Goods stowage path optimization method for international freight proxy

Abstract

The invention relates to the technical field of international freight agency, in particular to an international freight agency cargo stowage path optimization method, which comprises the following steps: firstly, collecting and preprocessing multi-source heterogeneous data, and constructing a comprehensive database comprising cargo characteristics, transportation tools, road condition policies and cost effectiveness; on the basis of an analytic hierarchy process-genetic hybrid algorithm, the cargo stowage priority is determined, intelligent stowage is completed, and accurate matching of the weight, the size, the characteristics and the shipping space of the transportation tool is achieved; then combining real-time dynamic data, adopting an improved Dijkstra algorithm to carry out initial path planning, and dynamically adjusting the path through a rolling window iteration mechanism; and finally, verifying the feasibility of the scheme through a cross-modal information interaction correction module, and generating an optimal stowage path combination. According to the method, collaborative optimization of stowage and paths is realized, the dynamic adaptability and the resource utilization rate are improved, the transportation cost is reduced, the transportation period is shortened, and the method is suitable for complex international freight scenes such as multimodal transportation and cross-border transportation.

Description

Technical Field

[0001] This invention relates to the field of international freight forwarding technology, specifically to a method for optimizing cargo loading routes in international freight forwarding. Background Technology

[0002] With the continued deepening of globalization, international freight forwarding, as a core link connecting supply and demand and coordinating cross-border transportation, directly impacts the cost and timeliness of international trade through its operational efficiency. Currently, the international freight landscape is increasingly complex, involving multiple modes of transport such as sea, air, and land, and facing multiple constraints including diverse cargo types, volatile cross-border policies, uncertain road conditions, and limited cargo space. Traditional loading and route planning methods are no longer sufficient to meet the industry's demands for efficient operation.

[0003] Current international freight forwarding cargo loading technologies primarily focus on the space utilization or load balancing of a single mode of transport, lacking a comprehensive consideration of multiple factors such as cargo characteristics, timeliness requirements, and value differences. For example, some methods only complete loading through simple weight-volume ratios without specifically zoning for fragile, perishable, or dangerous goods, which can easily increase the risk of cargo damage. Furthermore, loading plans and route planning are often independent; route planning is performed after loading is complete, often resulting in poor compatibility between the loading plan and route nodes. For instance, if some cargo is loaded onto a particular mode of transport, but the corresponding route is severely congested or subject to policy restrictions, reloading is necessary, increasing additional costs and time.

[0004] In terms of route planning, existing technologies mostly employ static route algorithms, such as the traditional Dijkstra algorithm and A* algorithm. These algorithms generate routes based solely on static data such as road conditions and port status at the initial planning stage, lacking the ability to adapt to dynamic factors in real time. International freight involves large cross-border distances and long transportation cycles, and is prone to unforeseen circumstances such as port congestion, flight delays, road construction, extreme weather, and customs policy adjustments. Static routes cannot be adjusted in a timely manner, easily leading to delays in transportation time and even the risk of contract breaches. Furthermore, some route planning methods do not fully integrate information from multiple cross-border stages, such as the customs clearance efficiency of different countries and the connectivity of transit nodes, resulting in poor overall route continuity, excessive transit waiting times, and further increased operating costs.

[0005] Meanwhile, existing technologies lack sufficient data utilization capabilities. International freight involves multi-source and heterogeneous data on goods, transportation vehicles, road conditions, and policies, making it difficult for traditional methods to effectively integrate and preprocess the data, resulting in incomplete decision-making basis for loading and route planning. For example, failure to obtain real-time operational status of transportation vehicles may lead to cargo being loaded onto vehicles at risk of malfunction; failure to keep abreast of changes in customs supervision policies may result in cargo being delayed during customs clearance. Furthermore, existing methods lack an effective feedback loop mechanism, meaning that data from actual operations cannot feed back into algorithm optimization, making it difficult to continuously improve the adaptability of loading and route solutions.

[0006] Furthermore, the optimization of loading routes in current multimodal transport scenarios is particularly prominent. Multimodal transport involves the connection of multiple modes of transport, and existing methods struggle to coordinate loading plans with transshipment routes, often resulting in cargo accumulation at transshipment nodes and poor connectivity between transport vehicles, significantly reducing transport efficiency. In summary, existing international freight forwarding cargo loading route planning methods suffer from defects such as a disconnect between loading and routes, poor dynamic adaptability, insufficient data utilization, and inadequate coordination across multiple constraints. This leads to high transport costs, low timeliness, low resource utilization, and high risk of cargo damage, failing to meet the efficient operational needs of the international freight forwarding industry in the context of globalized trade. Therefore, there is an urgent need for an optimization method that can achieve coordinated optimization of loading and routes, dynamically adapt to complex scenarios, and fully utilize multi-source data. Summary of the Invention

[0007] To address the technical problems in existing international freight forwarding cargo loading and routing planning methods, such as the disconnect between loading and routing, poor dynamic adaptability, insufficient data utilization, and inadequate coordination among multiple constraints, which lead to high transportation costs, low timeliness, low resource utilization, and high risk of cargo damage, this paper proposes an international freight forwarding cargo loading and routing optimization method. This method achieves coordinated optimization of loading and routing, improves dynamic adaptability and resource utilization, and reduces costs and risk of cargo damage.

[0008] The technical solution adopted by this invention to solve its technical problem is: an international freight forwarding cargo loading route optimization method, comprising the following steps: S1: Data collection and preprocessing. Collect cargo information, transportation vehicle information, cross-border road condition information, policy and regulation information, cost and timeliness information, and environmental information. Perform noise reduction, standardization, and completion processing on the collected data to build a comprehensive database. Cargo information includes weight, volume, characteristics, timeliness requirements, fragile and perishable attributes, and destination. Transportation vehicle information includes cabin capacity, load limit, transportation mode, and operation and maintenance status. S2: Intelligent loading scheme generation. Based on a comprehensive database, it adopts a hierarchical analysis-genetic hybrid algorithm to establish cargo loading priorities with the goals of lowest cost, highest space utilization, and optimal timeliness. Combined with the cargo hold parameters of the transport vehicle, it completes the precise loading of cargo and transport vehicle and generates an initial loading scheme. S3: Dynamic route planning, based on the initial load planning scheme, adopts an improved Dijkstra algorithm, and dynamically generates the initial route by combining real-time traffic conditions, port congestion, and flight / voyage information; through a rolling window iteration mechanism, dynamic data is updated at preset time intervals to iteratively adjust the route and avoid risk nodes. S4: Scheme verification and optimization. Through the cross-modal information interaction correction module, the rationality of load allocation, path feasibility and cost and timeliness are compared. If the preset threshold is not met, return to S2 or S3 to re-optimize until the optimal load allocation path combination is generated. S5: Solution Execution and Feedback. Outputs the optimal solution and guides its execution. Collects actual data during the execution process to form a feedback loop for optimizing algorithm parameters.

[0009] Specifically, in step S1, multi-source data acquisition is achieved through IoT sensors, cross-border transportation platform API interfaces, customs supervision systems, and third-party meteorological platforms. Data preprocessing uses Lagrange interpolation to fill in missing values ​​and standard deviation to remove outliers.

[0010] Specifically, in step S2, the priority evaluation index of the analytic hierarchy process-genetic hybrid algorithm includes cargo timeliness weight, cargo value weight, characteristic adaptation weight and spatial adaptation weight. The crossover probability of the genetic algorithm is set to 0.6-0.8 and the mutation probability is set to 0.01-0.03.

[0011] Specifically, in step S3, the improved Dijkstra algorithm corrects the path weights by introducing a risk cost coefficient, which includes congestion risk, policy change risk, and weather risk. The rolling window iteration interval is set to 1-4 hours.

[0012] Specifically, in step S4, the cross-modal information interaction correction module adopts a combination of IoT data and manual verification, with preset thresholds including space utilization rate ≥85%, cost deviation ≤±5%, and timeliness compliance rate ≥95%.

[0013] Specifically, in step S2, cargo characteristic adaptation includes the separate loading of fragile and heavy items, the matching of perishable items with refrigerated compartments, and the separate loading of dangerous goods in separate zones.

[0014] Specifically, in step S3, the dynamic data update content includes changes in port berthing time, road construction information, customs clearance efficiency, and extreme weather warnings.

[0015] Specifically, in step S5, the feedback closed-loop data includes actual transportation costs, time consumption, cargo damage rate, and number of route adjustments, which are used as weight parameters for iterative optimization algorithms.

[0016] Specifically, the modes of transportation include sea transport, air transport, land transport, and multimodal transport, and in the multimodal transport scenario, the optimal connection scheme of the transit nodes is automatically matched.

[0017] Specifically, in step S2, the initial loading scheme uses three-dimensional cabin modeling technology to simulate the stacking state of cargo and avoid the risk of center of gravity shift.

[0018] The beneficial effects of this invention are: 1. Achieve coordinated optimization of load allocation and route planning: Break through the limitations of traditional independent load allocation and route planning, and through data linkage and algorithm integration, make the load allocation scheme adapt to the route characteristics and the route planning fit the load allocation requirements, avoid secondary adjustments, significantly improve operational efficiency and reduce additional costs.

[0019] 2. Strong dynamic adaptability: Through the rolling window iteration mechanism and real-time updates of multi-source data, it can quickly respond to emergencies in cross-border transportation (congestion, weather, policy changes), dynamically adjust loading and routing schemes, effectively avoid risks, ensure transportation timeliness, and improve the stability and reliability of the scheme.

[0020] 3. Multi-constraint collaborative consideration: Taking into account multiple constraints such as cost, timeliness, space utilization, cargo characteristics, and risk control, the algorithm adopts a hybrid analytic hierarchy process (AHP)-genetic algorithm and an improved Dijkstra algorithm to achieve multi-objective optimization, thereby improving resource utilization, reducing cargo damage risk, and meeting the transportation needs of different types of goods.

[0021] 4. Full utilization of data: Integrate multi-source heterogeneous data, construct a comprehensive database through scientific preprocessing, and combine cross-modal information interaction correction and feedback loop to realize data-driven decision-making and algorithm iterative optimization, continuously improving the adaptability and optimization accuracy of the solution.

[0022] 5. Wide range of applicable scenarios: It can be adapted to various international freight scenarios such as sea freight, air freight, land freight and multimodal transport. It has good adaptability for complex needs such as special cargo and transshipment, and is highly versatile, which can meet the operational needs of freight forwarding companies of different sizes. Attached Figure Description

[0023] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] Figure 1 The flowchart of the international freight forwarding cargo loading route optimization method provided by the present invention. Detailed Implementation

[0025] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0026] like Figure 1 As shown, the international freight forwarding cargo loading route optimization method of the present invention includes the following steps: S1: Data Acquisition and Preprocessing. This involves collecting multi-source heterogeneous data, including cargo information (weight, volume, characteristics, timeliness requirements, fragile / perishable attributes, destination, cargo value), transport vehicle information (cabin capacity, load limits, transport mode, maintenance status, operating costs), cross-border road condition information (road congestion, flight route weather, port berthing status, transit node connectivity), policy and regulatory information (customs clearance rules, import / export control policies, tax standards), and environmental information (extreme weather warnings, regional security risks). Data acquisition is achieved through IoT sensors (real-time monitoring of cargo status and transport vehicle maintenance parameters), cross-border transport platform API interfaces (obtaining road conditions and flight / voyage dynamics), customs supervision systems (synchronizing policies and regulations, clearance efficiency), and third-party meteorological platforms (obtaining weather warnings). The collected data is preprocessed as follows: missing values ​​are filled in using Lagrange interpolation, outliers are removed using the standard deviation method (with a threshold of 3 times the standard deviation), and data of different formats are standardized to unify data units and formats. A structured comprehensive database is constructed to provide comprehensive and accurate data support for subsequent load allocation and route planning.

[0027] S2: Intelligent Loading Scheme Generation. Based on a comprehensive database, a hybrid analytic hierarchy process (AHP)-genetic algorithm is used to construct a loading optimization model. The core objectives are to minimize transportation costs, maximize cargo space utilization, and optimize transportation timeliness, while also considering constraints such as cargo characteristic adaptability and minimizing cargo damage risk. First, the AHP is used to establish cargo loading priorities, constructing an objective layer (loading optimization), a criterion layer (timeliness requirement, cargo value, characteristic adaptability, and spatial adaptability), and a scheme layer (each batch of cargo). The weights of each criterion are calculated, with timeliness requirement weighted at 0.3-0.4, cargo value weighted at 0.2-0.3, characteristic adaptability weighted at 0.2-0.3, and spatial adaptability weighted at 0.1-0.2. Then, a genetic algorithm is used to solve the loading problem, encoding the cargo loading scheme as chromosomes. The objective function value is used as the fitness, with crossover probabilities of 0.6-0.8 and mutation probabilities of 0.01-0.03. Through selection, crossover, and mutation iterations, the optimal loading combination is generated. Meanwhile, three-dimensional cargo space modeling technology is used to simulate the stacking state of cargo to ensure the balance of the cargo center of gravity and avoid the risk of center of gravity shift; targeted loading is carried out for special cargoes, fragile items are placed separately from heavy items, perishable items are matched with refrigerated compartments, and dangerous goods are loaded in separate zones to generate an initial loading plan.

[0028] S3: Dynamic Route Planning. Based on the initial loading plan and combined with route-related data in the comprehensive database, an improved Dijkstra algorithm is used to generate the initial route. The improved Dijkstra algorithm adjusts the route weight by introducing a risk cost coefficient. The route weight = distance cost + time cost + risk cost, where the risk cost coefficient includes congestion risk (calculated based on real-time congestion duration), policy change risk (based on policy stability score), and weather risk (based on weather warning level), achieving a synergistic consideration of cost and risk. Subsequently, a rolling window iteration mechanism is initiated, with an iteration interval of 1-4 hours. At each preset interval, dynamic data (port berthing time changes, road construction information, customs clearance efficiency, extreme weather warnings, etc.) is updated through the data acquisition module. Based on the updated data, the route is iteratively adjusted to avoid congested nodes and risky areas, optimize the connection scheme of transfer nodes, and ensure that the route is always in an optimal state. For multimodal transport scenarios, the optimal connecting transport vehicle for transfer nodes is automatically matched to shorten transfer waiting time.

[0029] S4: Scheme Verification and Optimization. The initial loading route scheme is verified through a cross-modal information interaction correction module. This module combines real-time IoT data with manual verification. Verification indicators include: cargo space utilization rate ≥85%, transportation cost deviation ≤±5%, timeliness compliance rate ≥95%, cargo characteristic adaptation compliance rate 100%, and route risk level ≤medium. If the scheme does not meet any preset threshold, it returns to the corresponding step for re-optimization based on the deviation type: if the loading indicator does not meet the standard, it returns to S2 to adjust the loading priority and algorithm parameters; if the route indicator does not meet the standard, it returns to S3 to optimize the route iteration strategy. If there are deviations in multiple indicators, the parameters of S2 and S3 are adjusted simultaneously until the optimal loading route combination that meets all threshold requirements is generated.

[0030] S5: Solution Execution and Feedback. The optimal loading route combination is output to the freight forwarding operation system to guide on-site loading, transportation execution, and transshipment coordination. During transportation execution, real-time operational data is collected, including actual transportation costs, time consumption, cargo damage rate, number of route adjustments, transshipment waiting time, and customer satisfaction, forming a feedback loop. The feedback data is input into the algorithm model to iteratively optimize the weight parameters of the hierarchical analysis, the crossover and mutation probability of the genetic algorithm, and the improved risk coefficient calculation method of the Dijkstra algorithm, thereby enhancing the adaptability and optimization effect of subsequent solutions.

[0031] Example 1: Multimodal Transport Scenario (Cross-border E-commerce Goods). This example applies to the transportation of cross-border e-commerce goods from Shenzhen, China to Hamburg, Germany. The transportation methods are sea transport (Shenzhen Yantian Port to Hamburg Port, Germany) + land transport (Hamburg Port to various destinations within Germany). The total number of goods is 50 batches, including three categories: ordinary parcels, fragile electronic components, and perishable food (requiring refrigeration). Among them, there are 30 batches of ordinary parcels (total weight 25 tons, total volume 80 cubic meters, delivery time requirement 15-20 days), 10 batches of fragile electronic components (total weight 5 tons, total volume 12 cubic meters, delivery time requirement 12-16 days, high value), and 10 batches of perishable food (total weight 8 tons, total volume 18 cubic meters, delivery time requirement 8-12 days, requiring refrigeration throughout). The objectives are to achieve a space utilization rate of ≥85%, cost deviation ≤±5%, delivery time compliance rate of ≥95%, and cargo damage rate of ≤0.5%.

[0032] The specific steps for optimizing the loading path using the method of this invention are as follows: S1: Data Acquisition and Preprocessing. Real-time cargo status (temperature of perishable foods, vibration values ​​of fragile electronic components) and transportation information are collected via IoT sensors (5 shipping containers, 40 cubic meters each, 20 tons each; 3 refrigerated trucks, 10 tons each, refrigeration temperature range -10℃ to 0℃; 5 ordinary trucks, 8 tons each). Port berthing status is obtained via API interfaces of Yantian Port and Hamburg Port (Yantian Port currently has low congestion levels, estimated berthing time error ±2 hours; Hamburg Port's transshipment node connection efficiency is 85%) and shipping route weather (no extreme weather in the North Pacific Ocean in the next 7 days, wind speed 3-5). Policies are synchronized through the customs supervision system (customs clearance time for Sino-German cross-border e-commerce goods is 48 hours, no new control clauses). Weather for land transport routes within Germany is obtained through a third-party meteorological platform (mainly sunny in the next 10 days, with light rain in some areas, no extreme weather warnings). The collected data were preprocessed as follows: Lagrange interpolation was used to complete the missing volume values ​​of two batches of goods, and one abnormal port congestion data (deviation exceeding 3 times the standard deviation) was removed using the standard deviation method. The weight unit was unified to tons, the volume to cubic meters, and the time to hours, and a comprehensive database was constructed.

[0033] S2: Intelligent Load Allocation Scheme Generation. A hybrid analytic hierarchy process (AHP)-genetic algorithm is employed, with the following weights set for the criteria layer: timeliness requirement weight 0.35, cargo value weight 0.25, characteristic adaptation weight 0.25, and spatial adaptation weight 0.15. First, the AHP is used to calculate the priority of each batch of goods: perishable food has the highest timeliness requirement, with a priority score of 0.85; fragile electronic components have high value and relatively high timeliness requirement, with a priority score of 0.78; and ordinary parcels have a priority score of 0.62. Then, a genetic algorithm is used to solve the load allocation problem, with a crossover probability of 0.7 and a mutation probability of 0.02. Chromosomes encode the cargo-transport vehicle correspondence, and the fitness function is (space utilization rate × 0.3 + cost saving rate × 0.4 + timeliness compliance rate × 0.3). Through 100 iterations, a loading plan was generated: Of the five ocean shipping containers, one carries perishable food (8 tons, 18 cubic meters, equipped with a temperature control device set to -5°C), two carry fragile electronic components (5 tons, 12 cubic meters, cushioned and placed in the middle layer of the container to avoid compression), and two carry general parcels (25 tons, 80 cubic meters, stacked appropriately to ensure a balanced center of gravity). During land transport, one refrigerated truck connects to the ocean shipping containers to transport perishable food, two general trucks transport fragile electronic components, and three general trucks transport general parcels, corresponding to different destinations within Germany. Through 3D cargo hold modeling simulation, the container space utilization rate is 88%, the truck space utilization rate is 86%, and the center of gravity offset is ≤3%, meeting the basic loading requirements.

[0034] S3: Dynamic Route Planning. Based on the initial loading plan, an improved Dijkstra algorithm is used to generate the initial routes: the sea route is Shenzhen Yantian Port → South China Sea → Strait of Malacca → Indian Ocean → Suez Canal → Mediterranean Sea → Strait of Gibraltar → North Sea → Hamburg Port, Germany, with an estimated sea travel time of 10 days; the land route is planned with 3 branch lines according to the destination of each batch of goods, namely Hamburg Port → Berlin (fragile electronic components, estimated travel time 3 hours), Hamburg Port → Munich (perishable food, estimated travel time 5 hours), and Hamburg Port → Frankfurt (general parcels, estimated travel time 4 hours). In the route weight calculation of the improved Dijkstra algorithm, the risk cost coefficients are set as follows: sea transport congestion risk coefficient 0.1 (current port congestion is low), weather risk coefficient 0.05 (no extreme weather), policy risk coefficient 0.05 (stable policy); land transport congestion risk coefficient 0.15 (local light rain may cause slight congestion), weather risk coefficient 0.1 (light rain has no impact). Set the rolling window iteration interval to 2 hours and update data in real time: On the 3rd day of the sea voyage, obtain information on temporary traffic control in the Suez Canal (a 20% decrease in traffic efficiency), iteratively adjust the sea route, detour around the Cape of Good Hope, and increase the estimated time by 1.5 days. At the same time, adjust the departure time of the land transport to avoid transfer waiting. On the 1st hour of the land transport phase, obtain information on construction on the Hamburg to Berlin section, adjust the land transport route, detour via alternative roads, and increase the estimated time by 0.5 hours to ensure that the timeliness standard is met.

[0035] S4: Solution Verification and Optimization. Verification via the cross-modal information interaction correction module showed: IoT data indicated container space utilization of 88% and truck space utilization of 86%, both ≥85%; cost accounting deviation of 3.2%, ≤±5%; timeliness compliance rate of 96%, ≥95%; cargo characteristic adaptation compliance rate of 100% (perishable foods were temperature-controlled normally, fragile items were placed separately); route risk level was moderate (no high-risk nodes after detours); all indicators met preset thresholds, requiring no secondary optimization, and the optimal loading route solution was generated.

[0036] S5: Implementation and Feedback. Transportation was executed according to the optimal plan: For the sea transport phase, after detouring around the Cape of Good Hope, the actual travel time was 11.4 days, a deviation of 0.1 days from the adjusted estimated time. For the land transport phase, after detouring, the actual travel times were 3.4 hours for Berlin, 4.8 hours for Munich, and 4.1 hours for Frankfurt, all meeting the timeliness requirements. The overall cargo damage rate was 0.3%, lower than the target value of 0.5%; the actual transportation cost deviated from the budget by 2.8%, within the threshold range. Feedback data was collected (timeliness deviation, cost deviation, cargo damage rate, and number of route adjustments twice). The weights of the hierarchical analysis were iteratively optimized (increasing the land transport risk weight to 0.2) and the crossover probability of the genetic algorithm was adjusted to 0.72, providing a basis for optimization of subsequent similar transportation operations.

[0037] Example 2: Air Freight Expedited Cargo Scenario (Medical Supplies). This example applies to the expedited transport of medical supplies from Beijing, China to Lagos, Nigeria, Africa. The mode of transport is air freight (Beijing Capital International Airport to Lagos Murtala Mohammed Airport). The cargo consists of COVID-19 vaccines and related medical equipment, totaling 30 batches. Among them, there are 20 batches of vaccines (total weight 12 tons, total volume 15 cubic meters, requiring a complete cold chain, temperature control at 2-8℃, delivery within 48 hours, extremely high value) and 10 batches of medical equipment (total weight 8 tons, total volume 10 cubic meters, fragile, delivery within 72 hours). The objectives are: 100% timeliness, damage rate ≤0.1%, cost deviation ≤±8%, and space utilization ≥80%.

[0038] The specific steps for optimizing the loading path using the method of this invention are as follows: S1: Data Acquisition and Preprocessing. Cargo status (real-time vaccine temperature, medical equipment vibration values) and transportation information are collected via IoT sensors (two Boeing 777 freighters are used, each with a cargo hold capacity of 45 cubic meters, a payload of 50 tons, equipped with cold chain units, and a temperature control accuracy of ±0.5℃). Flight status is obtained through API interfaces of Beijing Capital International Airport and Lagos International Airport (current flight punctuality rate at Beijing Capital International Airport is 92%, customs clearance efficiency at Lagos International Airport is 72 hours, and there is no need for connecting transfers), and flight route weather (no extreme weather in the North Indian Ocean and over Africa in the next 48 hours, with localized cloud cover and no thunderstorm warnings). Policies are synchronized through the customs supervision system (green channel for cross-border transportation of medical supplies, customs clearance time can be shortened to 48 hours, no additional approval required). Flight route safety information is obtained through a third-party security platform (no regional security risks, flight routes are open normally). Data preprocessing: Vaccine temperature data is complete, but there is one outlier in the medical equipment weight data (deviation of 3.2 times the standard deviation), which is removed using the standard deviation method. Average weight data for the same batch of equipment is supplemented. The time unit is standardized to hours and the temperature to °C, and a comprehensive database is constructed.

[0039] S2: Intelligent Loading Scheme Generation. A hybrid analytic hierarchy process (AHP)-genetic algorithm is employed, with the following weights set for the criteria layer: timeliness requirement weight 0.4 (prioritizing expedited scenarios), cargo value weight 0.3 (vegetables have extremely high value), characteristic adaptation weight 0.2 (cold chain, fragile requirements), and space adaptation weight 0.1. Priorities are calculated using AHP: vaccines receive a priority score of 0.92, and medical equipment a priority score of 0.75. Genetic algorithm parameters are set as follows: crossover probability 0.75, mutation probability 0.025, fitness function (timeliness compliance rate × 0.5 + cargo damage risk rate × 0.3 + space utilization rate × 0.2), and chromosome encoding represents the correspondence between cargo, cargo plane, and cargo hold locations. After 80 iterations, a loading plan was generated: two cargo planes will each carry 10 batches of vaccines (6 tons, 7.5 cubic meters) and 5 batches of medical equipment (4 tons, 5 cubic meters). The vaccines will be placed in the cold chain unit in the middle of the cargo plane, with a temperature set at 5°C, and surrounded by insulating material to isolate them from the medical equipment. The medical equipment will be placed in the front of the cargo plane, secured with shock-absorbing pallets, and stacked no more than two layers high to avoid compression. 3D cargo hold modeling shows that the space utilization rates of the two cargo planes are 83% and 82% respectively, both ≥80%; the center of gravity offset is ≤2%, and the temperature stability of the cold chain unit is ±0.3°C, meeting the loading requirements.

[0040] S3: Dynamic Route Planning. Based on the initial load planning scheme, an improved Dijkstra algorithm is used to generate the initial route: Beijing Capital International Airport → Urumqi → Islamabad, Pakistan → Nairobi, Kenya → Murtala Muhammad Airport, Lagos, Nigeria, with an estimated flight time of 22 hours, including 1-hour stopovers in Islamabad and Nairobi (for refueling and crew rotation). In the route weight calculation, risk cost coefficients are set as follows: weather risk coefficient 0.08 (no impact from localized cloud cover), policy risk coefficient 0.02 (stable green channel policy), and flight delay risk coefficient 0.1 (high on-time rate). A rolling window iteration interval of 1 hour is used to update data in real time: when the flight reaches Islamabad for the stopover, temporary flight control information from Nairobi Airport is obtained (extending the stopover time to 1.5 hours), and the flight speed is iteratively adjusted (increasing by 5% within a safe range) to shorten the duration of subsequent segments, ensuring the total flight time does not exceed 24 hours; 2 hours before arriving at Lagos Airport, information on improved customs clearance green channel efficiency is obtained (customs clearance time can be shortened to 36 hours), and the departure time of ground connecting vehicles is adjusted to avoid waiting.

[0041] S4: Solution Verification and Optimization. Verification of the cross-modal information interaction correction module: IoT data shows cargo aircraft space utilization rate of 82.5% (average), ≥80%; cost accounting deviation of 4.5%, ≤±8%; timeliness compliance rate of 100% (total flight time 23.5 hours, customs clearance 35 hours, total time 58.5 hours, meeting the requirements of 48-hour flight for vaccines and 72-hour full-process equipment); cargo damage rate of 0.05%, ≤0.1%; characteristic adaptation compliance rate of 100% (vaccine temperature stable, equipment shockproof compliance); low route risk level, all indicators meet the standards, generating the optimal solution.

[0042] S5: Solution Implementation and Feedback. In actual execution, the vaccine temperature remained stable between 4.8-5.2℃ throughout the process, with no cargo damage; vibration levels of medical equipment were within safe ranges, with no breakage. The total flight time was 23.5 hours, customs clearance 34.5 hours, and the total processing time was 58 hours, achieving a 100% timeliness compliance rate. The actual cost deviated from the budget by 4.2%, meeting the requirements. Feedback data included one adjustment to stopover time, improved customs clearance efficiency, no cargo damage, iterative optimization of the genetic algorithm to a mutation probability of 0.02, and improvements to the flight speed adjustment coefficient of the Dijkstra algorithm, enhancing timeliness adaptability in expedited scenarios.

[0043] Example 3: Multi-port transshipment sea freight scenario (industrial parts). This example applies to the transportation of industrial parts from Shanghai, China to São Paulo, Brazil. The transportation method is multi-port transshipment sea freight (Shanghai Port → Singapore Port → Durban Port, South Africa → Santos Port, Brazil) + land transportation (Santos Port to São Paulo). The goods are automotive parts, totaling 40 batches, including engine parts (15 batches, total weight 30 tons, total volume 25 cubic meters, fragile, high value, time requirement 30-35 days) and body parts (25 batches, total weight 45 tons, total volume 60 cubic meters, general characteristics, time requirement 35-40 days). The objectives are space utilization ≥85%, cost deviation ≤±6%, timeliness compliance rate ≥95%, and transshipment efficiency ≥90%.

[0044] The specific steps for optimizing the loading path using the method of this invention are as follows: S1: Data Acquisition and Preprocessing. Cargo status (engine component vibration values, appearance integrity) and transportation information are collected via IoT sensors (one bulk carrier with a cargo hold capacity of 120 cubic meters and a deadweight of 80 tons; efficiency of transshipment and hoisting equipment at Singapore, Durban, and Santos ports; four land transport trucks, each with a deadweight of 20 tons). Port dynamics are obtained via port API interfaces (Shanghai port loading efficiency 20 cubic meters / hour, Singapore port transshipment waiting time 24 hours, Durban port transshipment waiting time 18 hours, Santos port unloading efficiency 18 cubic meters / hour), and route weather (no typhoons, hurricanes, or other extreme weather in the Indian Ocean and Atlantic Ocean for the next 30 days, wind force 4-6). Customs policies are synchronized through the customs supervision system (customs clearance rules for industrial parts in China, Singapore, South Africa, and Brazil; no regulatory clauses; clearance times are 24 hours, 12 hours, 18 hours, and 24 hours respectively). Ship maintenance status is obtained through a third-party shipping platform (bulk carrier has no malfunctions and sufficient range). Data preprocessing: The volume data of three batches of body parts were completed using the Lagrange interpolation method. No outliers were found. The weight unit was standardized to tons, the volume to cubic meters, and the time to hours. A comprehensive database was then constructed.

[0045] S2: Intelligent loading scheme generation. Hierarchical Analysis-Genetic Hybrid Algorithm Criterion Layer Weights: Timeliness Demand Weight 0.3, Cargo Value Weight 0.25, Characteristic Adaptation Weight 0.25, Space Adaptation Weight 0.2. Hierarchical Analysis Calculation Priority: Engine Component Priority Score 0.82, Body Component Priority Score 0.68. Genetic Algorithm Parameters: Crossover Probability 0.65, Mutation Probability 0.03, Fitness Function: (Space Utilization × 0.3 + Transfer Connection Efficiency × 0.3 + Cost Saving Rate × 0.4), Chromosome Encoding: Cargo-Cargo Hold Location-Transfer Node Correspondence. After 120 iterations, the loading scheme is as follows: Engine components (30 tons, 25 cubic meters) are placed in the front of the bulk carrier's cargo hold, secured with shock-absorbing materials, stacked one layer high, isolated from the body components; body components (45 tons, 60 cubic meters) are placed in the rear of the cargo hold, stacked reasonably to ensure center of gravity balance. 3D cargo hold modeling shows that the cargo hold space utilization rate is 87.5%, ≥85%; the center of gravity offset is ≤2.5%, the engine component anti-vibration measures meet the standards, and the loading requirements are satisfied. Meanwhile, the lifting sequence for each transfer node is planned: engine components are lifted first to shorten transfer waiting time.

[0046] S3: Dynamic Route Planning. An improved Dijkstra algorithm generates the initial route: Shanghai Port → South China Sea → Strait of Malacca → Singapore Port (transit) → Indian Ocean → Durban Port (transit) → Atlantic Ocean → Santos Port (transit) → São Paulo, Brazil. The estimated total time is 32 days, including 28 days by sea, 4 days by transit (Singapore 24 hours, Durban 18 hours, Santos 21 hours), and 2 days by land. Route weight calculation: Congestion risk coefficient 0.12 (moderate congestion at Singapore and Durban ports), weather risk coefficient 0.08 (normal wind speed), and policy risk coefficient 0.05 (stable customs clearance policy). The rolling window iterates every 4 hours, updating data in real time: One day before the ship arrives at Singapore Port, information on increased congestion at Singapore Port is obtained (waiting time extended to 36 hours), and the ship's sailing speed is iteratively adjusted (increased by 3%) to shorten the transit time from Singapore Port to Durban Port and compensate for transit delays; After arriving at Durban Port, information on unloading equipment failure at Santos Port is obtained (unloading efficiency dropped to 12 cubic meters / hour), and the land transport departure time is adjusted, while coordinating backup unloading equipment at Santos Port to ensure transit connection efficiency ≥90%.

[0047] S4: Solution Verification and Optimization. Verification of the cross-modal information interaction correction module: Cargo hold space utilization rate 87.5%, ≥85%; cost deviation 3.8%, ≤±6%; timeliness compliance rate 97% (total actual time 33 days, meeting the 30-35 day requirement for engine parts, and all body parts within the 35-40 day range); transfer efficiency 92%, ≥90%; characteristic adaptation compliance rate 100%; route risk level medium; all indicators meet the standards, generating the optimal solution.

[0048] S5: Solution Execution and Feedback. In actual transport, engine parts were undamaged, and vehicle body parts remained intact. The transit wait time at Singapore Port was 35 hours, Durban Port 18 hours, and Santos Port unloading was completed using backup equipment, restoring efficiency to 18 cubic meters per hour, with a transit efficiency of 92%. The total duration was 33 days, with a cost deviation of 3.6%, meeting the target. Feedback data included one adjustment to transit congestion, one response to equipment failure, iterative optimization of the transit connection weight in the hierarchical analysis to 0.35, and improvements to the speed adjustment threshold in the Dijkstra algorithm, enhancing adaptability to multi-port transit scenarios.

[0049] Compared to the traditional single-method loading plus static route method in Example 1, this method uses existing technology and focuses solely on cargo hold space utilization, neglecting cargo characteristic priority. Route planning employs a static algorithm without dynamic adjustments. Applied to the scenario in Example 1: fragile electronic components were mixed with ordinary parcels during loading, and perishable food was not separately assigned refrigerated storage space; the route used a fixed route (not detouring around the Cape of Good Hope) without dynamic iteration. Results: Cargo hold space utilization was 82%, lower than the 88% in Example 1; cargo damage rate was 2.8%, far exceeding the target of 0.5%; due to Suez Canal regulations, transit time was extended to 25 days, with a timeliness compliance rate of 75%; actual cost deviation was 8.5%, exceeding the ±5% threshold, resulting in poor overall operational performance.

[0050] Compared to the loading and route separation optimization method in Example 2, the loading plan was completed first, and then the route planning was performed separately, without a collaborative optimization mechanism. Applied to the scenario in Example 2: flight stopover nodes were not considered during loading, vaccines and medical equipment were mixed and placed at the edge of the cold chain unit, resulting in large temperature fluctuations; after route planning was completed, it was not dynamically adjusted according to flight control. Results: Space utilization rate was 78%, lower than 82.5% in Example 2; vaccine temperature fluctuation range was 2-10℃, and the cargo damage rate was 0.5%, exceeding the 0.1% target; due to the extended stopover in Nairobi, the total duration was extended to 72 hours, and the vaccine's timeliness did not meet the standard; cost deviation was 9.2%, exceeding the ±8% threshold.

[0051] Compared to Example 3, which lacks cross-modal correction and feedback loop, this invention employs the loading and routing algorithm, but without cross-modal information interaction correction and feedback loop. Applied to the scenario of Example 3: the loading scheme was not validated through 3D modeling, resulting in a 5% center of gravity offset, posing a transportation safety risk; there was no data verification after route adjustment, and the delay in unloading at Santos Port was not addressed promptly. Results: Space utilization was 86%, close to the target, but the center of gravity offset increased ship navigation energy consumption, leading to a cost deviation of 7.1%; transshipment efficiency was 82%, lower than the 90% target; total duration was 38 days, with a timeliness compliance rate of 80%, and no feedback data was provided to optimize subsequent algorithms, preventing any improvement in the scheme's adaptability.

[0052] Compared to Example 4's single-algorithm loading and dynamic path method, this paper uses a single genetic algorithm for loading (without hierarchical analysis prioritization) combined with dynamic path planning. Applied to the scenario of Example 1: loading priorities were chaotic, perishable food was loaded into containers without cold chain, and fragile items were severely crushed; dynamic path adjustment was normal. Results: cargo damage rate 3.2%, timeliness compliance rate 80%, cost deviation 6.8%, space utilization rate 84%. Although dynamic path adjustment was effective, unreasonable loading led to poor overall results, confirming the necessity of the hierarchical analysis-genetic hybrid algorithm of this invention.

[0053] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing cargo loading routes in international freight forwarding, characterized in that, Includes the following steps: S1: Data collection and preprocessing. Collect cargo information, transportation vehicle information, cross-border road condition information, policy and regulation information, cost and timeliness information, and environmental information. Perform noise reduction, standardization, and completion processing on the collected data to build a comprehensive database. Cargo information includes weight, volume, characteristics, timeliness requirements, fragile and perishable attributes, and destination. Transportation vehicle information includes cabin capacity, load limit, transportation mode, and operation and maintenance status. S2: Intelligent loading scheme generation. Based on a comprehensive database, it adopts a hierarchical analysis-genetic hybrid algorithm to establish cargo loading priorities with the goals of lowest cost, highest space utilization, and optimal timeliness. Combined with the cargo hold parameters of the transport vehicle, it completes the precise loading of cargo and transport vehicle and generates an initial loading scheme. S3: Dynamic route planning, based on the initial load planning scheme, adopts an improved Dijkstra algorithm, and dynamically generates the initial route by combining real-time traffic conditions, port congestion, and flight / voyage information; By using a rolling window iteration mechanism to update dynamic data at preset time intervals, the path is iteratively adjusted to avoid risky nodes. S4: Scheme verification and optimization. Through the cross-modal information interaction correction module, the rationality of load allocation, path feasibility and cost and timeliness are compared. If the preset threshold is not met, return to S2 or S3 to re-optimize until the optimal load allocation path combination is generated. S5: Solution Execution and Feedback. Outputs the optimal solution and guides its execution. Collects actual data during the execution process to form a feedback loop for optimizing algorithm parameters.

2. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S1, multi-source data acquisition is achieved through IoT sensors, cross-border transportation platform API interfaces, customs supervision systems, and third-party meteorological platforms. Data preprocessing uses Lagrange interpolation to fill in missing values ​​and standard deviation to remove outliers.

3. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S2, the priority evaluation index of the analytic hierarchy process-genetic hybrid algorithm includes cargo timeliness weight, cargo value weight, characteristic adaptation weight and spatial adaptation weight. The crossover probability of the genetic algorithm is set to 0.6-0.8 and the mutation probability is set to 0.01-0.

03.

4. The international freight forwarding cargo loading route optimization method according to claim 1, characterized in that: In step S3, the improved Dijkstra algorithm corrects the path weights by introducing a risk cost coefficient, which includes congestion risk, policy change risk, and weather risk. The rolling window iteration interval is set to 1-4 hours.

5. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S4, the cross-modal information interaction correction module adopts a combination of IoT data and manual verification, with preset thresholds including space utilization rate ≥85%, cost deviation ≤±5%, and timeliness compliance rate ≥95%.

6. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S2, cargo characteristic adaptation includes the separate loading of fragile and heavy items, the matching of perishable items with refrigerated compartments, and the separate loading of dangerous goods in separate zones.

7. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S3, the dynamic data updates include changes in port berthing times, road construction information, customs clearance efficiency, and extreme weather warnings.

8. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S5, the feedback closed-loop data includes actual transportation costs, time consumption, cargo damage rate, and number of route adjustments, which are used as weight parameters for iterative optimization algorithms.

9. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: The transportation modes include sea freight, air freight, land freight, and multimodal transport. In multimodal transport scenarios, the optimal connection scheme for transit nodes is automatically matched.

10. The method for optimizing cargo loading routes in international freight forwarding according to claim 1, characterized in that: In step S2, the initial loading scheme uses three-dimensional cargo space modeling technology to simulate the cargo stacking state and avoid the risk of center of gravity shift.

Images