An improved graph neural network-based logistics efficient transportation method and system

Abstract

The present application relates to the technical field of logistics transportation, and more particularly to a logistics efficient transportation method and system based on an improved graph neural network, comprising: obtaining logistics basic data and collecting the same to form transportation data; analyzing transportation status to obtain a dynamic transportation relationship graph; performing a trust evaluation to obtain a trusted transportation relationship graph; inputting the improved graph neural network to obtain a transportation correlation result; identifying affected tasks to form a linkage task group; jointly adjusting to generate a candidate transportation scheme; combining the dynamic transportation relationship graph to determine a target scheme; sending an adjustment instruction and updating the transportation data according to feedback to generate a transportation result. The present application can preform a multi-vehicle joint execution state and exclude a new resource competition scheme, which is conducive to avoiding secondary congestion.

Description

Technical Field

[0001] This invention relates to the field of logistics and transportation, and in particular to an efficient logistics transportation method and system, electronic equipment and storage medium based on an improved graph neural network. Background Technology

[0002] In multi-warehouse collaborative delivery scenarios in urban core areas, multiple delivery nodes need to coordinate to complete delivery tasks to commercial outlets and community service points. Vehicle transportation is not only constrained by road restrictions, delivery windows, vehicle capacity, and loading and unloading conditions, but also affected by order priority and shared road and loading / unloading resources. When road congestion, parking space occupancy, or order adjustments occur during transportation, the original transportation plan may be inconsistent with the actual transportation status, requiring timely identification of affected vehicles, orders, and subsequent tasks.

[0003] Existing technologies typically construct transportation networks based on delivery nodes, road networks, vehicle locations, and order information, and calculate delivery routes with shorter transit times or lower costs based on road connectivity. When road conditions or order information change, the system generally replans the local routes of affected vehicles or updates the overall transportation plan according to a set schedule.

[0004] However, existing technologies primarily analyze road conditions and individual vehicle routes, making it difficult to accurately determine the impact of changes in transportation status on order execution and other related tasks. When rerouting routes for multiple affected vehicles, each vehicle may collectively choose the same alternative routes, loading / unloading berths, or delivery times, creating new resource competition and causing vehicle waiting and delays in subsequent tasks. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies, such as the difficulty in accurately identifying the impact of changes in transportation status on vehicles, orders, and subsequent tasks, and the difficulty in avoiding new conflicts of shared resources caused by partial detours of multiple vehicles. This invention proposes an efficient logistics transportation method and system, electronic equipment, and storage medium based on an improved graph neural network.

[0006] To at least solve some of the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] Firstly, this invention provides a logistics efficiency transportation method based on an improved graph neural network, including:

[0008] Acquire basic logistics data and perform aggregation processing on the basic logistics data to form transportation data;

[0009] Perform a transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph;

[0010] A reliability assessment is performed on the dynamic transportation relationship graph to obtain a reliable transportation relationship graph;

[0011] The reliable transportation relationship graph is input into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results;

[0012] Based on the transportation association results, the affected tasks are identified and grouped to obtain the linked task group;

[0013] The joint task group performs a joint adjustment analysis of the transportation plan to generate candidate transportation plans;

[0014] Based on the candidate transportation schemes and the dynamic transportation relationship diagram, a preliminary simulation is performed to determine the target scheme;

[0015] According to the target plan, generate and send adjustment instructions, obtain feedback data, update transportation data, and generate transportation results.

[0016] Preferably, a transportation status correlation analysis is performed on the transportation data to obtain a dynamic transportation relationship graph, including:

[0017] Perform state difference analysis on the transportation data to obtain transportation change events;

[0018] Perform association extraction centered on the object corresponding to the transportation change event to obtain the event association object;

[0019] Perform node mapping and relationship construction on the event-related objects to obtain an initial transportation relationship graph;

[0020] Resource competition quantification is performed on the initial transportation relationship graph to obtain a dynamic transportation relationship graph.

[0021] Preferably, a reliability assessment is performed on the dynamic transportation relationship graph to obtain a reliable transportation relationship graph, including:

[0022] The dynamic transportation relationship diagram is subjected to relationship basis extraction to obtain relationship basis data;

[0023] Based on the data, a timeliness and observation analysis is performed on the relationship to obtain relationship evaluation data;

[0024] Perform a credibility calculation on the relationship evaluation data to obtain reliable relationship data;

[0025] The trusted relationship data is written into the dynamic transportation relationship graph to obtain the trusted transportation relationship graph.

[0026] Preferably, the reliable transportation relationship graph is input into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results, including:

[0027] The reliable transportation relationship graph is processed to obtain the transportation graph input data;

[0028] Perform channel-specific aggregation on the input data of the transportation map to obtain a representation of the transportation impact;

[0029] The transportation impact is represented by the original plan preserved and merged to obtain the merged node representation;

[0030] Multi-task analysis is performed on the fusion node representation to obtain transportation association results.

[0031] Preferably, affected tasks are identified and grouped based on the transportation association results to obtain a coordinated task group, including:

[0032] The affected tasks are obtained by performing threshold filtering on the transportation association results;

[0033] Starting with the affected tasks, perform association expansion along the dynamic transportation relationship graph to obtain a task association set;

[0034] Perform an intersection determination on the task association set, and merge sets with common tasks into candidate task groups;

[0035] The candidate task groups are deduplicated and aggregated to obtain the linked task groups.

[0036] Preferably, the joint transportation plan adjustment analysis is performed on the coordinated task group to generate candidate transportation plans, including:

[0037] Adjustment options are generated for the linked task group to obtain adjustment candidate data;

[0038] The adjusted candidate data are combined to obtain an initial transportation plan;

[0039] Perform transportation constraint verification on the initial transportation plan to obtain a feasible transportation plan;

[0040] The feasible transportation schemes are deduplicated and aggregated to generate candidate transportation schemes.

[0041] Preferably, a preliminary simulation is performed based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme, including:

[0042] The candidate transportation schemes are written into the dynamic transportation relationship graph and the state is updated to obtain a temporary transportation relationship graph.

[0043] Perform joint state calculations on the temporary transportation relationship diagram to obtain scheme simulation data;

[0044] The simulation data of the proposed scheme is subjected to normalized evaluation to obtain the scheme evaluation data;

[0045] Threshold filtering is performed on the evaluation data of the proposed schemes to determine the target scheme.

[0046] Preferably, adjustment instructions are generated and sent according to the target plan, and after obtaining feedback data, the transportation data is updated to generate transportation results, including:

[0047] Based on the target scheme, an adjustment instruction is generated and sent, and instruction sending data is obtained;

[0048] Data collection and execution feedback are sent according to the instructions to obtain feedback data;

[0049] The feedback data is written into the transportation data execution status update to obtain the feedback transportation data;

[0050] The execution plan is determined based on the feedback transportation data, and a transportation result is generated.

[0051] Preferably, joint state calculations are performed on the temporary transport relationship diagram to obtain scheme simulation data, including:

[0052] Perform a task timing analysis on the temporary transportation relationship diagram to obtain the estimated arrival time and the estimated total time of the plan;

[0053] Perform a delivery discrepancy analysis on the estimated arrival time to obtain the estimated order timeout amount;

[0054] Perform competitive increment analysis on the temporary and dynamic transportation relationship diagrams to obtain the new resource competition volume;

[0055] Perform a scheme change statistics analysis on the candidate transportation schemes to obtain the scheme change quantity;

[0056] The estimated total time taken for the proposed solution, the estimated timeout of orders, the amount of new resource competition, and the amount of solution changes are aggregated to obtain the solution simulation data.

[0057] Secondly, the present invention provides a high-efficiency logistics transportation system based on an improved graph neural network, comprising:

[0058] The data sensing unit is used to acquire basic logistics data and perform aggregation processing on the basic logistics data to form transportation data;

[0059] The relationship identification unit is used to perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph; perform a credibility assessment on the dynamic transportation relationship graph to obtain a credible transportation relationship graph; and input the credible transportation relationship graph into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results.

[0060] The task grouping unit is used to identify affected tasks and group them according to the transportation association results to obtain a coordinated task group.

[0061] The scheme verification unit is used to perform joint adjustment and analysis of transportation schemes on the coordinated task group to generate candidate transportation schemes; perform a pre-rehearsal based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme; generate and send adjustment instructions according to the target scheme, update transportation data after obtaining feedback data, and generate transportation results.

[0062] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method in the first aspect.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] I. This invention performs a reliability assessment on the dynamic transportation relationship diagram, retaining transportation relationships that have not yet completely failed, while weakening unreliable relationships formed by outdated data, single-source records, and short-term abnormal fluctuations. This helps to avoid large-scale misadjustments of transportation tasks caused by occasional abnormal data.

[0065] Second, by handling the transportation execution relationship and the resource competition relationship separately, this invention can determine whether the task still has execution capacity, whether the delay will spread to subsequent tasks, and whether multiple tasks are competing for resources. This helps to avoid classifying different transportation anomalies as road delays and avoid local anomalies causing large-scale unnecessary changes to transportation plans.

[0066] Third, by calculating the transportation status of multiple vehicles jointly executing candidate transportation schemes before the adjustment instruction is issued, and excluding schemes that would further increase competition for road or loading and unloading resources, this invention helps to avoid secondary congestion after a single vehicle's feasible scheme is jointly executed. Attached Figure Description

[0067] Figure 1 This is a flowchart of the method of the present invention.

[0068] Figure 2 This is a comparison diagram of the effects of the present invention and the prior art. Detailed Implementation

[0069] To make the technical means, creative features, achieved objectives, and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention.

[0070] A specific embodiment of the present invention discloses an efficient logistics transportation method based on an improved graph neural network. Please refer to [link to relevant documentation]. Figure 1 ,include:

[0071] S1. Obtain basic logistics data and perform aggregation processing on the basic logistics data to form transportation data;

[0072] S2. Perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship diagram;

[0073] S3. Perform a reliability assessment on the dynamic transportation relationship diagram to obtain a reliable transportation relationship diagram;

[0074] S4. Input the credible transportation relationship graph into a pre-constructed improved graph neural network to analyze the transportation impact and obtain the transportation correlation results;

[0075] S5. Based on the transportation association results, identify the affected tasks and group them to obtain the linkage task group;

[0076] S6. Perform joint adjustment and analysis of transportation plans on the linked task group to generate candidate transportation plans;

[0077] S7. Perform a preliminary simulation based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme;

[0078] S8. Generate and send adjustment instructions according to the target plan, update transportation data after obtaining feedback data, and generate transportation results.

[0079] As an optional implementation, the specific steps in step S1 include:

[0080] The data sensing unit sequentially receives order records sent by the order management terminal, vehicle records sent by the vehicle terminal, road records sent by the road information platform, service resource records sent by the loading and unloading resource management terminal, and execution records sent by the vehicle terminal and the loading and unloading resource management terminal.

[0081] The order record includes the order identifier, pickup location, delivery location, goods demand, delivery window, estimated loading and unloading time, order status, and order execution sequence; the vehicle record includes the vehicle identifier, current location, remaining capacity, assigned order identifier, current travel segment sequence, planned arrival time, and estimated arrival time; the road record includes the road segment identifier, road segment connection relationship, baseline travel time, real-time travel time, road capacity, traffic restrictions, and update time; the service resource record includes the berth identifier, reservation order identifier, available number of berths, estimated release time, reservation start time, and original reservation release time; and the execution record includes the order execution status, order completion status, actual vehicle travel segment, actual travel time, and actual loading and unloading status.

[0082] The data sensing unit collects and processes the aforementioned basic logistics data. This collection and processing involves converting each record to the unified system time of the logistics scheduling server, and then matching the objects according to the order identifier, vehicle identifier, road segment identifier, and parking space identifier.

[0083] After collection, transportation data is generated. Each piece of transportation data records the object identifier, status time, status content, and data source. In this embodiment, one task corresponds to the transportation execution arrangement of one order. The estimated arrival time of the order refers to the estimated time when the vehicle will arrive at the order delivery location.

[0084] As an alternative implementation method, road congestion, parking delays, and order adjustments typically only affect a portion of the tasks. Continuing to use the existing relationships may easily miss the real changes, while rebuilding all relationships would expand the scope of adjustments. In view of this, the present invention proposes the following working method:

[0085] S2. Perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship diagram;

[0086] Transportation status correlation analysis refers to comparing the current status of the same object with the previous valid status to identify the object that can change the transportation schedule, and then extracting related tasks around the object and establishing transportation relationships. The previous valid status is the most recent status record with complete key fields for the same object before the current record.

[0087] S2.1 Perform state difference analysis on the transportation data to obtain transportation change events;

[0088] Specifically, the relationship identification unit first reads the real-time travel time and the baseline travel time of each road segment from the transportation data; when the real-time travel time is higher than the baseline travel time, the difference between the two is divided by the baseline travel time to obtain the travel change ratio; when the real-time travel time is not higher than the baseline travel time, the travel change ratio is recorded as 0; when the travel change ratio reaches the travel change threshold, the road segment identifier of the corresponding road segment is read, the event category is recorded as road change, and a road change event is formed.

[0089] Subsequently, the estimated release time and original scheduled release time of each loading and unloading berth are read from the transportation data; when the estimated release time is later than the original scheduled release time, the time difference between the two is taken as the resource delay; when the estimated release time is not later than the original scheduled release time, the resource delay is recorded as 0; when the resource delay reaches the resource change threshold, the berth identifier of the corresponding loading and unloading berth is read, the event category is recorded as resource change, and a resource change event is formed.

[0090] Next, read the estimated arrival time and planned arrival time of each vehicle from the transportation data; when the estimated arrival time is later than the planned arrival time, the time difference between the two is taken as the vehicle delay; when the estimated arrival time is not later than the planned arrival time, the vehicle delay is recorded as 0; when the vehicle delay reaches the execution change threshold, read the vehicle identifier of the corresponding vehicle, record the event category as vehicle change, and form a vehicle change event.

[0091] Then, the current order status record and the previous complete order status record for each order are read from the transportation data. The previous complete order status record is the one that is closest in time to the current order status record, and whose order identifier, delivery window, order status, order execution sequence, and completion status are all clearly recorded.

[0092] If there is no previous complete order status record, the current order is identified as a new order, the corresponding order identifier is read, the event category is recorded as order change, and an order change event is formed.

[0093] If a previous complete order status record exists, the delivery time window, order status, order execution sequence, and completion status of the two records are compared item by item. If at least one of them is different, the corresponding order identifier is read, the event category is recorded as an order change, and an order change event is formed.

[0094] The passage change threshold, resource change threshold, and execution change threshold represent the minimum level to which a change in the corresponding status needs to trigger a transport relationship update. Each threshold can be determined using the 95th percentile of the corresponding change in the most recent 30 days of normal transport records, ensuring that minor daily fluctuations do not trigger transport relationship updates.

[0095] Road change events, resource change events, vehicle change events, and order change events are aggregated to obtain transportation change events. Each transportation change event records the event category and the identifier of the object of change.

[0096] This approach only identifies changes in state that substantially alter transportation arrangements as events, which helps reduce unnecessary relationship adjustments caused by normal fluctuations.

[0097] S2.2. Perform association extraction centered on the object corresponding to the transportation change event to obtain the event association object;

[0098] The relationship identification unit reads the vehicle identifier and the vehicle-assigned order identifier from the transportation data, and writes the corresponding vehicle identifier and order identifier into the same record to form a carrier correspondence record.

[0099] Subsequently, orders carried by the same vehicle are identified based on the corresponding transport records. The order execution sequence is then read from the corresponding order records, and the order identifiers are arranged according to the order execution sequence to form a task sequence record.

[0100] Then, read the vehicle identifier and the road segment identifier in the current route from the transportation data, and write the vehicle identifier and the road segment identifier arranged in the order of passage into the same record to form a route correspondence record.

[0101] Then, the order identifier, berth identifier, reservation start time, and original reservation release time are read from the transportation data, and the above information is written into the same record to form a berth reservation record.

[0102] For road change events, the system searches for vehicles that include the changed road section based on the route corresponding record, and then searches for orders that the corresponding vehicles are carrying but have not yet been completed based on the carrier corresponding record. Subsequently, the system determines the corresponding order and its subsequent orders based on the task sequence record, and searches for the corresponding loading and unloading berth based on the berth reservation record.

[0103] For resource change events, the system searches for orders that reserved the changed berths based on berth reservation records, and then searches for the corresponding vehicles based on the corresponding transport records. Subsequently, the system determines the corresponding orders and their subsequent orders based on the task sequence records, and searches for the road segments traversed by the corresponding vehicles based on the route records.

[0104] For vehicle change events, the corresponding vehicle is used to find the orders that have not yet been completed based on the corresponding transport records. Then, the corresponding orders and their successors are determined based on the task sequence records. Subsequently, the corresponding road segments and loading / unloading berths are found based on the route records and berth reservation records.

[0105] For order change events, the corresponding carrier vehicle is found based on the carrier correspondence record, and then the corresponding order, adjacent preceding orders and all subsequent orders are found based on the task sequence record. Subsequently, the road segment and loading / unloading berth corresponding to the above orders are found based on the route correspondence record and berth reservation record.

[0106] The found vehicle identifiers, order identifiers, road segment identifiers, and parking space identifiers are deduplicated and aggregated, and the corresponding transportation records are written together to obtain the event-related objects.

[0107] By using carrier correspondence records, task sequence records, route correspondence records, and berth reservation records generated from transportation data to find related objects, the scope of the event's impact can be limited to objects that have an actual transportation relationship with the changed object.

[0108] When performing the transportation status association analysis for the first time, vehicles, orders, roads, and loading / unloading resources that have not yet been completed in the transportation data are mapped to corresponding nodes. If a previous dynamic transportation relationship graph already exists, the previous dynamic transportation relationship graph is copied, and the event-related objects are used as the scope of the current node and relationship update; nodes and relationships not included in the event-related objects retain the records from the previous dynamic transportation relationship graph.

[0109] S2.3 Perform node mapping and relationship construction on the event-related objects to obtain an initial transportation relationship graph;

[0110] The vehicle node is written with its current location, remaining capacity, and vehicle latency.

[0111] The order node records the delivery window, cargo demand, and remaining time window margin. Recording tasks in sequence, starting from the vehicle's current location, the real-time travel time of each road segment up to the corresponding order's delivery location and the estimated loading / unloading time of each preceding order are added sequentially to obtain the estimated arrival time of the corresponding order. Then, the estimated arrival time is obtained by subtracting the estimated arrival time from the latest delivery time of the order.

[0112] The road node contains real-time traffic time, road capacity, and update time. The loading / unloading resource node contains the number of services available, estimated release time, and scheduled time slot.

[0113] Subsequently, a transportation relationship is established between nodes. When the remaining capacity of the vehicle is not less than the order's cargo requirements, the vehicle meets the traffic restrictions of the road corresponding to the order, and the estimated arrival time is not later than the latest delivery time of the order, a transportation feasibility relationship is established between the vehicle node and the order node, indicating that the vehicle can execute the corresponding order under the current capacity, road, and time conditions.

[0114] When two road segments are connected one after the other in the current route and both are passable, a road passage relationship is established between the corresponding road nodes, indicating that vehicles can continuously enter the next road segment from the previous one.

[0115] When two orders are arranged sequentially in the task sequence record, a task priority relationship is established between the corresponding order nodes, indicating that the completion time of the previous order will affect the start time of the next order.

[0116] If the route records of two tasks contain the same road segment, or if the berth reservation records of two tasks contain the same loading / unloading berth, a resource sharing relationship is established between the corresponding order nodes, and the shared resource identifier is recorded.

[0117] When establishing each relationship, the relationship category and the identifiers of the objects at both ends of the relationship are written into the corresponding relationship record; for resource-sharing relationships, the identifier of the shared resource is also written into the corresponding relationship record.

[0118] By grouping the above nodes and relationships together, we obtain the initial transportation relationship diagram.

[0119] Different relationships indicate whether a vehicle can fulfill an order, whether roads can be used continuously, whether delays can propagate along the order of tasks, and whether tasks compete for resources. This helps to avoid confusing different transportation impacts with the same road connection relationship.

[0120] S2.4 Perform resource competition quantification on the initial transportation relationship diagram to obtain a dynamic transportation relationship diagram;

[0121] For road resources, the estimated time of the vehicle's entry into the target road segment is obtained by summing the vehicle's current location, current time, and the real-time passage time of the preceding road segment in the route. Then, the entry time is added to the real-time passage time of the target road segment to obtain the departure time, forming the estimated road occupancy period. The estimated road occupancy period is then written into the corresponding resource sharing relationship record.

[0122] For loading and unloading resources, the reservation start time is used as the start time of occupancy. The reservation start time is then added to the estimated loading and unloading time of the order to obtain the end time of occupancy, forming the estimated occupancy period of the berth. The estimated occupancy period of the berth is then written into the corresponding resource sharing relationship record.

[0123] Subsequently, the later start time of the two tasks is taken, and the earlier end time is taken. If the earlier end time is later than the later start time, the time difference between the two is taken as the overlap duration; otherwise, the overlap duration is recorded as 0.

[0124] The overlap ratio is obtained by dividing the overlap duration by the shorter resource occupation duration of the two tasks. The resource requirement for each road traffic task and each berth operation task is recorded as one occupation unit; the resource load ratio is obtained by dividing the sum of the resource requirements of the two tasks by the corresponding values ​​in the road capacity and the number of berths that can be served.

[0125] Multiplying the time period overlap ratio by the resource load ratio yields the resource contention value, which represents the degree of competition between two tasks using the same transportation resources during overlapping time periods. The greater the overlap in resource usage time periods, and the higher the combined resource demand of the two tasks relative to the available resource capacity, the higher the resource contention value.

[0126] When the overlap duration is 0, delete the corresponding resource sharing relationship; when the overlap duration is greater than 0, update the relationship category of the corresponding relationship to resource competition relationship, and write the resource competition value to the corresponding relationship record.

[0127] By processing other resource sharing relationships in the same way, a dynamic transportation relationship diagram is obtained.

[0128] For example, at 9:40 AM, the real-time passage time for section 17 increased from 8 minutes to 24 minutes, a change of 200%, exceeding the 50% passage change threshold, thus generating a road change event for section 17. Similarly, the estimated release time for berth 3 was delayed from 10:10 AM to 10:25 AM, a delay of 15 minutes, exceeding the 10-minute resource change threshold, thus generating a resource change event for berth 3.

[0129] Subsequently, by checking the route correspondence records and berth reservation records, it was found that the routes of vehicles A, B, and C all passed through segment 17, and their corresponding orders all reserved loading / unloading berth number 3, thus forming an event-related object. The three vehicles, their corresponding orders, segment 17, and loading / unloading berth number 3 were then mapped as nodes, and the transportation feasibility relationship between vehicles and orders, the task priority relationship between orders, and the resource-sharing relationship of sharing segment 17 and loading / unloading berth number 3 were established.

[0130] After calculating the resource occupancy period and overlap duration, update the resource sharing relationship with an overlap duration greater than 0 to a resource competition relationship and write the corresponding resource competition value.

[0131] Vehicle A's order reservation period is from 10:12 to 10:22, and Vehicle B's order reservation period is from 10:20 to 10:30. The overlap between the two tasks is 2 minutes, and the shorter resource occupation time is 10 minutes, resulting in a time overlap ratio of 0.2. The third loading / unloading berth can serve one vehicle at a time. The sum of the resource occupation units for the two tasks is 2, and the resource load ratio is 2, thus yielding a resource contention value of 0.4.

[0132] Write the order corresponding to vehicle A, the order corresponding to vehicle B, the third loading / unloading berth, and the resource competition value of 0.4 into the same relationship record, and update the relationship category from resource sharing relationship to resource competition relationship. Process other shared resources in the same way to obtain a dynamic transportation relationship diagram.

[0133] This approach helps to limit road anomalies, task delays, and shared resource competition to actual related objects, and uses resource competition values ​​to distinguish between general sharing relationships and resource competition that has formed obvious overlap.

[0134] As an alternative implementation, different data sources may provide inconsistent results for the same transportation status, and single location drifts and short-term communication interruptions may also lead to unreliable relationships. This implementation can solve the above problems, specifically as follows:

[0135] S3. Perform a reliability assessment on the dynamic transportation relationship diagram to obtain a reliable transportation relationship diagram;

[0136] Credibility assessment refers to examining whether the data forming each transportation relationship is timely, whether different sources corroborate each other, and whether the recent status has remained stable, in order to determine the degree to which the corresponding relationship participates in the transportation impact analysis.

[0137] S3.1. Perform relation basis extraction on the dynamic transportation relationship diagram to obtain relation basis data;

[0138] The relationship identification unit reads the relationship category and the object identifiers at both ends of a relationship from the dynamic transportation relationship graph; when the relationship is a resource competition relationship, it also reads the shared resource identifier.

[0139] Subsequently, based on the object identifiers at both ends of the relationship and the shared resource identifiers, the key status records that form the relationship are searched from the transportation data and the dynamic transportation relationship diagram.

[0140] For the original status records read from transportation data, extract the corresponding data source and status time; for the status content calculated from the estimated arrival time, remaining time window margin, and resource occupation period, find the original record on which the status content is based, and extract the corresponding data source and status time.

[0141] Write the relationship type, the identifiers of the objects at both ends of the relationship, the key status records, the data source, and the status time into the same record; for resource competition relationships, also write the shared resource identifier to obtain the relationship basis data.

[0142] For example, for transport feasibility relationships, key status records include vehicle remaining capacity, order cargo demand, road traffic restrictions, estimated order arrival time, and latest order delivery time. For road traffic relationships, key status records include road segment connectivity and road traffic status. For task priority relationships, key status records include task sequence records. For resource contention relationships, key status records include shared resource identifiers, resource occupancy periods for both tasks, resource occupancy units, and resource serviceability.

[0143] By using the object identifiers at both ends of the relationship to find the actual state record, it is possible to clearly identify which data forms each relationship, thus avoiding the use of irrelevant data for trust assessment.

[0144] S3.2. Perform timeliness and observation analysis on the relationship based on the data to obtain relationship evaluation data;

[0145] First, read the relationship category, object identifiers at both ends of the relationship, key status records, and status time from the relationship data; for resource competition relationships, also read the shared resource identifier.

[0146] Subsequently, key status items are determined according to relationship categories. Transportation feasibility relationships correspond to vehicle capacity, road traffic status, and estimated order arrival time; road traffic status corresponds to road segment connectivity and traffic status; task priority relationships correspond to the order execution priority; and resource competition relationships correspond to shared resources and the resource occupation periods of the two tasks.

[0147] For each key status item, select the record with the most recent status time and complete fields, and use the earliest status time among them as the relationship update time. Subtract the relationship update time from the current system time to obtain the data time difference. Records whose status time is no more than the time decay parameter from the current system time are determined as valid source records. The time decay parameter can be taken as 3 times the normal data reporting interval.

[0148] For the same key state item, the state support level is obtained by dividing the number of valid source records supporting the current relationship by the total number of valid source records; if no valid source records exist, the state support level is recorded as 0. The minimum state support level for each key state item is taken as the source consistency level.

[0149] Next, extract the complete state records of the 5 most recent key state items from the most recent to the oldest state time, and determine whether the current relationship is valid for each group according to the conditions for establishing the corresponding relationship. The continuous observation stability is obtained by dividing the number of record groups that support the current relationship by the actual number of extracted records; if no complete record group exists, the continuous observation stability is recorded as 0.

[0150] The relationship category, the identifiers of the objects at both ends of the relationship, the data time difference, the consistency of the source, and the stability of continuous observation are written into the same record; for resource competition relationships, the shared resource identifier is also written to obtain relationship evaluation data.

[0151] S3.3 Perform a credibility calculation on the relationship evaluation data to obtain reliable relationship data;

[0152] First, divide the data time difference by the time decay parameter, then take the negative of the ratio and substitute it into the exponential decay function to obtain the time reliability. The larger the data time difference, the lower the time reliability; when the data time difference is 0, the time reliability is 1.

[0153] Subsequently, the reliability of the time, the consistency of the source, and the stability of continuous observation are multiplied to obtain the reliability value of the relationship.

[0154] The formula for calculating the confidence value of a relationship is:

[0155]

[0156] in, This represents the reliability value of the relationship. For data time difference, For time decay parameters, To ensure consistency of origin, To measure the stability of continuous observation.

[0157] The relationship category, the identifiers of the objects at both ends of the relationship, and the relationship trust value are written into the same record; for resource competition relationships, the shared resource identifier is also written to obtain the relationship trust data.

[0158] S3.4. Write the trusted relationship data into the dynamic transportation relationship graph to obtain the trusted transportation relationship graph;

[0159] Based on the relationship category and the object identifiers at both ends of the relationship in the trusted relationship data, the corresponding relationship record is searched in the dynamic transportation relationship graph; for resource competition relationships, the corresponding relationship is also distinguished based on the shared resource identifier.

[0160] Write the relationship confidence value into the corresponding relationship record, and process other relationships in the same way to obtain a reliable transportation relationship graph. The reliable transportation relationship graph retains the original nodes and relationship fields from the dynamic transportation relationship graph. When the relationship confidence value is low, the corresponding relationship is not directly deleted, but its participation in the transportation impact analysis is reduced.

[0161] This method can preserve transportation relationships that are not completely invalid, while weakening unreliable relationships formed by outdated data, single-source records, and short-term abnormal fluctuations, which helps to avoid large-scale misadjustments of transportation tasks caused by occasional abnormal data.

[0162] As an alternative implementation, conventional road maps struggle to distinguish the propagation of delays in preceding orders to subsequent orders, as well as the impact of multiple vehicles competing for the same alternative resource. Therefore, this invention proposes the following working method:

[0163] S4. Input the credible transportation relationship graph into a pre-constructed improved graph neural network to analyze the transportation impact and obtain the transportation correlation results;

[0164] The improved graph neural network is a network model formed by adding node feature encoding layers, node update layers, trusted control channels, transport execution channels, resource contention channels, original plan retention channels, and multi-task output channels to the node aggregation structure of the general graph neural network. The node update layer is used to perform concatenation, linear transformation, and nonlinear processing on the current node representation and the aggregation result of the corresponding channel to form the corresponding node processing representation.

[0165] Before the system is put into operation, historical transportation relationship graphs and corresponding actual execution results can be used as training samples to determine network parameters through general supervised training and backpropagation methods. Actual execution results include actual order delivery time, actual resource occupancy status, and order completion status. After training is complete, the network parameters are fixed for analysis of the current reliable transportation relationship graph.

[0166] S4.1 Perform graph feature processing on the trusted transportation relationship graph to obtain transportation graph input data;

[0167] The relationship identification unit reads vehicle nodes, order nodes, road nodes, and loading / unloading resource nodes from the trusted transportation relationship graph, and forms the current node features according to the fixed order of the existing fields in each node.

[0168] Vehicle node characteristics include current location, remaining capacity, and vehicle delay; order node characteristics include delivery window, cargo demand, and remaining window margin; road node characteristics include real-time passage time, road capacity, and update time; loading and unloading resource node characteristics include available quantity, expected release time, and reserved time slot.

[0169] Subsequently, the relationship category, object identifiers at both ends of the relationship, and relationship trust value are read from the trusted transport relationship graph. For resource contention relationships, the shared resource identifier and resource contention value are also read.

[0170] Next, retrieve the previous valid state corresponding to the current state of each node from the transportation data, and organize them according to the field order with the same characteristics of the current node. If there are no calculation fields such as remaining time window margin in the previous valid state, recalculate the corresponding fields in the same way as in S2.3.

[0171] The node features corresponding to the previous valid state are input into the node feature encoding layer to obtain a stable node representation; the node features corresponding to the current state are input into the same node feature encoding layer to obtain the current node representation. The node feature encoding layer is used to perform normalization and linear transformation on various types of node data according to a fixed field order, forming node representations with consistent dimensions. A node representation refers to converting multiple state fields recorded in a node into a set of numerical data that can be computed by the improved graph neural network, according to a fixed order.

[0172] Regarding the order sequence, the estimated completion time of the preceding order is first calculated from its estimated arrival time and estimated loading / unloading time. Then, the latest delivery time is read from the delivery time window of the subsequent order, and the estimated loading / unloading time is subtracted from the latest delivery time to obtain the latest start time of the subsequent order.

[0173] The time window margin is obtained by subtracting the estimated completion time of the preceding order from the latest start time of the subsequent order. When the calculation result is less than 0, the time window margin is recorded as 0. The smaller the time window margin, the more likely the delay of the preceding order will affect the subsequent order.

[0174] The formula for calculating the propagation weight of task priority is:

[0175]

[0176] in, To propagate weight, This is the time window margin between the completion of the preceding order and the latest start time of the subsequent order. This is the time window margin decay parameter, which can be taken as the median of the positive time window margin of orders completed on time in the most recent 30 days.

[0177] The current node representation, stable node representation, relationship category, relationship credibility value, propagation weight, and resource competition value are aggregated according to the corresponding nodes and relationships to obtain the transportation graph input data.

[0178] S4.2 Perform channel-by-channel aggregation on the input data of the transportation map to obtain a representation of the transportation impact;

[0179] The trusted control channel reads the relationship trust value of each relationship and uses the relationship trust value to adjust the amount of information transmitted between the nodes at both ends of the relationship. The lower the relationship trust value, the lower the proportion of information from the corresponding adjacent nodes participating in the aggregation.

[0180] The transportation execution channel first processes the transportation feasibility relationship; for each adjacent node connected to the current node, the current node representation of the adjacent node is multiplied by the corresponding relationship confidence value, and then the calculation results of each adjacent node are summed.

[0181] Then, the task sequence relationship is processed; the current node representation of the preceding order node is multiplied by the propagation weight and the relationship confidence value in turn, and then the calculation results of each preceding order are summed.

[0182] The transportation execution channel concatenates the aggregation results corresponding to transportation feasibility relationships and task sequence relationships with the current node representation, and then inputs them into the pre-built node update layer to perform linear transformation and nonlinear processing to obtain the transportation execution representation. The transportation execution representation reflects the execution conditions between vehicles and orders, as well as the propagation of delays in previous orders to subsequent orders.

[0183] The resource competition channel handles resource competition relationships; it multiplies the current node representation of adjacent task nodes by the resource competition value and the relationship confidence value in turn, and then sums the calculation results corresponding to the same node.

[0184] Subsequently, the resource competition channel concatenates the aggregation results corresponding to the resource competition relationship with the current node representation, and then inputs them into the pre-built node update layer to perform linear transformation and nonlinear processing to obtain the resource competition representation. The resource competition representation reflects the shared occupation of road and loading / unloading resources between the corresponding task and other tasks.

[0185] The transportation execution representation and resource competition representation are concatenated according to the node correspondence, and then input into the node update layer to perform linear transformation and nonlinear processing to form a transportation impact representation with the same dimension as the stable node representation.

[0186] S4.3. Perform the original plan retention fusion on the transportation impact representation to obtain the fused node representation;

[0187] First, determine the directly affected tasks based on the event category and the identifier of the changed object of the transportation change event. The directly affected tasks corresponding to the road change event are those whose routes include the changed road segments; the directly affected tasks corresponding to the resource change event are those whose berth reservations have changed; the directly affected tasks corresponding to the vehicle change event are those whose current vehicle is carrying; and the directly affected tasks corresponding to the order change event are those whose orders have changed.

[0188] For nodes that are affected by changes and nodes that are directly affected by the changes, the adoption ratio of the current transportation impact representation is recorded as 1; for nodes that are not included in the event-related objects, the adoption ratio of the current transportation impact representation is recorded as 0.

[0189] For other task nodes in the event-related object, find the relationship path from the directly affected task node to the current task node along the task sequence relationship and resource competition relationship. Multiply the confidence values ​​of each relationship in the same path, and take the result with the largest product when multiple paths exist, as the adoption ratio of the current transportation impact representation.

[0190] The adoption ratio of the stable node representation is obtained by subtracting the adoption ratio of the current transportation impact representation from 1. Then, the stable node representation and the current transportation impact representation are weighted according to the corresponding ratios, and the two weighted results are added together to obtain the fused node representation.

[0191] When a newly inserted order does not have a previous valid state, the current transportation impact representation is directly used to form a merged node representation.

[0192] This process ensures that nodes directly affected by the change primarily adopt their current state, unaffected nodes continue to use their stable state before the event, and indirectly affected nodes gradually adjust according to the reliability of their relationships. This helps to avoid large-scale and unnecessary changes to transportation plans caused by local anomalies.

[0193] S4.4 Perform multi-task analysis on the fusion node representation to obtain transportation association results;

[0194] The multi-task output channels generate transportation feasibility margin, delay propagation risk, and resource competition risk, respectively.

[0195] First, subtract the estimated arrival time from the latest delivery time of the order to obtain the time margin; then, subtract the order's cargo requirements from the remaining vehicle capacity to obtain the capacity margin.

[0196] Subsequently, for each road and loading / unloading resource expected to be used in the order, the remaining amount of each resource is obtained by subtracting the number of tasks using the resource simultaneously during the corresponding time period from the number of available resources, and the minimum value is taken as the remaining amount of the order.

[0197] Negative values ​​in time margin, capacity margin, and resource margin are all recorded as 0. Then, they are normalized using the 95th percentile of the corresponding margin in the tasks that were completed normally in the last 30 days. If the normalization result exceeds 1, it is recorded as 1.

[0198] The minimum value among the three normalized results is taken as the transportation feasibility margin. Using the minimum value can reflect the main conditions that limit the continuation of the task in terms of time, vehicle capacity, and resource capacity.

[0199] The multi-task output channel performs delay propagation mapping and S-shaped processing on the fusion node representation to obtain the delay propagation risk between 0 and 1; then it performs resource contention mapping and S-shaped processing on the fusion node representation to obtain the resource contention risk between 0 and 1.

[0200] Resource competition value indicates the strength of the relationship between two tasks competing for the same resource during the current time period, while resource competition risk indicates the degree to which a task is affected by multiple resource competition relationships in subsequent execution.

[0201] Vehicle identification, order identification, transportation feasibility margin, delay propagation risk, and resource competition risk are written into the same record to form a transportation-related result.

[0202] Separating transportation execution relationships from resource competition relationships allows for the separate assessment of whether tasks still have execution capacity, whether delays will propagate to subsequent tasks, and whether multiple tasks are competing for resources. This helps avoid uniformly classifying different transportation anomalies as road delays.

[0203] As an alternative implementation method, simply dividing and adjusting the objects according to vehicles and delivery areas easily overlooks the connections formed between different vehicles due to shared alternative roads and loading / unloading berths. This implementation method can solve the above problems, specifically as follows:

[0204] S5. Based on the transportation association results, identify the affected tasks and group them to obtain the linkage task group.

[0205] S5.1 Perform threshold filtering on the transportation association results to obtain the affected tasks;

[0206] The task grouping unit sequentially reads the transportation feasibility margin, delay propagation risk, and resource competition risk of each task, and compares them with the corresponding thresholds respectively.

[0207] The feasible margin threshold can be taken as the 5th percentile of the transportation feasible margin corresponding to the timely completion of tasks in the most recent 30 days; the propagation risk threshold and competition risk threshold are calibrated using confirmed abnormal events, and the candidate value with the smallest sum of the number of misjudged tasks and the number of missed tasks is taken.

[0208] When at least one of the following conditions is met: the transportation feasibility margin is lower than the feasibility margin threshold, the delay propagation risk is higher than the propagation risk threshold, or the resource competition risk is higher than the competition risk threshold, the corresponding task will be identified as the affected task.

[0209] This screening method takes into account the task's own margin, delay propagation, and resource competition, which helps to avoid judging the scope of impact solely based on whether a vehicle is delayed.

[0210] S5.2. Starting from the affected tasks, perform association expansion along the dynamic transportation relationship graph to obtain the task association set;

[0211] For an affected task, first establish a task association set containing the task, then read the sequence relationship and resource competition relationship of the tasks connected to the task from the trusted transportation relationship graph, and read the delay propagation risk and resource competition risk of the corresponding adjacent tasks from the transportation association results.

[0212] Search for subsequent tasks along the task sequence. If the risk of delay propagation of a subsequent task exceeds the propagation risk threshold, add the subsequent task to the task association set.

[0213] Subsequently, adjacent tasks sharing roads or loading / unloading berths are searched along the resource competition relationship. If the resource competition risk of an adjacent task is higher than the competition risk threshold, the adjacent task is added to the task association set.

[0214] For each new task added, its subsequent tasks and resource-competing tasks are searched in the same way until no new tasks are added. Other affected tasks are processed in the same way to obtain the task association set corresponding to each affected task.

[0215] By expanding relationships layer by layer, tasks that are delayed or compete for the same resources in different vehicles can be included in the same association scope, while avoiding directly including all delivery tasks in the adjustment.

[0216] S5.3 Perform an intersection determination on the task association set, and merge the sets with common tasks into candidate task groups;

[0217] The task grouping unit compares the task identifiers in each task association set one by one. When two task association sets contain at least one identical task, the union of all tasks in the two sets is taken, and duplicate tasks are removed to form a merged task set.

[0218] The merged task set is then compared with the remaining task sets. If common tasks still exist, merging continues until no two task sets contain the same tasks.

[0219] Each merged task set is designated as a candidate task group to avoid the same task falling into multiple adjustment ranges simultaneously, thus reducing conflicting adjustment schemes for the same order.

[0220] S5.4 Perform deduplication and aggregation on the candidate task group to obtain the linkage task group;

[0221] Based on the corresponding carrier record, locate the carrier vehicle corresponding to each task in each candidate task group, and write the task identifier and the corresponding vehicle identifier into the same record.

[0222] Subsequently, duplicate task identifiers are deduplicated, and task records and vehicle identifiers in candidate task groups are aggregated separately to obtain linked task groups. When the same vehicle corresponds to multiple tasks, the transportation correspondence between each task and the vehicle is preserved.

[0223] The tasks in the joint task group have delay propagation relationships, resource competition relationships, and continuous associations formed by the connection of common tasks. Tasks that have not entered the joint task group maintain their original transportation plans.

[0224] This process can incorporate tasks and vehicles that affect each other into the adjustment, avoiding the transfer of delays or resource competition to other vehicles after adjusting only one vehicle.

[0225] As an alternative implementation, when multiple vehicles choose alternative routes with shorter travel times, they may simultaneously enter the same road segment and use the same loading / unloading berth. In view of this, the present invention proposes the following operating method:

[0226] S6. Perform joint adjustment and analysis of transportation plans on the linked task group to generate candidate transportation plans;

[0227] Joint adjustment analysis of transportation plans refers to coordinating vehicle routes, order execution sequence, loading and unloading service periods, and order carrier relationships within the same plan.

[0228] S6.1. Generate adjustment options for the linked task group to obtain adjustment candidate data;

[0229] The solution verification unit reads the task identifier and vehicle identifier from the linked task group, and then reads the current driving route sequence, order execution order, delivery location and loading and unloading service time from the transportation data according to the above identifiers.

[0230] For each vehicle, a multi-path search method is used to generate alternative routes that meet the road traffic restrictions, and the estimated travel time of the route is calculated according to the real-time travel time of each road segment. A preset number of routes with shorter estimated travel times are retained.

[0231] Subsequently, the execution order of adjacent orders carried by the same vehicle is exchanged, and the complete order order order formed after each exchange is determined as the candidate order order order. Then, based on the current time, the real-time travel time of the corresponding alternative route, and the estimated loading and unloading time of the preceding order, the estimated arrival time of each order in the candidate order order order order order order is recalculated.

[0232] Then, based on the reserved service time slots for berths and the estimated loading and unloading time of orders, candidate service time slots are determined from the time slots that have not yet reached the number of berths available for service.

[0233] For orders that have not yet been executed and need to be transferred, select vehicles with remaining capacity that meet the cargo requirements from the linked vehicles with the same pickup location. Then, calculate the additional travel time after inserting the order based on the current location, current route, and order delivery location of the candidate vehicles. If the additional travel time does not exceed the remaining time window margin of the order, the corresponding vehicle will be identified as a candidate carrier vehicle.

[0234] The alternative routes, candidate order order, candidate service time period, and candidate carrier vehicles are aggregated to obtain the candidate data for adjustment.

[0235] S6.2 Perform combination processing on the adjusted candidate data to obtain an initial transportation plan;

[0236] The scheme verification unit selects corresponding alternative routes, order sequences, service periods, and transport vehicles from the adjusted candidate data based on vehicle and task identifiers. For each combination, at least one of the following must differ from the original transportation plan: route, order execution sequence, loading / unloading service period, or transport vehicle within the linked task group; the remaining aspects must follow the original transportation plan.

[0237] When the carrier vehicle of an order changes, the order is removed from the original order execution sequence of the carrier vehicle, and then written into the corresponding position of the candidate carrier vehicle according to the candidate order, so that the order carrier relationship is consistent with the order execution sequence.

[0238] Subsequently, based on the combined transport vehicles and order execution order, the demand for unfulfilled orders carried by each vehicle was recalculated, and the estimated arrival time of each order was recalculated according to the combined route sequence, order execution order, and loading and unloading service period.

[0239] The initial transportation plan is obtained by writing the corresponding transport vehicles, order execution order, route sequence, loading and unloading service time, vehicle cargo requirements, and expected arrival time of the order for each task in the joint task group into the same plan record.

[0240] S6.3 Perform transportation constraint verification on the initial transportation plan to obtain a feasible transportation plan;

[0241] The remaining vehicle capacity, road traffic restrictions, order delivery window, number of available berths, and order allocation integrity in each initial transportation plan are verified sequentially.

[0242] In the initial transportation plan, the vehicle capacity check passes if the sum of all outstanding order demands for a vehicle does not exceed the vehicle's current available capacity. The road access check passes if the vehicle complies with the traffic restrictions of each road segment in the corresponding travel sequence.

[0243] The delivery window verification passes when the estimated arrival time of an order is no later than the latest delivery time. The loading and unloading resource verification passes when the number of tasks using the same berth during the same time period does not exceed the number of berths that can be served.

[0244] The order allocation integrity check passes when each order is assigned to a single vehicle and there are no omissions or duplicates in the allocation.

[0245] The initial transportation plan that passes all verifications is determined to be a feasible transportation plan.

[0246] S6.4. Perform deduplication and aggregation on the feasible transportation schemes to generate candidate transportation schemes;

[0247] Compare the vehicles, order execution sequence, route sequence, and loading / unloading service periods in each feasible transportation option.

[0248] If all the above contents are the same, the corresponding feasible transportation plan will be identified as a duplicate plan and only one will be retained. Then, the remaining feasible transportation plans will be collected into candidate transportation plans. The candidate transportation plans record the carrier correspondence, order execution order, route sequence and loading and unloading service time period, which will be used for subsequent rehearsal in combination with dynamic transportation relationship diagram.

[0249] Jointly changing routes, order sequences, loading and unloading service hours, and carrier relationships helps to keep road adjustments consistent with loading and unloading arrangements, reducing situations where routes meet traffic conditions but loading and unloading berths still conflict.

[0250] As an optional implementation method, the feasibility of the candidate transportation scheme for a single vehicle does not necessarily mean that it will still be feasible when multiple vehicles execute it simultaneously. This implementation method can solve the above problem, specifically as follows:

[0251] S7. Perform a preliminary simulation based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme;

[0252] The rehearsal refers to writing candidate transportation plans into a dynamic transportation relationship diagram and calculating the transportation status after all tasks are executed together, without sending adjustment instructions to vehicles and loading / unloading resources.

[0253] S7.1. Write the candidate transportation scheme into the dynamic transportation relationship diagram and perform a state update to obtain a temporary transportation relationship diagram;

[0254] The scheme verification unit replicates the dynamic transportation relationship diagram for each candidate transportation scheme, and then sequentially writes in the carrier correspondence, order execution order, driving segment sequence and loading and unloading service time period of the candidate transportation scheme.

[0255] For tasks where the carrier vehicle, route sequence, order execution order, and loading / unloading service period change, reassess whether the vehicle capacity, road traffic restrictions, and order delivery window meet the transportation conditions, and update the transportation feasibility relationship between vehicle nodes and order nodes accordingly.

[0256] When the order of order execution changes, delete the original task sequence and re-establish the task sequence according to the adjusted order execution order.

[0257] Subsequently, the estimated road occupancy period was recalculated according to the adjusted driving route sequence, and the estimated berth occupancy period was recalculated according to the adjusted loading and unloading service period.

[0258] Then, using the resource competition quantification method in S2.4, we re-determine whether there is a resource competition relationship between tasks and calculate the corresponding resource competition value. Nodes and relationships that have not been changed by the candidate transportation scheme retain their original records to obtain a temporary transportation relationship diagram.

[0259] S7.2 Perform joint state calculations on the temporary transportation relationship diagram to obtain scheme simulation data;

[0260] Starting from the current moment, the estimated travel time of each road segment is accumulated sequentially according to the sequence of each vehicle's travel route, and the estimated loading and unloading time of the preceding orders is accumulated according to the order of order execution to obtain the estimated arrival time of each task.

[0261] Subsequently, for each task, the estimated travel time and estimated loading / unloading time are added together to obtain the estimated time for the task; then, the estimated time for each task in the candidate transportation plan is added together to obtain the total estimated time for the plan.

[0262] Next, compare the estimated arrival time of each task with the latest delivery time of the order. If the estimated arrival time is later than the latest delivery time, record the time difference between the two as the estimated timeout for that task. Otherwise, record the estimated timeout as 0. Add the estimated timeouts of each task together to get the estimated timeout for the order.

[0263] Then, the resource competition value in the temporary transportation relationship diagram is compared with the resource competition value corresponding to the same task and the same shared resource in the dynamic transportation relationship diagram. When the temporary resource competition value is large, the difference between the two is recorded as the new competition value. When a new resource competition relationship is formed in the temporary transportation relationship diagram, the corresponding original resource competition value is recorded as 0. The new competition values ​​are added together to obtain the new resource competition amount.

[0264] Then, count the number of tasks where the carrier vehicle, order execution order, route sequence, and loading / unloading service period changes. If there are multiple changes for the same task, only count it once to obtain the number of changes in the plan.

[0265] Establish corresponding records for the total estimated time of the plan, the estimated order overtime, the amount of new resource competition, and the amount of plan changes with the corresponding candidate transportation plans to obtain the plan simulation data.

[0266] S7.3. Perform normalization evaluation on the simulation data of the above scheme to obtain the scheme evaluation data;

[0267] For the same batch of candidate transportation plans, range normalization was performed on the total estimated time of the task, the estimated timeout of the order, the amount of new resource competition, and the amount of plan changes.

[0268] When the maximum and minimum values ​​of the same indicator are the same, the normalized result of that indicator for each candidate transportation scheme is recorded as 0; in other cases, the corresponding value is subtracted from the minimum value and then divided by the difference between the maximum and minimum values ​​to obtain the normalized result.

[0269] Subsequently, a weighted summation was performed on the four normalized results to obtain the evaluation value of the scheme:

[0270]

[0271] in, This is the evaluation value of the scheme. For the normalized result of the k-th term, For the corresponding evaluation weights.

[0272] It should be noted that the evaluation weights corresponding to the estimated order timeout and the amount of new resource contention are higher than those corresponding to the estimated total task time and the amount of plan changes. Evaluation weights can be determined by replaying the most recent 30 confirmed abnormal events. Among candidate combinations where the sum of evaluation weights is 1, the selection scheme corresponding to each historical event is determined for each combination. Then, the actual timeout and the actual number of resource conflicts for the selected schemes are compared, and the evaluation weight combination with the smallest normalized sum of the two results is selected. For example, the four evaluation weights can be set to 0.2, 0.35, 0.35, and 0.1 respectively.

[0273] Write the scheme evaluation value, the newly added resource competition quantity, and the corresponding candidate transportation scheme into the same record to obtain the scheme evaluation data.

[0274] S7.4 Perform threshold filtering on the evaluation data of the proposed scheme to determine the target scheme;

[0275] First, compare the new resource competition amount and the competition allowance threshold of each candidate transportation scheme, and eliminate candidate transportation schemes whose new resource competition amount exceeds the competition allowance threshold.

[0276] If there are remaining candidate transportation schemes, the candidate transportation scheme with the smallest scheme evaluation value is selected as the target scheme.

[0277] If no remaining candidate transportation schemes exist, return to step S6 to regenerate and adjust the candidate data, and expand the selection range of alternative routes or candidate service periods; if no candidate transportation schemes that meet the competition allowance threshold still exist after regeneration, no adjustment instruction is sent, and the next transportation status correlation analysis is performed based on the latest transportation data.

[0278] The competition allowance threshold represents the degree of increased resource competition allowed after the scheme is adjusted. It can be determined by the 95th percentile of the newly added resource competition amount corresponding to the normal adjustment scheme in the most recent 30 days.

[0279] For example, the first candidate transportation plan causes three vehicles to simultaneously travel through the same road segment and use the same loading / unloading berth, resulting in a concentrated increase in resource competition. The second candidate transportation plan allocates vehicles to different routes and loading / unloading service periods, and the increased resource competition does not exceed the allowed threshold, with a smaller expected order overtime. Therefore, the second candidate transportation plan is selected as the target plan.

[0280] This process can calculate the transportation status after multiple vehicles jointly execute candidate transportation plans before the adjustment instruction is issued, and exclude plans that will further increase competition for road or loading and unloading resources, which helps to avoid secondary congestion after a single vehicle's feasible plan is jointly executed.

[0281] As an optional implementation, continuing to schedule according to the already executed state when the target solution has not been received by the terminal or actually executed will cause system state distortion. This implementation can solve the above problem, specifically as follows:

[0282] S8. Generate and send adjustment instructions according to the target plan, update transportation data after obtaining feedback data, and generate transportation results.

[0283] S8.1. Generate and send adjustment instructions based on the target scheme to obtain instruction sending data;

[0284] The scheme verification unit compares the target scheme with the current transportation data to determine the changed carrier correspondence, order execution sequence, route sequence, and loading and unloading service periods.

[0285] Subsequently, the order execution order and driving route sequence are written into the vehicle adjustment instruction according to the vehicle identification and sent to the corresponding vehicle terminal.

[0286] Then, according to the order identifier, the changed carrier correspondence is written into the order adjustment instruction and sent to the order management terminal.

[0287] Then, according to the berth identification, the berth and the corresponding loading and unloading service period are written into the resource adjustment instruction and sent to the loading and unloading resource management terminal.

[0288] All adjustment instructions are written to the same instruction identifier. The instruction identifier, the recipient, and the corresponding vehicle identifier, order identifier, and parking space identifier are written to the same record to obtain the instruction transmission data.

[0289] S8.2. Send data collection and execution feedback according to the instruction to obtain feedback data;

[0290] The data sensing unit receives execution feedback returned by each terminal according to the instruction identifier.

[0291] The vehicle terminal returns the instruction reception status, the vehicle's current location, the actual route sequence, the actual order execution order, and the order execution status. The order management terminal returns the instruction reception status, the carrier relationship confirmation status, and the order handover status. The loading and unloading resource management terminal returns the instruction reception status, the berth confirmation status, and the confirmed loading and unloading service time period.

[0292] Subsequently, the latest road records after the adjustment command was sent are read from the road information platform. The latest road records include the road segment traffic status and real-time traffic time.

[0293] Then, according to the receiving object and instruction identifier in the data sent by the instruction, the above execution feedback is collected with the corresponding vehicle identifier, order identifier, road segment identifier and parking space identifier to obtain feedback data.

[0294] S8.3. Write the feedback data into the transportation data execution status update to obtain the feedback transportation data;

[0295] The data sensing unit searches for corresponding records in the transportation data according to vehicle identification, order identification, road segment identification, and parking space identification.

[0296] Subsequently, the instruction reception status of each terminal, the current location of the vehicle, the actual driving route sequence, the actual order execution order, the order execution status, the carrier relationship confirmation status, the order handover status, the berth confirmation status, the confirmed loading and unloading service period, the latest road traffic status, and the real-time traffic time are written into the corresponding object records.

[0297] Then, the receiving time of each feedback is recorded as the corresponding status time, and the corresponding data source is retained to obtain the feedback transportation data.

[0298] S8.4 Execute the decision on the feedback transportation data and generate the transportation result;

[0299] The system sequentially checks whether the corresponding terminal has received the adjustment instruction, whether the vehicle is executing according to the target plan's corresponding route and order sequence, whether the carrier relationship adjustment has completed the order handover, and whether the loading and unloading berth has been confirmed.

[0300] Then, it is determined whether the road segment corresponding to the target route still meets the traffic conditions in the latest road records.

[0301] If at least one of these conditions is not met, the feedback transportation data will be used as the updated transportation data, and the transportation status correlation analysis will be re-executed.

[0302] When all conditions are met, the carrier correspondence, the actual vehicle travel route sequence, the actual order execution order, and the confirmed loading and unloading service time period are retrieved from the feedback transportation data.

[0303] Then, based on the vehicle's current location, the real-time traffic time of the remaining road segment, and the estimated loading and unloading time of previous orders, the latest estimated delivery time for each order is calculated.

[0304] The transportation result is generated by writing the carrier correspondence, the actual route sequence of the vehicle, the actual order execution order, the latest estimated delivery time, and the confirmed loading and unloading service time into the same record.

[0305] This process can update the transportation status based on the actual feedback from the terminal, avoiding the direct assumption that unreceived or unexecuted adjustment instructions are already completed, and re-triggering the transportation relationship analysis when the road conditions change again.

[0306] As a second aspect of the present invention, the present invention also provides a high-efficiency logistics transportation system based on an improved graph neural network, comprising:

[0307] The data sensing unit is used to acquire basic logistics data and perform aggregation processing on the basic logistics data to form transportation data;

[0308] The relationship identification unit is used to perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph; perform a credibility assessment on the dynamic transportation relationship graph to obtain a credible transportation relationship graph; and input the credible transportation relationship graph into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results.

[0309] The task grouping unit is used to identify affected tasks and group them according to the transportation association results to obtain a coordinated task group.

[0310] The scheme verification unit is used to perform joint adjustment and analysis of transportation schemes on the coordinated task group to generate candidate transportation schemes; perform a pre-rehearsal based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme; generate and send adjustment instructions according to the target scheme, update transportation data after obtaining feedback data, and generate transportation results.

[0311] like Figure 2 As shown, under the same order size, number of vehicles, road conditions, and loading / unloading resources, this invention has a higher accuracy rate in identifying affected tasks compared to existing technologies, while reducing the proportion of incorrectly adjusted tasks, the amount of new resource competition, and the number of secondary congestion occurrences. These results demonstrate that this invention, by performing a reliable assessment of transportation relationships, distinguishing between delay propagation and resource competition, and conducting joint rehearsals of candidate transportation schemes before issuing adjustment instructions, helps to more accurately determine the scope of affected tasks, reduce unnecessary adjustments to unaffected tasks, and lower the possibility of new resource conflicts and secondary congestion arising from multiple vehicles jointly implementing adjustment schemes.

[0312] The embodiments of the present invention described above are subject to modification and change of method by those skilled in the art without departing from the embodiments and broader aspects of the present invention. The appended claims are intended to include all such modifications and changes of method that do not depart from the present invention.

[0313] Those skilled in the art should understand that the embodiments of the present invention can be implemented using a pure hardware architecture, a pure software architecture, or an integrated hardware and software architecture. The present invention can be prepared as a computer program product, which can be stored in various non-volatile computer-readable storage media, including but not limited to solid-state drives, flash memory chips, mobile storage devices, optical discs, cloud storage servers, and other standardized storage media, and is not limited to traditional storage media.

[0314] Based on the foregoing description in conjunction with the accompanying drawings, those skilled in the art will understand that the embodiments of this application can also be implemented by software programs. Therefore, this application also provides a computer-readable storage medium. This computer-readable storage medium stores computer-readable instructions thereon, which, when executed by one or more processors, implement the method described above in conjunction with the accompanying drawings.

[0315] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0316] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0317] It should be understood that when the terms "first," "second," "third," and "fourth," etc., are used in the claims, specification, and drawings of this application, they are used only to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" as used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0318] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0319] Although the embodiments of this application are described above, the content is merely an example adopted for the purpose of facilitating understanding of this application and is not intended to limit the scope and application scenarios of this application. Any person skilled in the art described in this application may make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in this application, but the scope of patent protection of this application shall still be determined by the scope defined in the appended claims.

Claims

1. A logistics high-efficiency transportation method based on an improved graph neural network, characterized in that, include: Acquire basic logistics data and perform aggregation processing on the basic logistics data to form transportation data; Perform a transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph; A reliability assessment is performed on the dynamic transportation relationship graph to obtain a reliable transportation relationship graph; The reliable transportation relationship graph is input into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results; Based on the transportation association results, the affected tasks are identified and grouped to obtain the linked task group; The joint task group performs a joint adjustment analysis of the transportation plan to generate candidate transportation plans; Based on the candidate transportation schemes and the dynamic transportation relationship diagram, a preliminary simulation is performed to determine the target scheme; According to the target plan, generate and send adjustment instructions, obtain feedback data, update transportation data, and generate transportation results.

2. The efficient logistics transportation method according to claim 1, characterized in that, Perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph, including: Perform state difference analysis on the transportation data to obtain transportation change events; Perform association extraction centered on the object corresponding to the transportation change event to obtain the event association object; Perform node mapping and relationship construction on the event-related objects to obtain an initial transportation relationship graph; Resource competition quantification is performed on the initial transportation relationship graph to obtain a dynamic transportation relationship graph.

3. The efficient logistics transportation method according to claim 1, characterized in that, A reliability assessment is performed on the dynamic transportation relationship graph to obtain a reliable transportation relationship graph, including: The dynamic transportation relationship diagram is subjected to relationship basis extraction to obtain relationship basis data; Based on the data, a timeliness and observation analysis is performed on the relationship to obtain relationship evaluation data; Perform a credibility calculation on the relationship evaluation data to obtain reliable relationship data; The trusted relationship data is written into the dynamic transportation relationship graph to obtain the trusted transportation relationship graph.

4. The efficient logistics transportation method according to claim 1, characterized in that, The reliable transportation relationship graph is input into a pre-constructed improved graph neural network to analyze the transportation impact, resulting in transportation correlation results, including: The reliable transportation relationship graph is processed to obtain the transportation graph input data; Perform channel-specific aggregation on the input data of the transportation map to obtain a representation of the transportation impact; The transportation impact is represented by the original plan preserved and merged to obtain the merged node representation; Multi-task analysis is performed on the fusion node representation to obtain transportation association results.

5. The efficient logistics transportation method according to claim 1, characterized in that, Based on the transportation correlation results, affected tasks are identified and grouped to obtain a coordinated task group, including: The affected tasks are obtained by performing threshold filtering on the transportation association results; Starting with the affected tasks, perform association expansion along the dynamic transportation relationship graph to obtain a task association set; Perform an intersection determination on the task association set, and merge sets with common tasks into candidate task groups; The candidate task groups are deduplicated and aggregated to obtain the linked task groups.

6. The efficient logistics transportation method according to claim 1, characterized in that, The coordinated task group performs joint adjustment analysis of transportation plans to generate candidate transportation plans, including: Adjustment options are generated for the linked task group to obtain adjustment candidate data; The adjusted candidate data are combined to obtain an initial transportation plan; Perform transportation constraint verification on the initial transportation plan to obtain a feasible transportation plan; The feasible transportation schemes are deduplicated and aggregated to generate candidate transportation schemes.

7. The efficient logistics transportation method according to claim 1, characterized in that, Based on the candidate transportation schemes and the dynamic transportation relationship diagram, a preliminary simulation is performed to determine the target scheme, including: The candidate transportation schemes are written into the dynamic transportation relationship graph and the state is updated to obtain a temporary transportation relationship graph. Perform joint state calculations on the temporary transportation relationship diagram to obtain scheme simulation data; The simulation data of the proposed scheme is subjected to normalized evaluation to obtain the scheme evaluation data; Threshold filtering is performed on the evaluation data of the proposed schemes to determine the target scheme.

8. The efficient logistics transportation method according to claim 1, characterized in that, According to the target plan, generate and send adjustment instructions, obtain feedback data, update transportation data, and generate transportation results, including: Based on the target scheme, an adjustment instruction is generated and sent, and instruction sending data is obtained; Data collection and execution feedback are sent according to the instructions to obtain feedback data; The feedback data is written into the transportation data execution status update to obtain the feedback transportation data; The execution plan is determined based on the feedback transportation data, and a transportation result is generated.

9. The efficient logistics transportation method according to claim 7, characterized in that, Perform joint state calculations on the temporary transport relationship diagram to obtain scheme simulation data, including: Perform a task timing analysis on the temporary transportation relationship diagram to obtain the estimated arrival time and the estimated total time of the plan; Perform a delivery discrepancy analysis on the estimated arrival time to obtain the estimated order timeout amount; Perform competitive increment analysis on the temporary and dynamic transportation relationship diagrams to obtain the new resource competition volume; Perform a scheme change statistics analysis on the candidate transportation schemes to obtain the scheme change quantity; The estimated total time taken for the proposed solution, the estimated timeout of orders, the amount of new resource competition, and the amount of solution changes are aggregated to obtain the solution simulation data.

10. A high-efficiency logistics transportation system based on an improved graph neural network, applicable to the high-efficiency logistics transportation method described in any one of claims 1 to 9, characterized in that, include: The data sensing unit is used to acquire basic logistics data and perform aggregation processing on the basic logistics data to form transportation data; A relationship identification unit is used to perform transportation status correlation analysis on the transportation data to obtain a dynamic transportation relationship graph. A reliability assessment is performed on the dynamic transportation relationship graph to obtain a reliable transportation relationship graph; The reliable transportation relationship graph is input into a pre-constructed improved graph neural network to analyze the transportation impact and obtain transportation correlation results; The task grouping unit is used to identify affected tasks and group them according to the transportation association results to obtain a coordinated task group. The scheme verification unit is used to perform joint adjustment and analysis of transportation schemes on the coordinated task group to generate candidate transportation schemes; perform a pre-rehearsal based on the candidate transportation schemes and the dynamic transportation relationship diagram to determine the target scheme; generate and send adjustment instructions according to the target scheme, update transportation data after obtaining feedback data, and generate transportation results.

Images