Logistics optimization method and system based on AI collaboration

Abstract

The invention discloses a logistics optimization method and system based on AI collaboration, and relates to the technical field of logistics dispatch, and the method comprises the steps: constructing a hierarchical collaboration architecture, and obtaining multi-source heterogeneous logistics data through data layer integration; in the model layer, training a demand prediction model, performing logistics demand prediction in combination with the multi-source heterogeneous logistics data, generating a logistics demand prediction result, transmitting the logistics demand prediction result to the decision-making layer, performing logistics scheduling optimization by the decision-making layer, generating a global optimization strategy, and transmitting the global optimization strategy to the application layer; and the application layer outputs an intelligent scheduling instruction for logistics management. The technical problems that in an existing logistics system, due to the fact that multi-source heterogeneous data is difficult to integrate effectively and privacy protection is insufficient, logistics demand prediction is inaccurate, scheduling efficiency is low and supply chain cooperation capacity is weak are solved, and the purposes that privacy protection and efficient integration of the multi-source data are achieved through a layered cooperation architecture, and the logistics demand prediction efficiency is improved are achieved. And the technical effects of logistics demand prediction precision, scheduling efficiency and supply chain cooperation capability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics scheduling, and particularly relates to a logistics optimization method and system based on AI collaboration. Background Art

[0002] With the rapid development of the logistics industry, intelligent and digital transformation has become the key direction to improve logistics efficiency and reduce costs. However, there are still many deficiencies in the existing logistics systems in aspects such as data integration, demand forecasting, scheduling optimization, and supply chain collaboration. First of all, traditional logistics systems rely on manual scheduling, which is inefficient and error-prone, and it is difficult to cope with complex and changing market demands. Secondly, logistics data is scattered among different entities, and data sharing has privacy risks and trust barriers, resulting in the difficulty of effectively integrating multi-source heterogeneous data. In addition, the existing scheduling systems lack the ability of real-time dynamic adjustment and are difficult to handle emergencies and market fluctuations. These problems limit the overall efficiency and collaboration ability of the logistics system and increase the operating costs. Summary of the Invention

[0003] This application provides a logistics optimization method and system based on AI collaboration, which is used to solve the technical problems in the existing logistics system that multi-source heterogeneous data is difficult to effectively integrate and privacy protection is insufficient, resulting in inaccurate logistics demand forecasting, low scheduling efficiency, and weak supply chain collaboration ability.

[0004] In the first aspect of this application, a logistics optimization method based on AI collaboration is provided. The method includes: constructing a hierarchical collaboration architecture, which includes a data layer, a model layer, a decision layer, and an application layer; integrating and obtaining multi-source heterogeneous logistics data through the data layer; in the model layer, using federated learning to collaboratively train a demand forecasting model, and combining the multi-source heterogeneous logistics data to conduct logistics demand forecasting to generate a logistics demand forecasting result; transmitting the logistics demand forecasting result to the decision layer, where the decision layer conducts logistics scheduling optimization to generate a global optimization strategy and transmit it to the application layer; outputting intelligent scheduling instructions through the application layer to the logistics management system for logistics management.

[0005] In the second aspect of the present application, a logistics optimization system based on AI collaboration is provided. The system includes: a hierarchical collaboration architecture construction module for constructing a hierarchical collaboration architecture, which includes a data layer, a model layer, a decision layer, and an application layer; a logistics data acquisition module for integrating and acquiring multi-source heterogeneous logistics data through the data layer; a logistics demand prediction module for, in the model layer, using federated learning to collaboratively train a demand prediction model and performing logistics demand prediction in combination with the multi-source heterogeneous logistics data to generate a logistics demand prediction result; a logistics scheduling optimization module for transmitting the logistics demand prediction result to the decision layer, where the decision layer performs logistics scheduling optimization to generate a global optimization strategy and transmits it to the application layer; and a logistics scheduling management module for outputting intelligent scheduling instructions to the logistics management system through the application layer for logistics management.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0007] A logistics optimization method and system based on AI collaboration provided in the present application relate to the technical field of logistics scheduling. By constructing a hierarchical collaboration architecture including a data layer, a model layer, a decision layer, and an application layer, integrating multi-source heterogeneous logistics data through the data layer, using federated learning in the model layer to collaboratively train a demand prediction model and generate a prediction result, the decision layer performing logistics scheduling optimization based on the prediction result and generating a global optimization strategy, and finally the application layer outputting intelligent scheduling instructions to the logistics management system, it solves the technical problems in the existing logistics system that multi-source heterogeneous data is difficult to effectively integrate and privacy protection is insufficient, resulting in inaccurate logistics demand prediction, low scheduling efficiency, and weak supply chain collaboration ability. It realizes the technical effect of achieving privacy protection and efficient integration of multi-source data through a hierarchical collaboration architecture, improving the accuracy of logistics demand prediction, scheduling efficiency, and supply chain collaboration ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0009] Figure 1 It is a schematic flowchart of a logistics optimization method based on AI collaboration provided in an embodiment of the present application;

[0010] Figure 2 It is a schematic structural diagram of a logistics optimization system based on AI collaboration provided in an embodiment of the present application.

[0011] Description of the reference numerals: The hierarchical collaborative architecture construction module 11, the logistics data acquisition module 12, the logistics demand prediction module 13, the logistics scheduling optimization module 14, and the logistics scheduling management module 15. Detailed implementation manners

[0012] This application provides a logistics optimization method and system based on AI collaboration, which is used to solve the technical problems in the existing logistics system, where multi-source heterogeneous data is difficult to effectively integrate and privacy protection is insufficient, resulting in inaccurate logistics demand prediction, low scheduling efficiency, and weak supply chain collaboration ability.

[0013] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.

[0014] It should be noted that the terms "first", "second", etc. in the specification of this application and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances, so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0015] Embodiment 1, as Figure 1 shown, this application provides a logistics optimization method based on AI collaboration, and the method includes:

[0016] P10: Construct a hierarchical collaborative architecture, and the hierarchical collaborative architecture includes a data layer, a model layer, a decision layer, and an application layer.

[0017] Specifically, constructing a hierarchical collaborative architecture is the core foundation for implementing the logistics optimization method. The hierarchical collaborative architecture is an architecture model that divides the system into multiple layers, each layer is responsible for specific functions, and the layers interact through interfaces. This architecture model helps to reduce the complexity of the system, improve the maintainability and scalability of the system.

[0018] Specifically, the hierarchical collaborative architecture includes a data layer, a model layer, a decision layer, and an application layer. As the basic layer of the architecture, the data layer is responsible for integrating multi-source heterogeneous logistics data, including but not limited to transportation data, warehousing data, order data, and environmental data, etc. It ensures the integrity and availability of the data through data cleaning, standardization, and storage technologies (such as distributed databases or data lakes).

[0019] The model layer is located above the data layer. Its core function is to collaboratively train a logistics demand prediction model based on Federated Learning technology. This technology can achieve joint modeling of multi-party data while protecting data privacy, thereby improving the accuracy and generalization ability of the prediction model. The decision layer further receives the logistics demand prediction results output by the model layer and uses optimization algorithms (such as linear programming, genetic algorithms, or reinforcement learning) to optimize the global logistics scheduling and generate optimal strategies.

[0020] Finally, the application layer converts the optimization strategies of the decision layer into specific intelligent scheduling instructions and outputs them to the logistics management system through interface technologies (such as APIs or message queues) to achieve real-time scheduling and management of logistics resources. Through the design of the above hierarchical architecture, this method not only realizes the effective integration and utilization of logistics data but also solves the contradiction between data privacy and collaborative optimization through technical means such as federated learning, providing an intelligent and global solution for logistics management.

[0021] P20: Obtain multi-source heterogeneous logistics data through integration by the data layer.

[0022] Furthermore, step P20 of the embodiment of the present application further includes:

[0023] P21: Based on the internal logistics system, obtain order data, inventory data, transportation data, and warehousing data, and obtain weather data, traffic condition data, and market demand data based on external data sources; jointly form the original multi-source logistics data; P22: Preprocess the original multi-source logistics data, including data cleaning, data conversion, and data fusion, to obtain multi-source heterogeneous logistics data.

[0024] It should be understood that the core task of the data layer is to integrate and obtain multi-source heterogeneous logistics data, providing a data basis for subsequent logistics demand prediction and scheduling optimization.

[0025] First, the data layer obtains multi-source logistics data through the internal logistics system and external data sources. Specifically, order data, inventory data, transportation data, and warehousing data are collected from the enterprise's internal logistics system. These data reflect the internal operating status of the logistics system and are the basic information for logistics optimization. For example, order data records customers' purchase behaviors and demands, inventory data reflects the storage situation of goods, transportation data covers the real-time status of logistics distribution, and warehousing data involves the operational efficiency of warehouses. The collection of these data provides a direct basis for the refined management of the logistics system. At the same time, weather data, traffic condition data, and market demand data are obtained from external data sources. Although these external data come from different data sources, they are closely related to the operation of the logistics system. For example, weather data affects the timeliness of transportation and the management cost of warehousing, traffic condition data is directly related to the selection of transportation routes and distribution efficiency, and market demand data provides a macro-level reference for logistics demand forecasting. By combining internal and external data, a comprehensive original multi-source logistics data set is constructed, providing rich data support for subsequent processing and analysis.

[0026] Next, preprocess the collected original multi-source logistics data. This process includes three steps: data cleaning, data transformation, and data fusion. Data cleaning is the first step of preprocessing, aiming to remove noise, errors, and duplicate records in the data. Since the sources of logistics data are extensive and complex, the data may contain missing values, inconsistent formats, or incorrect information. Through data cleaning techniques, such as rule-based verification, interpolation methods, or machine learning algorithms, these problems can be effectively corrected to ensure the integrity and accuracy of the data. Data transformation is the process of converting data from one format or structure to another. Since the types of multi-source data collected are diverse, such as text, numerical values, time series, etc., data transformation can unify these data into a format suitable for subsequent processing. For example, convert timestamp data into a unified time format, and convert unstructured data (such as text descriptions) into structured data (such as keyword extraction) through natural language processing techniques. In addition, data transformation also includes data standardization and normalization processes to eliminate the dimensional differences between different data sources and improve the consistency of the data. Data fusion is to integrate data from different sources to form a unified data view. In the logistics scenario, there may be correlations between multi-source data, such as order data and transportation data, weather data and warehousing data, etc. Through data fusion techniques, such as association rule mining or model-based fusion methods, the potential relationships between these data can be mined to provide more comprehensive information for subsequent model training and decision-making. For example, fusing order data with weather data can analyze the impact of weather conditions on order delivery; fusing transportation data with traffic condition data can optimize transportation route planning. After data cleaning, transformation, and fusion processing, the data layer finally generates high-quality multi-source heterogeneous logistics data. These data not only cover all aspects of the logistics system but also improve the usability and consistency of the data through preprocessing, providing a solid foundation for subsequent model training and decision-making.

[0027] P30: In the model layer, use federated learning to co-train a demand prediction model and combine the multi-source heterogeneous logistics data to conduct logistics demand prediction and generate a logistics demand prediction result.

[0028] Furthermore, step P30 of the embodiment of the present application further includes:

[0029] P31: Obtain the logistics data of multiple participants and store it distributively. Each participant only processes the data locally; P32: Build a federated learning framework. Under the federated learning framework, each participant conducts model co-training, shares the update of model parameters, and trains to generate a demand prediction model.

[0030] Optionally, the model layer uses federated learning technology to co-train a logistics demand prediction model and combines multi-source heterogeneous logistics data to generate a prediction result.

[0031] First, the model layer obtains the logistics data of multiple participants and stores it distributively. Each participant only processes the data locally without sharing the original data with other nodes. This distributive storage method not only protects data privacy but also reduces data transmission costs. The core of distributive storage lies in dispersing the data across multiple computing nodes, with each node responsible for processing local data, thereby achieving efficient data management and utilization.

[0032] Next, the model layer constructs a federated learning framework, under which the participants conduct collaborative model training. Federated learning is a distributed machine learning technique that allows participants to independently train models on local data and collaboratively optimize the global model by sharing model parameter updates. Specifically, the basic process of federated learning includes: the central server initializes the model and distributes it to each participant; the participants independently train on local data and calculate model updates; the updated model parameters are transmitted to the central server via encrypted communication for aggregation; the aggregated global model is then distributed back to each participant to enter the next round of training. In this way, federated learning can make full use of the characteristics of multi-source heterogeneous data and improve the prediction accuracy of the model while protecting data privacy.

[0033] In the scenario of logistics demand prediction, the construction and application of the federated learning framework are of great significance. Logistics demand prediction involves multi-source heterogeneous data, such as order data, inventory data, transportation data, weather data, etc. These data sources are extensive and the structures are complex, making it difficult for traditional centralized model training methods to effectively utilize them. Federated learning, through distributed training and parameter sharing, can give full play to the advantages of multi-source data while avoiding data privacy leakage. In addition, federated learning also supports a variety of deep learning models and can adapt to complex logistics demand prediction tasks. This process not only improves the generalization ability and prediction accuracy of the model but also provides reliable data support for the global optimization of the logistics system.

[0034] Furthermore, step P30 of the embodiment of the present application further includes:

[0035] P34: Configure the demand prediction cycle; P35: Input the multi-source heterogeneous logistics data into the demand prediction model, and according to the demand prediction cycle, predict the future logistics demand to generate a logistics demand prediction result, where the logistics demand prediction result includes the prediction time range, predicted demand quantity, and prediction confidence level.

[0036] In a possible embodiment of the present application, the configuration of the demand prediction cycle and the logistics demand prediction based on this cycle are further introduced to ensure the systematicness and practicality of demand prediction and provide more accurate and operable prediction results for the optimization of the logistics system.

[0037] First, the configuration of the demand forecasting period is one of the key links in realizing logistics demand forecasting. The demand forecasting period refers to the time range within which the model forecasts logistics demand, and it can be in time units such as hours, days, weeks, or months. The purpose of configuring the demand forecasting period is to determine the most suitable forecasting time span according to the actual needs and application scenarios of the logistics business. For example, in e-commerce logistics, short-term forecasting may be required on an hourly or daily basis to cope with rapidly changing order demands; while in supply chain logistics, medium- and long-term forecasting on a weekly or monthly basis may be more appropriate to optimize inventory management and resource allocation.

[0038] When configuring the demand forecasting period, multiple factors need to be considered comprehensively, including the urgency of business needs, the availability and timeliness of data, as well as the complexity and computational cost of the forecasting model. For example, short-term forecasting usually requires more frequent data updates and faster model responses, while medium- and long-term forecasting can utilize a wider range of historical data for analysis, but may require a more complex model structure to capture long-term trends.

[0039] Next, the preprocessed multi-source heterogeneous logistics data is input into the demand forecasting model, and based on the configured demand forecasting period, the future logistics demand is predicted to generate logistics demand forecasting results, including the forecasting time range and forecasting demand volume, and also including the forecasting confidence level to evaluate the reliability of the forecasting results.

[0040] Among them, the forecasting time range is determined according to the demand forecasting period, clarifying the specific time interval covered by the forecasting results. For example, if the demand forecasting period is configured as "weekly", the forecasting time range may be each day or each hour within the next week. The forecasting demand volume is the specific value of the logistics demand output by the model, reflecting the scale of logistics activities expected to occur within the forecasting time range. For example, the number of orders, the volume of goods transported, or inventory requirements, etc. The forecasting demand volume is an important basis for resource allocation and scheduling optimization in the logistics system. The forecasting confidence level is a quantitative evaluation of the reliability of the forecasting results by the model, usually expressed as a percentage or probability value. The higher the forecasting confidence level, the more confident the model is in the reliability of the forecasting results. For example, if the forecasting confidence level is 90%, it means that the model has a high level of confidence that the forecasting results are close to the actual demand. The calculation of the forecasting confidence level can be achieved through the uncertainty estimation of the model, historical forecasting error analysis, or statistical methods.

[0041] Through collaborative training under the federated learning framework, combined with the configuration of multi-source heterogeneous logistics data and the demand forecasting period, accurate forecasting of future logistics demand is achieved, and forecasting results including the forecasting time range, forecasting demand volume, and forecasting confidence level are generated. This process not only improves the accuracy and reliability of logistics demand forecasting, but also provides a solid data foundation and decision support for the global optimization of the logistics system.

[0042] P40: Transmit the logistics demand prediction result to the decision-making layer. The decision-making layer performs logistics scheduling optimization and generates a global optimization strategy to be transmitted to the application layer.

[0043] Further, step P40 of the embodiment of the present application further includes:

[0044] P41: The decision-making layer is built-in with a scheduling optimization model; P42: Transmit the logistics demand prediction result to the decision-making layer to match and generate multiple alternative logistics scheduling plans; P43: According to the scheduling optimization model, combined with the logistics optimization goal, perform global scheduling optimization on the multiple alternative logistics scheduling plans, generate a global optimization strategy and transmit it to the application layer.

[0045] It should be understood that the decision-making layer receives the logistics demand prediction result from the model layer and generates a global optimization strategy based on this to achieve efficient scheduling of the logistics system and optimal allocation of resources.

[0046] First, the decision-making layer is built-in with a scheduling optimization model, which is the core foundation for realizing logistics scheduling optimization. The scheduling optimization model is constructed based on mathematical programming, heuristic algorithms or machine learning techniques, and can handle complex logistics scheduling problems, such as vehicle routing optimization, warehouse resource allocation and delivery plan formulation. By considering various constraints, such as vehicle capacity, time window and traffic restrictions, and optimization goals, such as minimizing transportation costs, maximizing delivery efficiency or balancing resource utilization rate, the model generates an efficient scheduling plan. By having a built-in scheduling optimization model, the decision-making layer can provide a solid theoretical basis and technical support for subsequent global scheduling optimization.

[0047] Next, the decision-making layer receives the logistics demand prediction result and generates multiple alternative logistics scheduling plans based on this data. This process is a prerequisite step for generating the global optimization strategy, aiming to provide diverse choices for the optimization algorithm. The logistics demand prediction result provides detailed information on future logistics demand, including the prediction time range, predicted demand volume and prediction confidence level. The decision-making layer generates multiple alternative scheduling plans based on this information and in combination with the actual operation of the logistics system. For example, according to the demand volume within different time ranges, the decision-making layer can generate different transportation route plans, vehicle scheduling plans, inventory allocation strategies or order allocation plans. These alternative plans consider different business scenarios and potential risks, providing a rich candidate set for subsequent global scheduling optimization. By generating multiple alternative plans, the decision-making layer can ensure the flexibility and adaptability of the optimization process, thereby improving the quality of the global optimization strategy.

[0048] Furthermore, based on the built-in scheduling optimization model and in combination with the logistics optimization objectives, the decision-making layer conducts global scheduling optimization for multiple alternative logistics scheduling plans. This process is a key link in the entire logistics optimization method, aiming to select the optimal scheduling strategy from multiple alternatives to achieve the global optimization of the logistics system. Multiple optimization algorithms can be used for global scheduling optimization, such as linear programming, integer programming, genetic algorithms, ant colony algorithms, or simulated annealing algorithms, etc. These algorithms gradually optimize the scheduling plan through iterative calculations to meet the logistics optimization objectives. For example, if the optimization objective is to minimize the transportation cost, the scheduling plan with the lowest cost is found by adjusting the transportation route, vehicle allocation, and transportation time. During the process of global scheduling optimization, the decision-making layer comprehensively considers various constraints of the logistics system to ensure that the generated global optimization strategy is feasible and executable in actual operations. After being screened and adjusted by the optimization algorithm, the global optimization strategy is finally generated and transmitted to the application layer. The global optimization strategy usually includes transportation route planning, vehicle scheduling, inventory allocation, and order distribution, etc., providing clear guidance for the actual operation of the logistics system.

[0049] This process fully demonstrates the technical advantages of the scheduling optimization model, the diversity of alternative plans, and the flexibility of global scheduling optimization, ensuring the efficient operation and optimal allocation of resources of the logistics system in complex and changing business scenarios.

[0050] Furthermore, step P43 of the embodiment of this application further includes:

[0051] P43-1: Obtain the logistics optimization objectives, where the logistics optimization objectives include resource availability, transportation cost, and delivery time; P43-2: Based on the logistics optimization objectives, construct an equation for plan optimization; P43-3: Use the equation for plan optimization to calculate the fitness of the multiple alternative logistics scheduling plans, generating multiple plan fitness values; P43-4: According to the multiple plan fitness values, screen and extract the alternative logistics scheduling plan with the maximum fitness value as the global optimization strategy.

[0052] Specifically, the process of global scheduling optimization can be further refined. By clarifying the logistics optimization objectives, constructing an equation for plan optimization, calculating the plan fitness, and screening the optimal plan, it is ensured that the generated global optimization strategy can accurately meet the actual needs of the logistics system.

[0053] First, obtain the logistics optimization objectives, which usually include resource availability, transportation cost, and delivery time, that is, the key factors that need to be prioritized in the actual operation of the logistics system. For example, ensuring the efficient utilization of logistics resources, reducing transportation costs, and completing the delivery of goods on time.

[0054] Based on these logistics optimization goals, an equation for scheme optimization is constructed. This equation is a mathematical model used to quantify the performance of each alternative logistics scheduling scheme. By comprehensively considering factors such as resource availability, transportation cost, and delivery time, the equation converts these goals into quantifiable fitness metrics. For example, resource availability can be measured by resource utilization rate, transportation cost can be calculated by the length of the transportation route and the cost of using transportation tools, and delivery time can be evaluated by the degree of compliance with the delivery time window.

[0055] Subsequently, using the constructed equation for scheme optimization, the fitness of multiple alternative logistics scheduling schemes is calculated to generate the fitness value of each scheme. The fitness calculation is performed by substituting the characteristics of each alternative scheme into the equation for scheme optimization to obtain a numerical value that comprehensively reflects the performance of the scheme. The higher the fitness value, the better the scheme performs in meeting the logistics optimization goals. For example, a scheme with a higher fitness may be more efficient in resource utilization, have a lower transportation cost, and a more compliant delivery time.

[0056] Finally, based on the calculated fitness of multiple schemes, the alternative logistics scheduling scheme with the maximum fitness is selected and used as the global optimization strategy. This screening process ensures that the global optimization strategy can achieve the best balance among key goals such as resource availability, transportation cost, and delivery time. In this way, the decision-making layer can not only select the optimal scheduling strategy from multiple alternative schemes but also ensure that the strategy has the highest feasibility and efficiency in the actual logistics system.

[0057] P50: Output intelligent scheduling instructions to the logistics management system through the application layer for logistics management.

[0058] Furthermore, step P50 of the embodiment of the present application further includes:

[0059] P51: Through the application layer, convert the global optimization strategy into intelligent scheduling instructions, where the intelligent scheduling instructions include transportation task allocation, vehicle scheduling, warehouse operation arrangement, and personnel scheduling information; P52: Send the intelligent scheduling instructions to the logistics execution unit through the logistics management system interface for real-time logistics control.

[0060] Optionally, the functions of the application layer can be further refined to ensure that the global optimization strategy can be converted into specific intelligent scheduling instructions and real-time control of logistics operations can be achieved through the logistics management system.

[0061] First, through the application layer, the global optimization strategy is converted into intelligent scheduling instructions. These instructions cover multiple core aspects of logistics management, including transportation task allocation, vehicle scheduling, warehouse operation arrangement, and personnel scheduling information. Transportation task allocation ensures that the transportation routes and schedules of goods meet the requirements of the optimization strategy; vehicle scheduling instructions specify the tasks, routes, and time nodes of each vehicle to achieve efficient utilization of transportation resources; warehouse operation arrangement optimizes the storage and flow of goods to improve warehousing efficiency; personnel scheduling information reasonably allocates manpower according to task requirements to ensure the smooth execution of logistics operations. Through these detailed instruction contents, the application layer materializes the global optimization strategy, enabling it to directly guide the actual operation of the logistics system.

[0062] Furthermore, through a dedicated interface that connects the application layer with the logistics management system, the generated intelligent scheduling instructions are sent to the logistics execution unit to achieve real-time control of logistics operations. The logistics management system interface serves as a bridge for information transmission, ensuring that intelligent scheduling instructions can be accurately and efficiently conveyed to execution units such as transport vehicles, warehousing equipment, and operators. During the execution process, the logistics management system monitors the progress of each operation in real time and dynamically adjusts the scheduling instructions based on feedback data to handle possible emergencies, such as traffic congestion, equipment failures, or demand changes. This real-time control mechanism not only improves the response speed and flexibility of the logistics system but also ensures the dynamic consistency between logistics operations and the global optimization strategy, thus realizing the intelligent and efficient management of the logistics system.

[0063] In summary, the embodiments of this application have at least the following technical effects:

[0064] This application constructs a hierarchical collaborative architecture including a data layer, a model layer, a decision layer, and an application layer. The data layer integrates multi-source heterogeneous logistics data, the model layer uses federated learning to jointly train a demand prediction model and generate prediction results, the decision layer optimizes logistics scheduling based on the prediction results and generates a global optimization strategy, and finally the application layer outputs intelligent scheduling instructions to the logistics management system.

[0065] It achieves the technical effects of realizing the privacy protection and efficient integration of multi-source data, improving the accuracy of logistics demand prediction, scheduling efficiency, and supply chain collaboration ability through the hierarchical collaborative architecture.

[0066] Embodiment 2, based on the same inventive concept as the logistics optimization method based on AI collaboration in the foregoing embodiment, as Figure 2 shown, this application provides a logistics optimization system based on AI collaboration. The system in the embodiments of this application and the method embodiments are based on the same inventive concept. Among them, the system includes:

[0067] Hierarchical collaborative architecture construction module 11, which is used to construct a hierarchical collaborative architecture, and the hierarchical collaborative architecture includes a data layer, a model layer, a decision-making layer, and an application layer.

[0068] Logistics data acquisition module 12, which is used to integrate and acquire multi-source heterogeneous logistics data through the data layer.

[0069] Logistics demand forecasting module 13, which is used to adopt federated learning in the model layer to collaboratively train a demand forecasting model, and combine the multi-source heterogeneous logistics data to conduct logistics demand forecasting and generate a logistics demand forecasting result.

[0070] Logistics scheduling optimization module 14, which is used to transmit the logistics demand forecasting result to the decision-making layer, and the decision-making layer conducts logistics scheduling optimization and generates a global optimization strategy and transmits it to the application layer.

[0071] Logistics scheduling management module 15, which is used to output intelligent scheduling instructions through the application layer to a logistics management system for logistics management.

[0072] Furthermore, the logistics data acquisition module 12 is also used to perform the following steps:

[0073] Based on the internal logistics system, obtain order data, inventory data, transportation data, and warehousing data, and based on external data sources, obtain weather data, traffic condition data, and market demand data; jointly form the original multi-source logistics data; preprocess the original multi-source logistics data, including data cleaning, data conversion, and data fusion, to obtain multi-source heterogeneous logistics data.

[0074] Furthermore, the logistics demand forecasting module 13 is also used to perform the following steps:

[0075] Obtain the logistics data of multiple participants and perform distributed storage, and each participant only processes the data locally; construct a federated learning framework, and under the federated learning framework, each participant conducts model collaborative training and shares model parameter updates to train and generate a demand forecasting model.

[0076] Furthermore, the logistics demand forecasting module 13 is also used to perform the following steps:

[0077] Configure the demand forecasting cycle; input the multi-source heterogeneous logistics data into the demand forecasting model, and according to the demand forecasting cycle, forecast the future logistics demand to generate a logistics demand forecasting result, and the logistics demand forecasting result includes the forecast time range, forecast demand quantity, and forecast confidence level.

[0078] Further, the logistics scheduling optimization module 14 is further configured to perform the following steps:

[0079] The decision-making layer is built with a scheduling optimization model; the logistics demand prediction result is transmitted to the decision-making layer, and multiple alternative logistics scheduling plans are generated through matching; according to the scheduling optimization model and in combination with the logistics optimization objective, global scheduling optimization is performed on the multiple alternative logistics scheduling plans, and a global optimization strategy is generated and transmitted to the application layer.

[0080] Further, the logistics scheduling optimization module 14 is further configured to perform the following steps:

[0081] Obtain the logistics optimization objective, where the logistics optimization objective includes resource availability, transportation cost, and delivery time; based on the logistics optimization objective, construct an equation for optimizing the plan; use the equation for optimizing the plan to calculate the fitness of the multiple alternative logistics scheduling plans, and generate multiple plan fitness values; according to the multiple plan fitness values, screen and extract the alternative logistics scheduling plan with the maximum fitness as the global optimization strategy.

[0082] Further, the logistics scheduling management module 15 is further configured to perform the following steps:

[0083] Through the application layer, convert the global optimization strategy into intelligent scheduling instructions, where the intelligent scheduling instructions include transportation task allocation, vehicle scheduling, warehouse operation arrangement, and personnel scheduling information; through the logistics management system interface, send the intelligent scheduling instructions to the logistics execution unit for real-time logistics control.

[0084] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification is provided. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0086] This specification and the drawings are only exemplary descriptions of the present application and are considered to have covered any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A logistics optimization method based on AI collaboration, characterized in that: The method comprises: Constructing a layered collaborative architecture, which includes a data layer, a model layer, a decision layer, and an application layer; Acquire multi-source heterogeneous logistics data through the data layer integration; In the model layer, a demand forecasting model is trained collaboratively using federated learning, and logistics demand forecasting is performed in combination with the multi-source heterogeneous logistics data to generate a logistics demand forecasting result; The logistics demand forecast result is transmitted to the decision layer, and the decision layer performs logistics scheduling optimization, generates a global optimization strategy and transmits it to the application layer; The intelligent scheduling instructions are output through the application layer to the logistics management system for logistics management.

2. A logistics optimization method based on AI collaboration as claimed in claim 1, characterized in that: The data layer is integrated to obtain multi-source heterogeneous logistics data, including: Based on the internal logistics system, order data, inventory data, transportation data and warehousing data are obtained, and weather data, traffic data and market demand data are obtained based on external data sources; together they constitute the original multi-source logistics data; The original multi-source logistics data is preprocessed, including data cleaning, data conversion and data fusion, to obtain multi-source heterogeneous logistics data.

3. The logistics optimization method based on AI collaboration as claimed in claim 1, characterized in that: In the model layer, a federated learning collaborative training demand prediction model is adopted, including: Obtain logistics data from multiple participants and store them in a distributed manner, with each participant only processing the data locally; Build a federated learning framework, under which all participants conduct collaborative model training, share model parameter updates, and train to generate demand forecasting models.

4. A logistics optimization method based on AI collaboration as claimed in claim 3, characterized in that: Combining the multi-source heterogeneous logistics data to perform logistics demand forecasting includes: Configure demand forecast period; The multi-source heterogeneous logistics data is input into the demand forecasting model, and the future logistics demand is predicted according to the demand forecasting cycle to generate a logistics demand forecasting result, which includes a forecasting time range, a forecasting demand amount and a forecasting confidence level.

5. The logistics optimization method based on AI collaboration as claimed in claim 4, characterized in that: The logistics demand forecast result is transmitted to the decision layer, and the decision layer performs logistics scheduling optimization, including: The decision layer has a built-in scheduling optimization model; Transmitting the logistics demand forecast results to the decision-making layer to match and generate multiple alternative logistics scheduling solutions; According to the scheduling optimization model and in combination with the logistics optimization goal, a global scheduling optimization is performed on the multiple alternative logistics scheduling solutions, and a global optimization strategy is generated and transmitted to the application layer.

6. A logistics optimization method based on AI collaboration as claimed in claim 5, characterized in that: Performing global scheduling optimization on the multiple alternative logistics scheduling solutions includes: obtaining a logistics optimization target, wherein the logistics optimization target includes resource availability, transportation cost, and delivery time; Based on the logistics optimization goal, construct a solution optimization equation; Using the solution optimization equation, calculating the fitness of the multiple alternative logistics scheduling solutions to generate multiple solution fitnesses; According to the fitness of the multiple solutions, the alternative logistics scheduling solution with the largest fitness is screened and extracted as the global optimization strategy.

7. The logistics optimization method based on AI collaboration as claimed in claim 1, characterized in that: Outputting intelligent dispatching instructions to the logistics management system through the application layer to perform logistics management includes: Through the application layer, the global optimization strategy is converted into intelligent scheduling instructions, and the intelligent scheduling instructions include transportation task allocation, vehicle scheduling, warehouse operation arrangement and personnel scheduling information; The intelligent dispatching instructions are sent to the logistics execution unit through the logistics management system interface for real-time logistics management and control.

8. A logistics optimization system based on AI collaboration, characterized in that: The system comprises: A layered collaborative architecture building module, wherein the layered collaborative architecture building module is used to build a layered collaborative architecture, wherein the layered collaborative architecture includes a data layer, a model layer, a decision layer, and an application layer; A logistics data acquisition module, the logistics data acquisition module is used to acquire multi-source heterogeneous logistics data through the data layer integration; A logistics demand forecasting module, which is used to adopt federated learning to collaboratively train a demand forecasting model in the model layer, and to perform logistics demand forecasting in combination with the multi-source heterogeneous logistics data to generate a logistics demand forecasting result; A logistics scheduling optimization module, which is used to transmit the logistics demand forecast result to the decision layer, and the decision layer performs logistics scheduling optimization, generates a global optimization strategy and transmits it to the application layer; The logistics scheduling management module is used to output intelligent scheduling instructions to the logistics management system through the application layer to perform logistics management.

Images