A supply chain processing method, an agent network and a system

Abstract

The application provides a supply chain processing method, an agent network and a system. The method is applied to an agent network, and the agent network includes N second agents of participants. The method includes the following steps: determining M candidate agents related to a demand intention input by a demand party from the N second agents based on the demand intention; M is less than or equal to N; combining and competing with the M candidate agents based on a resource contract of each candidate agent, determining a target participant combination of the candidate agents that pass the competition; and determining the target participant combination as a target supply chain that meets the demand intention of the demand party. The scheme realizes the determination of the target supply chain in the manner of the agent, improves the intelligent degree of the supply chain determination, and improves the efficiency of the determination of the target supply chain.

Description

Technical Field

[0001] This application relates to agent technology, and more particularly to a supply chain processing method, agent network, and system. Background Technology

[0002] Currently, product delivery relies on collaborative efforts within the supply chain. In current supply chain collaboration practices, products often require the joint efforts of multiple upstream and downstream entities to achieve complete delivery. Common technical approaches to determining a supply chain combination that meets product requirements include: bidding and selection based on pre-set templates, linear programming solutions relying on a central planning system, or multi-round manual negotiations and step-by-step confirmation through paper contracts.

[0003] However, these methods often have inherent problems such as information silos, process fragmentation, and delayed response. For example, bidding and selection rely on a static supplier list and cannot dynamically respond to changes in production capacity. The central optimizer needs to collect a large amount of raw data and assume that the behavior of each link is controllable, while manual negotiation faces repeated wrangling and low matching efficiency.

[0004] Overall, the relevant technologies often operate in a centralized or semi-manual manner, which is based on "collecting data first and then calculating". They are difficult to adapt to highly volatile and constrained supply chain environments and have significant drawbacks such as low efficiency, high cost and slow response to disturbances. Summary of the Invention

[0005] To address the aforementioned issues, this application provides at least one supply chain processing method, intelligent agent network, and system. This solution utilizes intelligent agents to determine the target supply chain, thereby improving the intelligence level and efficiency of supply chain determination.

[0006] The technical solution of this application is implemented as follows: Firstly, this application provides a supply chain processing method applied to an agent network, which includes N participating second agents. The method includes: determining M candidate agents related to the demand intent input by the demander; M being less than or equal to N; combining and electing the M candidate agents based on the resource contract of each candidate agent, and determining the combination of candidate agents that passes the election as the target participant combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used; and determining the target participant combination as the target supply chain that satisfies the demander's demand intent.

[0007] Secondly, this application provides an intelligent agent network, comprising: a first determining unit, configured to determine M candidate intelligent agents related to the demand intent from N second intelligent agents based on the demand intent input by the demander; M is less than or equal to N; the intelligent agent network includes N participating second intelligent agents; a processing unit, configured to combine and elect M candidate intelligent agents based on the resource contract of each candidate intelligent agent, and determine the elected candidate intelligent agent combination as the target participating agent combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used; and a second determining unit, configured to determine the target participating agent combination as the target supply chain that satisfies the demander's demand intent.

[0008] Thirdly, this application provides a supply chain processing system, comprising a first intelligent agent of the demand side and an intelligent agent network, the intelligent agent network comprising N second intelligent agents of participating parties; the first intelligent agent is used to: obtain the demand intent input by the demand side; the intelligent agent network is used to: based on the demand intent, determine M candidate intelligent agents related to the demand intent from the N second intelligent agents; M is less than or equal to N; based on the resource contract of each candidate intelligent agent, combine and elect the M candidate intelligent agents, and determine the combination of candidate intelligent agents that passes the election as the target participating party combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used; and the target participating party combination is determined as the target supply chain that satisfies the demand intent of the demand side.

[0009] Fourthly, this application also provides a storage medium storing a computer program or instructions that, when executed by a processor, implement any of the methods provided in the first aspect above.

[0010] Fifthly, this application also provides a computer program product comprising a computer program or instructions that, when executed by a processor, implement any of the methods provided in the first aspect above.

[0011] In this scheme, multiple second-level agents form an agent network. The target supply chain is determined through interactions between these agents, improving the intelligence and response speed of supply chain determination, increasing efficiency, and reducing costs. Furthermore, the fact that the target supply chain is determined by an agent network, rather than a single agent, enhances reliability through decentralization. Attached Figure Description

[0012] Figure 1 A schematic diagram of an optional supply chain scenario provided in an embodiment of this application; Figure 2 A schematic diagram of a first optional process for a supply chain processing method provided in an embodiment of this application; Figure 3 A second optional flowchart illustrating the supply chain processing method provided in this application embodiment; Figure 4 A schematic diagram of a third optional process for processing the supply chain provided in an embodiment of this application; Figure 5 A schematic diagram of a fourth optional process for processing the supply chain provided in an embodiment of this application; Figure 6 A fifth optional flowchart illustrating the supply chain processing method provided in the embodiments of this application; Figure 7 A sixth optional flowchart illustrating the supply chain processing method provided in this application embodiment; Figure 8 A seventh optional flowchart illustrating the supply chain processing method provided in this application embodiment; Figure 9 An eighth optional process diagram of the supply chain processing method provided in the embodiments of this application; Figure 10 A schematic diagram of an optional structure for the overall architecture of an autonomous intelligent agent ecosystem based on data asset flow, provided in an embodiment of this application; Figure 11 This is an optional flowchart illustrating the internal structure of the autonomous intelligent agent and the data asset contract generation process provided in an embodiment of this application. Figure 12 This is an optional flowchart illustrating the intent-driven consumption and value realization process of data assets, provided in an embodiment of this application. Figure 13 An optional flowchart illustrating the ecological emergence and self-evolution process provided in an embodiment of this application; Figure 14 This is a schematic diagram of an optional structure of the supply chain processing system provided in an embodiment of this application.

[0013] It should be noted that the terms "first" and "second" mentioned above are only used to distinguish between different options and do not represent the degree of superiority or inferiority of the options or their priority in the implementation process. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of the application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0015] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0016] In the following description, the terms "first," "second," and "third" are used only to distinguish different objects and do not represent a specific order of objects, nor are they constituting a chronological order. It is understood that "first," "second," and "third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0018] This application provides a supply chain processing method, system, storage medium, and program product. The following describes various embodiments of the supply chain processing method, system, storage medium, and program product provided in this application.

[0019] To make it easier to understand, let's first explain the application scenarios of the supply chain.

[0020] In one example, reference Figure 1 The scenario depicted, a supply chain, includes at least one demander 10 and multiple participants 20. Demander 10 is the entity providing the demand. Participants 20 are the entities providing resources to demander 10.

[0021] For example, in the electric vehicle supply chain scenario, demand side 10 can be a car manufacturer, and multiple participants 20 can include: battery manufacturer A, battery manufacturer B, battery manufacturer C, logistics service provider D, logistics service provider E, door manufacturer F, glass provider G, etc.

[0022] In a first aspect, embodiments of this application provide a supply chain processing method, which can be executed by an agent network in a supply chain processing system. The system includes a first agent of the demand side and an agent network, the agent network including second agents of N participating parties.

[0023] The first intelligent agent is used to analyze the requirements of the demand side. That is, an intelligent agent is deployed on the demand side, called the first intelligent agent.

[0024] The second agent is an agent deployed among the participating parties. Here, one agent is deployed for each participating party. Multiple second agents from multiple participating parties constitute an agent network.

[0025] An intelligent agent is an intelligent module capable of autonomously perceiving its environment, understanding its goals, planning, and taking action to accomplish complex tasks. A general-purpose intelligent agent system operates internally according to a closed-loop logic of "Perception-Planning-Action".

[0026] Each intelligent agent is assigned a role and function at the beginning, thus enabling it to perform specific tasks later.

[0027] In some embodiments, reference Figure 2 The processing methods for the supply chain, as shown, may include, but are not limited to, the following S201 to S203.

[0028] S201. Based on the demand intent input by the demand side, the agent network determines M candidate agents related to the demand intent from among N second agents.

[0029] Demand intent refers to the constraints imposed by the demand side on the supply chain. For example, demand intent could be "to ensure that the next-generation flagship product is always in stock during its initial launch season."

[0030] M is less than or equal to N.

[0031] The N second agents are the second agents corresponding to all participants, and the M candidate agents are the M candidate agents related to the demand intention. That is, the agents belonging to the participants related to the demand intention are the candidate agents.

[0032] For example, the N second intelligent agents could include: battery manufacturer A, battery manufacturer B, battery manufacturer C, logistics service provider D, logistics service provider E, car door manufacturer F, glass supplier G, etc. The M candidate intelligent agents could include: battery manufacturer A, battery manufacturer B, battery manufacturer C, logistics service provider D, logistics service provider E.

[0033] Once the first agent obtains the demand intent from the demander, it will quickly filter and match among the existing N second agents based on this intent. This usually involves a combination of intent parsing and capability matching: First, the demand intent is semantically decomposed to extract key elements (such as task type, required tools, domain knowledge, etc.); then, each second agent is searched; finally, M agents are selected as candidates, ensuring that these candidate agents have the core capabilities required to handle the demand, thereby laying the foundation for subsequent accurate task allocation or collaborative negotiation.

[0034] S202. Based on the resource contract of each candidate agent, the agent network combines and elects M candidate agents, and determines the combination of candidate agents that passes the election as the target participant combination.

[0035] Resource contracts include resource data and value rules; the value rules are the dynamic pricing method when resource data is used.

[0036] In this agent network, the resource contract for each candidate agent is a standardized digital contract, which mainly includes the following two core elements: Resource data refers to the tangible or intangible resources that the intelligent agent can provide, such as computing power, storage space, API access, specialized models (e.g., translation, recognition), sensor data streams, knowledge graph fragments, and physical execution capabilities (e.g., robotic arm control). Resource data clarifies what the intelligent agent "has" and "can provide."

[0037] Value rules: This specifically refers to the dynamic pricing method used by an intelligent agent when its resource data is used by other intelligent agents or systems. This means that the cost of resource usage is not fixed but fluctuates in real-time based on factors such as market supply and demand, task urgency, call frequency, data quality, energy consumption, and time of day. For example, an intelligent agent providing computing power may raise its price during peak periods and lower it during off-peak hours; a data acquisition agent may use different pricing curves depending on the real-time requirements of the requested data. Dynamic pricing gives resource contracts market sensitivity and economic game-theoretic capabilities.

[0038] In summary, the resource contract is a "quotation and capability specification" presented by candidate intelligent agents when participating in the election, and serves as the economic and technical basis for subsequent combination optimization and election decisions.

[0039] Based on the resource contract of each candidate agent, the agent network first arranges and combines the M candidate agents according to the structured decomposition of the required task to form a variety of possible collaborative combinations (for example, for the task of "image annotation + speech synthesis + result upload", try different annotation agents and synthesis agents to pair up). Then, an election mechanism is launched for each combination to compete, and the candidate agent combination that passes the election is determined as the target participant combination.

[0040] In short, the campaign rules and the desired intentions are aligned.

[0041] For example, it is possible to not only evaluate the complementarity of resource data and functional synergy among the agents within the combination, as well as whether there are resource conflicts (such as two agents competing for the same exclusive data source), but also, based on the dynamic pricing rules in their respective contracts, combined with factors such as the current market environment, task deadline, and system load, to calculate in real time the comprehensive utility value of each combination in executing the task, including the total expected cost, completion time, and success rate. For example, if the dynamic pricing value of one agent is low at this time, and the reliability of another agent is high but the cost is slightly higher, a multi-objective optimization model (such as weighted sum, Pareto front) or auction-style consensus algorithm can be used to select the agent combination. Finally, the candidate agent combination with the best comprehensive utility (such as the shortest time within the budget, or the lowest cost and the reliability standard) is selected as the target participant combination. This combination will be formally authorized to execute the required task, and the fee will be settled according to their respective dynamic pricing rules after the task is completed.

[0042] S203, the intelligent agent network identifies the target supply chain by combining the target participants to meet the demand intentions of the demanders.

[0043] After completing the combination and selection of candidate agents, the agent network will formally establish the final selected target participant combination as the target supply chain.

[0044] This means that the agents in this combination will be viewed as a dynamic, on-demand virtual chain from resource acquisition and processing to final delivery. Each agent plays a role in the supply chain (such as a resource provider, intermediate processor, or final delivery provider), and they collaborate to meet the original needs of the demand side. This supply chain is not fixed, but is generated temporarily for specific demand tasks. It has a clear starting node (demand input), intermediate nodes (the connection of resource and value rules between agents), and ending node (delivery of results to the demand side). It can also adaptively adjust according to environmental factors such as dynamic pricing and real-time load. Once the task is completed, the supply chain can be disbanded or retained as a template for future reuse, thus achieving the goal of "customizing an optimal collaborative chain for every demand moment".

[0045] In this scheme, the entire agent network adopts a decentralized multi-agent collaborative architecture: each second agent corresponds to an independent participant (such as a data source, computing power node, service module, etc.), and numerous second agents together constitute a dynamic, autonomous agent network. When determining the target supply chain that meets the demand intention, it does not rely on a single central node for global scheduling, but automatically completes the screening, combination, and utility evaluation of candidate agents through point-to-point interaction, election negotiation, and resource contract matching between agents, ultimately converging to the optimal combination of target participants.

[0046] This mechanism offers significant advantages: First, because the entire selection and selection process is completed autonomously and in a distributed manner by the intelligent agent network, it avoids the bottlenecks of human intervention or centralized decision-making, thereby greatly improving the intelligence and responsiveness of supply chain determination, and generating collaborative solutions for dynamic needs within milliseconds to seconds. Second, dynamic pricing and multi-objective optimization based on resource contracts enable the system to select the most suitable combination of intelligent agents at a lower overall cost (including monetary cost, time cost, and energy consumption), significantly improving the efficiency of supply chain determination. Finally, the decentralized decision-making method eliminates the risk of single points of failure. Even if some intelligent agents in the network fail or withdraw, the remaining intelligent agents can still quickly recombine and reach a consensus, thereby greatly enhancing the reliability and robustness of the system. In summary, this solution transforms the traditional static, centralized, and manually driven determination process of the supply chain into a dynamic, distributed, and autonomous intelligent agent collaboration process, realizing "on-demand, flexible, and autonomous" supply chain construction.

[0047] The following describes the process by which the agent network in S201 determines M candidate agents related to the demand intention from among N second agents based on the demand intention.

[0048] In some embodiments, the demand intent input by the demander can be obtained by the first agent of the participating party, and then processed by the first agent to obtain L resource demand lists. The L resource demand lists are then given to the agent network, and the agent network determines M candidate agents based on the L resource demand lists.

[0049] refer to Figure 3 The process may include, but is not limited to, S301 to S303 described below.

[0050] S301. The first intelligent agent performs task planning based on the demand intention and determines L sub-tasks.

[0051] L is less than or equal to M.

[0052] After receiving a high-level business intent, the first intelligent agent (i.e., the demand-side intelligent agent) will automatically decompose the intent into L interrelated, independently executable, and resource-quantifiable subtasks through its built-in task planning engine (usually combined with semantic parsing, rule reasoning, or optimization algorithms). Each subtask corresponds to a specific set of data asset requirements or resource requirements, thereby providing clear input for subsequent intelligent agent election and contract matching.

[0053] For example, a car manufacturer, as the first intelligent agent, has the following demand: "to obtain 5,000 sets of high energy density batteries in the next four weeks, with at least 3 suppliers, a total cost of no more than 8 million yuan, and to ensure supply resilience ≥99%."

[0054] Based on task planning, the agent may identify the following four sub-tasks: Sub-task 1 (procure 3000 battery sets, requiring suppliers with a credit rating ≥ 0.9 and located in East China); Sub-task 2 (procure 2000 battery sets, requiring suppliers with a credit rating ≥ 0.85 and located in South China); Sub-task 3 (coordinate logistics services to ensure batteries are delivered in three batches at the beginning of the 2nd, 3rd, and 4th weeks); Sub-task 4 (subscribe to "capacity fluctuation early warning data assets," updated weekly, for dynamic adjustment of subsequent procurement). These four sub-tasks will trigger corresponding data asset requirements, which will be executed by participating agents in the network through a competitive selection process.

[0055] S302, The first intelligent agent maps L subtasks into L resource requirement lists.

[0056] After completing the task planning for the demand intent and obtaining L sub-tasks, the first intelligent agent will further transform each sub-task into a structured resource requirement list. This list clearly describes the type of data assets required to execute the sub-task (such as "capacity commitment assets" and "logistics timeliness assets"), the specific quantity or capacity range, quality requirements (such as confidence level not lower than a certain threshold), spatiotemporal constraints (such as geographical location and delivery window), and acceptable cost or budget ceiling, so that the subsequent auction and negotiation rules can accurately match candidate participating intelligent agents based on these lists.

[0057] Example: Continuing from the previous first intelligent agent of the automotive OEM, its planned four sub-tasks are mapped to the following four resource requirement lists: Subtask 1 (Procurement of 3000 sets of batteries in East China) → Resource Requirement List A: Requires "Capacity Commitment Data Assets", quantity 3000 sets, confidence level ≥95%, supplier located in East China, budget unit price ≤1550 yuan / set, delivery window is before the end of the first week.

[0058] Subtask 2 (Procurement of 2000 sets of batteries in South China) → Resource Requirement List B: Requires "Capacity Commitment Data Assets", quantity 2000 sets, confidence level ≥90%, supplier located in South China, budget unit price ≤1480 yuan / set, delivery window within the second week.

[0059] Subtask 3 (Batch Logistics) → Resource Requirement List C: Requires "Logistics Timeliness Data Assets", to be transported in three batches with a total volume of 5,000 sets, requiring a 48-hour delivery rate of ≥99%, and a total budget of ≤200,000 yuan.

[0060] Subtask 4 (Capacity Fluctuation Early Warning) → Resource Requirement List D: Requires "Capacity Fluctuation Early Warning Data Asset", updated weekly, predicting the probability of capacity fluctuation in the coming week, with a confidence level of ≥85% and a budget unit price of ≤5000 yuan / time.

[0061] S303. Based on L resource requirement lists, the agent network identifies the second agents related to the L resource requirement lists as candidate agents from among N second agents, resulting in M ​​candidate agents.

[0062] After receiving L resource requirement lists generated by the first agent, the agent network initiates a round of relevance screening: for each resource requirement list describing the data asset type, geographical region, capability scope, and quality requirements, the registered nodes in the network (i.e., N second agents) quickly compare the data asset type with the metadata index (such as pre-published asset summaries and capability tags) using semantic matching algorithms (such as ontology-based asset category mapping); only those second agents that can provide at least one data asset contract that matches the requirements in the list and whose current status (such as online, unlocked, and meeting the reputation threshold) allows them to participate in the election will be retained.

[0063] Subsequently, the network deduplicates and merges the agents selected from all resource demand lists to form the final set of candidate agents, denoted as M (M≤N). For example, for the demand list of "3000 sets of battery production capacity in East China", the system will select the second agent that has all production capacity data assets covering East China and has available production capacity ≥3000 sets; for the "logistics timeliness" list, it will select logistics agents with corresponding transportation capabilities. The union of the two sets yields M candidate agents.

[0064] For example, suppose the first intelligent agent (automobile OEM) generates three resource demand lists: List 1: East China region, battery production capacity commitment data assets, quantity ≥ 3000 sets, confidence level ≥ 95%; List 2: South China region, battery production capacity commitment data assets, quantity ≥ 2000 sets, confidence level ≥ 90%; List 3: Nationwide, logistics timeliness data assets, transporting 5000 sets of batteries in three batches, with a 48-hour delivery rate ≥ 99%.

[0065] At this point, there are N=5 second agents in the agent network: A (East China Battery Factory): can provide capacity assets, region East China, current available capacity of 5000 sets; B (South China Battery Factory): can provide capacity assets, region South China, available capacity of 3000 sets; C (Central China Battery Factory): only covers the Central China region, cannot match East China or South China; D (Logistics Company): can provide logistics timeliness assets, nationwide network, delivery rate ≥99.5%; E (Data Service Provider): can provide capacity fluctuation early warning assets, but there are no early warning assets in this demand list.

[0066] The network filters data through semantic matching and metadata indexing: for list 1, A is matched (matches region and capacity); for list 2, B is matched (matches region and capacity); for list 3, D is matched (matches logistics requirements); C is excluded due to region mismatch, and E is excluded due to asset type mismatch.

[0067] After deduplication and merging, we get M=3 candidate agents: A, B, and D.

[0068] In this way, by automatically breaking down high-order intents into sub-tasks and resource requirement lists, and quickly filtering relevant candidate agents based on semantic matching and metadata indexing, the response speed and resource matching accuracy of supply chain collaboration are significantly improved. Compared with traditional manual or centralized methods, it avoids full broadcasting and information overload, and reduces communication and computing overhead.

[0069] The following section explains the process by which the agent network in S202 combines and elects M candidate agents based on the resource contract of each candidate agent, and determines the combination of candidate agents that passes the election as the target participant combination.

[0070] In some embodiments, reference Figure 4 The process may include, but is not limited to, S401 to S403 described below.

[0071] S401, The agent network obtains the resource contracts of the participants among the M candidate agents.

[0072] Resource contracts include resource data and value rules; the value rules are the dynamic pricing method when resource data is used.

[0073] After identifying M candidate agents, the agent network requests or receives resource contracts submitted by each agent. These contracts contain two core components: first, resource data, which is the specific data assets or capability commitments the agent is willing to provide (e.g., "supply capacity commitment data assets for the next 72 hours," typically processed for privacy or aggregated to avoid exposing the original data); and second, value rules, which define the dynamic pricing method for the resource data when it is actually used—the pricing can change in real time based on usage, time window, real-time confidence level, market supply and demand index, or the urgency of the demand, rather than a fixed unit price.

[0074] For example, in a resource contract submitted by a logistics agent, the resource data is "cold chain transportation timeliness data asset (guaranteed delivery within 24 hours, confidence level ≥ 95%)", and the value rule is "basic freight fee of 2000 yuan / time; if the demander requests to increase the confidence level to 98%, the total price will increase by 15%; if it is called during peak hours (17:00-19:00), an additional 10% dynamic congestion surcharge will be charged per order; the same demander will enjoy a 10% tiered discount starting from the 5th call in a single month". In this way, when the first agent (such as a fresh food e-commerce platform) actually uses the data asset, the cost will be automatically calculated and adjusted according to the conditions at that time.

[0075] S402, the agent network determines the target election rules based on demand intentions.

[0076] Demand intent refers to the high-level business objectives proposed by the demand side (the first intelligent agent), such as "ensuring the resilience of the supply of a certain key material." It needs to be automatically broken down into demands for specific data assets.

[0077] The intelligent agent network automatically selects or configures a set of target election rules based on the specific characteristics of the demand intent (such as the type of asset required, timeliness requirements, risk appetite, budget constraints, etc.). For example, if the intent emphasizes "low cost priority," the weight of the auction rules is increased; if it emphasizes "supply resilience," the threshold of the reputation rules is strengthened and risk premiums are allowed in the negotiation process; if the intent is extremely sensitive to delivery time, the negotiation rounds are shortened and a fast combined auction is adopted. These rules are solidified in the form of smart contracts or on-chain parameters to ensure that all candidate intelligent agents are aware of the evaluation criteria.

[0078] S403. The agent network combines and elects resource contracts of participants among M candidate agents through target election rules, and determines the combination of candidate agents that passes the election as the target participant combination.

[0079] Target election rules are a comprehensive mechanism that is dynamically selected to achieve demand intentions. They typically combine at least one of auction rules, negotiation rules, and reputation rules to balance intention satisfaction, participant benefits, and supply chain efficiency.

[0080] Resource contracts are standardized digital commitments submitted by candidate agents, which include resource data (such as supply capacity data assets) and value rules (dynamic pricing methods, such as prices that fluctuate with confidence level, usage, or time period).

[0081] Target participant portfolio: The set of all winning candidate agents selected by the election rules, which work together to satisfy the intent of the demander.

[0082] The network takes resource contracts submitted by M candidate agents as input and executes the combination and election process defined by the target election rules. Specifically: First, sealed bidding and combination optimization are performed through auction rules to select a preliminary subset of contracts that meet the demand vector and have high overall cost-effectiveness; then, under negotiation rules, the demand side and agents within the candidate combinations are allowed to make a limited number of rounds of alternating proposals and counter-proposals on terms such as quantity, confidence level, and price to achieve Pareto improvement; simultaneously, reputation rules dynamically adjust the scoring weights and collateral requirements of each agent and eliminate participants with insufficient reputation. Finally, the agents corresponding to the selected contract subset are determined as the target participant combination, and the network automatically generates multi-party smart contracts to lock commitments and collateral, completing decentralized resource allocation.

[0083] In this way, by determining the target participant combination through the target election rules, it can be flexibly adjusted based on the needs and intentions, thus improving the accuracy of the target participant combination.

[0084] In some embodiments, the target election rule combines auction rules, negotiation rules, and reputation rules, and is used to achieve a dynamic equilibrium of demander intent satisfaction, participant revenue, and supply chain efficiency.

[0085] The target election rules, by integrating auction rules, negotiation rules, and reputation rules, dynamically balance the core demands of the three parties in supply chain collaboration: Auction rules achieve efficient initial resource matching through competitive pricing and portfolio optimization, prioritizing the cost and quantity intentions of demanders; Negotiation rules allow for limited rounds of multi-attribute negotiation, enabling demanders to adjust terms according to actual needs (such as increasing confidence or the number of suppliers), while ensuring that participants receive a reasonable premium, thereby balancing intention satisfaction and individual gains; Reputation rules continuously adjust the election weights and collateral requirements of each agent based on historical performance and asset quality, giving high-reputation parties more opportunities and forming long-term cooperation incentives, thereby reducing default risk and monitoring costs, and improving overall supply chain efficiency. These three elements are mutually reinforcing and nested—auctions generate benchmark combinations, negotiation optimizes the Pareto boundary, and reputation corrects subsequent election thresholds—ultimately causing the system to spontaneously converge to an acceptable level of satisfaction for demanders, considerable gains for participants, and a highly efficient non-cooperative game equilibrium across the entire network.

[0086] In other implementations, the target selection rules also include risk diversification rules, timeliness priority rules, and sustainability rules to further enhance the robustness of supply chain collaboration.

[0087] Risk diversification rule: During portfolio optimization, the geographical region, industry category, or technology route of the agent is scored for diversification, favoring candidate combinations with broader sources and higher redundancy. Data effect: Quantifiable as a reduction in supply chain disruption risk (e.g., after simulating an attack on a single region, the capacity loss ratio drops from 40% to 15%), and resilience indicators (e.g., recovery time) are shortened by more than 30%.

[0088] Timeliness Priority Rule: Based on the response time, delivery speed, or data update frequency promised in the resource contract, the faster the response, the higher the score weight, but this must be weighed against cost or confidence level. Data Effects: Overall order delivery cycle is shortened by an average of 20% to 35%, and the matching success rate of urgent intentions is increased by more than 50%.

[0089] Sustainability Rules: Environmental indicators such as carbon emissions, energy consumption, or waste utilization rates disclosed by smart agents will be included in the scoring, with additional points or subsidies awarded for low-carbon contracts. Data Effects: The overall carbon footprint of the supply chain will decrease by 15% to 25%, compliance costs for meeting regulatory requirements will be reduced, and long-term benefits from enhanced brand reputation will be achieved.

[0090] These rules can be computed in parallel with auction, negotiation, and reputation rules (e.g., through weighted summation or Pareto ranking), enabling the election results to spontaneously exhibit macroeconomic characteristics such as resilience, rapid response, and low emissions, in addition to satisfying efficiency and profitability.

[0091] The following section explains the process by which the agent network in S202 combines and elects M candidate agents based on the resource contract of each candidate agent, and determines the combination of candidate agents that passes the election as the target participant combination.

[0092] In some embodiments, reference Figure 5 The process may include, but is not limited to, S501 to S504 described below.

[0093] S501. The intelligent network processes the resource contracts of the participants among the M candidate intelligent agents through auction rules, and determines multiple combinations of candidate intelligent agents and the supply efficiency under each combination.

[0094] Each combination can meet the intended requirements.

[0095] The agent network first optimizes all possible subsets of contracts based on resource contracts (including resource data and dynamic pricing rules) submitted by M candidate agents using a combined auction mechanism (such as multi-attribute sealed bids or VCG auctions). Under the premise of satisfying demand intentions (such as hard constraints like total quantity, geographical coverage, and confidence level), the auction rules select multiple feasible combinations and calculate the supply efficiency of each combination—usually defined as the ratio of the committed total quantity of contracts within the combination to the total cost, or a comprehensive performance index incorporating factors such as confidence level and response speed. This step ensures the comparability of candidate solutions in terms of pure efficiency in resource matching, providing a foundation for subsequent multi-objective decision-making.

[0096] S502. The intelligent network processes the resource contracts of the participants among the M candidate intelligent agents through negotiation rules, and determines the participants' revenue under each combination.

[0097] For each feasible combination generated during the auction phase, the agent network initiates negotiation rules, allowing candidate agents within the combination to negotiate with the demand side on price, quantity, additional terms, and other attributes in a limited number of rounds. The negotiation process can employ alternating proposals or game-theoretic bargaining models (such as the Rubinstein model), ultimately reaching a mutually acceptable payment scheme. This scheme determines the actual revenue (such as token income, reputation rewards, or future priority) that each participating agent can obtain under that combination. Because the bargaining power of the same agent differs across combinations (for example, an agent is a scarce resource in combination A but can be replaced in combination B), its revenue will dynamically change, reflecting the distribution of benefits across different combinations.

[0098] S503, the intelligent network determines the demander's intention and satisfaction through reputation rules.

[0099] Reputation rules do not directly address contract prices or quantities; instead, they assess the overall degree to which each feasible combination satisfies the higher-order intentions of the demander. Specifically, the system calculates the combination's intention satisfaction score based on the weights of the demander's stated intentions (e.g., "resilience first," "cost second," "delivery timeliness third") and the historical reputation scores of each agent (performance rate, data quality, cooperation resilience, etc.). For example, if the demander requires "ensuring supply resilience," the reputation rules will focus on the default history and collateral adequacy of the agents in the combination; agents with higher reputations will receive higher scores. Simultaneously, implicit indicators such as supplier diversification and redundancy in the combination will also translate into higher satisfaction. This step ensures that the final decision relies not only on economic efficiency but also incorporates trust and risk appetite.

[0100] S504. The intelligent network determines the target participant combination based on the supply efficiency under each combination, the participant benefits under each combination, and the demander's intention satisfaction.

[0101] The agent network uses three dimensions for each feasible combination: supply efficiency (provided by auction rules), participant revenue (provided by negotiation rules, which can be summarized as total revenue or revenue Gini coefficient), and demander intention satisfaction (provided by reputation rules). It employs multi-objective decision-making methods (such as weighted summation, Pareto ranking, or analytic hierarchy process) for comprehensive scoring. The system can dynamically adjust the weights of each dimension based on the demanders' preset preferences (e.g., prioritizing satisfaction and then balancing efficiency and revenue), ultimately selecting the combination with the highest comprehensive score as the target participant combination. If multiple incomparable (Pareto front) combinations exist, a lightweight vote or randomized selection can be added to break the tie. This step achieves a dynamic equilibrium of the three objectives, ensuring that the election result neither simply pursues the lowest cost nor blindly chases high reputation, but rather spontaneously generates decisions that align with the overall interests.

[0102] In this way, through the hierarchical collaboration of auctions, negotiations, and reputation rules, multiple feasible combinations are efficiently selected from a massive pool of candidate agents, the supply efficiency and participant benefits of each combination are quantified, and multi-objective comprehensive decision-making is achieved by integrating the satisfaction of demanders. Compared with traditional centralized or single-dimensional optimization methods, this mechanism does not require the exposure of raw data and can simultaneously take into account the economy (efficiency) of resource matching, the fairness (benefit) of benefit distribution, and the robustness (satisfaction) of long-term cooperation. Through decentralized combination elections, it spontaneously converges to the Pareto optimal target participant combination, thereby significantly improving the supply chain's response speed, risk resistance, and overall collaborative efficiency in dynamic environments.

[0103] In some embodiments, reference Figure 6 As shown in S202, the agent network combines and elects M candidate agents based on the resource contract of each candidate agent. The process of determining the combination of candidate agents that has passed the election as the target participant combination may include, but is not limited to, the following S601 and S602.

[0104] S601. Through each second agent in the agent network, the resource contracts of the M candidate agents are combined and contested; the combination of participants supported by each second agent is obtained.

[0105] In the agent network, each second agent (i.e., a candidate participant) acts as an independent decision-making node, running target election rules locally based on its own information (such as publicly available resource contracts of all M candidate agents, historical reputation data, and network broadcast demand intentions). These rules include sub-mechanisms such as auctions, bargaining, and reputation management. Each second agent optimizes the combination of M resource contracts according to its own interests (e.g., maximizing its own probability of being selected, or maximizing the overall benefit of the combination), generating a participant combination it "supports." Due to differences in local information and interest functions among agents, their resulting combinations may differ—some favor high-reputation combinations, some favor low-cost combinations, and some prioritize their own inclusion.

[0106] S602. Determine the target participant combination based on the participant combination supported by each second agent in the agent network.

[0107] The agent network collects the "supported combinations" submitted by each second agent and merges these local candidate combinations using a decentralized consensus protocol (such as weighted voting, Byzantine fault-tolerant aggregation, or median mechanism). The weights can be dynamically adjusted based on each second agent's reputation score or its contribution in historical elections. The system ultimately outputs a globally optimal or majority-consensus target combination, which typically lies on the Pareto boundary of all locally supported combinations, thus avoiding single-point computational bias or malicious manipulation. If initial consensus cannot be reached, an additional round of coordination negotiation or randomization to break a tie can be triggered.

[0108] This method distributes the decision-making pressure of portfolio elections among each participating agent, eliminating the need for a centralized optimizer or trusted third party. Each agent independently computes based on local information, and the results are then aggregated into a global portfolio through a consensus network. This significantly improves the system's fault tolerance, Byzantine attack resistance, and scalability (because the computational load is distributed). Furthermore, each agent submits honest support portfolios to maximize its own interests, which in turn drives the final result towards global Pareto optimality. This achieves the ideal characteristic of "emergent self-interest and global collaboration" in a decentralized environment, effectively avoiding single points of failure and black-box decision-making problems.

[0109] The following describes the process in S601 of combining and electing resource contracts for M candidate agents to obtain the combination of participants supported by each second agent.

[0110] refer to Figure 7 The process may include, but is not limited to, S701 to S704 described below.

[0111] S701. Based on the resource contract of each candidate intelligent agent, determine the function score, price score, reputation score and collaborative efficiency score of each participant.

[0112] Each second agent (acting as an independent decision-making node) reads the resource contracts submitted by all candidate agents and extracts or calculates individual scores across four dimensions. Functionality Score: Measures the degree of matching between the resource data provided by the participant and the corresponding sub-tasks in the demand intent, such as capacity quantity, geographical coverage, and whether the confidence level meets the requirements. Price Score: Based on the value rules (dynamic pricing method) in the resource contract and current market conditions, it is converted into competitiveness relative to the budget; generally, the lower the price or the higher the cost-effectiveness, the higher the score. Reputation Score: Derived from historical feedback data maintained by the agent network, including fulfillment rate, data quality error, and cooperation resilience, and normalized after exponential decay or Bayesian updates. Collaboration Efficiency Score: Measures the participant's historical performance when cooperating with other agents, such as synchronization delays, information transmission losses, or the success rate of joint fulfillment in past joint deliveries; also dynamically calculated based on historical feedback.

[0113] S702. Calculate the functional score, price score, reputation score, and collaborative efficiency score for each combination method based on the functional score, price score, reputation score, and collaborative efficiency score of each participant.

[0114] Reputation score and collaboration efficiency score are determined based on historical feedback from participants.

[0115] After generating multiple feasible combinations (each combination includes several participants and the overall combination satisfies the hard constraint of the desired intent), the second agent aggregates four combination-level scores for each combination based on the individual scores of the participants.

[0116] The combined functional score can be obtained by weighting the functional scores of each participant in the combination (with the weights being commitments or confidence levels), reflecting the overall capability satisfaction.

[0117] The combined price score is obtained by summing the dynamic pricing of each participant according to the actual usage within the combination, and then comparing it with the budget to obtain a standardized score.

[0118] Combined reputation score: Common methods include taking the lowest reputation score (weakest link principle), weighted average or product, emphasizing the impact of the least reliable node in the combination on the overall trust.

[0119] Combined Collaboration Efficiency Score: This score assesses the smoothness of collaboration within the combined entity based on the average of historical pairwise collaboration efficiencies among the participants or the total delay of the minimum spanning tree.

[0120] Both the reputation score and the collaborative efficiency score rely on historical feedback data stored in the agent network to ensure that the evaluation is objective and timely.

[0121] S703. Determine the total score for each combination method based on the function score, price score, reputation score, and collaborative efficiency score of each combination method.

[0122] The second agent, based on its own preferences or the implicit weights in the demander's intentions (e.g., increasing the reputation score weight for "resilience priority," and increasing the price score weight for "cost sensitivity"), merges the four combined-level scores into a single total score through linear weighting, multiplicative normalization, or the analytic hierarchy process (AHP). The formula can be expressed as: The total score is calculated as follows: α1·Function Score + β1·Price Score + γ1·Reputation Score + δ1·Synergy Efficiency Score, where α1 + β1 + γ1 + δ1 = 1. α1, β1, γ1, and δ1 can be dynamically adjusted according to the intended demand, representing the weighting coefficients of the function score, price score, reputation score, and synergy efficiency score, respectively. If some indicators are incommensurable, Pareto ranking can be used to first screen for non-inferior combinations, and then the best combination within the boundary is selected based on the total score.

[0123] S704. Based on the total score of each combination, determine the combination of participants supported by the second agent.

[0124] The second agent sorts all feasible combinations from highest to lowest total score and selects the highest-scoring combination as its "supported participant combination." If multiple agents have the same highest score, randomization can be applied or the "minimum maximum regret" rule can be used to break the tie. This supported combination is then signed by the second agent and submitted to the consensus layer of the agent network for subsequent global fusion. By independently calculating their own optimal combinations, each second agent, while pursuing local interests (such as its own selection probability or combination efficiency), also provides the network with a diversified decision-making perspective.

[0125] This method decomposes the complex multi-objective combination election into four interpretable steps: individual scoring, combination aggregation, weighted summation, and optimal selection. This allows each second agent to independently and transparently compute its own supported optimal combination using historical feedback data (reputation and collaborative efficiency) and current resource contracts. Because the four dimensions of function, price, reputation, and collaboration mutually constrain each other, the ultimately supported combination naturally balances capability, cost, trust, and smooth collaboration, avoiding biases caused by single indicators. Furthermore, all computations are based on localized rules and public data, eliminating the need for centralized scheduling. This significantly improves the robustness, resistance to manipulation, and scalability of decision-making, providing an efficient, fair, and auditable election foundation for autonomous collaboration in decentralized supply chains.

[0126] The method provided in this application embodiment can also update the collaborative efficiency score and reputation score of the participants based on the current data, for reference in subsequent supply chain determination.

[0127] refer to Figure 8 The process may include, but is not limited to, S801 to S803 described below.

[0128] S801, When the intelligent agent network provides services according to the combination of target participants in the target supply chain, the feedback data of each participant includes: performance credit score, data quality score, value creation score, and collaboration stability score.

[0129] Once the target participant group (i.e. the set of successfully elected agents) actually provides services according to the resource contract, the agent network will collect the posterior performance data of each participant in this collaboration and form a feedback score in four dimensions.

[0130] Performance Credit Score: Measures whether the participant has delivered in accordance with the contractual commitments in terms of quantity, quality, and time (e.g., the ratio of actual delivery to committed quantity, delay time, etc.).

[0131] Data quality score: assesses the degree of deviation between the data assets it provides (such as supply capacity signals) and the actual situation, provided by a third-party oracle or post-audit.

[0132] Value Creation Score: Reflects the actual contribution of the participant's services to the overall intent of the demander (such as cost savings and resilience improvement), calculated by comparing the system performance difference with and without the participant.

[0133] Collaborative stability score: measures the stability of the participant's behavior in the process of interacting with other intelligent agents, such as fluctuations in information transmission delays, frequency of abnormal exits, and degree of cooperation in collaborative decision-making.

[0134] These feedback data are generated by demanders, other participants, or automated auditing mechanisms, and are written into the network ledger after being verified through consensus.

[0135] S802, the intelligent agent network updates the credit score of participants based on their performance credit score, data quality score, and value creation score.

[0136] The agent network employs a weighted moving average or Bayesian update mechanism to integrate each participant's performance credit score, data quality score, and value creation score obtained in the current round with their historical reputation score to arrive at a new reputation score. For example: The new reputation score is calculated as follows: New Reputation Score = λ × Historical Reputation Score + (1-λ) × (α²·Performance Credit Score + β²·Data Quality Score + γ²·Value Creation Score), where α² + β² + γ² = 1, λ is the forgetting factor (typically 0.7~0.9), and α², β², and γ² represent the weighting coefficients of the performance credit score, data quality score, and value creation score, respectively. This design allows the reputation score to comprehensively reflect the reliability (performance), honesty (data quality), and actual contribution (value creation) of the participants, avoiding bias from a single indicator. The updated reputation score will directly affect the reputation weight, collateral requirements, and entry threshold in subsequent election rules.

[0137] S803, the agent network updates the collaborative efficiency score of the participants based on the collaborative stability score.

[0138] The collaborative efficiency score specifically measures the quality of interaction among participants in multi-agent collaborative scenarios. It is maintained separately from the reputation score to avoid conflating individual reliability with team fit. Based on the collaborative stability score obtained in the current round, combined with historical collaboration records (e.g., historical stability scores with other specific agent combinations), the agent network updates the participant's collaborative efficiency score using an exponential moving average or a two-way reputation propagation algorithm. Furthermore, the network can maintain a collaborative efficiency matrix for agent pairs; when two participants jointly serve multiple times, their pairing efficiency score is individually reinforced. The updated collaborative efficiency score is used in subsequent elections to evaluate the smoothness of collaboration within the combination (e.g., calculating the combination's collaborative efficiency score), thereby guiding the network to prioritize agent combinations that are both reliable and compatible.

[0139] This feedback update mechanism decomposes the post-hoc performance of supply chain collaboration into four independent dimensions: performance credit, data quality, value creation, and collaboration stability. These dimensions are then mapped to the dynamic updates of reputation scores and collaboration efficiency scores, achieving decoupled evaluation of the "individual reliability" and "team suitability" of participants. Through weighted moving averages and historical data fusion, the system can adaptively track changes in participant behavior, quickly penalizing malicious or low-quality nodes while incentivizing long-term stable collaborative behavior. This provides a real-time and accurate credit foundation for decentralized target election rules, enabling subsequent auction, negotiation, and reputation rules to continuously optimize the selection quality of target participant combinations based on constantly evolving reputation and collaboration efficiency data. Ultimately, this drives the entire supply chain ecosystem to autonomously evolve towards greater efficiency, reliability, and resilience.

[0140] The method provided in this application embodiment may further include the process of each smart agent issuing a resource contract for its corresponding participant.

[0141] refer to Figure 9 The content shown is executed for each second agent in the agent network, and the following steps S901 to S905 are performed.

[0142] S901. Obtain the original resource data, inventory data, and historical transaction data of the participants corresponding to the second intelligent agent.

[0143] Each second agent (corresponding to a participant in the supply chain, such as a manufacturer or logistics provider) first collects three types of basic information from local data assets or off-chain systems: raw resource data (such as equipment capacity, man-hours, and types of raw materials), inventory data (current available inventory, in-transit inventory, and safety stock threshold), and historical transaction data (transaction prices of past contracts, actual delivery volume, and credit records of counterparties).

[0144] S902. Based on the original resource data, inventory data, and historical transaction data, generate resource forecast data; the resource forecast data is used to characterize the resource data that the forecasting participants can provide.

[0145] The second agent uses built-in predictive models (such as time series analysis, machine learning regression, or federated learning-based joint prediction) to fuse raw resource data, inventory data, and historical transaction data to generate resource forecast data. This forecast data characterizes the resource capacity that participants can stably provide within a specific future time window (such as the next 72 hours, next week, or next month), and typically includes the predicted supply, confidence interval, and risk factors that may affect supply (such as equipment maintenance plans). The resource forecast data is both a quantitative expression of its own capabilities and the basis for commitments in subsequent contracts.

[0146] S903, Determine the value rules for participating parties.

[0147] The value rule defines how to dynamically price the resource data when an external agent uses it. The second agent develops a programmable pricing function based on its own cost structure (production costs, storage costs), market supply and demand expectations (such as historical transaction price trends), and risk appetite (such as additional compensation requirements for emergency calls). Common value rules include: a base unit price + confidence premium (an additional price is charged if the demander requests higher confidence) + time surcharge (charged during peak hours) + bulk discount (discounts for large orders), etc. The value rule can be partially disclosed to attract bidders.

[0148] S904. Organize the predicted resource data and value rules according to the set format to obtain the resource contract of the participants of the second intelligent agent.

[0149] The second agent encapsulates the generated resource forecast data (including predicted supply, time windows, confidence levels, etc.) and the determined value rules (dynamic pricing logic) in a structured manner according to a predefined standard format (such as JSON or smart contract templates) of the agent network, forming a complete resource contract. The contract may also include clauses such as validity period, collateral requirements, and liability for breach of contract to ensure machine readability and automatic execution capabilities. This step is equivalent to transforming private capabilities into tradable data assets.

[0150] S905, Publish resource contracts to the agent network.

[0151] The second agent signs the encapsulated resource contract and broadcasts it to the agent network (e.g., to a distributed ledger or message bus). Other agents in the network (including demanders and verification nodes) can retrieve and verify the contract's integrity and validity (e.g., correct signature, sufficient collateral). Once published, the contract enters the bidding pool for use in auctions, negotiations, and other target bidding rules. Agents can update the contract at any time (e.g., adjust prices), but must ensure that historical versions are traceable to prevent malicious modification.

[0152] This method enables supply chain participants to transform private data (raw resources, inventory, historical transactions) into standardized resource forecasting data through internal forecasting models, bind dynamic value rules, and ultimately encapsulate it into publishable, searchable, and programmable resource contracts. The entire process does not require exposing the raw data, but only outputs value signals with confidence intervals. Thus, while protecting the privacy of data assets, it realizes the transformation from "passively responding to inquiries" to "actively publishing asset capabilities," providing a unified, transparent, and highly liquid data element foundation for subsequent decentralized elections, pricing, and trust assessments.

[0153] The following example illustrates the supply chain processing solution.

[0154] Current supply chain management and data asset applications face a dual bottleneck: "fractured value loop" and "rigid collaboration paradigm". 1. Closed path to realize the value of data assets: Existing technologies mainly use data assets for internal process optimization (such as demand forecasting and inventory control). Their value creation is limited to the enterprise and lacks a market mechanism to enable them to circulate securely and reliably within the supply chain network and generate cross-organizational value-added, making it difficult to capitalize data assets.

[0155] 2. Collaboration relies on centralized planning and static rules: Existing collaboration models depend on plans or fixed contracts issued by a central platform, which cannot adapt to dynamic and changing environments. Participants lack economic incentives to share high-value data assets, resulting in insufficient depth and flexibility of collaboration, essentially remaining weak connections between "data silos."

[0156] 3. The system lacks adaptive and evolutionary capabilities: The rules and models of the existing system are pre-set and static, and it cannot learn and optimize the collaborative rules autonomously from dynamic interactions, nor can it reconstruct the collaborative mode autonomously when business goals change. It appears fragile and sluggish when faced with sudden disturbances.

[0157] 4. Technical solutions focus on local optimization and neglect ecosystem building: Most solutions focus on algorithm-level improvements (such as more accurate prediction models) or simply add blockchain and other technologies to achieve information storage, failing to design from the fundamental level of building an incentive-compatible and self-evolving data asset value ecosystem, thus showing limited creativity.

[0158] Therefore, there is an urgent need for a new paradigm that can activate the cross-organizational circulation of data assets, drive autonomous collaboration through market mechanisms, and possess ecosystem-level evolution capabilities.

[0159] This embodiment aims to overcome the shortcomings of existing technologies, such as high barriers to data asset circulation, low levels of automation and intelligence in supply chain collaboration, and a lack of self-optimization capabilities. It provides a method and system for the collaborative evolution of supply chain intentions based on data asset flows and an autonomous intelligent agent ecosystem. This establishes a market-based pricing and circulation mechanism for data assets, achieves business intention-driven automated collaboration, and ultimately forms an intelligent supply chain organism capable of continuous learning, adaptation, and evolution.

[0160] This embodiment involves constructing a data asset-driven autonomous collaborative ecosystem for the supply chain. In this ecosystem, each supply chain entity is abstracted as an autonomous intelligent agent encapsulating its private data assets and decision-making logic. Agents do not directly exchange raw data; instead, they exchange privacy-processed value signals (such as "supply capacity commitment data assets" and "demand intention data assets") through standardized, tradable data asset contracts. The entire system operates under high-level business intents (such as "ensuring the resilience of a key material supply"), which are automatically broken down into demands for a series of data assets and dynamically matched and negotiated through a decentralized value consensus network. The macro-level supply chain optimization (such as improved efficiency and reduced risk) is not generated through central computation but rather emerges spontaneously as all agents interact to maximize the value of their own data assets. This system achieves three fundamental transformations: data assets from static resources to tradable production factors; collaboration from pre-defined processes to intent-driven processes; and the ecosystem from mechanical execution to autonomous evolution.

[0161] This embodiment constructs a four-layer autonomous value cycle system: "autonomous intelligent agents encapsulate data assets - intent drives the generation of asset demand - value networks facilitate asset transactions - emergent result feedback drives ecological evolution".

[0162] 1. Autonomous Intelligent Agent and Data Asset Encapsulation Layer: Each supply chain participant (enterprise, warehouse, etc.) is instantiated as an autonomous intelligent agent. The agent encapsulates its core private data assets (such as production capacity, inventory, and quality data) into a standardized data asset contract. This contract defines the right to use the data assets, the method of value measurement (such as fixed fees and performance sharing), and access conditions, without exposing the raw data.

[0163] 2. Intent Resolution and Asset Demand Mapping Layer: Users or the system submit high-level business intents (such as "minimize specific supply chain disruption risks"). The intent resolution engine leverages domain knowledge to automatically deconstruct abstract intents into a series of specific, actionable tasks, and further maps these tasks into a list of consumption demands for one or more data asset contracts in the network.

[0164] 3. Value Consensus Network and Dynamic Market Layer: This is a decentralized network supporting the publication, discovery, valuation, and trading of data asset contracts. Smart agents can publish their supply intentions for data asset contracts (e.g., "Selling a visual insight into production capacity for the next week"). The network forms a dynamic price and reputation system based on game theory mechanisms and historical performance records. Intention-driven demand lists automatically match the best supply within this network, completing transactions through smart contracts.

[0165] 4. Emergent Collaboration and Ecosystem Evolution Layer: Countless intelligent agents interact and compete within the value network based on their own interests (profiting from trading data assets or achieving business goals). This leads to the spontaneous emergence of efficient resource allocation models on a macro level (such as natural inventory balancing and accelerated emergency response). Simultaneously, the network continuously analyzes successful collaborative models, automatically discovering and suggesting new, more efficient data asset combinations or trading rules. These are implemented through community governance, thereby achieving the self-evolution of ecosystem rules.

[0166] The core of this embodiment includes, but is not limited to: 1. An "autonomous intelligent agent" was constructed as the basic unit for the value-creating and game-playing of data assets.

[0167] Value unlocking under privacy protection: Each intelligent agent is the absolute sovereign of its internal data assets, providing processed value services (such as insights, commitments, and predictions) to the outside world through data asset contracts, rather than raw data. This fundamentally solves the problems of data privacy and trade secrets, and lays the foundation for large-scale trusted collaboration.

[0168] Compared with existing technologies: Unlike traditional solutions where data needs to be exchanged in a centralized or standardized manner, this invention achieves "data moving without moving its value" through intelligent agent encapsulation and contractual interfaces, which is a fundamental innovation in architectural philosophy.

[0169] 2. A new collaborative paradigm of "intent-driven - automatic matching of data assets" was proposed.

[0170] From process automation to intent automation: System input has shifted from detailed work orders and plans to high-level business intents. The intent parsing engine automatically transforms these into market demands for data assets, and completes automatic discovery, negotiation, combination, and procurement within the value network, achieving a leap from "human-driven processes" to "intent-driven resources (data assets)."

[0171] Dynamic asset portfolio and creation: The system can dynamically combine multiple basic data asset contracts from different intelligent agents into a complex, customized "solution package" based on real-time intent, which greatly improves the reusability and scenario adaptability of data assets and stimulates new value creation.

[0172] 3. It has achieved emergent optimization and ecosystem-level self-evolution based on market game theory.

[0173] Decentralized emergent optimization: The globally optimal state (such as the lowest total cost and fastest response) is not calculated by a central algorithm, but rather a stable state that naturally emerges after all agents in the value network market interact and adaptively adjust to maximize their own data asset returns. This model has extremely strong robustness and adaptability.

[0174] Self-evolution of ecological rules: The system has a built-in evolution mechanism that can automatically identify more effective data asset interaction patterns or contract templates by analyzing historical collaborative data, and update network rules after community consultation, enabling the entire ecosystem to have the ability to continuously learn and improve, surpassing static programming systems.

[0175] The following section describes the overall architecture of the autonomous intelligent agent ecosystem based on data asset flows. (Reference) Figure 10 The contents shown include: autonomous intelligent agent layer 1001, value consensus network layer 1002, and business intent layer 1003.

[0176] The bottom layer (Autonomous Intelligent Agent Layer 1001): Each supply chain entity (supplier, manufacturer, etc.) is instantiated as an autonomous intelligent agent, serving as the sovereign unit of data assets. The intelligent agent internally encapsulates private data and decision-making logic, and externally interacts with value only through standardized data asset contracts, realizing "data remains still, value moves".

[0177] Middle Layer (Value Consensus Network Layer 1002): This is a decentralized virtual value exchange market. Data asset contracts issued by various intelligent agents flow, are discovered, and matched within this network. The network has a built-in dynamic pricing and reputation feedback mechanism, which determines the fair value of contracts based on market competition, establishing a trustworthy trading environment.

[0178] Top layer (Business Intent Layer 1003): Users or the system input high-level business intents (such as ensuring supply resilience). The intent parsing engine automatically deconstructs abstract intents into a specific list of data asset requirements and projects them into the value network to find matching contracts.

[0179] Macro-level optimization states (such as resilience and efficiency) are not designed by a central algorithm, but rather are the result of collective wisdom spontaneously emerging from the interactions and games played by countless intelligent agents within a value network based on their own interests. A three-layered separation architecture of "intent-network-intelligent agent" is constructed to clearly delineate the boundaries between business objectives, value exchange rules, and individual behaviors. Value flow visualization: clearly demonstrating how data assets are transformed from a private state into tradable contracts, matching intent needs within the network, and ultimately realizing value. Intuitively demonstrating how system-level optimization characteristics arise from underlying local interactions.

[0180] The following section explains the internal structure of the autonomous intelligent agent and the data asset contract generation process.

[0181] refer to Figure 11 The contents shown include the internal structure of the autonomous intelligent agent 1101, the standardized data asset contract 1102, and the data asset contract structure 1103.

[0182] The internal structure 1101 of the autonomous intelligent agent includes: a data resource pool 11011, a value encapsulator 11012, and a contract generator 11013. The data resource pool 11011 includes: raw generated data A01, real-time inventory data A02, historical transaction data A03, and device sensor data A04. The value encapsulator 11012 includes: data cleaning and labeling B01, model analysis B02, and privacy computation B03. Model analysis B02 is used to generate insights, predictions, and commitments. Privacy computation B03 is used for encryption and decryption through federated learning and differential privacy. The contract generator 11013 is used to: define usage rights and scope C01, set up the computational model C02, and write verification logic C03.

[0183] The data asset contract structure 1103 includes: metadata 11031, access policy 11032, pricing model 11033, and smart contract 11034. Metadata 11031 includes name, description, provider, and validity period. Access policy 11032 includes permissions, frequency, and invocation method. Pricing model 11033 includes fixed fees, revenue sharing, and auctions. Smart contract 11034 is used for automated execution, verification, and settlement.

[0184] The data resource pool is used to store various raw and private data owned by intelligent agents, which are the raw materials for value creation. The value encapsulator is the core processing unit that transforms raw data into valuable "products" through three steps: Data processing: cleaning and labeling to ensure data quality; Value extraction: using models to analyze data and generate external "products," such as capacity forecast insights, delivery date commitments, and demand forecast reports.

[0185] Privacy Protection: Applying privacy-preserving computing technologies (such as federated learning) ensures that raw data does not leave the domain during the generation of "products," fundamentally protecting privacy. Contract Generator: Packages valuable "products" into standardized goods, clearly defining their usage rules, prices, and verification methods.

[0186] The generated contract is a machine-readable digital contract comprising four key elements: Metadata: Describes the basic information of the contract for easy discovery and understanding. Access Policy: Defines how the asset is used, such as access permissions and frequency limits. Pricing Model: Clarifies the value exchange method, supporting flexible business models. Smart Contract: Enables the automatic and immutable execution of the contract terms, ensuring transaction trustworthiness.

[0187] Value Encapsulation Pipeline: Showcasing the standardized production process from "raw data" to "tradable data assets." Embedded Privacy Protection: Emphasizing privacy-preserving computational processing in the value extraction stage, a prerequisite for trusted circulation. Standardized Contract Structure: Defining a standardized digital contract format for data assets as tradable commodities, the foundation for building the market.

[0188] One of the core aspects of this embodiment is transforming the private and heterogeneous business data within each participant (smart agent) of the supply chain into securely circulated and directly accessible "data asset contracts" through a standardized "value encapsulation" process. This process is not simply about putting data on the blockchain or establishing ownership; rather, it involves processing raw data into standardized services or models with specific business value, while ensuring that the raw data is not exposed. The following example illustrates this further: Example 1: Packaging of a Tier-1 Supplier's "Productivity Health" Service. Background: A key component supplier (Agent S) providing parts to an OEM. Raw Data: Real-time equipment status codes, quality inspection records, and material inventory data from its internal Manufacturing Execution System (MES). Value Packaging Process: Local Value Calculation: S's local "value packager" loads a lightweight AI model to analyze the raw data in real time, outputting a comprehensive "Productivity Health Index" (PHI, range 0-100) along with a 24-hour capacity fluctuation prediction range. Privacy and Boundary Protection: The entire calculation is performed on S's local edge server. The publicly released contract does not contain any specific equipment numbers, production batches, or defect codes; it only provides a PHI index query interface and prediction range. Service Packaging: The "Productivity Health Index Query and Prediction Service" is defined as a standardized data asset. Its value lies in providing downstream manufacturers with accurate insights into supply chain resilience, rather than the raw production data itself. Generate standardized contract: The system automatically generates a standardized contract for the data asset. The "smart contract" terms stipulate that authorized customers can make no more than 1,000 queries per month; each query returns the PHI value and prediction range; and the pricing method is a "monthly subscription fee".

[0189] Example 2: Encapsulation of "Dynamic Routing Reliability" service for 3PL logistics providers. Background: A third-party logistics company (Agent L). Raw Data: GPS tracks from its fleet management system, historical order delivery times, and average speeds across different road segments. Value Encapsulation Process: Aggregation and Insight Generation: L's "value encapsulator" uses a spatiotemporal aggregation algorithm to generate a "dynamic route reliability heatmap for East China region across different time periods." This heatmap displays the historical average on-time delivery rate for different regions at different time periods. Privacy Anonymization: Differential privacy processing of vehicle trajectories ensures that the heatmap cannot be used to deduce the travel path of any individual vehicle. Service Encapsulation: The "Regional Route Reliability Insight Service" is encapsulated as a data asset. Its value lies in providing shippers with intelligent route selection suggestions based on historical big data, rather than providing real-time vehicle monitoring. Standardized Contract Generation: The contract stipulates that subscribers can subscribe to the heatmap data stream by region and by time granularity (e.g., updated every 15 minutes). The data stream is accessed by "query area". A dynamic pricing model based on "time".

[0190] Contract generator: Packages valuable "products" into standardized commodities, clearly defining their usage rules, prices, and verification methods.

[0191] Usage Rules (Execution Constraints): This is the "operation manual" for the data asset contract, defined on-chain as structured data. **Calling Credentials:** Specifies the "digital key" required to call this asset, potentially linked to the caller's digital identity or specific credentials (such as a "long-term partner credential"). **Calculation Quota:** Defines the maximum allowed computation, query count, or data throughput within a specific period (e.g., per second, per day). **Derivative Usage Restrictions:** Specifies the scope of use for new insights generated based on this data asset (e.g., generated reports) (e.g., "for internal decision-making only, resale prohibited"). **Pricing Model (Value Exchange Mechanism):** This solution supports complex pricing models reflecting the dynamic value and supply-demand relationship of data assets, crucial for their marketization. **Tiered Subscription System:** Fixed fees for basic functions, with additional charges for advanced functions (e.g., higher frequency, finer-grained data). **Dynamic Revenue Sharing Based on Value Contribution (Core Innovation):** For assets that directly improve business benefits (e.g., optimization algorithms), a "basic access fee + performance sharing" model is adopted. The revenue sharing ratio and verification logic are directly written into the smart contract. Real-time bidding and options: For scarce or high-value assets (such as key capacity reservations), support short-term spot bidding on the on-chain trading market, or the sale of "data service options" (such as priority access rights). Verification methods (effect verification): Ensure that the contract is faithfully executed and that value delivery is measurable. Verifiable computational proof: The supply-side intelligent agent can generate a short mathematical proof confirming that it has correctly executed the calculation as required by the contract (such as generating a valid predictive model), without revealing the calculation details and raw data. The demand-side or a third party can quickly verify the proof. On-chain oracle + automatic settlement: For the "effect-sharing" model, inputting business result data (such as the actual cost savings amount signed by multiple parties) through a trusted on-chain oracle triggers the smart contract to automatically complete the profit-sharing settlement, realizing a closed loop of value delivery.

[0192] During intent resolution, the engine not only resolves "what functionality is needed," but also "the cost structure that is willing to bear" (such as accepting only a subscription model or being willing to try revenue sharing).

[0193] During market matching, the algorithm precisely matches the demand side's "budget and pricing preferences" with the supply side's "pricing model." A manufacturer wishing to adopt a "performance-based revenue sharing" model will only be matched with optimization service providers who support this model and are confident in the effectiveness of their own algorithms.

[0194] During contract execution and feedback, the on-chain defined "verification method" is automatically triggered to complete value settlement. At the same time, the "pricing result" (such as the actual share of the profits) and "verification result" (such as the accuracy of the service) of this transaction are recorded, serving as core inputs for the market valuation of this data asset and the reputation of the supplier, driving value discovery within the ecosystem.

[0195] Deepening Packaging Strategies for Different Intelligent Agents: For brand owners / core manufacturers: Packaging focuses on demand-side insights and market influence. For example, packaging services like "cross-channel real-time sales trend maps" or "new product concept market popularity testing services." Contract pricing is often linked to incremental revenue generated by the insights (performance-based revenue sharing), and usage rules strictly limit the use of insights to ensure they do not harm brand value. For small and medium-sized manufacturers / suppliers: Packaging focuses on operational transparency and capability standardization. Examples include "product quality consistency analysis as a service" or "standard working hour database access service." Pricing is mostly a fixed subscription fee, aiming to reduce customer verification costs. For logistics and warehousing service providers: Packaging focuses on process commitment. Examples include "end-to-end carbon footprint tracking and verification services" or "warehouse capacity utilization prediction and reservation services." Pricing may fluctuate based on the achievement of committed key performance indicators (KPIs) (such as on-time delivery rate and carbon emission reduction). For financial institutions: As ecosystem participants, their packaging focuses on risk pricing capabilities based on real-time supply chain data. Examples include "dynamic risk assessment services for in-transit cargo pledges based on multi-source logistics data." This is a classic example of how a combination of cross-entity data assets can generate new value.

[0196] This deep differentiation allows each intelligent agent to find a unique "niche" within the ecosystem, based on its core data capabilities, providing irreplaceable value rather than engaging in homogeneous data exchange.

[0197] The following section explains the process of intention-driven data asset consumption and value realization.

[0198] refer to Figure 12 The process may include, but is not limited to, S1201 to S1217 described below.

[0199] S1201. Enter the business intent, such as "reduce costs by 5% next quarter".

[0200] S1202, the intent parsing engine performs the parsing.

[0201] S1203, invoke the domain knowledge graph.

[0202] S1204, Call the machine learning model.

[0203] S1205, Output a list of data asset requirements. For example: 1. Supplier A's capacity optimization proposal contract; 2. Logistics provider B's multimodal transport solution contract.

[0204] S1206, Enter the value consensus network market.

[0205] S1207, Intelligent Matching and Game Theory.

[0206] S1208, Demand side: Demand list.

[0207] S1209. Attacker: Can use Discretionary Access Control List (DAC) list.

[0208] S1210, Multi-party game and bargaining.

[0209] S1211. A transaction is reached, a composite data asset package is generated, and multiple DACs are combined into a solution.

[0210] S1212, Execute the smart contract and call the service.

[0211] S1213. Generate business results. For example, generate a detailed report on reducing the impact of this implementation plan.

[0212] S1214, Results Evaluation and Value Feedback.

[0213] If the result meets the standard, proceed to S1215 below; if the result does not meet the standard, proceed to S1216 below.

[0214] S1215, Positive Feedback: Enhances the provider's reputation and stabilizes or increases contract estimates.

[0215] S1216. Negative feedback: Reduces the provider's reputation, triggering claims or price negotiations.

[0216] S1217. Conclusion: Business objectives achieved, network value system updated.

[0217] Analysis of the key steps in this process: Intent Input and Parsing (S1201-S1203): Users input high-level business objectives. The intent parsing engine (combining knowledge graphs and AI models) automatically translates these into a list of requirements for specific data asset contracts, realizing the transformation from "what to do" to "what digital resources are needed".

[0218] This solution's "intent parsing engine" achieves a fundamental paradigm shift: from "task decomposition and allocation" to "resource requirement generation and configuration." This solution parses requests based on both "goals" and "resources."

[0219] a. Semantic Enhancement of Intent: Input "Ensure uninterrupted supply of the next-generation flagship product during its launch season". The engine first combines the supply chain knowledge graph to deconstruct "uninterrupted supply" into multiple interrelated business objectives: maximizing demand satisfaction rate during the launch period, minimizing long-tail inventory risk, and ensuring the supply flexibility of core components.

[0220] b. Inference Mapping from Goals to Data Asset Requirements (Core Innovation): Based on machine learning and graph reasoning, the engine answers: "To achieve these goals, what information advantages and decision support capabilities do I need to acquire or utilize?" The output is not a task list, but a machine-readable "Data Asset Procurement Requirements List," for example: Requirement Item D1: "Prediction of the probability distribution of pre-sale sales of flagship products by region and channel" service. (Corresponding goal: Precise inventory preparation); Requirement Item D2: "Monthly capacity elasticity and risk assessment of key chip suppliers" service. (Corresponding goal: Ensuring supply); Requirement Item D3: Access to "Dynamic safety stock optimization model based on real-time sales signals." (Corresponding goal: Balancing supply and demand).

[0221] c. Structured Description of the Demand List: Each demand item in the list contains rich metadata: a functional description of the required service, non-functional requirements (e.g., prediction confidence > 90%, data latency < 1 hour), preferred business terms (e.g., acceptance of performance sharing, budget cap), and the source of the preference (e.g., from a certain type of agent with a good history of cooperation). This list directly enters the value consensus network as a "purchase order." This process shifts the focus of collaboration from "commanding agents to do what" to "configuring resources (data assets) for agents," realizing the "servitization" and "marketization" of supply chain collaboration.

[0222] Market Matching and Game Theory (S1204-S1206): The demand list enters the value network market and is automatically matched and game-theorized with the contract supply issued by various agents. The system may combine multiple contracts (such as purchasing "capacity optimization" and "logistics solution" contracts separately) to form a customized "composite data asset package" to meet the intent with the best cost performance.

[0223] A more detailed description of matching and game theory rules: Matching and game theory take place in a “value consensus network,” which is a simulated, multi-agent data asset market.

[0224] a. Multi-attribute bidirectional matching: The system matches the demand list with the supply contracts across the entire network in multiple dimensions: capability matching (whether the functions meet the requirements), quality matching (whether the performance indicators meet the standards), economic matching (whether the price is within the budget), reputation matching (the supplier's historical performance), and terms compatibility matching (whether the pricing model and usage rules are acceptable).

[0225] b. Combinatorial optimization and collaborative bargaining: Package Building: For complex requirements, the system acts as a "virtual integrator," attempting to combine multiple independent supply contracts into a "solution package." For example, it can combine A's forecasting service, B's inventory optimization model, and C's logistics capacity to meet the requirement of "end-to-end fulfillment optimization."

[0226] Game Theory Mechanism: The system initiates multiple rounds of negotiation. In the first round, an initial combination plan and total price are formed based on the bids from each contract. In the second round, "bundled pricing" is initiated with relevant suppliers: "If a bundled purchase is made, are all parties willing to offer discounts?" This incentivizes previously independent data service providers to spontaneously coordinate in order to win bundled orders. At the same time, competition also arises between different combination plans.

[0227] Comprehensive utility assessment: For each candidate solution (including single contracts and packages), calculate a comprehensive utility score: U=w1 Functionality score +w2 (1 / Total Cost) + w3 Average reputation score +w4 Synergy bonus. U represents the overall utility score, and w1, w2, w3, and w4 represent the weighting coefficients of the functional score, cost score, average reputation score, and synergy bonus, respectively. The "synergy bonus" assesses the complementarity between contracts within the portfolio and the smoothness of data flow.

[0228] c. Market Clearing and Dynamic Pricing: The winning bid and its transaction price constitute the "market clearing price" for this round. High-frequency trading data dynamically influences the "market valuation" of similar data assets, forming a transparent price discovery mechanism.

[0229] Implementation Example: Intent: "To achieve an ultimate logistics experience in core cities of North China, enabling 'order before bed, receive upon waking' for the upcoming online shopping festival." Analyzed Requirement List: D1: City-level real-time order density prediction (requires 1 hour advance notice, accuracy >85%). D2: Integration of real-time aggregation and scheduling capabilities for last-mile crowdsourced delivery. D3: Real-time inventory visibility and dynamic allocation suggestions for forward warehouses.

[0230] Matching and Game Theory Process: i. D1 matches with the e-commerce platform's own prediction service P. ii. D2 no existing service is found, but the system discovers that a crowdsourcing platform Q provides a "regional real-time delivery capacity demand release interface," while thousands of individual rider agents R1, R2, etc., provide "my real-time location and willingness to accept orders" contracts. iii. Dynamic Combination and Market Creation: The system automatically combines Q's interface contract with the delivery capacity contracts of multiple Rs, temporarily creating a composite data asset called "North China Instant Crowdsourced Delivery Capacity Market." It represents the demand side, playing a game with Q regarding "transaction commission rates," while the overall service price of this "market" (including estimated rider fees) also competes with the prices of traditional logistics companies. iv. D3 matches with the inventory query service of warehousing service provider W. v. Ultimately, the system may select a "super solution package" consisting of P's prediction + (Q's interface + the delivery capacity of numerous Rs) dynamic market + W's inventory service. This package is not pre-existing, but dynamically assembled through real-time matching and game theory, satisfying complex business intentions in near real-time with optimal overall utility.

[0231] Contract Execution and Result Generation (S1207-S1208): Transactions are executed automatically through smart contracts. The requesting party receives the data services stipulated in the contract (such as insight reports and optimization solutions) and applies them to actual business operations to generate measurable business results.

[0232] Performance Evaluation and Feedback Loop (S1209-S1210): Business results are evaluated and returned to the network as value feedback. Contracts with good performance and their providers receive reputation rewards and value enhancements; those with poor performance are penalized. This mechanism ensures the quality and credibility of the network and completes the loop from "consumption" to "value recognition".

[0233] The reputation and value system in this solution is a tightly coupled, bidirectionally reinforcing, positive feedback flywheel driving the survival of the fittest within the ecosystem. The process is described in detail as follows: a. Multi-dimensional dynamic reputation profile: Each agent (supply side) possesses a dynamically updated reputation profile vector, such as: [performance credit score, data quality score, value creation score, collaboration stability score]. The "value creation score" is a key innovation, calculated based on the actual revenue sharing amount from historical transactions using the "performance-based revenue sharing" model or the quantified business improvement value reported by the user. b. Dynamic correlation between value and price: Base pricing: Set by the supply side, reflecting its subjective valuation. Reputation discount / premium: In the matching algorithm, contracts of high-reputation agents receive a "reputation multiplier," resulting in a higher utility score at the same price, equivalent to receiving an implicit premium from market trust. Low-reputation agents need to compensate for their risk discount by offering lower prices. Market valuation discovery: The historical transaction price of a data asset contract (especially the revenue generated from performance-based revenue sharing) and transaction frequency are publicly recorded and form its market valuation curve. A data asset that can continuously create high value for users will see its market valuation rise, attracting more demand and creating a positive cycle of "premium for high-quality assets." c. Closed-loop feedback and flywheel effect: i. After contract execution, verifiable business results and satisfaction feedback are recorded on the blockchain. ii. Based on this feedback, the system updates the corresponding dimensions of the supplier's reputation profile in real time. For example, if the actual cost savings exceed the promise, its "value creation score" is significantly increased. iii. The transaction data (transaction price, profit sharing amount) of this round of the contract is incorporated into its market valuation model, driving dynamic valuation adjustments. iv. High-reputation, high-valuation agents gain a significant advantage in the next round of matching: they are more likely to be selected, can maintain higher pricing, and attract more cooperation requests. This incentivizes all agents to continuously optimize the quality of the data assets and services they provide. Conversely, low-reputation, low-value agents are gradually marginalized by the market. This mechanism ensures that ecosystem resources are continuously allocated to nodes that can create real value, driving the entire supply chain ecosystem to "evolve" towards greater efficiency and agility.

[0234] Fully Automated Process: Demonstrates a fully automated, collaborative process from intent declaration to goal achievement, requiring no human intervention. Dynamic Composite Asset Packages: Highlights the system's ability to dynamically combine multiple basic contracts to form complex solutions based on real-time needs. Strong Feedback Loop: Clearly defines value assessment and reputation feedback as the core driving forces for healthy ecosystem development, enabling the system to self-purify and optimize.

[0235] The following section explains the process of ecological emergence and self-evolution.

[0236] refer to Figure 13 The process may include, but is not limited to, S1301 to S1308 described below.

[0237] S1301. Environmental inputs: market changes, new policies, and new technologies.

[0238] S1302, Data asset transactions and interactions between intelligent agents.

[0239] S1303, Ecological Output: Continuously enhanced adaptability, resilience, and creativity.

[0240] S1304. Generate macro-emergent patterns. For example: new collaborative network structures, efficient inventory distribution patterns.

[0241] S1305, the pattern analysis and rule discovery module is based on federated learning and complex network analysis.

[0242] S1306. Identify success patterns and propose evolutionary proposals. For example: "Combined transactions of Class A and Class B DACs can improve performance efficiency by 20%."

[0243] S1307, Distributed Community Governance Voting (DAO) / Intelligent Agent Voting.

[0244] S1308. Update network consensus rules, such as solidifying the success model into a new recommended contract template or standard transaction protocol.

[0245] Evolutionary closed-loop analysis can include: Micro-level Interaction (A): Agents engage in daily data asset transactions and collaborations within the value network, forming the foundation for all complex patterns. Macro-level Emergence (B): Numerous local interactions spontaneously form stable and efficient macro-level patterns (e.g., stable agile supply chain clusters among specific enterprises). Pattern Recognition (CD): System-level pattern analysis and rule discovery modules (based on technologies such as federated learning) continuously monitor the network, automatically identifying these successful and efficient emergent patterns and abstracting them into generalizable "evolutionary proposals." Community Governance (E): Proposals are not enforced by a central authority but submitted to a distributed community of all agents for voting, reflecting the democratic nature of the ecosystem. Rule Update (F): Once a proposal is passed, it is solidified by updating network consensus rules (e.g., adding new standard contract templates or optimizing matching algorithms). Closed-Loop Feedback: New rules, in turn, guide and optimize the agents' next round of interactions, thus initiating a new and more advanced evolutionary cycle.

[0246] Environmental impact and output: Environmental Inputs (G): Changes in external markets and policies act as "selective pressures," forcing the ecosystem to adapt and evolve. Ecosystem Outputs (H): Through a continuous evolutionary loop, the entire ecosystem outputs continuously improving adaptability, resilience, and collective innovation capabilities.

[0247] A complete evolutionary loop: Depicting the complete evolutionary cycle of "interaction → emergence → identification → governance → solidification → guiding interaction," this is key to the system's "vitality." Rules generated from the bottom up: Emphasizing that ecosystem rules (such as best practices) are not preset by designers, but automatically discovered from numerous successful interactions and adopted through community consensus, achieving true "self-organization" and "self-evolution." Ecosystem-level adaptability: Analogizing the entire system to a living organism, its final output is not a specific business outcome, but the continuous growth of the ecosystem's overall ability to adapt to environmental changes.

[0248] Example scenario: The evolution of supply chain emergency response models for cross-border e-commerce promotions.

[0249] Micro-level Interactions and Emerging Patterns: In the cross-border e-commerce supply chain, during major promotional events such as "Black Friday," multiple intelligent agents, including brand owners, overseas warehouses, customs clearance service providers, and last-mile delivery companies, face immense order pressure and uncertainty. After running this system for several major promotional cycles, an efficient emergency collaboration model has been observed: whenever order forecasts indicate congestion at a certain port, relevant intelligent agents (such as customs clearance agent A, port warehouse B, and specific route logistics C) automatically and quickly form a temporary "emergency alliance." By combining their data asset contracts (A's real-time customs clearance status, B's flexible storage capacity, and C's backup transportation capacity), they provide integrated "customs clearance-warehousing and distribution" support services for high-priority orders. The success rate of this model is significantly higher than random collaboration.

[0250] Pattern Mining and Rule Abstraction: The ecosystem layer's "Pattern Mining Module" (employing federated learning to protect the commercial privacy of each agent) analyzes historical successful interaction data from all "Emergency Alliances." It automatically identifies key pattern characteristics of success, such as: alliance members' data asset contracts must include real-time status updates; the pricing model between contracts must be compatible with premium mechanisms for short-term, high-priority services; and members need to exchange basic trusted credentials beforehand to quickly establish temporary trust.

[0251] The module generates an "evolutionary proposal": This module abstracts and standardizes these features, producing an evolutionary proposal called "Cross-border Promotion Port Emergency Collaboration Alliance Framework Contract Template". This proposal defines a set of interrelated standardized sub-contract templates, as well as recommended data flow, payment flow, and dispute resolution rules among them.

[0252] Community Governance and Ecosystem Rule Updates: The proposal is submitted to the on-chain DAO (Distributed Autonomous Organization) for governance voting. All agents in the ecosystem (not just historical participants) vote based on their own interests. Since this model has proven to improve the overall ecosystem's resilience and customer satisfaction, it is likely to be approved. After the vote is passed, this new "framework contract template" is automatically added to the value consensus network's "recommended template library." Network Layer Rule Adaptation: The network layer matching algorithm is also upgraded synchronously. When the system detects features such as "major promotions" or "port congestion warnings" again, it will proactively recommend the use of this evolved framework template for rapid network formation to relevant agents. Simultaneously, during the intent parsing phase, for intents such as "ensuring timely delivery of high-value orders," the parsing engine will prioritize generating a list of requirements that conform to this framework template. A New Evolutionary Cycle: When the next major promotional season arrives, more agents can adopt this superior, standardized collaboration model, thereby improving the overall emergency response level of the ecosystem. The system continues to monitor the template's performance under the new model, collecting new data to fuel the next possible evolution (such as integrating air transport resources or adding insurance services).

[0253] This embodiment has the following technical effects: 1. Pioneering a new path for the market-based circulation and value realization of data assets: Providing standardized value carriers and active trading markets for data assets, enabling them to circulate across organizations and directly generate economic benefits while ensuring privacy, thus truly activating the value of data elements.

[0254] 2. Achieve ultimate automation and intelligence in supply chain collaboration: Through the "intent-driven" model, complex multilateral collaboration is simplified to declaring business objectives. The system automatically completes resource discovery, negotiation and combination, greatly improving the agility and responsiveness of the supply chain.

[0255] 3. Endow the supply chain system with biological-like adaptability and evolutionary power: Based on the emergence mechanism and rule evolution capability, the system can adapt to environmental changes and learn from successful experiences, continuously evolving better collaborative models, possessing unprecedented resilience and long-term vitality.

[0256] 4. Build an incentive-compatible sustainable development ecosystem: Align individual interests (profiting from selling data assets) with overall interests (overall optimization of the supply chain) through market mechanisms, incentivize all participants to actively contribute high-quality data assets and collaborate, forming a powerful network effect and ecological prosperity.

[0257] Secondly, this application provides an intelligent agent network, comprising: a first determining unit, configured to determine M candidate intelligent agents related to the demand intent from N second intelligent agents based on the demand intent input by the demander; M is less than or equal to N; the intelligent agent network includes N participating second intelligent agents; a processing unit, configured to combine and elect M candidate intelligent agents based on the resource contract of each candidate intelligent agent, and determine the elected candidate intelligent agent combination as the target participating agent combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used; and a second determining unit, configured to determine the target participating agent combination as the target supply chain that satisfies the demander's demand intent.

[0258] Thirdly, embodiments of this application provide a supply chain processing system, with reference to... Figure 14 The supply chain processing system 140 shown includes a first intelligent agent 1401 on the demand side and an intelligent agent network 1402. The first intelligent agent 1401 is used to: obtain the demand intent input by the demander; The agent network 1402 is used to: determine M candidate agents related to the demand intention from N second agents based on the demand intention; M is less than or equal to N; the agent network 1402 includes N participating second agents 14021; The agent network 1402 is used to: combine and elect M candidate agents based on the resource contract of each candidate agent, determine the combination of candidate agents that passes the election as the target participant combination, and determine the target participant combination as the target supply chain that satisfies the demand intention of the demand side.

[0259] In some embodiments, the first intelligent agent 1401 is further configured to: perform task planning on the demand intention and determine L sub-tasks; L is less than or equal to M; the first intelligent agent maps the L sub-tasks to L resource demand lists; Correspondingly, the agent network 1402 is also used to: based on L resource demand lists, the agent network identifies the second agents related to the L resource demand lists as candidate agents among N second agents, thereby obtaining M candidate agents.

[0260] In some embodiments, the supply chain processing system 140 is also used to implement the method provided in the first aspect.

[0261] It should be noted that the communication device of the application provided in this application embodiment includes all the units included, which can be implemented by a processor in an electronic device; of course, it can also be implemented by specific logic circuits; in the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field-programmable gate array (FPGA), etc.

[0262] The description of the system embodiments above is similar to that of the method embodiments above, and has similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the description of the method embodiments of this application for understanding.

[0263] It should be noted that, in the embodiments of this application, if the above methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of software products. These computer software products are stored in a storage medium and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0264] Fourthly, embodiments of this application provide a storage medium, namely a computer-readable storage medium, on which a computer program or instructions are stored, which, when executed by a processor, implement the steps of any of the methods provided in the first aspect of the above embodiments.

[0265] Fifthly, embodiments of this application provide a computer program product, which includes a computer program or instructions that, when executed by a processor, implement the steps of any of the methods provided in the first aspect of the above embodiments.

[0266] It should be noted that the descriptions of the above embodiments of storage media, systems, and program products are similar to the descriptions of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the embodiments of storage media, devices, apparatuses, and program products of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0267] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0268] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0269] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another electronic device, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0270] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0271] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0272] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0273] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0274] The above are merely embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A supply chain processing method, characterized in that, The method is applied to an agent network, which includes N participating second agents; the method includes: Based on the demand intent input by the demand side, M candidate agents related to the demand intent are determined from N second agents; where M is less than or equal to N. Based on the resource contract of each candidate agent, the M candidate agents are combined and selected for competition, and the combination of candidate agents that passes the competition is determined as the target participant combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used. The target participant combination is identified as the target supply chain that satisfies the demand intentions of the demander.

2. The method according to claim 1, characterized in that, The resource contract based on each candidate agent involves combining and electing the M candidate agents, and determining the winning combination of candidate agents as the target participant combination, including: Obtain the resource contracts of the participants among the M candidate intelligent agents; The target election rules are determined based on the stated needs and intentions; The resource contracts of the participants among the M candidate intelligent agents are combined and selected through the target election rules, and the combination of candidate intelligent agents that passes the election is determined as the target participant combination.

3. The method according to claim 2, characterized in that, The target election rules combine auction rules, negotiation rules, and reputation rules, and are used to achieve a dynamic balance between the demand side's intention satisfaction, the participants' revenue, and the efficiency of the supply chain.

4. The method according to claim 3, characterized in that, The process of combining and electing resource contracts among the M candidate agents using the target election rules, and determining the elected candidate agent combination as the target participant combination, includes: The resource contracts of the participants among the M candidate intelligent agents are processed by auction rules to determine multiple combinations of candidate intelligent agents and the supply efficiency under each combination; each of the combinations can satisfy the demand intention. The resource contracts of the participants among the M candidate intelligent agents are processed by the negotiation rules to determine the participant's income under each of the combined methods; Determine the client's intended satisfaction through reputation rules; The target participant combination is determined based on the supply efficiency under each combination, the participant benefits under each combination, and the demander's intention satisfaction.

5. The method according to any one of claims 2-4, characterized in that, The resource contract based on each candidate agent involves combining and electing the M candidate agents, and determining the winning combination of candidate agents as the target participant combination, including: Through each second agent in the agent network, the resource contracts of the M candidate agents are combined and contested; thus, the combination of participants supported by each second agent is obtained. The target participant combination is determined by each second agent in the agent network based on the participant combination supported by each second agent.

6. The method according to claim 5, characterized in that, The resource contracts of the M candidate smart agents are combined and selected; The combination of participants supported by each of the second agents includes: Based on the resource contract of each candidate agent, determine the function score, price score, reputation score and collaborative efficiency score of each participant; Based on the functional score, price score, reputation score, and collaboration efficiency score of each participant, the functional score, price score, reputation score, and collaboration efficiency score of each combination are calculated; the reputation score and collaboration efficiency score are determined based on the historical feedback of the participants. The total score for each combination is determined based on the functional score, price score, reputation score, and collaborative efficiency score for each combination. Based on the total score of each combination, the combination of participants supported by the second agent is determined.

7. The method according to claim 6, characterized in that, The method further includes: When providing services according to the target participant combination in the target supply chain, obtain feedback data for each participant; the feedback data includes: performance credit score, data quality score, value creation score, and collaboration stability score; The credit score of the participating party is updated based on its performance credit score, data quality score, and value creation score. The collaborative efficiency score of the participating parties is updated based on the collaborative stability score.

8. The method according to any one of claims 2-4, characterized in that, Before the agent network obtains the resource contracts of the participants among the M candidate agents, the method further includes: For each of the second agents in the agent network, execute: Obtain the original resource data, inventory data, and historical transaction data of the participants corresponding to the second intelligent agent; Based on the original resource data, inventory data, and historical transaction data, resource prediction data is generated; the resource prediction data is used to characterize and predict the resource data that the participating party can provide. Determine the value rules for the participating parties; The predicted resource data and the value rules are organized according to a set format to obtain the resource contract of the participants of the second intelligent agent; The resource contract is published to the intelligent agent network.

9. An intelligent agent network, characterized in that, The agent network includes: The first determining unit is configured to determine M candidate agents related to the demand intent input by the demander from N second agents; wherein M is less than or equal to N; and the agent network includes N participating second agents. A processing unit is configured to: In the agent network, which comprises N participating agents, a second agent; based on the resource contract of each candidate agent, combine and elect the M candidate agents, and determine the winning combination of candidate agents as the target participating agent combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used. The second determining unit is used to determine the target participant combination as a target supply chain that satisfies the demand intentions of the demander.

10. A supply chain processing system, characterized in that, The system includes a first intelligent agent of the demand side and an intelligent agent network, wherein the intelligent agent network includes second intelligent agents of N participating parties; The first intelligent agent is used to: obtain the demand intent input by the demander; The agent network is used to: determine M candidate agents related to the demand intent from N second agents based on the demand intent; where M is less than or equal to N; combine and elect the M candidate agents based on the resource contract of each candidate agent, and determine the elected candidate agent combination as the target participant combination; the resource contract includes resource data and value rules; the value rules are the dynamic pricing method when the resource data is used; and determine the target participant combination as the target supply chain that satisfies the demand intent of the demander.

11. The system according to claim 10, characterized in that, The first intelligent agent is further configured to: perform task planning on the demand intent, and determine L sub-tasks; wherein L is less than or equal to M; and map the L sub-tasks to L resource demand lists. Correspondingly, the agent network is further configured to: based on the L resource requirement lists, determine the second agents related to the L resource requirement lists from among the N second agents as candidate agents, thereby obtaining the M candidate agents.

Images