An AI algorithm-based supply chain performance path optimization and collaborative decision-making method

Abstract

The application discloses a supply chain performance path optimization and collaborative decision-making method based on an AI algorithm, and particularly relates to the field of supply chain performance path optimization, and comprises the following steps: firstly, order goods information and warehouse resource data are acquired, and it is judged whether a single warehouse can independently perform the performance by matching; if yes, a comprehensive evaluation function containing total transportation time, time redundancy, path complexity and resource utilization rate is constructed, and an optimal path from the warehouse to a delivery place is generated; if multiple warehouses need to be cooperated, a centralized point warehouse is dynamically selected, the goods transfer cost and historical performance risk value of each warehouse are comprehensively calculated, a comprehensive early warning score is generated, and an optimal centralized point is selected, finally, the transfer path from the supply warehouse to the centralized point and the performance path from the centralized point to the terminal are planned, and a full-link collaborative scheme is formed; the application breaks through the traditional single-warehouse decision-making mode, realizes global scheduling of resources through a double-path optimization mechanism, significantly reduces the transportation delay risk, and improves the operation efficiency of the supply chain.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply chain fulfillment path optimization, more specifically, the present application relates to a supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm. BACKGROUND

[0002] The current supply chain path optimization technology has formed a method system with data-driven as the core. The mainstream scheme focuses on single warehouse intelligent scheduling, multi-warehouse basic collaboration and risk response mechanism: in the single warehouse scene, the AI algorithm generates a local optimal path through historical data analysis, which significantly improves the efficiency of warehouse operation; in the multi-warehouse collaboration field, the centralized system uses digital twin technology to realize the preliminary matching of transport capacity and tasks; in the risk prevention and control aspect, the head enterprises integrate meteorological, traffic and other multi-source data to build prediction model, which enhances the prediction ability of delay events. Emerging technologies such as large models begin to penetrate the logistics decision system, and promote the evolution of supply chain to dynamic and visual. However, these technologies are still limited by centralized architecture, and the optimization range is concentrated in local links. The ability of global resource collaboration and real-time risk linkage has not been broken through.

[0003] However, it still has some disadvantages in actual use:

[0004] Decision logic is fragmented: single warehouse optimization and multi-warehouse collaboration are handled in isolation, and most systems cannot dynamically switch decision modes according to order demand, resulting in low efficiency of cross-warehouse resource allocation;

[0005] Evaluation dimension is single: path optimization relies too much on transportation time and distance cost, ignoring key implicit indicators such as path complexity, time buffer redundancy, and actual execution deviation is significant;

[0006] Risk response lags: traditional risk models rely on static historical data and are difficult to dynamically integrate sudden environmental disturbances, risk prediction and cost optimization are disconnected, and emergency decision-making relies on manual intervention.

[0007] These defects together lead to systematic bottlenecks in supply chain, such as high resource mismatch rate, weak resistance to sudden risks, and insufficient adaptability to complex scenarios. SUMMARY

[0008] In order to overcome the above-mentioned defects of the prior art, the present application provides a supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm, which solves the problems in the background art by the following scheme.

[0009] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm, comprising:

[0010] S1: data acquisition: acquire target order information and each warehouse site information, and preprocess;

[0011] S2: Feasibility judgment of single-warehouse fulfillment: match the target order information with each warehouse site information to determine whether there is a single warehouse that can meet the demand of all goods of the order; if so, determine the candidate warehouse and execute step S3, otherwise execute step S4;

[0012] S3: Single-warehouse optimal path generation: take the candidate warehouse as the starting point, the delivery address in the order information as the end point, and construct a comprehensive evaluation index function; evaluate each path from the starting point to the end point according to the comprehensive evaluation index function, and obtain the optimal path; if there are two or more candidate warehouses, determine the optimal candidate warehouse according to the comprehensive evaluation index of the optimal path of each warehouse;

[0013] S4: Multi-warehouse collaborative decision and path optimization:

[0014] When multiple warehouses are needed to coordinate to meet the demand of all goods of the target order, the multiple warehouses are taken as candidate centralized warehouses, and a transfer warehouse is determined, and then a function is established to analyze and obtain a multi-warehouse collaborative fulfillment path, the specific steps of which are as follows:

[0015] S401: Constructing a transfer cost function: calculate the total transportation cost and total transportation time of the goods transferred from other warehouses to the candidate warehouse as the centralized point;

[0016] S402: Constructing a risk prediction function: based on the historical fulfillment time and external environmental risk of the warehouse, calculate the default risk value of the warehouse as the centralized point;

[0017] S403: According to the transfer cost function and the default risk prediction function, generate a comprehensive early warning function of each candidate warehouse by weighted calculation, and select the candidate warehouse with the optimal comprehensive evaluation value as the final centralized point;

[0018] S404: Based on the determined centralized point, plan the transfer path of each warehouse to the centralized point and the fulfillment path from the centralized point to the order delivery address to form a multi-warehouse collaborative fulfillment path.

[0019] Preferably, the data acquisition comprises acquiring target order information and warehouse site information for preprocessing;

[0020] The acquired target order information includes: cargo type, cargo quantity, delivery deadline, and delivery address; and the warehouse site information includes: inventory data and transportation resource data of each warehouse;

[0021] The inventory cargo data includes: inventory cargo types, inventory cargo quantity, and unit volume of each cargo; the transportation resource data includes: vehicle resources, vehicle capacity, vehicle load, and loading and unloading efficiency; and the order volume, order weight, and estimated loading and unloading time are calculated according to the target order information and the inventory cargo data.

[0022] Preferably, the feasibility judgment of the single-warehouse fulfillment includes:

[0023] The cargo types and cargo quantity based on the target order information are matched with the inventory cargo quantity and inventory cargo types in each warehouse site information. If there is a single warehouse that meets all conditions of the target order information, all single warehouses that meet the conditions are taken as candidate warehouses, and S3 is performed. If there is no single warehouse that meets the conditions, but the total quantity and types of inventory cargos of multiple single warehouses meet all matching conditions of the target order information, each single warehouse that meets the conditions is taken as a candidate warehouse of a centralized point, and S4 is entered.

[0024] Preferably, the single-warehouse optimal path generation includes:

[0025] The candidate warehouse is taken as a starting point, and the delivery address in the order information is taken as an end point. All passable path scheme sets from the starting point to the end point are counted, and the estimated driving time of all passable paths is predicted.

[0026] In combination with the order information and the transportation resource data, the estimated arrival time and all available transportation resource sets are obtained, and an evaluation vector including four maximization indexes is generated for each candidate path scheme: total transportation time, time redundancy, path complexity, and resource utilization rate. The evaluation indexes of all candidate path schemes are normalized, and a comprehensive evaluation index is obtained by calculation. The specific steps are as follows:

[0027] S301 defines a multi-target vector: a four-dimensional evaluation vector including total transportation time, time redundancy, path complexity, and resource utilization rate is constructed for each candidate path scheme.

[0028] S302 constructs a standardized decision matrix: an m*4-dimensional standard matrix is generated based on vector normalization method.

[0029] S303 determines ideal values and anti-ideal values: the maximum value of each index in the scheme is calculated to form an ideal solution, and the minimum value is calculated to form a negative ideal solution.

[0030] S304 calculates distance measure: a weight coefficient is introduced to strengthen the influence of key indexes by weighted processing, and the deviation degree of the scheme from the ideal solution and the negative ideal solution is quantified.

[0031] S305 calculates a single warehouse comprehensive evaluation index: the ratio of anti-ideal distance and total distance, violates the time limit for filing or buffer threshold, all scheme scores are mapped to the interval [0, 1], and the optimal path scheme is obtained according to the comprehensive evaluation index value;

[0032] S306 if there are multiple candidate warehouses, calculate the comprehensive evaluation index of the optimal path of each warehouse, and determine the optimal candidate warehouse according to the comprehensive evaluation index value;

[0033] Preferably, the transport cost function is constructed, comprising:

[0034] First, calculate the transport cost of each warehouse that needs to be transported to the designated centralized warehouse, then calculate the total cost of all single transport warehouses, and obtain the transport cost function; wherein the cost of a single transport warehouse is composed of three parts: distance cost, time cost and cargo characteristics;

[0035] The distance cost function is constructed based on the transportation distance parameter, which combines the basic linear coefficient and the nonlinear penalty coefficient to obtain the adjusted basic distance cost and the converged long-distance implicit cost, and then the toll generated in the traffic is calculated to obtain the distance cost function;

[0036] The time cost function is constructed based on the timeliness parameter, which combines the basic time cost, the emergency transportation coefficient and the time sensitivity coefficient to obtain the basic time cost and the cargo characteristic cost adjusted according to the characteristics of the goods, and then the time cost function is obtained;

[0037] The comprehensive transport cost function affected by the characteristics of the goods is obtained by adjusting the above two types of function sums through the calculation of the goods density factor and the transport goods quantity.

[0038] Preferably, the risk prediction function is constructed, comprising:

[0039] The risk is quantified from two dimensions, namely the timeliness risk and the external environment risk. The timeliness risk is based on the historical transportation task data of the candidate centralized warehouse, analyzes the deviation between the actual time length and the planned time length of the past transportation tasks, and calculates the weighted average deviation value by combining the deviation rate of all historical tasks of the warehouse and the task weight. The higher the deviation value is, the greater the probability of distribution delay of the warehouse as a centralized point is; the external environment risk dynamically integrates the influence of sudden external factors on the transportation path, including the implementation of meteorology and geographical location sensitivity, and improves the implicit risk of border or port warehouse due to geographical location;

[0040] The two types of risk values are normalized to the same dimension, the historical deviation weighted average value is directly mapped to a probability value, the external environment risk is inhibited by a decay coefficient to prevent the excessive dominance of external risk, and the standardized score is converted by combining the adjustment of the geographical position sensitive coefficient.

[0041] According to the business requirements, the two types of risks are assigned weights, and the weighted sum is generated to generate a comprehensive risk quantitative value.

[0042] Preferably, the comprehensive evaluation function of each candidate warehouse comprises:

[0043] The transfer cost and risk prediction value of each candidate warehouse are standardized respectively, so as to be mapped to the interval [0, 1], the weights are assigned according to the importance of the transfer cost and risk prediction value in decision-making, the weighted sum of the normalized transfer cost and risk prediction value is calculated, and finally the comprehensive early warning values of all candidate centralized point warehouses are sorted, and the warehouse with the lowest early warning value is selected as the final centralized point.

[0044] Preferably, the multi-warehouse collaborative fulfillment path comprises:

[0045] The single-warehouse optimal path generation method is called to generate the optimal fulfillment path from the determined centralized point warehouse to the order delivery address;

[0046] The single-warehouse optimal path generation method is called to generate the optimal transfer path from each transfer warehouse to the centralized point;

[0047] The generated optimal transfer path and optimal fulfillment path are combined to form a complete multi-warehouse collaborative fulfillment path of transfer warehouse-centralized point warehouse-order delivery address.

[0048] The technical effects and advantages of the present application are:

[0049] 1. Innovative dual-mode dynamic decision-making architecture: Breakthrough the limitation of traditional decision-making logic independence, realize adaptive switching of decision-making mode through single-warehouse fulfillment feasibility judgment: In the single-warehouse scenario, a four-dimensional evaluation system including total transportation time, time redundancy, path complexity and resource utilization is constructed, the path advantages and disadvantages are quantified through standardization processing and distance measurement, and the comprehensive optimal path is generated; In the multi-warehouse scenario, the transfer network is generated by combining the distance of the transfer warehouse and the inventory matching rule through the dynamic centralized point selection model, the global resource scheduling is realized, and the cross-warehouse collaboration efficiency is improved.

[0050] 2、Constructing risk and cost joint optimization mechanism. The transfer cost function integrates distance cost and time cost, including basic distance coefficient, long distance cost, toll point additional cost, basic time cost, emergency coefficient and sensitivity coefficient, accurately quantifying the differences of different goods transfer; the risk prediction function integrates historical performance time limit deviation and dynamic external environment risk to generate normalized risk value; the two are weighted and integrated through comprehensive early warning score to balance cost and risk, reduce the probability of default and invalid transfer.

[0051] 3、Explicitly quantify implicit factors to enhance adaptability in complex scenarios. Factors such as path complexity and time redundancy, which are traditionally ignored, are included in the decision-making system to improve the adaptability to complex road networks and urgent orders and optimize resource utilization.

[0052] 4、Global real-time response mechanism to improve system agility. Multi-warehouse scenario reuses single-warehouse path optimization algorithm to reduce repeated calculations; when external environment changes, parameters can be dynamically updated to adjust the path, realizing the coordination of local decision and global target, enhancing the anti-interference ability and operation efficiency of the supply chain. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 The overall structure of the present application is shown in the figure.

[0054] Figure 2 The overall flowchart of the present application is shown in the figure.

[0055] Figure 3 The single-warehouse performance structure of the present application is shown in the figure.

[0056] Figure 4 The multi-warehouse collaborative decision structure of the present application is shown in the figure. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0058] Reference Figure 1 - Figure 4 The supply chain performance path optimization and collaborative decision method based on AI algorithm shown in the figure comprises:

[0059] S1: Obtain data: obtain target order information and each warehouse site information, and preprocess;

[0060] Through the deployment of standardized API interface system and AI algorithm, data automatic processing is realized:

[0061] Obtain target order information through the system API interface, automatically capture order key fields, and read local database warehouse site information;

[0062] The target order information obtained includes: cargo type, cargo quantity, delivery deadline, and delivery address;

[0063] The warehouse site information includes: warehouse inventory data and transportation resource data;

[0064] The inventory data includes: inventory cargo type, inventory cargo quantity, unit volume of each cargo, and unit weight of each cargo; the transportation resource data includes: vehicle resources, vehicle capacity, vehicle load, and loading and unloading efficiency;

[0065] Call the local database for each cargo unit volume parameter, and calculate the order cargo volume and order cargo weight based on the order cargo type, cargo quantity, and unit weight of each cargo, and calculate the expected loading and unloading time based on the loading and unloading efficiency.

[0066] S2: Feasibility judgment of single warehouse fulfillment: match the target order information with the warehouse site information to determine whether a single warehouse can meet all the needs of the order goods, and determine the candidate warehouse; if a single warehouse can meet all the needs of the order goods, execute step S3, otherwise execute step S4.

[0067] Based on the cargo type and cargo quantity of the target order information, match the inventory cargo quantity and inventory cargo type in the warehouse site information to determine whether a single warehouse can meet the requirements of the cargo type and cargo quantity of the target order information;

[0068] If there is a single warehouse that meets all the conditions of the target order information, all single warehouses that meet the conditions are used as candidate warehouses, and step S3 is executed;

[0069] If there is no single warehouse that meets the conditions, but the total quantity and type of inventory goods in multiple single warehouses meet all the matching conditions of the target order information, each single warehouse that meets the conditions is used as a candidate warehouse, and step S4 is entered.

[0070] S3: Single warehouse optimal path generation: use the candidate warehouse as the starting point, the delivery address in the order information as the end point, construct a graph neural network model based on AI algorithm, and count all the path scheme sets that can be passed from the starting point to the end point, denoted as: wherein represents the i-th path scheme, and predicts the expected driving time of all passable paths; and based on the transportation resource data in the order information and warehouse site information, all available transportation resource sets are given, denoted as: where h represents the total number of all transport resources; accordingly, a comprehensive evaluation index of the path is generated, and the optimal path is selected, with the specific steps as follows:

[0071] S301: Define a multi-objective vector: construct a four-dimensional evaluation vector for each candidate path scheme, and unify the four key optimization objectives into a maximization problem:

[0072] ;

[0073] wherein: T represents the sum of the expected travel time and the expected loading and unloading time;

[0074] represents the time redundancy, i.e., the margin of the arrival time being earlier than the order delivery deadline, wherein, represents the order delivery deadline, represents the expected delivery arrival time of the scheme i;

[0075] represents the negative value of the path complexity, wherein, edges represents the road section type, including: expressway, trunk road, congested road section, construction road section, and the corresponding complexity coefficients are 0.1, 0.3, 0.8, and 1.5;

[0076] represents the resource utilization rate, ;

[0077] S302: Construct a standardized decision matrix

[0078] Define an initial decision matrix: combine the evaluation vectors of the m candidate schemes into an m*4 matrix:

[0079] ;

[0080] wherein m is the total number of candidate schemes, represents the jth index value of the ith scheme, the first list represents the negative value of the transport time, the second list represents the time redundancy, the third list represents the negative value of the path complexity, and the fourth list represents the resource utilization rate; then, standardization processing is performed to eliminate the unit difference of the index and avoid large order of magnitude indexes dominating the calculation results, and the matrix Z is obtained, and the specific mathematical function of the standardization processing is:

[0081] ,

[0082] wherein represents the standardized value of the jth index of the ith scheme; and the standardized matrix is obtained.​ ;

[0083] S303: Determine the ideal value and anti-ideal value

[0084] On the basis of the standardized decision matrix Z, the most ideal solution and the worst

[0085] ideal solution as a reference point for subsequent distance calculation; ideal solution: a virtual optimal scheme, which takes the maximum value in each index of all candidate schemes, and its calculation formula is:

[0086] ;

[0087] Where represents the maximum value of the jth index in m schemes;

[0088] anti-ideal solution: a virtual worst scheme, which takes the minimum value in each index of all candidate schemes, and its calculation formula is:

[0089] ;

[0090] Where represents the minimum value of the jth index in m schemes;

[0091] S304: Calculate the distance measure

[0092] Calculate the weighted Euclidean distance of each candidate scheme i to the ideal solution and the anti-ideal solution , quantify the deviation degree of the scheme from the optimal and worst solutions; the specific mathematical function of the weighted Euclidean distance of scheme i to the ideal solution is:

[0093] ;

[0094] The specific mathematical function of the weighted Euclidean distance of scheme i to the ideal solution is:

[0095] ;

[0096] Where represents the standardized value of the ith scheme and the jth index, is the weight of the jth index, according to the actual situation analysis, the priority order should be according to efficiency, reliability, road complexity, cost; therefore , its constraints: ;

[0097] S305: Calculate the single-bin comprehensive evaluation index

[0098] By calculating the comprehensive evaluation index of each scheme i , the closeness to the ideal solution and the distance from the anti-ideal solution are evaluated, and a single bin comprehensive evaluation index in the interval [0, 1] is generated, and the specific mathematical function is:

[0099] ;

[0100] It should be further pointed out that all the passable paths need to meet the following conditions when being determined as the optimal path: 1. The expected transportation time of the path should not exceed the order delivery deadline; 2. The redundancy time should not be less than the preset minimum buffer time.

[0101] According to the above steps, the optimal path scheme is obtained: , which represents the selection of the scheme P with the highest comprehensive evaluation as the optimal path scheme under the transportation resource;

[0102] S306: Determine the optimal candidate warehouse

[0103] If there are two or more candidate warehouses, repeat the above steps to calculate the optimal path scheme of each candidate warehouse, and sort the candidate warehouses according to the comprehensive evaluation index of their optimal path schemes from large to small, and select the candidate warehouse with the largest comprehensive evaluation index as the optimal warehouse.

[0104] S4: Multi-warehouse collaborative decision and path optimization: Take each warehouse that needs to be collaboratively dispatched by multiple warehouses as a centralized point candidate warehouse, construct the transfer cost function and the transfer risk function of each centralized point candidate warehouse, obtain the comprehensive feasibility function, and determine the optimal centralized point warehouse according to the function value. The specific steps are as follows:

[0105] Define the candidate centralized point warehouse as y, and the transfer warehouse set as , x represents the xth transfer warehouse, the set of transfer warehouses is defined as follows: based on the quantity of goods lacking in the candidate central point warehouse and the goods inventory of all transfer warehouses, the goods are allocated from near to far according to the distance between the warehouses, that is, the distance between all transfer warehouses and candidate central point warehouses is first sorted from short to long, if the inventory of the first candidate transfer warehouse meets the quantity of goods lacking in the candidate central point warehouse, only the candidate transfer warehouse is used as the transfer warehouse; if not, it is judged whether the total inventory of the first and second transfer warehouses meets the quantity of goods lacking in the candidate central point warehouse, if yes, the first and second candidate transfer warehouses are used as the transfer warehouses; if not, it is calculated whether the total inventory of the first, second and third transfer warehouses meets the requirement, until the total inventory of the candidate transfer warehouses meets the quantity of goods lacking in the candidate central point warehouse, the candidate transfer warehouses meeting the requirement are used as the transfer warehouses of the candidate central point warehouse;

[0106] It should be further explained that the candidate central point is defined as the terminal point and the transfer warehouse is defined as the starting point here, steps S301-S305 are repeated to obtain the optimal path scheme Pi of the transfer warehouse to the candidate central point warehouse, and the actual distance of the transfer warehouse x to the candidate central point y is obtained according to the optimal path , the expected time , and the proportion of the perishable goods in the transfer goods is counted;

[0107] S401: Constructing a transfer cost function: define the transfer cost as , and the specific mathematical function is as follows:

[0108]

[0109] wherein is the transfer goods quantity of the warehouse x, is the goods density factor, ; the distance cost function is: , wherein is a basic linear coefficient, representing the basic operation cost per unit distance, reflecting the fuel consumption and vehicle wear per unit distance; is a nonlinear penalty coefficient, representing the marginal cost increasing factor of long-distance transportation, reflecting the implicit cost, but not dominant in the distance cost, therefore, the value is 0.02, is a distance index, taking the value of 1.3, is the number of toll points, is the single toll fee;

[0110] The time cost function is: , wherein is the basic time cost, is the emergency coefficient, when transporting ordinary goods, is the emergency coefficient, when transporting fragile goods, , is the time sensitivity, taking a value of 0.4;

[0111] S402: Constructing a risk prediction function: quantifying the risk value of candidate point warehouse y as a concentration point when the order fulfillment fails due to transportation time delay or external factors, outputting a normalized score R, the mathematical function of which is as follows:

[0112]

[0113] wherein, is the timeliness risk weight, is the environmental risk weight, the environment during transportation can better reflect the risk of the entire process, so the timeliness risk weight is 0.4 and the environmental risk weight is 0.6;

[0114] represents the timeliness risk of fulfillment, based on the historical transportation task timeliness deviation of warehouse y, the delivery delay probability of warehouse y as a concentration point is predicted, which is represented as:

[0115]

[0116] wherein, is the historical transportation task sample quantity of warehouse y, is the actual transportation duration of task j, is the planned transportation duration of task j, is the task weight, when transporting fragile goods , when transporting ordinary goods , the historical task quantity is the historical completed task within 90 days;

[0117] represents the external environmental risk, dynamically integrating the influence of unexpected factors on the transportation path, which is represented as: wherein is the attenuation coefficient, avoiding the dominance of external risk in the total score, , is the geographical location sensitivity coefficient, ordinary warehouse: , border or port warehouse , is the safety threshold, taking a value of 2, is the regional implementation risk index, according to the access meteorological platform API, dynamically integrating real-time meteorological factors, updating the real-time safety threshold , taking a value of [0, 4].

[0118] S403: Constructing a candidate centralized point comprehensive early warning function: by weighted fusion of the transfer cost function and the risk prediction function, a candidate centralized point comprehensive early warning function is constructed, the design logic is: the transfer cost and the risk prediction value of each candidate warehouse are standardized respectively, so as to be mapped to the [0, 1] interval, and then the weights are allocated according to the importance of the transfer cost and the risk prediction value in decision-making; then the normalized transfer cost and risk prediction value are multiplied by the corresponding weight and added, to obtain the comprehensive early warning value of each candidate centralized point warehouse; finally, the comprehensive early warning values of all candidate centralized point warehouses are sorted, and the warehouse with the lowest early warning value is selected as the final centralized point, and the specific steps are as follows:

[0119] Normalization processing: ,

[0120] The combined comprehensive early warning function: , wherein , are the weight coefficients of the transfer cost and the risk prediction value, respectively, since the transfer is preferentially transferred to the warehouse closest to the candidate centralized point warehouse, the transfer cost can more highlight the comprehensive early warning value of the candidate centralized point warehouse, therefore , ; according to the above steps, the comprehensive early warning values of all candidate centralized point warehouses are calculated, and are sorted from low to high according to the value, and the candidate centralized point warehouse with the lowest comprehensive early warning value is selected as the centralized point warehouse.

[0121] S404: Based on the determined centralized point warehouse, the transfer path of each transfer warehouse to the centralized point warehouse and the fulfillment path of the centralized point to the order delivery address are planned to form a multi-warehouse collaborative fulfillment path.

[0122] Taking the determined centralized point warehouse as the starting point and the delivery address in the target order information as the terminal, repeat step S3 to determine the optimal path from the centralized point warehouse to the order delivery address;

[0123] Taking the centralized point warehouse as the terminal, and respectively taking the warehouses that need to be transferred to the centralized point warehouse as the starting point, repeat step S3 to determine the optimal path from the transfer warehouse to the centralized point warehouse;

[0124] Based on the optimal path from the transfer warehouse to the centralized point warehouse and the optimal path from the centralized point warehouse to the order delivery address, a multi-warehouse collaborative optimal fulfillment path from the transfer warehouse to the centralized point warehouse to the order delivery address is formed.

[0125] The present application solves the core defects of resource scheduling fragmentation, risk response lag and single evaluation dimension in supply chain path optimization by constructing a single-warehouse fulfillment and multi-warehouse collaborative dual-path decision framework. Its innovative implementation embodies three levels:

[0126] The base layer: based on the AI algorithm to dynamically judge the order fulfillment mode: single warehouse feasibility and multi-warehouse cooperation necessity, realize the intelligent switching of resource scheduling mode;

[0127] Optimization layer: single warehouse scene adopts multi-dimensional path evaluation system, combines transportation efficiency, time redundancy, path complexity and resource utilization to generate local optimal path;

[0128] In the multi-warehouse scene, a comprehensive early warning scoring model is innovatively designed, the transfer cost function and the risk prediction function are weighted and fused, and the global optimal centralized point is dynamically selected;

[0129] The execution layer: the single warehouse path optimization algorithm is reused to generate a full-link collaborative scheme from the transfer warehouse to the centralized point and finally to the terminal user, realizing the seamless connection of resource allocation and fulfillment path.

[0130] This method fundamentally breaks through the limitations of static planning and experience decision in traditional technology, dynamically predicts the risk to drive global resource optimization, significantly improves the adaptability, robustness and execution efficiency of the supply chain in complex scenarios, and provides a scalable intelligent decision basis for high volatility logistics environment.

[0131] Secondly: the drawings of the disclosed embodiments only involve the structures involved in the disclosed embodiments, other structures can refer to the usual design, and in the case of no conflict, the same embodiment and different embodiments of the present application can be combined with each other;

[0132] Finally: the above only describes the preferred embodiments of the present application and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithms, characterized in that, include: S1: Data Acquisition: Acquire target order information and information about each warehouse site, and perform preprocessing; S2: Feasibility assessment of single warehouse fulfillment: Match the target order information with the information of each warehouse site to determine whether there is a single warehouse that can meet all the requirements of the order's goods; If a candidate repository exists, determine the candidate repository and proceed to step S3; otherwise, proceed to step S4. S3: Single-Warehouse Optimal Route Generation: Using the candidate warehouse as the starting point and the delivery address in the order information as the destination, a comprehensive evaluation index function is constructed. Each path from the starting point to the destination is evaluated based on the comprehensive evaluation index function, and the optimal path is obtained. If there are two or more candidate warehouses, the optimal candidate warehouse is determined based on the comprehensive evaluation index of the optimal path for each warehouse. The specific steps are as follows: S301: Define a multi-objective vector: Construct a four-dimensional evaluation vector for each candidate route, including total transportation time, time redundancy, path complexity, and resource utilization. S302: Constructing a standardized decision matrix: Based on vector normalization, an m*4 dimensional standard matrix is ​​generated, where m is the total number of candidate paths; S303: Determine the ideal and anti-ideal values: Calculate the maximum value of each index in the scheme to form the ideal solution, and calculate the minimum value to form the negative ideal solution; S304: Calculate distance measure: Introduce weighting coefficients, weighted processing to enhance the influence of key indicators, and quantify the degree of deviation between the scheme and the ideal solution and the negative ideal solution; S305: Calculate the comprehensive evaluation index for a single warehouse: use the ratio of the anti-ideal distance to the total distance, and directly eliminate those that violate the timeliness or buffer threshold. Map the scores of all schemes to the interval [0, 1], and obtain the optimal path scheme based on the comprehensive evaluation index value. S306: If there are multiple candidate warehouses, calculate the comprehensive evaluation index of the optimal path for each warehouse, and determine the optimal candidate warehouse based on the comprehensive evaluation index value. S4: Multi-warehouse collaborative decision-making and path optimization: When multiple warehouses need to be coordinated to meet the full demand for a target order, these warehouses are selected as candidate central warehouses, and transshipment warehouses are determined. Then, a function is established, and the multi-warehouse collaborative fulfillment path is obtained after analysis. The specific steps are as follows: S401: Construct a transshipment cost function: Calculate the total transportation cost and total transportation time for goods to be transshipped from other warehouses to each candidate warehouse when each candidate warehouse is the central point. S402: Construct a risk prediction function: Calculate the default risk value of the warehouse as a concentration point based on the warehouse's historical performance timeliness and external environmental risks; S403: Based on the transshipment cost function and the default risk prediction function, a comprehensive early warning function for each candidate warehouse is generated through weighted calculation, and the candidate warehouse with the best comprehensive evaluation value is selected as the final concentration point; S404: Based on the determined central point, plan the transfer routes from each warehouse to the central point and the fulfillment routes from the central point to the order delivery address to form a multi-warehouse collaborative fulfillment path.

2. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm as described in claim 1, characterized in that: The data acquisition includes acquiring target order information and information about each warehouse site, and performing preprocessing. The target order information obtained includes: goods type, goods quantity, delivery deadline, and delivery address; the warehouse site information includes: inventory data and transportation resource data for each warehouse; The inventory data includes: inventory types, inventory quantities, and unit volume of each item; the transportation resource data includes: vehicle resources, vehicle capacity, vehicle load capacity, and loading and unloading efficiency; and the order volume, order weight, and estimated loading and unloading time are calculated based on the target order information and inventory data.

3. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm according to claim 1, characterized in that: The feasibility assessment for single-warehouse fulfillment includes: Based on the goods type and quantity of the target order information, it is matched with the inventory goods quantity and inventory goods type in the information of each warehouse site. If there is a single warehouse that meets all the conditions of the target order information, then all single warehouses that meet the conditions are selected as candidate warehouses and S3 is executed. If there are no single warehouses that meet the conditions, but the total inventory goods quantity and type of multiple single warehouses meet all the matching conditions of the target order information, then each single warehouse that meets the conditions is selected as a centralization point candidate warehouse and proceeds to S4.

4. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm as described in claim 1, characterized in that: The construction of the transport cost function includes: First, calculate the transshipment cost for each warehouse that needs to be transferred to the designated central warehouse. Then, sum the costs of all individual transshipment warehouses to obtain the transshipment cost function. The cost of an individual transshipment warehouse consists of three parts: distance cost, time cost, and cargo characteristics. A distance cost function is constructed based on transportation distance parameters. This function combines basic linear coefficients and nonlinear penalty coefficients to obtain the adjusted basic distance cost and the converged long-distance implicit cost. Then, the toll fees generated during the journey are statistically analyzed, and the distance cost function is obtained by combining them. A time cost function is constructed based on timeliness parameters. This function combines the basic time cost, the emergency transportation coefficient, and the time sensitivity coefficient to obtain the basic time cost and the cargo characteristic cost adjusted according to the characteristics of the cargo itself. The combination of these two factors yields the time cost function. By calculating the cargo density factor and the volume of transshipped cargo, the sum and value of the above two types of functions are adjusted to obtain the comprehensive transshipment cost function after the influence of cargo characteristics.

5. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm according to claim 1, characterized in that: The construction of the risk prediction function includes: Risk is quantified from two dimensions: delivery time risk and external environment risk. Delivery time risk is based on historical transportation task data of candidate central warehouses, analyzing the deviation between the actual and planned delivery times of past tasks. By statistically analyzing the deviation rate of all historical tasks of the warehouse and combining the task weights, a weighted average deviation value is calculated. The higher the deviation value, the greater the probability of delivery delays when the warehouse is used as a central warehouse. External environment risk dynamically integrates the impact of sudden external factors on transportation routes, including implementing weather and geographical location sensitivity, and improving the hidden risks brought about by the geographical location of border or port warehouses. The two types of risk values ​​are normalized and unified to the same dimension. The weighted average of historical deviations is directly mapped to the probability value. The external environmental risk is suppressed by the attenuation coefficient to suppress the excessive dominance of external risk. Combined with the adjustment of the geographical location sensitivity coefficient, it is converted into a standardized score. Based on business needs, the two types of risks are assigned weights, and the weighted sum is used to generate a comprehensive risk quantification value.

6. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm according to claim 1, characterized in that: The comprehensive early warning function for each candidate warehouse includes: The transshipment cost and risk prediction value of each candidate warehouse are standardized and mapped to the interval [0, 1]. Then, weights are assigned according to the importance of transshipment cost and risk prediction value in decision-making. The normalized transshipment cost and risk prediction value are weighted and summed. Finally, the comprehensive early warning value of all candidate concentration point warehouses is sorted from low to high, and the warehouse with the lowest early warning value is selected as the final concentration point.

7. The supply chain fulfillment path optimization and collaborative decision-making method based on AI algorithm according to claim 1, characterized in that: The multi-warehouse collaborative fulfillment path includes: Starting from a defined central warehouse and ending at the order delivery address, the optimal fulfillment path from the central warehouse to the delivery address is generated by calling the single warehouse optimal path generation method. Starting from each transshipment warehouse and ending at the central warehouse, the optimal transshipment route from each transshipment warehouse to the central warehouse is generated by calling the single warehouse optimal route generation method. The generated optimal transit route is combined with the optimal fulfillment route to form a complete multi-warehouse collaborative fulfillment route from the transit warehouse to the central warehouse and finally to the order delivery address.

Images