention discloses a logistics transportation personalized recommendation path planning method and system based on a cloud platform, and belongs to the technical field of logistics transportation, and the method comprises the steps: constructing a heterogeneous traffic map model fusing road sections, geographic interest points and traffic event information; collecting historical transportation task records of a plurality of users, generating user semantic intention tags based on delivery behaviors, and mapping the user semantic intention tags to traffic map nodes and edge attributes to form a semantic constraint graph structure; performing joint modeling on the user intention and the path reachability by using a heterogeneous graph neural network to generate a path scoring sequence; after the transportation task is completed, abnormal events are collected, and the semantic intention label mapping relation is dynamically updated based on reverse alignment errors; the model and the mapping module are deployed on a cloud platform, the traffic state is obtained in real time in combination with edge equipment, and path recommendation online generation and rapid pushing are achieved; according to the method, the personalized matching degree and the real-time response capability of path recommendation can be effectively improved, and the method is suitable for intelligent logistics path planning in a complex scene.

Description

Technical Field

[0001] The present invention relates to the field of logistics and transportation technology, and in particular to a cloud platform-based logistics and transportation personalized recommendation route planning method and system. Background Art

[0002] Currently, mainstream logistics route recommendation methods mostly rely on static road networks and current traffic information for route planning, and lack in-depth modeling of users' transportation intentions and individual preferences. Especially in interference scenarios involving cross-regional, complex road network structures, and abnormal weather, the recommendation results often show delayed responses or do not meet user preferences, resulting in reduced transportation efficiency.

[0003] Some existing methods attempt to incorporate machine learning or graph neural networks (GNNs) to process dynamic traffic data, but these approaches are often limited to route optimization within a single-region graph structure and struggle to effectively scale to cross-regional routing scenarios. Furthermore, current models often focus solely on temporal and spatial variations in features, ignoring the varying needs of users for route semantics in different task contexts (e.g., avoiding specific locations, preferring a specific delivery sequence, etc.), resulting in a lack of personalized, "intent-driven" routing in recommended routes.

[0004] Therefore, there is an urgent need for a personalized route recommendation method that can integrate user semantic intent labels, abnormal event feedback mechanism and cross-regional heterogeneous graph modeling capabilities, especially to realize large-scale heterogeneous data training and real-time route push based on cloud platform architecture, so as to improve the intelligence level and robustness of the system. Summary of the Invention

[0005] The purpose of the present invention is to provide a cloud platform-based logistics transportation personalized recommendation path planning method and system to address the shortcomings of the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a cloud-based logistics transportation personalized recommendation route planning method, comprising: Construct a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Collect historical transportation task records of multiple users, generate user semantic intent labels based on their delivery behaviors and historical preferences, and map the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Based on the constructed semantically constrained traffic graph, a heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task, generating a path score sequence. After each transport task is completed, abnormal transport events that occur during its execution are collected, and the causal deviation relationship between them and the recommended route is analyzed. Based on the reverse alignment error between the deviation path and the true path, the semantic intent mapping relationship is dynamically updated. The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

[0007] Preferably, constructing a heterogeneous traffic graph model including multiple transport area nodes and path edges includes: Obtain vector road data, point of interest distribution data, and historical traffic event data covering multiple transportation areas, analyze their spatial location and time attributes, and initialize the corresponding road segment nodes, geographic point of interest nodes, and traffic event nodes; Based on the road topology and the distribution density of points of interest, the geographical adjacent edges between road nodes and points of interest nodes are constructed, and the edge weights are calculated using the regional proximity function based on multi-scale grid division; Based on the time lag relationship between historical traffic events and path execution results, event impact edges are constructed between event nodes and road nodes, and a decreasing weighting mechanism is used to characterize the impact of event persistence on path accessibility. Multi-dimensional attribute encoding is performed on all nodes and edge structures, including road grade, point of interest type, and event urgency, to generate a graph structure feature matrix that supports heterogeneous graph neural network input.

[0008] Preferably, generating a user semantic intent tag based on the user's delivery behavior and historical preferences includes: Slice the user's historical transport trajectory and construct a multi-dimensional behavior sequence by combining the task time period, transport type, and delivery scenario context labels. A behavioral feature encoding network with weighted attention mechanism is used to represent each behavioral sequence, generating a high-dimensional embedding vector that includes spatial path preference, temporal response characteristics, and behavioral frequency. A contrastive learning mechanism is introduced, using the behavior sequences of the same user in different scenarios as positive sample pairs and the behavior sequences of different users as negative sample pairs; Dense clustering is performed in the embedding space to identify multiple cluster centers with stable transportation preferences, and each center is assigned an interpretable semantic intent label.

[0009] Preferably, generating a path scoring sequence includes: The various nodes in the traffic graph are divided into multiple subgraphs according to their structural roles, and embedding spaces are constructed for each subgraph. The node feature alignment and semantic association initial modeling are achieved through meta-path guidance. The semantic intention label of the user's current transportation task is used as the graph state initialization factor, and the node set strongly related to the intention label is assigned a context activation weight to form the starting point of the semantically driven graph structure. A multi-view heterogeneous graph attention mechanism is used to aggregate different types of neighbor information from the structural view, preference view, and event interference view, dynamically capturing the semantic accessibility and behavioral fit of the path. The historical score dependency of candidate paths is regularized, and a user-customized path score sequence is finally output for sorting.

[0010] Preferably, dynamically updating the semantic intent mapping relationship based on the reverse alignment error between the deviation path and the true path includes: Perform node-level sequence alignment between the recommended path and the actual execution path, construct a space-time alignment matrix, identify path segments with significant deviations, and extract the corresponding path deviation node sets; For each node in the deviation node set, the frequency of its cumulative participation in the deviation path in historical tasks is calculated, and a node deviation confidence model is established based on the task context to measure the credibility of its semantic mapping; For nodes whose confidence is lower than the threshold, their original bound positive preference labels are replaced with context-sensitive labels, and the node semantic attribute vector is updated.

[0011] Preferably, the edge computing device is combined to obtain the local traffic status in real time, and the online generation of the recommended route results that meet the user's preferences includes: The local traffic status reported by edge devices is integrated into the cloud-based semantic traffic graph in the form of graph patches, and the graph structure is kept synchronized with the current traffic environment through a node-level incremental update mechanism. Based on the user intent label, the associated nodes in the graph are activated and the subgraph input of the task is constructed. The heterogeneous graph neural network model is used for multi-hop message transmission and path candidate generation. User context dynamic factors, including task urgency and cargo characteristics, are introduced to weight and rank candidate route scores to generate real-time route recommendation results.

[0012] The present invention also provides a cloud platform-based logistics transportation personalized recommendation route planning system, including: Heterogeneous graph construction module: Constructs a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Semantic intent generation module: collects historical transportation task records of multiple users, generates user semantic intent labels based on their delivery behaviors and historical preferences, and maps the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Heterogeneous graph modeling module: Based on the constructed semantically constrained traffic graph, the heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task and generate a path score sequence; Route Execution Feedback Module: After each transport task is completed, it collects abnormal transport events that occur during its execution, analyzes the causal deviation relationship between the deviation path and the recommended path, and dynamically updates the semantic intent mapping relationship based on the reverse alignment error between the deviation path and the actual path; Cloud-edge collaborative path recommendation module: The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

[0013] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. This invention constructs a heterogeneous traffic graph structure that integrates roads, points of interest, and traffic events, and introduces user semantic intent labels to constrain the graph structure. This allows the route recommendation model to not only consider physical accessibility but also deeply reflect user personalized preferences. Joint modeling through heterogeneous graph neural networks enables multidimensional route evaluation from the structural, semantic, and behavioral levels, significantly improving the matching accuracy of recommended routes and user satisfaction.

[0014] 2. This invention introduces a transport execution feedback mechanism, enabling closed-loop optimization of the route recommendation system through reverse alignment error analysis and adaptive updating of semantic intent labels. Combining the heterogeneous modeling capabilities of the cloud with the real-time perception capabilities of edge devices, the system enables low-latency, highly adaptable, personalized online route recommendations, significantly improving the intelligence and responsiveness of logistics and transportation in complex traffic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0016] Figure 1 This is a mind map of the method of the present invention.

[0017] Figure 2 This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 As shown, the embodiment of the present invention provides a method for planning personalized recommended routes for logistics transportation based on a cloud platform, including: Construct a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Collect historical transportation task records of multiple users, generate user semantic intent labels based on their delivery behaviors and historical preferences, and map the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Based on the constructed semantically constrained traffic graph, a heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task, generating a path score sequence. After each transport task is completed, abnormal transport events that occur during its execution are collected, and the causal deviation relationship between them and the recommended route is analyzed. Based on the reverse alignment error between the deviation path and the true path, the semantic intent mapping relationship is dynamically updated. The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

[0020] In this example, to achieve personalized recommendations for logistics transportation routes, the system first constructs a heterogeneous transportation graph model. This model is used to integrate multi-source transportation data and semantic information across multiple transportation areas, serving as the input for the subsequent graph neural network model. Specifically, it includes the following steps: The system first integrates OpenStreetMap road vector data, geographic point of interest (POI) data, and structured event data (such as accidents, construction, and traffic control) from a historical traffic event platform. It then spatially normalizes this data to a unified projected coordinate system.

[0021] For road vector data, it is cut into line segments according to its topological structure and defined as road segment nodes; For point of interest data, mark them as geographic point of interest nodes according to their types (such as warehouses, gas stations, distribution stations, transportation hubs, etc.); For traffic event data, metadata such as event occurrence time, affected area, and event type are extracted to generate traffic event nodes, and spatial mapping is performed based on the event center point location.

[0022] Each type of node is assigned a set of initial attribute vectors, including spatial coordinates, category code, timestamp, etc.

[0023] When constructing the edge structure between road nodes and POI nodes, a multi-scale regional gridding mechanism is introduced. The system divides the target city or region into multiple grid areas (e.g., 500m×500m, 1km×1km), calculates the density gradient of POIs and the similarity of road node distribution between grids, and forms a "geographical adjacency."

[0024] The edge weights are defined using the following proximity function: ; Where d(i,j) is the Euclidean distance between nodes i and j, is the density difference of interest points between grids, and α and β are adjustment parameters. The system connects node pairs with weights greater than the set threshold as geographic adjacency edges.

[0025] In view of the relationship between traffic event nodes and road nodes, the system adopts a time window-based hysteresis response modeling method. The specific operation is as follows: Traverse each traffic event node and calculate its impact on the traffic speed of surrounding roads and the actual execution status of the route (such as delays and diversions) within several time windows after its occurrence (such as 0-15 minutes, 15-30 minutes, etc.); Add event-affecting edges to the road segments affected by the event in each path, and use the following dynamic weighting function: ;in, The type and level are determined, and λ is the attenuation coefficient. This mechanism can be used to simulate the long-term and short-term interference of the event's persistence and propagation range on the traffic state.

[0026] Finally, to support the training of heterogeneous graph neural network models, the system uniformly encodes all nodes and edges into a graph structure feature matrix. Each node is embedded with the following attributes: Road nodes: road grade (highway / national highway / branch road), average speed, number of lanes; Point of interest nodes: POI type, service density, service time; Event node: event type (accident / construction), impact level, and duration; Each edge is embedded with a type label (adjacency / influence) and its corresponding weight. All graph structure data is input into the model training module in Tensor format, providing underlying graph data support for personalized path recommendations.

[0027] To enhance the personalization of logistics route recommendations, this example proposes a semantic constraint graph modeling method that combines user transportation behavior characteristics and historical preferences. This method generates user semantic intent labels and embeds them into the nodes and edges of the transportation graph. This method explicitly guides the model to capture user behavior patterns during route planning, achieving personalized route optimization.

[0028] The system collects behavioral records of multiple users in past transportation tasks from the logistics platform's historical order database, including but not limited to: The geographical coordinates of the origin and destination; Transport mission time (time period, working days / holidays); Actual execution path (GPS trajectory data); Transfer station and stop selection; User behavior that actively deviates from the recommended path; Task type (express delivery / cold chain / bulk materials, etc.) and cargo attributes.

[0029] The system cleans, denoises and standardizes the above data to form a set of transportation task sequences that can be used for behavioral analysis.

[0030] Each transport task is represented as a structured sequence containing multiple fields, enhanced with task context (such as weather, region, and road grade). To extract latent semantic features that can distinguish user preferences, the following user transport intention embedding model is constructed: Use a multi-layer Transformer structure to model the contextual features of each behavior sequence, capturing the temporal relationship and preference dependence between behaviors; Introducing a trajectory similarity attention mechanism to assign higher attention weights to key path nodes such as "repeated passage through a specific road section" and "non-main road preference"; The sequence embedding results of each user are summarized to obtain their transportation preference vector in multiple contextual scenarios.

[0031] The contrastive learning method is used to construct the embedding space, with "the behavior of the same user in different scenarios" as the positive sample pair and "similar task behaviors of different users" as the negative sample pair. The optimization model enables the embedding space to have the ability to distinguish users.

[0032] After embedding training is completed, dense clustering is performed in the embedding space. For each cluster center, interpretable semantics are extracted based on the representative behavior sequence, such as: “Prefer to detour via urban arterial roads”; “Prioritize highways + suburban transit warehouses”; “Delivery at a fixed time period every day + dependence on specific routes”.

[0033] Define a semantic intent label for each type of intent and establish a label index library.

[0034] The generated semantic intent labels are embedded into the heterogeneous traffic map, specifically: If the user prefers a certain type of road structure (e.g., avoiding urban roads), their intention label is mapped to the "negative preference" attribute of a specific road node; If users often select a specific transit warehouse or point of interest as a docking location, the corresponding point of interest node will be added with a "high priority" semantic intent label; If certain event nodes (such as construction and traffic restrictions) have a significant impact on user path deviation, a "sensitive label" is embedded on the event edge.

[0035] In this way, the original traffic map is enhanced with semantic intent to form a semantic constraint graph structure, which can be used as the structured input of the graph neural network model to realize path optimization reasoning explicitly driven by user behavior.

[0036] To effectively jointly model user semantic intent and path accessibility, this example proposes a path score generation method based on a multi-view heterogeneous graph neural network. This method not only fully utilizes the structural and semantic information in the traffic map, but also guides graph state initialization through user behavior and introduces a path decoupling mechanism to address the problem of redundant path inertia and improve the quality of personalized recommendations.

[0037] First, the semantically constrained traffic graph constructed above is divided into several subgraph structures, each representing a different structural role relationship: Structural subgraph: contains road segment nodes and their geographical adjacent edges, used to represent the physical accessibility structure; Preference subgraph: Contains interest point nodes and attribute edges related to user preferences, such as "preferred warehouse" edges and semantic matching edges; Disturbance subgraph: contains traffic event nodes and their “event impact edges” associated with path disturbances.

[0038] The above subgraphs are input into the graph embedding encoding module of different dimensions respectively. A meta-path guided method is adopted, such as the "point of interest-road-point of interest" or "event-road-warehouse" path, to construct a multi-semantic intersection space, and semantically align the node features to generate an initial embedding vector shared across subgraphs.

[0039] While traditional routing models use a starting point or location pair as input, this method uses the semantic intent labels of the user's current transportation task as guidance. The system first activates tags based on task requirements, such as "avoid urban areas," "prioritize highways and fixed warehouses," and "prioritize timeliness."

[0040] For nodes that highly match the label (such as road segments and points of interest that frequently appear in the preference category), activation weights are assigned in the graph, and semantic biases are superimposed on the initial embedding vector to form a task-specific semantic activation graph state, thereby skipping the traditional "position-space" mapping process and directly forming the initial state of the graph structure driven by user preferences.

[0041] After building a unified graph structure, the system introduces a three-view heterogeneous attention aggregation mechanism: Structural view: encodes physical adjacency relationships and calculates the structural consistency of road topology accessibility; Preference view: encodes semantic edges of interest point nodes and extracts the influence of interest regions consistent with user preferences; Perturbation view: Encodes the dynamic relationship between traffic event nodes and paths, and is used to identify risk areas that the mission may encounter.

[0042] Each node calculates the type attention weight and performs weighted fusion based on the different types of features of its neighbors in three views to obtain a node representation with comprehensive "structure-semantics-disturbance" features.

[0043] To prevent the scoring model from being affected by historical path inertia, the system designs a path dependency decoupling module: Compare the performance of candidate paths in historical tasks to identify potential redundancy bias (e.g., a certain path is always recommended first). Introducing a regularization penalty term to reduce the weight of high-frequency paths and low-quality paths in path score calculation; Perform the cumulative scoring function on each candidate path: ;in is the final representation of the i-th node in the path, is the weight coefficient, is the path history deviation score, λ is the adjustment parameter, and n is the total number of nodes.

[0044] Finally, a personalized path scoring sequence is generated for the user's current task, and the paths with the highest scores are output as recommendation results.

[0045] To enhance the self-correction capabilities of route recommendation systems within a closed-loop user feedback loop, this embodiment proposes a dynamic optimization method for semantic intent based on a reverse alignment error feedback mechanism. This method analyzes behavioral differences between recommended and actual routes, identifies nodes with semantic mapping deviations, and adaptively updates their corresponding labels and graph structure, thereby enhancing the recommendation system's ability to model users' true preferences.

[0046] After each transport task is completed, the system records two path sequences: System recommended path: obtained by heterogeneous graph neural network scoring and sorting; Actual execution path: obtained by restoring vehicle trajectory data (such as GPS).

[0047] The system uses a node-level dynamic time warping algorithm to perform path comparison, construct a space-time alignment matrix between the recommended path and the actual path, and identify locations where there are obvious deviations between path segments.

[0048] For areas where persistent deviation occurs (e.g., users detour around the same road section or point of interest in multiple tasks), the corresponding node set is extracted and recorded as the path deviation node set. , and record relevant context information (time, area, event impact, etc.).

[0049] right For each node n in the , count the frequency of its participation in the deviation path in the past N tasks , and combined with the consistency measurement of the task context (such as whether it is the same time period, the same type of goods, and the same intention label) to build a node deviation confidence model: ;in, is the context consistency coefficient, σ is the Sigmoid normalization function, and the output result , represents the semantic credibility of the node. If the value is lower than the set threshold (such as 0.3), it is considered that the semantic intent of the current binding is deviated.

[0050] For low-confidence nodes, the system introduces a semantic intent label adaptive transfer mechanism, which includes: Label de-weighting: Reduce the weight of the original bound label (such as "preferred route") in the graph embedding, and adjust its influence on the recommendation model by updating the semantic attribute vector of the node; Semantic level transfer: If the original label of a node is "high preference", it will be downgraded to "conditional preference" or "context-sensitive" if there is repeated avoidance behavior, reflecting its non-universal adaptability; Category reclassification: Combining the semantic embedding space with the cluster center changes, re-evaluate the category of the node in the semantic space and synchronously update its classification identification in the label index library.

[0051] These updates are automatically performed using the K-nearest neighbor relocalization algorithm in the embedding space, ensuring that label transfers are both semantically consistent and behaviorally compatible.

[0052] In addition to updating node attributes, the system also dynamically adjusts the structural relationship of deviation nodes in the semantic graph: If a node frequently misleads the path score propagation, the edge weight weakening operation is performed to reduce the intensity of information flow between it and other preferred nodes; If a node has been bypassed for a long time and has lost its semantic representativeness, a local node blocking mechanism can be triggered to temporarily remove it from the candidate path score in future tasks; The updated node and edge structures are synchronously written into the heterogeneous graph database and used as input for the next round of task path planning model, thereby realizing a semantic graph closed-loop self-optimization mechanism based on reverse alignment error.

[0053] This embodiment proposes a cloud-based route scoring and recommendation push method based on real-time graph structure updates and user preference tags. It combines a heterogeneous graph neural network model with a cloud-edge collaborative communication mechanism to achieve efficient and personalized route recommendation capabilities. It is particularly suitable for complex scenarios with drastic changes in traffic dynamics or significant user individual preferences.

[0054] To ensure that the input graph structure of the recommendation model is synchronized with the actual traffic status, the system designs an edge graph patch mechanism. Traffic data collected by edge devices (such as vehicle density, road congestion, event broadcasts, etc.) is structured into local graph patches. The format includes: New or modified nodes (such as new accident nodes and speed reduction nodes); Changes in edge weights (e.g., increased road travel time); Node status changes (such as changes in the business status of a point of interest).

[0055] The cloud graph database uses a node-level hash index matching mechanism to compare the patch with the original graph structure and perform incremental node and edge updates to avoid the overhead of full graph reconstruction, thereby achieving millisecond-level graph synchronization.

[0056] When a user initiates a route request, the system reads the current transport task parameters and preference intent labels (such as "avoid the city center," "prioritize expressways," and "fixed delivery stations") and constructs a task subgraph by filtering nodes and edges in the real-time graph structure that meet the following conditions: Points of interest or road segments that are highly correlated with the preference tags; The starting and ending points and the main relay nodes on the reachable paths; Traffic anomaly nodes currently affected by edge graph patches.

[0057] This subgraph serves as the input of the heterogeneous graph neural network model, ensuring that the model calculation range is limited to areas that are strongly related to the task, thereby improving real-time performance and computing efficiency.

[0058] To achieve more refined path sorting, the system introduces a user task context factor weighting mechanism in the path scoring stage. This factor includes but is not limited to: The urgency of the task (e.g., whether it is a time-limited delivery); Type of cargo (e.g., perishable, heavy load); Records of user's current behavior deviations (such as recent continuous modifications to recommended routes).

[0059] The system dynamically adds adjustment factors to the scoring function of candidate paths: ;in Score the model base, is the influence function of the i-th context factor on the path, The weight coefficient is used. After sorting the scoring results, the top-N paths are selected as the recommended output.

[0060] In order to quickly transmit the recommendation results to the user terminal, the system designs an adaptive communication channel selection mechanism: If the current network condition is stable and the latency is low, the system will prioritize pushing the route data to the vehicle terminal via the 5G direct link; If the network fluctuates greatly or the edge device is offline, the system will cache the path results in the local edge transit node and automatically transmit them back when the last connection is made, or wait until the device comes online before pushing them. The recommended data format includes node sequence, steering instructions, expected time and semantic annotations, ensuring that the navigation system can directly parse and execute it.

[0061] Ultimately, after receiving route recommendations, end users can obtain context-related, real-time and efficient personalized route guidance without having to wait for large-scale graph updates and complex model reasoning.

[0062] Example 2, please refer to Figure 2 As shown, the embodiment of the present invention provides a personalized recommended route planning system for logistics transportation based on a cloud platform, including: Heterogeneous graph construction module: Constructs a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Semantic intent generation module: collects historical transportation task records of multiple users, generates user semantic intent labels based on their delivery behaviors and historical preferences, and maps the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Heterogeneous graph modeling module: Based on the constructed semantically constrained traffic graph, the heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task and generate a path score sequence; Route Execution Feedback Module: After each transport task is completed, it collects abnormal transport events that occur during its execution, analyzes the causal deviation relationship between the deviation path and the recommended path, and dynamically updates the semantic intent mapping relationship based on the reverse alignment error between the deviation path and the actual path; Cloud-edge collaborative path recommendation module: The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

[0063] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A cloud-based logistics and transportation personalized recommendation route planning method, characterized by: include: Construct a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Collect historical transportation task records of multiple users, generate user semantic intent labels based on their delivery behaviors and historical preferences, and map the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Based on the constructed semantically constrained traffic graph, a heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task, generating a path score sequence. After each transport task is completed, abnormal transport events that occur during its execution are collected, and the causal deviation relationship between them and the recommended route is analyzed. Based on the reverse alignment error between the deviation path and the true path, the semantic intent mapping relationship is dynamically updated. The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

2. The method for personalized recommended route planning for logistics transportation based on a cloud platform according to claim 1, characterized in that: The construction of a heterogeneous transportation graph model containing multiple transportation area nodes and path edges includes: Obtain vector road data, point of interest distribution data, and historical traffic event data covering multiple transportation areas, analyze their spatial location and time attributes, and initialize the corresponding road segment nodes, geographic point of interest nodes, and traffic event nodes; Based on the road topology and the distribution density of points of interest, the geographical adjacent edges between road nodes and points of interest nodes are constructed, and the edge weights are calculated using the regional proximity function based on multi-scale grid division; Based on the time lag relationship between historical traffic events and path execution results, event impact edges are constructed between event nodes and road nodes, and a decreasing weighting mechanism is used to characterize the impact of event persistence on path accessibility. Multi-dimensional attribute encoding is performed on all nodes and edge structures, including road grade, point of interest type, and event urgency, to generate a graph structure feature matrix that supports heterogeneous graph neural network input.

3. The method for personalized recommended route planning for logistics transportation based on a cloud platform according to claim 1, characterized in that: Generating user semantic intent labels based on user delivery behavior and historical preferences includes: Slice the user's historical transport trajectory and construct a multi-dimensional behavior sequence by combining the task time period, transport type, and delivery scenario context labels. A behavioral feature encoding network with weighted attention mechanism is used to represent each behavioral sequence, generating a high-dimensional embedding vector that includes spatial path preference, temporal response characteristics, and behavioral frequency. A contrastive learning mechanism is introduced, using the behavior sequences of the same user in different scenarios as positive sample pairs and the behavior sequences of different users as negative sample pairs; Dense clustering is performed in the embedding space to identify multiple cluster centers with stable transportation preferences, and each center is assigned an interpretable semantic intent label.

4. The method for personalized recommended route planning for logistics transportation based on a cloud platform according to claim 1, characterized in that: Generating a path scoring sequence includes: The various nodes in the traffic graph are divided into multiple subgraphs according to their structural roles, and embedding spaces are constructed for each subgraph. The node feature alignment and semantic association initial modeling are achieved through meta-path guidance. The semantic intention label of the user's current transportation task is used as the graph state initialization factor, and the node set strongly related to the intention label is assigned a context activation weight to form the starting point of the semantically driven graph structure. A multi-view heterogeneous graph attention mechanism is used to aggregate different types of neighbor information from the structural view, preference view, and event interference view, dynamically capturing the semantic accessibility and behavioral fit of the path. The historical score dependency of candidate paths is regularized, and a user-customized path score sequence is finally output for sorting.

5. The method for personalized recommended route planning for logistics transportation based on a cloud platform according to claim 1, characterized in that: The dynamic updating of the semantic intent mapping relationship based on the reverse alignment error between the deviation path and the true path includes: Perform node-level sequence alignment between the recommended path and the actual execution path, construct a space-time alignment matrix, identify path segments with significant deviations, and extract the corresponding path deviation node sets; For each node in the deviation node set, the frequency of its cumulative participation in the deviation path in historical tasks is calculated, and a node deviation confidence model is established based on the task context to measure the credibility of its semantic mapping; For nodes whose confidence is lower than the threshold, their original bound positive preference labels are replaced with context-sensitive labels, and the node semantic attribute vector is updated.

6. The method for personalized recommended route planning for logistics transportation based on a cloud platform according to claim 1, characterized in that: The combination of edge computing devices to obtain real-time local traffic status and generate online recommended route results that meet user preferences includes: The local traffic status reported by edge devices is integrated into the cloud-based semantic traffic graph in the form of graph patches, and the graph structure is kept synchronized with the current traffic environment through a node-level incremental update mechanism. Based on the user intent label, the associated nodes in the graph are activated and the subgraph input of the task is constructed. The heterogeneous graph neural network model is used for multi-hop message transmission and path candidate generation. User context dynamic factors, including task urgency and cargo characteristics, are introduced to weight and rank candidate route scores to generate real-time route recommendation results.

7. A cloud-based personalized recommendation route planning system for logistics and transportation, used to implement the cloud-based personalized recommendation route planning method for logistics and transportation according to any one of claims 1 to 6, characterized in that: include: Heterogeneous graph construction module: Constructs a heterogeneous traffic graph model containing multiple transport area nodes and path edges. The traffic graph nodes include road segment nodes, geographical point of interest nodes, and traffic event nodes. The path edges include time-sensitive relationship edges, geographical adjacency edges, and event impact edges. Semantic intent generation module: collects historical transportation task records of multiple users, generates user semantic intent labels based on their delivery behaviors and historical preferences, and maps the semantic intent labels to the node and edge attributes of the transportation graph to form a semantic constraint graph structure; Heterogeneous graph modeling module: Based on the constructed semantically constrained traffic graph, the heterogeneous graph neural network model is used to jointly model the user intention and path accessibility of each transportation task and generate a path score sequence; Route Execution Feedback Module: After each transport task is completed, it collects abnormal transport events that occur during its execution, analyzes the causal deviation relationship between the deviation path and the recommended path, and dynamically updates the semantic intent mapping relationship based on the reverse alignment error between the deviation path and the actual path; Cloud-edge collaborative path recommendation module: The heterogeneous graph neural network model and semantic intent mapping module are deployed on the core computing nodes of the cloud platform, and combined with edge computing devices to obtain local traffic status in real time, so as to generate recommended path results that meet user preferences online.

Images