A cross-regional transport capacity collaborative scheduling and order intelligent order optimization method

Abstract

The application provides a cross-regional transport capacity collaborative scheduling and order intelligent order optimization method, multi-source transport capacity data is accessed through an Internet of Things gateway, a distributed virtual transport capacity pool is constructed after standardization, and a dynamic digital certificate is created for each vehicle; orders are multi-scale space-time coded to generate feature vectors; order matching adaptability is calculated, the order matching threshold is dynamically adjusted according to the order backlog rate, and candidate combinations are generated according to double-mode clustering of administrative levels; a joint optimization mathematical model with the target of total transportation cost is constructed, an improved genetic algorithm is used to determine the order combination and transport capacity assignment at the same time, and an integrated scheduling plan is output; order flow, transport capacity state and road condition changes are monitored, and incremental optimization is performed when the rescheduling condition is met, the application realizes joint optimization of order matching and order dispatching, adaptively adjusts the order matching threshold, and introduces regional load balancing punishment, thereby reducing transportation cost and improving transport capacity utilization.

Description

Technical Field

[0001] This invention relates to the field of logistics-related technologies, specifically a method for cross-regional collaborative scheduling of transportation capacity and intelligent order grouping optimization. Background Technology

[0002] In the logistics and transportation industry, the empty-running rate and idle rate of vehicles have remained high for a long time. The main technical reasons for this problem include: Information silos in transportation capacity: The transportation capacity data of different logistics companies and individual vehicle owners are isolated from each other. The dispatching system can only dispatch orders based on its own transportation capacity, and cannot achieve cross-enterprise and cross-regional transportation capacity sharing and collaboration, resulting in a large number of duplicate vehicle dispatches and route overlaps. Order grouping methods are static and rigid: Most existing grouping technologies rely on fixed rules such as routes and time periods for matching, lacking the ability to adapt to real-time order backlog pressure and changes in traffic conditions. This results in severe order backlog during peak hours and low-quality grouping during off-peak hours. The order-grouping and order-dispatch processes are disconnected: In current practice, orders are generally grouped first and then vehicles are assigned. The two stages are executed independently, which means that the optimal combination in the order-grouping stage cannot be effectively executed in the order-dispatch stage, and the overall solution deviates from the global optimum. The lack of a multi-level coordination mechanism for cross-regional capacity allocation often results in significant differences in load between regions, with some regions having idle capacity while other regions have backlogged orders.

[0003] Therefore, there is an urgent need for a cross-regional collaborative scheduling method that can break down capacity silos, achieve dynamic adaptive order consolidation, and jointly optimize order combinations and capacity assignment. Summary of the Invention

[0004] The purpose of this invention is to provide a cross-regional capacity collaborative scheduling and intelligent order grouping optimization method to solve the problems of isolated capacity information, lack of dynamic adaptive capability in grouping, and global suboptimal problems caused by the serial optimization of order combination and capacity assignment in cross-regional logistics scheduling.

[0005] To achieve the above objectives, the present invention provides a method for cross-regional collaborative scheduling of transportation capacity and intelligent order grouping optimization, comprising: Step S1: Access the capacity data of multiple capacity providers through the Internet of Things gateway, perform standardized processing, and construct a distributed virtual capacity pool. Each vehicle in the virtual capacity pool corresponds to a dynamic digital certificate, which records the real-time availability of the vehicle and the committed tasks. Step S2: Receive customer order requests, perform multi-scale spatiotemporal coding on the order's shipping and receiving locations according to at least two administrative levels, and generate an order feature vector by combining the order's time parameters and cargo feature parameters; Step S3: Calculate the grouping suitability based on the spatial overlap and time window overlap of the transportation routes of the two orders, dynamically adjust the grouping attraction threshold based on the current order backlog rate of the system, and cluster the orders based on whether the goods types are compatible to generate a candidate order combination set; wherein, different clustering rules are selected according to the administrative level of the origin and destination of the orders during clustering. Step S4: Construct a joint optimization mathematical model with total transportation cost as the evaluation index, and use a genetic algorithm to iteratively solve the mathematical model to reduce the total transportation cost. At the same time, determine the order combination and capacity assignment scheme, and output an integrated scheduling plan. The genetic algorithm prioritizes matching capacity according to the value score of the order combination during initialization, and adjusts the crossover probability according to the dispersion of the population fitness and the mutation probability according to the number of generations during the evolution process. Step S5: Monitor changes in order flow, capacity status, and road conditions. When the preset rescheduling trigger conditions are met, perform incremental optimization on unexecuted order combinations and update the integrated scheduling plan.

[0006] A further technical solution, the standardization process includes at least: unifying the vehicle location coordinates of different transport providers to the same coordinate system, unifying the weight unit to kilograms, unifying the volume unit to cubic meters, and unifying the time format to the same timestamp format.

[0007] A further technical solution, the multi-scale spatiotemporal coding specifically includes: performing hierarchical coding of the order's shipping location and receiving location according to three granularities: national level, regional level, and local level, forming a starting point coding vector and an ending point coding vector; and combining the start and end times of the expected shipping time window, the weight of the goods, the volume of the goods, and the goods type label vector with the starting point coding vector and the ending point coding vector to form an order feature vector.

[0008] A further technical solution is that the calculation method for the group-buying suitability in step S3 is as follows: For any two orders, the spatial gravity value is determined based on the shared mileage ratio of their transportation paths and the distance between the geometric centers of the two paths; the temporal gravity value is determined based on the ratio of the overlap length of their expected time windows to the total time span; and the compatibility coefficient is determined based on whether there is a conflict between their cargo types. The weighted spatial gravity value and the temporal gravity value are added together and then multiplied by the compatibility coefficient to obtain the group-buying suitability. The compatibility coefficient is 0 when there is a conflict and 1 when there is no conflict. The distance between the geometric centers of the two paths refers to: taking the average coordinates of the starting point, ending point, and all planned transit points of each path as the geometric center of the path, and then calculating the spherical distance between the two geometric centers.

[0009] A further technical solution, the method of dynamically adjusting the group-buying attraction threshold in step S3 is as follows: set multiple different order backlog rate intervals, each interval corresponding to a preset attraction threshold; obtain the current order backlog rate of the system, determine the interval in which the backlog rate is located, and select the attraction threshold corresponding to the interval as the current group-buying attraction threshold.

[0010] A further technical solution, the clustering method in step S3 is as follows: extract the spatial codes of the starting point and the ending point from the order feature vector; when the country-level codes or region-level codes of the starting point and the ending point are different, it is determined to be a cross-regional order, and the path overlap priority mode is adopted: calculate the path overlap degree between every two orders, and cluster the orders with the path overlap degree greater than the preset overlap degree threshold into the same group; when the country-level codes and region-level codes of the starting point and the ending point are the same, it is determined to be a same-region order, and the origin / destination aggregation mode is adopted: calculate the distance between the shipping location and the receiving location of every two orders, and cluster the orders with the shipping location distance less than the preset shipping distance threshold and the receiving location distance less than the preset receiving distance threshold into the same group.

[0011] A further technical solution is provided, wherein the total transportation cost includes: a fixed cost per trip, variable costs based on route planning, order timeout penalty costs, and regional load balancing penalty costs. The regional load balancing penalty cost is determined as follows: the scheduling range is divided into multiple regions; for each region, the ratio of the number of allocated vehicles in that region to the total number of vehicles in that region is obtained; the average of the ratios for all regions is calculated; the deviations between the ratios for each region and the average are squared, summed, and then multiplied by a dynamic penalty coefficient to obtain the regional load balancing penalty cost; the dynamic penalty coefficient is positively correlated with the variance of the number of uncompleted orders in each region during the current scheduling period.

[0012] A further technical solution, the specific execution method of the genetic algorithm in step S4 includes: Step S41: Obtain the combination value score for each order combination, which is a weighted sum of four parameters: total order weight, total volume, total freight revenue, and order timeliness; match transportation capacity to each order combination in descending order of combination value score to generate an initial population; Step S42: Use the reciprocal of the total transportation cost as the fitness value for each individual; Step S43: Obtain the standard deviation of the fitness values ​​of all individuals in the current population and obtain the current generation number; when the standard deviation is less than a preset lower threshold, increase the crossover probability by a preset increment value; when the standard deviation is greater than a preset upper threshold, decrease the crossover probability by a preset decrement value; when the ratio of the current generation number to the preset maximum generation number exceeds a preset ratio threshold, increase the mutation probability generation by generation according to a preset step size until the preset maximum generation number is reached. Step S44: After each generation of evolution is completed, the top K individuals with the highest fitness values ​​are retained for the next generation, where K is a preset positive integer; and simulated annealing perturbation operation is performed on the retained individuals with a preset perturbation probability. The perturbation operation is to randomly swap an order combination with a vehicle matching pair. Step S45: Repeat steps S42 to S44 until the current generation reaches the preset maximum generation, and output the integrated scheduling plan corresponding to the individual with the highest fitness value.

[0013] In a further technical solution, the preset rescheduling triggering condition in step S5 includes at least one of the following: When a new order arrives, the grouping suitability between the new order and any allocated but not yet loaded order in the current scheduling plan is obtained. If the grouping suitability is greater than the current gravity threshold corresponding to the allocated but not yet loaded order, rescheduling is triggered. When a capacity status change event occurs, an alternative capacity matching process is triggered within a preset response time threshold. The preset time window has expired.

[0014] A further technical solution is that the distributed virtual capacity pool uses distributed ledger technology to maintain dynamic digital credentials for each vehicle, and shares anonymized capacity data among the capacity providers through smart contract protocols. Each time a status change event of a scheduling plan is triggered, the smart contract automatically generates a record containing vehicle identifier, order combination identifier and cost sharing coefficient. This record is written into the blockchain after being confirmed by consensus among all participants, and is used for automatic settlement of fees and determination of responsibility after the task is completed.

[0015] In summary, compared with the prior art, the present invention has the following beneficial effects: By building a distributed virtual capacity pool and dynamic digital credentials, the information silos of capacity are broken down, enabling trusted sharing of capacity across enterprises and regions, reducing duplicate dispatch rates, and improving capacity utilization. The spatiotemporal gravity model is used to calculate the suitability of group buying, and the gravity threshold is dynamically adjusted according to the order backlog rate to achieve adaptive relaxation or tightening of group buying conditions and reduce order backlog during peak periods. By placing order combination and capacity allocation under the same mathematical model for joint optimization, the suboptimal defects of serial processing can be overcome, the total transportation cost can be reduced, and the vehicle loading rate can be improved. Introducing regional load balancing penalty costs and dynamic penalty coefficients guides the rational distribution of transportation capacity resources across regions and reduces load differences between regions. By using blockchain smart contracts to achieve scheduling status notarization and automatic fee settlement, cross-enterprise collaboration trust is enhanced and the settlement cycle is shortened. The improved genetic algorithm employs combinatorial value-first initialization, dynamic crossover mutation probability, and simulated annealing local search, significantly improving solution efficiency and solution quality. Attached Figure Description

[0016] The accompanying drawings, which are provided to further illustrate this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0017] Figure 1 This is a flowchart illustrating the method of this application. Detailed Implementation

[0018] To more clearly illustrate the overall concept of this application, a detailed explanation is provided below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0020] like Figure 1 As shown, a method for cross-regional collaborative scheduling of transportation capacity and intelligent order grouping optimization includes: Step S1: Access the capacity data of multiple capacity providers through the Internet of Things gateway, perform standardized processing, and construct a distributed virtual capacity pool. Each vehicle in the virtual capacity pool corresponds to a dynamic digital certificate, which records the real-time availability of the vehicle and the committed tasks. Step S2: Receive customer order requests, perform multi-scale spatiotemporal coding on the order's shipping and receiving locations according to at least two administrative levels, and generate an order feature vector by combining the order's time parameters and cargo feature parameters; Step S3: Calculate the grouping suitability based on the spatial overlap and time window overlap of the transportation routes of the two orders, dynamically adjust the grouping attraction threshold based on the current order backlog rate of the system, and cluster the orders based on whether the goods types are compatible to generate a candidate order combination set; wherein, different clustering rules are selected according to the administrative level of the origin and destination of the orders during clustering. Step S4: Construct a joint optimization mathematical model with total transportation cost as the evaluation index, and use a genetic algorithm to iteratively solve the mathematical model to reduce the total transportation cost. At the same time, determine the order combination and capacity assignment scheme, and output an integrated scheduling plan. The genetic algorithm prioritizes matching capacity according to the value score of the order combination during initialization, and adjusts the crossover probability according to the dispersion of the population fitness and the mutation probability according to the number of generations during the evolution process. Step S5: Monitor changes in order flow, capacity status, and road conditions. When the preset rescheduling trigger conditions are met, perform incremental optimization on unexecuted order combinations and update the integrated scheduling plan.

[0021] Detailed explanation: Step S1: Access the capacity data of multiple capacity providers through the Internet of Things gateway, perform standardized processing, and construct a distributed virtual capacity pool. Each vehicle in the virtual capacity pool corresponds to a dynamic digital certificate, which records the real-time availability of the vehicle and the committed tasks. Specific implementation method: Connect multiple transportation providers (including individual vehicle owners, small and medium-sized fleets, logistics company subsidiaries, or alliance members) to their vehicle terminals via IoT gateways (such as gateway devices that support MQTT, CoAP, or HTTP protocols), such as GPS / BeiDou positioning devices, OBD interfaces, and mobile apps.

[0022] The accessed capacity data should include at least: unique vehicle identifier (license plate number or device ID), real-time latitude and longitude, current load, available volume, vehicle status (idle / on duty / under maintenance), driver's continuous driving time, and vehicle type (van / refrigerated / flatbed, etc.).

[0023] Standardization: Unify data from different sources into an internal standard format. For example, convert all location coordinates to the WGS84 coordinate system, unify weight units to kilograms, volume units to cubic meters, and timestamps to Unix seconds or milliseconds.

[0024] Construct a distributed virtual transportation capacity pool: Use a distributed ledger to record the dynamic digital credentials of each vehicle.

[0025] Distributed ledgers include Hyperledger Fabric—an open-source, licensed consortium blockchain framework—or Ethereum private chains—permissioned blockchains built on the Ethereum protocol.

[0026] Each credential contains: vehicle ID, current available time period (locked task timeline), and real-time status hash value. Credential updates are triggered by smart contracts and verified by consensus nodes (each capacity provider or alliance node).

[0027] The role of dynamic digital vouchers: When making decisions, the scheduling system only reads the time period of "committed tasks" in the voucher to avoid duplicate assignments; at the same time, the voucher is tamper-proof and can be used for post-event auditing.

[0028] Preferred Implementation: In the consortium blockchain model, the digital certificate for each vehicle also includes the vehicle's credit score (based on historical on-time rate, cargo loss rate, etc.), and the dispatch system can give priority to vehicles with high credit scores when assigning vehicles.

[0029] Step S2: Receive customer order requests, perform multi-scale spatiotemporal coding on the order's shipping and receiving locations according to at least two administrative levels, and generate an order feature vector by combining the order's time parameters and cargo feature parameters; Unstructured order information (address text, time requirements) is converted into computable numerical feature vectors, providing a foundation for subsequent similarity calculation and clustering.

[0030] Specific implementation method: The place of shipment and the place of receipt should be coded in a hierarchical manner according to at least two administrative levels (e.g., country → province → city → district / county). For example, orders from China can be coded as follows: the country level code uses the ISO 3166-1 numerical code (e.g., 156 represents China), the region level code uses the provincial administrative division code (e.g., 110000 represents Beijing), and the local level code uses the city level code (e.g., 110100 represents the municipal districts of Beijing).

[0031] The encoding uses a fixed-length integer string, such as 6 or 9 bits, and allows bitwise comparison of hierarchical matching.

[0032] Time parameters: The expected shipping time window is converted into two timestamps (start time and end time), or into a center time plus the time window radius.

[0033] Cargo characteristic parameters: weight (kg), volume (m³), cargo type label (predefined label set, such as "general", "cold chain", "dangerous goods", "fragile", etc., using unique hot coding or category index).

[0034] The final order feature vector is represented as: [start point code, end point code, shipping start timestamp, shipping end timestamp, weight, volume, goods type code], or each level of the hierarchical coding is used as a vector element.

[0035] Preferred embodiment: For long-distance cross-province orders, only national and regional codes (two levels) are sufficient to distinguish directions; for intra-city orders, local three-level codes are used for accurate positioning; the system can be configured to use combinations of different levels.

[0036] Step S3: Calculate the grouping suitability based on the spatial overlap of the transportation routes and the overlap of the time windows of the two orders, and dynamically adjust the grouping attraction threshold according to the current order backlog rate of the system. Cluster the orders based on whether the goods types are compatible to generate a candidate order combination set. During clustering, different clustering rules are selected according to the administrative level of the origin and destination of the orders.

[0037] The system can evaluate the economics and feasibility of any two order combinations in real time and adapt to system load pressure to generate high-quality candidate order combinations.

[0038] Specific implementation method: Group buying compatibility: Taking into account spatial overlap (degree of route sharing), temporal overlap (degree of time window matching) and cargo compatibility.

[0039] Dynamic threshold: The system monitors the order backlog rate (the ratio of the number of currently unprocessed orders to the system's processing capacity) in real time, setting multiple backlog rate ranges (e.g., [0, 0.3), [0.3, 0.6), [0.6, 1.0), >1.0), with a preset gravity threshold for each range (e.g., 0.8, 0.6, 0.4, 0.2). The higher the backlog rate, the lower the threshold, allowing more orders to be combined to quickly clear the backlog.

[0040] Dual-mode clustering: Differentiate between cross-regional orders (where the origin and destination belong to different provincial regions) and intra-regional orders, and adopt different merging strategies; cross-regional orders prioritize route overlap (e.g., the same trunk line direction), while intra-regional orders prioritize the distance between the origin and destination (to facilitate cargo collection and delivery).

[0041] Preferred embodiment: During the morning rush hour, when the order backlog rate soars, the system automatically lowers the attraction threshold from 0.7 to 0.3, allowing orders that could barely be combined to be combined, thus reducing the overall waiting time.

[0042] Step S4: Construct a joint optimization mathematical model with total transportation cost as the evaluation index, and use a genetic algorithm to iteratively solve the mathematical model to reduce the total transportation cost. At the same time, determine the order combination and capacity assignment scheme, and output an integrated scheduling plan. When the genetic algorithm is initialized, it prioritizes matching capacity according to the value score of the order combination. During the evolution process, it adjusts the crossover probability according to the dispersion of the population fitness and the mutation probability according to the number of generations.

[0043] Related explanation: By solving the two decision variables of order combination (which orders are transported together) and capacity assignment (which vehicle performs which combination) in the same optimization problem, suboptimal solutions caused by sequential decision-making are avoided.

[0044] Objective function: Minimize total transportation costs (including fixed vehicle dispatch costs, variable route costs, timeout penalties, and regional load balancing penalties).

[0045] Decision variables: x_{v,c} represent whether vehicle v executes order combination c. Constraints include: an order can only belong to one combination, a combination can only be assigned to one vehicle, vehicle load / volume limits, driver working hour limits, time window constraints, etc.

[0046] Improved genetic algorithm: adopts combination value score priority initialization (instead of completely random), dynamic crossover / mutation probability, elite retention + simulated annealing local search.

[0047] Preferred Implementation: The system receives 50 orders and has 20 available vehicles. First, step S3 generates 15 candidate order combinations. Then, a genetic algorithm finds the optimal vehicle-combination matching scheme within 100 generations, resulting in a lower overall cost compared to the traditional serial method.

[0048] Step S5: Monitor changes in order flow, capacity status, and road conditions. When the preset rescheduling trigger conditions are met, perform incremental optimization on unexecuted order combinations and update the integrated scheduling plan.

[0049] In actual execution, order flow and capacity status change in real time, requiring a rapid response without having to optimize from scratch.

[0050] Triggering conditions: The new order can be combined with an existing task and is better; vehicle breakdown or traffic disruption; fixed time window expires (e.g., every 30 minutes).

[0051] Incremental optimization: Local rescheduling is performed only on affected tasks (such as existing tasks that may be added by new orders, or tasks originally assigned to faulty vehicles), and heuristic algorithms are used to solve the problem quickly.

[0052] Preferred embodiment: If a vehicle suddenly reports a malfunction before loading, the system finds another available vehicle in the nearby area within 5 seconds and updates the electronic waybill. The driver immediately receives the new task, and the delay is controlled within 10 minutes.

[0053] The standardization process includes at least the following: unifying the vehicle location coordinates of different transport providers to the same coordinate system, unifying the weight unit to kilograms, unifying the volume unit to cubic meters, and unifying the time format to the same timestamp format.

[0054] Ensure that heterogeneous data from different capacity providers can be understood and calculated in a unified manner.

[0055] Specific implementation method: Coordinate unification: The coordinates output by all GPS devices may use WGS84, GCJ-02 (Mars coordinate system), or BD09 (Baidu coordinate system); the system uses a pre-built conversion library (such as open-source coordinate conversion algorithms) to convert all input coordinates to the WGS84 coordinate system for distance calculation and map matching.

[0056] Standardize weight units: Different companies may use tons (t), kilograms (kg), or pounds (lb). The system will uniformly convert these to kilograms (kg) and retain two decimal places.

[0057] The volume unit is standardized to cubic meters (m³). For refrigerated trucks, which may use units such as "plates," the system needs to convert them according to the standard (e.g., 1 plate ≈ 1.2 m³).

[0058] Standardized time format: Different systems may use different formats (such as "2025-01-01 10:30:00" or "2025 / 01 / 01T10:30:00Z"). The system will uniformly convert it to a Unix timestamp (the number of seconds since 1970-01-01 00:00:00 UTC) to facilitate comparison of time windows and calculation of time differences.

[0059] Preferred Implementation: The system is configured with a data adaptation layer, and specific access plugins (such as reading API JSON fields or parsing CSV files) are written for each capacity provider. The plugins handle unit conversion internally. The converted data is stored in the standard data table of the distributed virtual capacity pool.

[0060] The multi-scale spatiotemporal coding specifically includes: hierarchically coding the origin and destination of the order according to three granularities: national level, regional level, and local level, forming a starting point coding vector and an ending point coding vector; and combining the start and end times of the expected delivery time window, the weight of the goods, the volume of the goods, and the goods type label vector with the starting point coding vector and the ending point coding vector to form an order feature vector.

[0061] Geographic location information is converted into a code that can be hierarchically matched, making it easier to determine whether two orders are "on the same route" and the granularity of clustering.

[0062] Specific implementation method: Country-level coding: Uses the UN M49 code or the ISO 3166-1 three-digit code. For example, China's code is 156, and the United States' code is 840.

[0063] Regional coding: China uses provincial administrative division codes (GB / T 2260), for example, Beijing 110000, Guangdong 440000.

[0064] Local-level coding: Use the prefecture-level city / district code, for example, Shenzhen 440300, Guangzhou 440100.

[0065] Encoding vector: The starting encoding vector is [country code, province code, city code], and the ending encoding vector is the same.

[0066] Order Feature Vector: Based on the above encoding, additional time window start and end times, weight, volume, and cargo type are added. For example: [156,440000,440300, 1743508800,1743512400, 500, 2.5, 1] ​​indicates a shipment from Shenzhen, Guangdong Province, China, with a time window of 2025-04-01 12:00-13:00, a weight of 500kg, and a volume of 2.5m³. 3 The type of goods is ordinary.

[0067] Preferred embodiment: For cross-border orders, a four-level coding system (continent-country-province-city) can be extended; the system allows configuration of the number of coding levels, for example, domestic orders only use the province-city level, and international orders use the country-province level.

[0068] The calculation method for the group-buying suitability in step S3 is as follows: For any two orders, the spatial gravity value is determined based on the shared mileage ratio of their transportation paths and the distance between the geometric centers of the two paths; the temporal gravity value is determined based on the ratio of the overlap length of their expected time windows to the total time span; and the compatibility coefficient is determined based on whether there is a conflict between their cargo types. The weighted spatial gravity value and the temporal gravity value are added together and then multiplied by the compatibility coefficient to obtain the group-buying suitability. The compatibility coefficient is 0 when there is a conflict and 1 when there is no conflict. The distance between the geometric centers of the two paths refers to the average coordinates of the starting point, ending point, and all planned transit points of each path as the geometric center of the path, and then the spherical distance between the two geometric centers is calculated.

[0069] Quantify the degree to which two orders are "on the same route" as the core criterion for deciding whether to combine them.

[0070] Specific implementation method: Spatial Gravity Value: First, calculate the shared mileage ratio of the two transportation routes. Method for obtaining shared mileage: Perform route planning for each route (e.g., using an open map API to obtain a series of latitude and longitude points of the actual driving route), then find the length of the shared road segment between the two routes through string matching or buffer analysis. Shared mileage ratio = Shared road segment length / (Total length of route 1 + Total length of route 2). Simultaneously calculate the distance between the geometric centers of the two routes: Take the latitude and longitude coordinates of all points along each route (including the start point, end point, and all inflection points on the planned route), calculate their arithmetic mean to obtain the center point, and then calculate the spherical distance between the two center points (using the Haversine formula); Spatial gravity value = Shared mileage ratio / (Center distance + 1) or multiplied by a coefficient inversely proportional to the distance.

[0071] Time gravity value: Time window overlap length = max(0, min(end time 1, end time 2) - max(start time 1, start time 2)). Total time span = max(end time 1, end time 2) - min(start time 1, start time 2). Time gravity value = overlap length / total time span.

[0072] Compatibility coefficient: A predefined set of conflict rules, such as "dangerous goods" and "food" cannot be transported together, and "refrigerated" and "room temperature" cannot be transported together. If any conflict rule is triggered, the compatibility coefficient = 0; otherwise, it = 1.

[0073] Group buying suitability: The spatial gravity value is multiplied by the weight α (e.g., 0.6), the temporal gravity value is multiplied by the weight β (e.g., 0.4), the sum is then multiplied by the compatibility coefficient. The final result is between 0 and 1, with a larger value indicating a better suitability for group buying.

[0074] Further explanation regarding "geometric center" and "spherical distance": The calculation of the geometric center does not require perfectly equidistant sampling of waypoints; it is sufficient to record key nodes (such as intersections and highway entrances / exits) in the order of the planned route. In practice, the route point set returned by the route planning interface of the map service can be called.

[0075] The formula for spherical distance is: d = R * acos{sin(lat1)*sin(lat2) + cos(lat1)*cos(lat2)*cos(lon2-lon1)}, where R is the average radius of the Earth (approximately 6371 km). This calculation is standard practice in Geographic Information Systems (GIS), and its accuracy meets the requirements for logistics scheduling.

[0076] Preferred Implementation: Two orders, Route 1 from Beijing to Shanghai (via Jinan), Route 2 from Tianjin to Shanghai (via Jinan), sharing a mileage (Jinan-Shanghai segment) of approximately 800km. The total route lengths are 1200km and 1000km respectively, with a sharing ratio of 800 / (1200+1000) = 0.363. The geometric center distance is approximately 200km; spatial gravity ≈ 0.363 / 201 ≈ 0.0018 (if divided by distance, the value is very small; the actual formula can be adjusted to sharing ratio / (1+distance / 100)); the time windows are 9-12 and 10-13, overlapping by 2 hours, with a total span of 4 hours, and a time gravity of 0.5. α = 0.5, β = 0.5, group buying compatibility = 0.5*0.0018 + 0.5*0.5 ≈ 0.2509; if the goods are compatible, continue; otherwise, set to 0.

[0077] The method for dynamically adjusting the group-buying attraction threshold in step S3 is as follows: set multiple different order backlog rate intervals, each interval corresponding to a preset attraction threshold; obtain the current order backlog rate of the system, determine the interval in which the backlog rate is located, and select the attraction threshold corresponding to the interval as the current group-buying attraction threshold.

[0078] This allows the system to automatically loosen or tighten group-buying conditions based on real-time load pressure, thus smoothing out order backlogs.

[0079] Specific implementation method: Pre-divide the order backlog rate into N consecutive intervals. For example: Interval 1: [0, 0.3) → Gravitational threshold = 0.7 Interval 2: [0.3, 0.6) → Gravitational threshold = 0.5 Interval 3: [0.6, 1.0) → Gravitational threshold = 0.3 Interval 4: [1.0, ∞) → Gravitational threshold = 0.1 The order backlog rate is defined as the current number of orders pending processing divided by (system's designed hourly processing capacity × observation duration). For example, if the system can process 100 orders per hour and there are currently 150 backlogged orders, then the backlog rate is 1.5.

[0080] The system acquires the backlog rate in real time, determines the corresponding interval, and retrieves the appropriate attraction threshold. Only order pairs with a matching degree greater than this threshold are allowed to be combined.

[0081] Preferred Implementation Example: During the Double 11 shopping festival, a logistics platform experienced a surge in orders, with the backlog rate reaching 2.0. The system automatically adopted the minimum threshold of 0.1, allowing almost all orders to participate in group buying. Although the quality of group buying decreased, the overall throughput increased, avoiding large-scale order timeouts.

[0082] The clustering method described in step S3 is as follows: extract the spatial codes of the origin and destination from the order feature vector. When the national or regional codes of the origin and destination are different, they are determined to be cross-regional orders, and the path overlap priority mode is adopted: calculate the path overlap degree between each two orders, and group the orders with a path overlap degree greater than a preset overlap degree threshold into the same group; when the national and regional codes of the origin and destination are the same, they are determined to be orders within the same region, and the origin / destination aggregation mode is adopted: calculate the distance between the shipping location and the receiving location of each two orders, and group the orders with a shipping distance less than a preset shipping distance threshold and a receiving distance less than a preset receiving distance threshold into the same group.

[0083] Different types of orders are clustered using the most suitable strategy to improve the effectiveness of the combination.

[0084] Specific implementation method: For cross-regional orders (where the origin and destination are not in the same provincial region): the route overlap priority mode is adopted.

[0085] For every two orders, the actual driving route is planned using the map API, and the proportion of the common road segment length of the two routes to the total route length (path overlap) is calculated.

[0086] If the overlap is greater than a preset threshold (e.g., 0.7), the two orders will be grouped into the same candidate combination.

[0087] Greedy or graph clustering algorithms (such as disjoint-set data structure) are used to connect orders that meet the conditions into a group.

[0088] Orders within the same region (origin and destination belong to the same prefecture-level city): adopt the origin / destination aggregation model.

[0089] Calculate the Euclidean or Manhattan distance between the shipping coordinates of two orders, and the distance between the receiving coordinates.

[0090] If both distances are less than a preset threshold (e.g., a delivery distance threshold of 5 kilometers and a receiving distance threshold of 5 kilometers), then they are merged.

[0091] This model is conducive to forming a front-end collection combination or a back-end delivery combination in the "collection-trunk line-distribution" process.

[0092] Preferred Implementation: In a certain city, multiple orders are shipped from different residential areas to the same shopping mall. The distance between the shipping locations is within 3 kilometers, and the delivery locations are the same (distance 0). By using an aggregation model, these orders are combined into one batch and picked up by a single truck for direct delivery, saving multiple round trips.

[0093] The total transportation cost includes: fixed cost per trip, variable cost based on route planning, order timeout penalty cost, and regional load balancing penalty cost. The regional load balancing penalty cost is determined as follows: the scheduling range is divided into multiple regions; for each region, the ratio of the number of allocated vehicles in that region to the total number of vehicles in that region is obtained; the average of the ratios for all regions is calculated; the deviations between the ratios for each region and the average value are squared, summed, and then multiplied by a dynamic penalty coefficient to obtain the regional load balancing penalty cost; the dynamic penalty coefficient is positively correlated with the variance of the number of uncompleted orders in each region during the current scheduling period.

[0094] The optimization objective incorporates the balance of regional transport capacity distribution to avoid excessive concentration of transport capacity in some areas while other areas suffer from insufficient vehicles.

[0095] Specific implementation method: Fixed costs per trip: These include the driver's basic salary, vehicle depreciation and amortization, insurance, and other fixed expenses incurred for each trip.

[0096] Variable costs based on route planning: fuel consumption, tolls, etc. are calculated based on the optimal route for the vehicle to execute order combinations (considering congestion, highway priority, etc.), and are proportional to the mileage traveled.

[0097] Order late delivery penalty cost: If the delivery time is later than the expected time window, the penalty unit price is multiplied by the minute or hour (the penalty coefficient varies depending on the time sensitivity level of the order).

[0098] Regional load balancing penalty cost: First, the dispatch scope is divided into multiple regions (such as urban districts or provincial regions).

[0099] For each region r, calculate the ratio of the number of allocated vehicles to the total number of available vehicles in the region. r .

[0100] Calculate the average (avg) of the ratios across all regions. ratio = (Σratio r ) / R.

[0101] Calculate the sum of squared deviations Σ(ratio) r - avg ratio )^2.

[0102] Multiplied by the dynamic penalty coefficient λ dynamic They will incur the cost of punishment.

[0103] Dynamic penalty coefficient λ dynamic The variance σ of the number of unfulfilled orders in each region during the current scheduling period 2 Positive correlation. For example, λ.dynamic = λ base * (1 + σ 2 / ); σ 2 This represents the variance of the number of unfulfilled orders in each region within the current scheduling period, used to quantify the unevenness of order backlog between regions. Variance σ 2 The calculation method is as follows: First, calculate the number of unfulfilled orders in each region, then calculate the average number of unfulfilled orders in all regions, then calculate the square of (number of unfulfilled orders in the region - average number) for each region, and finally calculate the average of these squares (or divide by the number of regions R).

[0104] σ0 2 This is a preset variance normalization constant, whose function is to make the variance dimensionless, so that λ dynamic Dimensions of λ base Consistent. σ0 2 It can take the value 1, or it can take the historical average variance or other empirical values, such as σ0. 2 = 100; λ base The preset base penalty coefficient represents the penalty intensity when the regional load is perfectly balanced. Its value can be set by the system administrator according to actual business needs, for example, λ. base = 500 or λ base = 1000; The larger the variance, the more uneven the backlog between regions, and the larger the coefficient, forcing the scheduling scheme to be more inclined to balanced distribution.

[0105] When the number of unfulfilled orders is perfectly balanced across regions (σ) 2 = 0), λ dynamic = λ base The penalty cost is minimized; when the degree of imbalance between regions increases (σ 2 When λ increases, dynamic This also increases the penalty weight for regional load imbalance in the total cost, guiding the scheduling algorithm to prioritize solutions that make the regional capacity distribution more balanced.

[0106] Preferred Implementation: A city has an East District and a West District. The East District has more orders but also more vehicles, while the West District has fewer orders but also fewer vehicles. The variance of uncompleted orders in the current scheduling cycle is 2.5 (relatively high). The system automatically increases the regional balancing penalty weight, causing the optimization algorithm to prioritize dispatching some vehicles from the East District to the West District to perform tasks, ultimately bringing the completion rates of the two districts closer together.

[0107] The specific execution method of the genetic algorithm in step S4 includes: Step S41: Obtain the combination value score for each order combination, which is a weighted sum of four parameters: total order weight, total volume, total freight revenue, and order timeliness; match transportation capacity to each order combination in descending order of combination value score to generate an initial population; Step S42: Use the reciprocal of the total transportation cost as the fitness value for each individual; Step S43: Obtain the standard deviation of the fitness values ​​of all individuals in the current population and obtain the current generation number; when the standard deviation is less than a preset lower threshold, increase the crossover probability by a preset increment value; when the standard deviation is greater than a preset upper threshold, decrease the crossover probability by a preset decrement value; when the ratio of the current generation number to the preset maximum generation number exceeds a preset ratio threshold, increase the mutation probability generation by generation according to a preset step size until the preset maximum generation number is reached. Step S44: After each generation of evolution is completed, the top K individuals with the highest fitness values ​​are retained for the next generation, where K is a preset positive integer; and simulated annealing perturbation operation is performed on the retained individuals with a preset perturbation probability. The perturbation operation is to randomly swap an order combination with a vehicle matching pair. Step S45: Repeat steps S42 to S44 until the current generation reaches the preset maximum generation, and output the integrated scheduling plan corresponding to the individual with the highest fitness value.

[0108] We design an efficient search algorithm that is not prone to getting trapped in local optima for joint optimization problems.

[0109] Explanation of each sub-step: Step S41 (Initialize the population): The combined value score for each order combination is calculated as follows: w1 * Total weight + w2 * Total volume + w3 * Total shipping revenue + w4 * Time urgency. Time urgency can be expressed as (the reciprocal of the current time and the latest shipping time, or exponential decay).

[0110] The combinations are sorted in descending order of their scores, and each combination is matched with the most suitable vehicle (e.g., the vehicle that meets the weight and capacity requirements and has the lowest cost). The remaining capacity is randomly allocated to the remaining combinations, forming the initial population of individuals (each individual represents a complete "combination-vehicle" matching scheme).

[0111] Population size is usually set to 100-200.

[0112] Step S42 (fitness value): Each individual corresponds to a complete scheduling plan, and its total transportation cost (including fixed, variable, timeout, and regional equilibrium penalties) is calculated. Fitness = 1 / (total cost + 1e-6). Higher fitness indicates a better solution.

[0113] Step S43 (Dynamic Crossover Mutation): Calculate the standard deviation of fitness (std) of all individuals in the population. fit If std fit If the value is less than the lower threshold (e.g., 0.05), it indicates that the population is converging, and the crossover probability should be increased to introduce new genes; if std fit If the value is greater than the upper limit threshold (e.g., 0.2), it indicates that the population is too dispersed. Reduce the crossover probability to speed up convergence.

[0114] The initial crossover probability is 0.8, and the step size is adjusted by 0.05 each time.

[0115] When the number of generations gen is equal to the maximum number of generations max gen When the ratio exceeds 0.7, the mutation probability increases with each generation. The initial mutation probability is 0.01, increasing by 0.002 each generation until it reaches the maximum mutation probability of 0.1.

[0116] Step S44 (Elite Preservation and Simulated Annealing): The top K (e.g., 10) individuals in each generation are retained and directly enter the next generation to ensure that the optimal solution is not lost.

[0117] These elite individuals are perturbed with a probability of 0.1: two different order combinations are randomly selected, their currently assigned vehicles are swapped (if capacity allows), or an order is moved from one vehicle to another. Simulated annealing mechanism: a solution worse than the current one is accepted with a certain probability, the initial temperature is set to 1000, and the temperature is reduced by 0.95 per generation.

[0118] Step S45 (Iteration Termination): When the preset maximum number of generations (e.g., 500 generations) is reached, the scheduling plan corresponding to the historical best individual is output.

[0119] Preferred embodiment: In actual testing, compared with the standard genetic algorithm, this improved genetic algorithm has a faster convergence speed and a lower quality of the final solution (i.e., the total transportation cost).

[0120] The preset rescheduling triggering condition mentioned in step S5 includes at least one of the following: When a new order arrives, the grouping suitability between the new order and any allocated but not yet loaded order in the current scheduling plan is obtained. If the grouping suitability is greater than the current gravity threshold corresponding to the allocated but not yet loaded order, rescheduling is triggered. When a capacity status change event occurs, an alternative capacity matching process is triggered within a preset response time threshold. The preset time window has expired.

[0121] Define which dynamic events require interrupting the current scheduling plan and initiating a re-optimization.

[0122] Specific implementation method: Triggering condition 1: A new order arrives, and its suitability for grouping with all assigned but not yet dispatched orders is calculated. If an existing order exists with a suitability greater than its current attraction threshold (note: this threshold may vary depending on the backlog rate), rescheduling is triggered. In this case, the new order and existing orders are merged and re-optimized.

[0123] Triggering condition 2: A change in transport capacity status, such as vehicle malfunction, driver leave, or temporary road closure. The system has a response time threshold (e.g., 5 seconds), within which the alternative transport capacity matching process must be triggered, i.e., another available vehicle must be selected from the virtual transport capacity pool to replace it.

[0124] Triggering condition 3: Timed triggering, for example, automatically performing incremental optimization every 30 minutes, re-clustering and assigning new orders with unexecuted orders.

[0125] Preferred Implementation: If a vehicle experiences a tire blowout while in motion, the system detects the fault via the onboard IoT device and immediately triggers a re-dispatch. A replacement vehicle is located within a 5-kilometer radius, and the electronic waybill is automatically updated. Both the original driver and the replacement driver are notified, and the overall delay is controlled within 15 minutes.

[0126] The distributed virtual capacity pool uses distributed ledger technology to maintain dynamic digital credentials for each vehicle and shares anonymized capacity data among capacity providers through smart contract protocols. Each time a status change event of a scheduling plan is triggered, the smart contract automatically generates a record containing vehicle identifier, order combination identifier, and cost sharing coefficient. This record is written into the blockchain after being confirmed by consensus among all participants and is used for automatic settlement of fees and determination of responsibility after the task is completed.

[0127] Provides trusted sharing and automated settlement mechanisms across enterprises.

[0128] Specific implementation method: A consortium blockchain is built using permissioned blockchains (such as Hyperledger Fabric), with each capacity provider acting as a node, sharing anonymized capacity data (e.g., only the region, vehicle type, and available time period are disclosed, while specific customer information is not).

[0129] Dynamic digital credentials exist as on-chain assets, including vehicle ID, available time slot, current task hash, etc.

[0130] When the dispatch system generates an integrated dispatch plan and assigns vehicles to execute order combinations, it invokes the smart contract. Smart contract logic: Verify whether the available time slots for vehicles conflict with the mission time.

[0131] Lock the time slot corresponding to this vehicle.

[0132] Generate a record: {vehicle_id, order_combination_id, departure_time, route_hash, cost_allocation_coefficient}.

[0133] The record is broadcast to the consensus nodes, and after verification by all nodes, it is written into a new block.

[0134] After the task is completed, the driver submits proof of completion (such as a signed photo or GPS track) through the app, and the smart contract automatically transfers the freight from the shipper's pre-deposited account to the accounts of each transportation provider according to the cost sharing coefficient.

[0135] Preferred Implementation Example: Company A's vehicle executes an order combination from Company B. The smart contract automatically completes the payment within 1 minute after the task is completed, based on the pre-agreed profit-sharing ratio (e.g., Company A 80%, platform 20%). No manual reconciliation is required, and disputes can be traced through on-chain records.

[0136] The following description is based on specific embodiments: Example 1: Same-city instant delivery scenario: Background: A local freight platform has 2,000 registered drivers and an average of 5,000 orders per day. The platform uses the method of this invention for dispatching.

[0137] Steps and procedures: Capacity Access (S1): Drivers register via a mobile app and authorize the app to access information such as GPS location, vehicle load, and capacity. After standardizing this data, the platform creates a dynamic digital credential for each driver on the Hyperledger Fabric blockchain, recording the time period during which they have accepted orders today.

[0138] Order Generation (S2): Customer A places an order to deliver a box of documents from location A (Shenzhen Nanshan Science and Technology Park) to location B (Futian Civic Center). The box weighs 5kg and has a volume of 0.1m³. Delivery is expected within 1 hour. The system generates feature vectors for locations A and B using three-level coding: Nanshan District 440305 and Futian District 440304.

[0139] Group Buying Clustering (S3): The system currently has 10 unassigned orders. Calculate the group buying fit between the new order and existing orders. Order B travels from location C (Nanshan District Coastal City) to location D (Futian District Shopping Park), with high path overlap (both follow Nanhai Avenue-Binhe Avenue), overlapping time windows, and compatible goods (both are files), resulting in a fit of 0.82. The current backlog rate is 0.4 (relatively low), corresponding to a gravity threshold of 0.6. Since 0.82 > 0.6, group buying is allowed. Use a same-area aggregation mode to combine A and B.

[0140] Joint Optimization (S4): The system has 15 candidate order combinations and 20 available vehicles. The objective function considers fixed costs (50 yuan per trip), variable costs (2 yuan per kilometer), timeout penalties (2 yuan per minute), and regional equilibrium (vehicle ratio between Nanshan District and Futian District). The genetic algorithm finds the optimal solution after 80 generations: Combination A+B is assigned to driver 1 (Jinbei van, 1-ton load capacity), the route is planned as A→C→B→D, the total distance is 35 kilometers, the cost = 50 + 70 = 120 yuan, of which 70 yuan comes from variable costs (2 yuan / km × 35km), there is no timeout, and the regional ratio deviation is small.

[0141] Execution and Monitoring (S5): After driver 1 departs, the platform monitors their trajectory. During the journey, a new order C is added from Nanshan Bookstore to Futian Station. The system calculates the suitability of unexecuted orders in the combination A+B and finds a high degree of path overlap with B, triggering a rescheduling. C is added to the combination, the route is replanned, and the driver is notified to add a pickup point. The driver confirms and continues.

[0142] Settlement (S10): After the task is completed, the smart contract allocates the freight according to the mileage ratio based on the order combination (A+B+C) actually executed by driver 1, and transfers the fee from the customer account to the driver account, without the need for manual intervention.

[0143] Example 2: Inter-provincial trunk line transportation scenario: Background: A logistics alliance comprises five small and medium-sized logistics companies, each owning 10-20 trucks. The alliance uses the method of this invention to establish a cross-enterprise capacity-sharing platform.

[0144] Steps and procedures: Transportation capacity access: Each enterprise uploads real-time vehicle data to the consortium blockchain via API to create dynamic digital certificates, which indicate the enterprise to which the vehicle belongs, the available departure time, and the current city.

[0145] Order Posting: The shipper posted an order for electronic products, weighing 8 tons and measuring 25 cubic meters in volume, from Pudong New Area, Shanghai to Daxing District, Beijing. 3Delivery is expected within two days. The system uses spatiotemporal coding: National 156, Provincial 310000 (Shanghai), Beijing 110000, Local 310115 (Pudong), Daxing 110115.

[0146] Group buying clustering: The platform already has 3 other orders (Suzhou → Beijing, Hangzhou → Tianjin, etc.). Fit calculation: The Shanghai → Beijing and Suzhou → Beijing routes share a high mileage ratio (Beijing-Shanghai Expressway), have a high spatial attraction value, can be coordinated within a time window, and are compatible with each other (electronic products and general goods do not conflict), resulting in a fit of 0.75. The current backlog rate is 0.2 (low), and the threshold is 0.6, allowing group buying. A cross-regional route overlap priority mode is adopted to form combinations.

[0147] Joint Optimization: There are 8 available inter-provincial vehicles (belonging to 4 companies). The objective function includes regional equilibrium (vehicle usage ratio in East China and North China). A genetic algorithm selects the optimal solution: assigning the combination to a 17.5-meter van belonging to company B, which is currently empty in Shanghai, and company B has return cargo in Beijing. Cost Calculation: Fixed cost 800 yuan, variable cost 1.5 yuan per kilometer, Shanghai to Beijing is approximately 1200 kilometers, variable cost 1800 yuan, timeout penalty 0, regional equilibrium penalty low, total cost 2600 yuan.

[0148] Execution and Rescheduling: After the vehicle departed, due to road closures caused by heavy fog in Shandong, the system detected the change in road conditions and immediately triggered rescheduling, planning an alternative route (via the Rongwu Expressway), adding 150 kilometers to the journey but avoiding prolonged congestion. Simultaneously, the recipient was notified of an estimated 2-hour delay, and the overdue penalty coefficient was automatically adjusted.

[0149] Settlement and Liability Determination: Upon completion of the task, the smart contract automatically calculates the cost allocation: the order freight is 5000 yuan, and after deducting fuel and toll fees, the actual profit is 2000 yuan, which is allocated proportionally to Enterprise B (80%) and the platform (20%). Additional costs incurred due to detours are covered by the platform's insurance. The blockchain record is complete and dispute-free.

[0150] Example 3: A large logistics platform has integrated with three types of transportation capacity providers: Company A: Owns 20 vans, mainly serving the East China region; Team B: 15 refrigerated trucks, serving the entire country; C. Individual vehicle owner: 50 individual vehicles scattered across the country.

[0151] The platform uses the method described in this invention to perform cross-regional capacity collaborative scheduling and intelligent order grouping optimization.

[0152] Step S1: Capacity Access and Virtual Capacity Pool Construction Data access and standardization: Company A pushes vehicle data via API (coordinates GCJ-02, weight in tons, volume in liters, time format YYYY-MM-DD HH:MM:SS).

[0153] Team B uploads OBD data (coordinates WGS84, weight in kilograms, volume in cubic meters, timestamp in Unix milliseconds) via the MQTT protocol.

[0154] Individual vehicle owners (C) upload GPS data via a mobile app (coordinates BD09, weight in kilograms, volume not specified, time format ISO8601).

[0155] Standardization process: Coordinate unification: Call convert_to_wgs84 to convert all coordinates to WGS84.

[0156] The unit of weight is standardized: Company A's ton → kilogram (×1000).

[0157] The volume unit is standardized: for Company A, liters → cubic meters (÷1000); for fleet B, it remains unchanged; and for individual vehicle owners in C, the volume is estimated based on the vehicle model by default.

[0158] Time format is standardized: all timestamps are converted to Unix second-level timestamps.

[0159] Constructing a distributed virtual capacity pool: On the Hyperledger Fabric consortium blockchain, dynamic digital credentials are created for each vehicle. For example, the credential for vehicle Guangdong B12345 includes: vehicle_id, owner (Company A), available_time_slots (initially all available), committed_task_hashes (empty), credit_score (based on historical on-time rate).

[0160] In each subsequent scheduling assignment, the smart contract automatically updates the credential, locking the corresponding time period. Other alliance nodes verify the credential through consensus.

[0161] Step S2: Multi-scale spatiotemporal coding of orders: The platform received the following three orders:

[0162] Multi-scale spatiotemporal coding: Country-level coding: For example, China → 156 Regional codes: for example, Guangdong Province 440000, Beijing 110000, Tianjin 120000 Local-level codes: for example, Nanshan District 440305, Tianhe District 440106, Futian District 440304, Chaoyang District 110105, Heping District 120101 Order feature vector (taking O1 as an example): [156,440000,440305, 156,110000,110105, 1743512400,1743523200, 2500,8.5, 3] (Goods type 3 = Electronic products) Step S3: Group Buying Clustering Group buying suitability calculation: O1 and O2: Route planning (using Gaode API) shows a shared mileage (Guangzhou-Beijing section) of approximately 2100km; O1's total length is 2300km, and O2's total length is 2050km, with a sharing ratio of 2100 / (2300+2050)=0.483. The geometric center distance is 22.2km, and the spatial gravity value is 0.483 / (1+22.2 / 100)=0.395. The time window overlaps by 3 hours, with a total span of 4 hours, resulting in a time gravity of 0.75. Both are electronic products, with a compatibility coefficient of 1. Taking α=0.6 and β=0.4, the fit is 0.6×0.395+0.4×0.75=0.537.

[0163] O1 vs O3: Shenzhen → Beijing vs Shenzhen → Tianjin, shared mileage approximately 0 (path fork), spatial attraction ≈0, compatibility ≈0.15, very low.

[0164] O2 vs. O3: Guangzhou → Beijing vs. Shenzhen → Tianjin, low shared mileage, compatibility ≈ 0.22.

[0165] Dynamically adjust the gravity threshold: The system currently has 50 orders pending processing, with a processing capacity of 2 orders per minute. The observation period is 5 minutes, and the backlog rate is 50 / (2×5) = 5.0. Predefined interval: [0, 0.5) → threshold 0.7; [0.5, 1.0) → 0.5; [1.0, 2.0) → 0.3; [2.0,∞)→0.1; The current backlog rate is 5.0, and the threshold is 0.1. Therefore, the fit is 0.537 > 0.1, and O1 and O2 are allowed to combine orders.

[0166] Bimodal clustering: O1 originates in Shenzhen (440305), and O2 originates in Guangzhou (440106). They share the same regional address (440000) but differ in local address, thus classifying them as cross-regional orders. A path overlap priority mode is applied. The path overlap is calculated as: Shared Mileage / Total Length of O1 = 2100 / 2300 = 0.913, which exceeds the preset threshold of 0.7. Therefore, O1 and O2 are grouped into combination C1. O3 is grouped separately as combination C2 (due to its low overlap with O1 and O2).

[0167] Step S4: Joint Optimization and Genetic Algorithm Total transportation cost breakdown: Available vehicles: V1 (Shenzhen, 5 tons, fixed cost 300 yuan, variable cost 2.5 yuan / km), V2 (Guangzhou, 10 tons, fixed cost 350 yuan, variable cost 2.5 yuan / km), V3 (Shenzhen, 8 tons, fixed cost 320 yuan, variable cost 2.8 yuan / km).

[0168] The total weight of unit C1 is 5500 kg, and the total volume is 18.5 m³. 3 This can only be performed by V2 (10 tons). Combined unit C2 weighs 4000 kg and has a capacity of 12 m³. 3 It can be executed by V1 or V3.

[0169] Regional load balancing penalty: The regions are divided into East China (Shanghai, Jiangsu, Zhejiang, etc.), North China (Beijing, Tianjin, Hebei, etc.), and South China (Guangdong, etc.). The variance σ of unfulfilled orders in the current period. 2 =2.5, λ dynamic = 1000 × (1 + 2.5) = 3500. Calculate the sum of squares of the deviations between the allocated vehicle percentage and the average percentage for each region, and multiply by λ. dynamic .

[0170] Execution of the improved genetic algorithm: S41 Initialization: Calculate the combined value score for each order combination.

[0171] C1 (O1+O2): Total weight 5500, total volume 18.5m³ 3 The total shipping revenue is 1200 yuan + 1500 yuan = 2700 yuan. The timeliness (current time to the latest shipping time) is set to 0.8. The score is 0.3×5500 + 0.2×18.5 + 0.4×2700 + 0.1×0.8 = 1650 + 3.7 + 1080 + 0.08 = 2733.78.

[0172] Combination C2 (O3): Total weight 4000kg, total volume 12m³ 3The total shipping revenue is 2000 yuan. The timeliness is set at 0.7. The score is 0.3×4000 + 0.2×12 + 0.4×2000 + 0.1×0.7 = 1200 + 2.4 + 800 + 0.07 = 2002.47.

[0173] Sort by score from highest to lowest, first match vehicle V2 (the only feasible option) for C1, then match vehicle V1 (lower cost) for C2; generate an initial population of 100 individuals.

[0174] S42 Fitness: Calculate the total transportation cost for each individual (fixed + variable + timeout + regional equilibrium penalty) and take the reciprocal.

[0175] S43 Dynamic Crossover Mutation: Population fitness standard deviation = 0.08 (greater than the lower limit of 0.05, less than the upper limit of 0.15), crossover probability remains at 0.8. Current generation gen = 350, maximum generation 500, ratio 0.7, initiation mutation probability increases with each generation: increasing by 0.002 per generation, from 0.02 to 0.1.

[0176] S44 Elite Retention + Simulated Annealing: The top 10 individuals are retained in each generation, and the vehicle assignments of C1 and C2 are randomly swapped with a probability of 0.05 (if feasible), and the probability of accepting a worse solution decreases with temperature.

[0177] S45 Termination: Output the optimal individual after 500 generations: V2 executes C1, V1 executes C2, total cost = fixed (350+300) + variable (2.5×2100+2.5×1900) = 650+10000 = 10650 yuan, timeout 0, area penalty 500 yuan, total 11150 yuan.

[0178] Step S5 Rescheduling: After vehicle V2 departed (and entered the Beijing-Hong Kong-Macau Expressway), the platform received a new order O4: Zhengzhou → Beijing, weighing 1000kg, with a volume of 3m³. 3 The desired completion time is before 17:00 on the same day. The system calculates the compatibility between O4 and O1 / O2 in C1: O4 and O2 have extremely high route overlap (Zhengzhou-Beijing section), with a compatibility of 0.88, exceeding the current threshold (threshold 0.5 due to reduced backlog rate to 0.3), triggering rescheduling. The system adds O4 to C1, adjusts the V2 route to "Shenzhen → Guangzhou → Zhengzhou → Beijing," increases the load capacity to 6500kg, still meeting the 10-ton limit. The system updates the V2's digital voucher via smart contract, locks the new time period, and sends an electronic waybill to the driver. The rescheduling record is written to the blockchain.

[0179] Blockchain smart contracts and automated settlement: After the task is completed, the driver uploads a receipt. The smart contract automatically calculates the actual mileage allocation: O1 accounts for 30%, O2 for 40%, and O4 for 30%. The shipper's pre-deposited account is deducted from the total freight (O1's 1200 + O2's 1500 + O4's 800 = 3500 yuan), which is then allocated to the company owning V2 (2800 yuan) and the platform service fee (700 yuan). All operations are recorded on the blockchain and are traceable by all parties.

[0180] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0181] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A cross-regional transport capacity collaborative scheduling and order intelligent order optimization method, characterized in that, include: Step S1: Access the capacity data of multiple capacity providers through the Internet of Things gateway, perform standardized processing, and construct a distributed virtual capacity pool. Each vehicle in the virtual capacity pool corresponds to a dynamic digital certificate, which records the real-time availability of the vehicle and the committed tasks. Step S2: Receive customer order requests, perform multi-scale spatiotemporal coding on the order's shipping and receiving locations according to at least two administrative levels, and generate an order feature vector by combining the order's time parameters and cargo feature parameters; Step S3: Calculate the grouping suitability based on the spatial overlap and time window overlap of the transportation routes of the two orders, dynamically adjust the grouping attraction threshold based on the current order backlog rate of the system, and cluster the orders based on whether the goods types are compatible to generate a candidate order combination set; wherein, different clustering rules are selected according to the administrative level of the origin and destination of the orders during clustering. Step S4: Construct a joint optimization mathematical model with total transportation cost as the evaluation index, and use a genetic algorithm to iteratively solve the mathematical model to reduce the total transportation cost. At the same time, determine the order combination and capacity assignment scheme, and output an integrated scheduling plan. The genetic algorithm prioritizes matching capacity according to the value score of the order combination during initialization, and adjusts the crossover probability according to the dispersion of the population fitness and the mutation probability according to the number of generations during the evolution process. Step S5: Monitor changes in order flow, capacity status, and road conditions. When the preset rescheduling trigger conditions are met, perform incremental optimization on unexecuted order combinations and update the integrated scheduling plan.

2. The method according to claim 1, characterized in that, The standardization process includes at least the following: unifying the vehicle location coordinates of different transport providers to the same coordinate system, unifying the weight unit to kilograms, unifying the volume unit to cubic meters, and unifying the time format to the same timestamp format.

3. The method according to claim 1, characterized in that, The multi-scale spatiotemporal coding specifically includes: hierarchically coding the origin and destination of the order according to three granularities: national level, regional level, and local level, forming a starting point coding vector and an ending point coding vector; and combining the start and end times of the expected delivery time window, the weight of the goods, the volume of the goods, and the goods type label vector with the starting point coding vector and the ending point coding vector to form an order feature vector.

4. The method according to claim 1, characterized in that, The calculation method for the group-buying suitability in step S3 is as follows: For any two orders, the spatial gravity value is determined based on the shared mileage ratio of their transportation paths and the distance between the geometric centers of the two paths; the temporal gravity value is determined based on the ratio of the overlap length of their expected time windows to the total time span; and the compatibility coefficient is determined based on whether there is a conflict between their cargo types. The weighted spatial gravity value and the temporal gravity value are added together and then multiplied by the compatibility coefficient to obtain the group-buying suitability. The compatibility coefficient is 0 when there is a conflict and 1 when there is no conflict. The distance between the geometric centers of the two paths refers to the average coordinates of the starting point, ending point, and all planned transit points of each path as the geometric center of the path, and then the spherical distance between the two geometric centers is calculated.

5. The method according to claim 1, characterized in that, The method for dynamically adjusting the group-buying attraction threshold in step S3 is as follows: set multiple different order backlog rate intervals, each interval corresponding to a preset attraction threshold; obtain the current order backlog rate of the system, determine the interval in which the backlog rate is located, and select the attraction threshold corresponding to the interval as the current group-buying attraction threshold.

6. The method according to claim 1, characterized in that, The clustering method described in step S3 is as follows: extract the spatial codes of the origin and destination from the order feature vector. When the national or regional codes of the origin and destination are different, they are determined to be cross-regional orders, and the path overlap priority mode is adopted: calculate the path overlap degree between each two orders, and group the orders with a path overlap degree greater than a preset overlap degree threshold into the same group; when the national and regional codes of the origin and destination are the same, they are determined to be orders within the same region, and the origin / destination aggregation mode is adopted: calculate the distance between the shipping location and the receiving location of each two orders, and group the orders with a shipping distance less than a preset shipping distance threshold and a receiving distance less than a preset receiving distance threshold into the same group.

7. The method according to claim 1, characterized in that, The total transportation cost includes: fixed cost per trip, variable cost based on route planning, order timeout penalty cost, and regional load balancing penalty cost. The regional load balancing penalty cost is determined as follows: the scheduling range is divided into multiple regions; for each region, the ratio of the number of allocated vehicles in that region to the total number of vehicles in that region is obtained; the average of the ratios for all regions is calculated; the deviations between the ratios for each region and the average value are squared, summed, and then multiplied by a dynamic penalty coefficient to obtain the regional load balancing penalty cost; the dynamic penalty coefficient is positively correlated with the variance of the number of uncompleted orders in each region during the current scheduling period.

8. The method according to claim 1, characterized in that, The specific execution method of the genetic algorithm in step S4 includes: Step S41: Obtain the combination value score for each order combination, which is a weighted sum of four parameters: total order weight, total volume, total freight revenue, and order timeliness; match transportation capacity to each order combination in descending order of combination value score to generate an initial population; Step S42: Use the reciprocal of the total transportation cost as the fitness value for each individual; Step S43: Obtain the standard deviation of the fitness values ​​of all individuals in the current population and obtain the current generation number; when the standard deviation is less than a preset lower threshold, increase the crossover probability by a preset increment value; when the standard deviation is greater than a preset upper threshold, decrease the crossover probability by a preset decrement value; when the ratio of the current generation number to the preset maximum generation number exceeds a preset ratio threshold, increase the mutation probability generation by generation according to a preset step size until the preset maximum generation number is reached. Step S44: After each generation of evolution is completed, the top K individuals with the highest fitness values ​​are retained for the next generation, where K is a preset positive integer; and simulated annealing perturbation operation is performed on the retained individuals with a preset perturbation probability. The perturbation operation is to randomly swap an order combination with a vehicle matching pair. Step S45: Repeat steps S42 to S44 until the current generation reaches the preset maximum generation, and output the integrated scheduling plan corresponding to the individual with the highest fitness value.

9. The method according to claim 1, characterized in that, The preset rescheduling triggering condition mentioned in step S5 includes at least one of the following: When a new order arrives, the grouping suitability between the new order and any allocated but not yet loaded order in the current scheduling plan is obtained. If the grouping suitability is greater than the current gravity threshold corresponding to the allocated but not yet loaded order, rescheduling is triggered. When a capacity status change event occurs, an alternative capacity matching process is triggered within a preset response time threshold. The preset time window has expired.

10. The method according to claim 1, characterized in that, The distributed virtual capacity pool uses distributed ledger technology to maintain dynamic digital credentials for each vehicle and shares anonymized capacity data among capacity providers through smart contract protocols. Each time a status change event of a scheduling plan is triggered, the smart contract automatically generates a record containing vehicle identifier, order combination identifier, and cost sharing coefficient. This record is written into the blockchain after being confirmed by consensus among all participants and is used for automatic settlement of fees and determination of responsibility after the task is completed.

Images