Multi-agent cooperative behavior evidence storage and credible auditing method and system

Abstract

The invention provides a multi-agent cooperative behavior evidence storage and credible auditing method and system, and the method comprises the steps: firstly, creating a unique on-chain digital identity for each AI Agent through a decentralized identity technology, and binding historical reputation data; decomposing a complex task submitted by a user and encoding the complex task into an intelligent contract to define a task process, a cooperation rule and a reward and punishment mechanism, and storing key behaviors of each Agent to a block chain in a transaction form when the Agent executes the task; and finally, automatically propelling a cooperation stage based on the state machine logic of the intelligent contract, and triggering a built-in arbitration logic to process an abnormal condition when a dispute occurs. According to the method, the problems of trust missing, non-auditing of behaviors and difficulty in value distribution in multi-Agent collaboration are solved, and the method has the characteristics of high collaboration credibility, high process traceability and fair and automatic value distribution.

Description

Technical Field

[0001] This invention relates to the field of computer application technology, specifically to a method and system for evidence storage and trusted auditing of multi-agent collaborative behavior, and more specifically to a method and system for evidence storage and trusted auditing of multi-agent collaborative behavior based on blockchain smart contracts, which is particularly suitable for scenarios involving trusted collaboration, behavior tracing, and value distribution of multiple independent AI agents when performing complex tasks. Background Technology

[0002] With the rapid development of artificial intelligence technology, multi-agent systems (MAS) have become an important technological direction for solving complex tasks. Typical applications include autonomous intelligent agent systems such as AutoGPT and BabyAGI, which complete complex tasks such as data collection, analysis, and decision-making through the division of labor and cooperation of multiple AI agents. However, the traditional collaborative mode of multi-agent systems usually relies on a centralized controller, such as a task scheduling server or a master agent, to decompose tasks, assign roles, and aggregate results. This architecture has significant technical shortcomings.

[0003] To address the issues of trust deficiency and uncontrollable processes in multi-agent collaboration, centralized scheduling algorithms or API gateway technologies are typically used to distribute tasks and aggregate results, and predict future execution states. Effective management of computing and communication resources can improve system efficiency to some extent. However, existing centralized collaboration technologies are insufficient to resolve the following core contradictions: different agents may be provided by different developers, such as enterprise-level AI services, open-source models, or third-party plugins. Their code logic, data sources, and decision-making processes are all encapsulated as "black boxes," lacking an inherent foundation of trust between them; agent decision-making and execution occur in local environments, such as local model inference and private database queries, making their specific behavior unobservable from the outside; when tasks fail or results are abnormal, it is difficult to trace which agent made the error at which stage; in complex task chains involving multiple contributions, how to fairly evaluate the value of each agent and complete payment remains an unsolved problem. Therefore, existing multi-agent collaborative management suffers from at least the following technical problems: Lack of trust in collaboration: As the only trusted party, the centralized controller will collapse if it is attacked or malfunctions; and there is no native trust mechanism between different agents, making it impossible to ensure that each party performs the assigned tasks accurately. Unauditable behavior: The agent's key operations, such as data collection sources, parameter calls, and intermediate result transmission, do not leave immutable records, making it impossible to identify the responsible party when the task fails; Difficulty in value exchange: The lack of transparent and automated contribution assessment and incentive mechanisms makes it difficult to fairly distribute task benefits based on the actual contributions of each agent.

[0004] Therefore, it can be seen that existing methods have technical problems such as low credibility of multi-agent collaboration, difficulty in process traceability, and unfair value distribution. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for evidence storage and trusted auditing of multi-agent collaborative behavior.

[0006] A method for evidence storage and trusted auditing of multi-agent cooperative behavior provided by the present invention includes: Step S1: Create a unique on-chain digital identity for each intelligent agent, and bind and store the digital identity with its historical interaction records and reputation evaluation data on the blockchain; Step S2: Receive the target task submitted by the user, decompose the target task into multiple target sub-tasks, and encode the task flow of each target sub-task, the intelligent agent capability requirements of the target sub-task, the cooperation rules between target sub-tasks, and the reward and punishment mechanism into a smart contract. Step S3: Select the appropriate intelligent agent based on the intelligent agent's capability requirements for the target sub-task and the corresponding reputation evaluation data of the intelligent agent. During the execution of the target sub-task by the selected intelligent agent, submit the key behavioral data of the selected intelligent agent that meets the preset requirements to the blockchain in the form of a transaction for evidence storage. Step S4: Based on the collaboration rules between target sub-tasks, automatically advance the collaboration to the next stage according to the evidence verification results of the previous stage, and trigger the preset arbitration logic to handle abnormal situations when disputes occur.

[0007] Preferably, step S1 includes: Step S1.1: Generate the on-chain digital identity of the smart agent using a decentralized identity identifier (DID) or a blockchain wallet address; Step S1.2: Bind and store the on-chain digital identity with its historical interaction records and reputation evaluation data on the blockchain; The historical interaction records include: the number of times the agent successfully completed tasks, the average response time, and the historical collaboration success rate. The reputation evaluation data is generated and updated by other intelligent agents or users based on historical interaction records.

[0008] Preferably, the key behavioral data that meets the preset requirements includes: The data acquisition agent submits the data storage address and data content hash value; The key feature hashes of the external API name, request parameters, response timestamp, and returned results called by the data analysis agent meet the preset requirements; The result is the hash value of the final report submitted by the generating agent and a summary of the intermediate parameters of the generation process; The key behavioral data that meets the preset requirements is recorded in the blockchain ledger in an immutable transaction form through the smart contract's write interface.

[0009] Preferably, step S4 includes: Step S4.1: Maintain the state of the agent in the current collaboration phase, including: waiting for data collection to complete, data analysis in progress, and report generation completed; Step S4.2: When the handover conditions between agents in adjacent stages are met, the state transition to the next stage is triggered, and the corresponding reward tokens are released according to the reward and punishment mechanism; Step S4.3: When a downstream agent raises an objection to the upstream action, a preset verification script is invoked for automatic arbitration. If the verification fails, the responsible party's deposit is deducted according to the reward and punishment mechanism and the task is redistributed. If the preset verification script cannot complete the automatic arbitration, the current dispute is considered to be a complex situation, and a community voting mechanism is initiated for users to decide the final state. The preset verification scripts include: data content hash consistency verification, API call log integrity check, and collaboration rule parameter range verification.

[0010] Preferably, the method further includes: the blockchain is a distributed ledger that supports smart contracts, and each smart agent interacts with the blockchain node through a decentralized network to achieve behavior storage and state synchronization.

[0011] A multi-agent cooperative behavior evidence storage and trusted auditing system provided by the present invention includes: Module M1: Creates a unique on-chain digital identity for each intelligent agent and binds and stores the digital identity with its historical interaction records and reputation evaluation data on the blockchain; Module M2: Receives the target task submitted by the user, decomposes the target task into multiple target sub-tasks, and encodes the task flow of each target sub-task, the intelligent agent capability requirements of the target sub-task, the cooperation rules between target sub-tasks, and the reward and punishment mechanism into a smart contract. Module M3: Based on the agent capability requirements of the target sub-task and the corresponding reputation evaluation data of the agent, select the appropriate agent. During the execution of the target sub-task by the selected agent, submit the key behavioral data of the agent that meets the preset requirements to the blockchain in the form of transactions for storage. Module M4: Based on the collaboration rules between target subtasks, it automatically advances the collaboration to the next stage according to the evidence verification results of the previous stage, and triggers preset arbitration logic to handle abnormal situations when disputes occur.

[0012] Preferably, the module M1 includes: Module M1.1: Generates the on-chain digital identity of the smart agent through decentralized identity identifier DID or blockchain wallet address; Module M1.2: Binds and stores the on-chain digital identity with its historical interaction records and reputation evaluation data on the blockchain; The historical interaction records include: the number of times the agent successfully completed tasks, the average response time, and the historical collaboration success rate. The reputation evaluation data is generated and updated by other intelligent agents or users based on historical interaction records.

[0013] Preferably, the key behavioral data that meets the preset requirements includes: The data acquisition agent submits the data storage address and data content hash value; The key feature hashes of the external API name, request parameters, response timestamp, and returned results called by the data analysis agent meet the preset requirements; The result is the hash value of the final report submitted by the generating agent and a summary of the intermediate parameters of the generation process; The key behavioral data that meets the preset requirements is recorded in the blockchain ledger in an immutable transaction form through the smart contract's write interface.

[0014] Preferably, the module M4 includes: Module M4.1: Maintains the state of the agents in the current collaboration phase, including: waiting for data collection to complete, data analysis in progress, and report generation completed; Module M4.2: When the handover conditions between agents in adjacent stages are met, the state transition to the next stage is triggered, and the corresponding reward tokens are released according to the reward and punishment mechanism; Module M4.3: When a downstream agent raises an objection to the upstream action, a preset verification script is invoked for automatic arbitration. If the verification fails, the responsible party's deposit is deducted according to the reward and punishment mechanism and the task is reassigned. If the preset verification script cannot complete the automatic arbitration, the current dispute is considered to be a complex situation, and a community voting mechanism is initiated for users to decide the final state. The preset verification scripts include: data content hash consistency verification, API call log integrity check, and collaboration rule parameter range verification.

[0015] Preferably, the method further includes: the blockchain is a distributed ledger that supports smart contracts, and each smart agent interacts with the blockchain node through a decentralized network to achieve behavior storage and state synchronization.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention encodes complex task intentions into smart contracts, utilizes the distributed ledger characteristics of blockchain to achieve transparent evidence storage of agent behavior, automatically advances the collaboration process through a state machine mechanism, and incorporates dispute resolution logic to handle abnormal situations, thereby constructing a trustworthy and efficient multi-agent collaboration environment.

[0017] 2. This invention replaces the fragile scheduling mechanism of traditional centralized controllers with the immutability of blockchain and the rigid constraints of smart contracts, ensuring that the collaboration between agents is based on open and transparent rule execution, without relying on a single trusted third party. 3. All key operations of this invention are recorded indelibly on the blockchain, which can quickly locate the responsible party when a task fails, greatly improving the reliability and security of the system. 4. The smart contract of this invention clearly defines the contribution evaluation criteria and reward amount for each stage. Payment is automatically triggered upon completion of the task, avoiding the problems of vague contribution evaluation, payment delay or dispute in the traditional model. 5. The built-in dispute resolution mechanism of this invention can handle abnormal situations in complex collaborations, ensuring that the task process can continue to proceed or be reasonably terminated when some agents fail.

[0018] 6. This invention, through the deep integration of blockchain and smart contracts, constructs the first technical framework that supports trusted collaboration among multiple independent AI agents, traceable behavior, and fair value distribution, providing a reliable technical foundation for the automated execution of complex tasks. Attached Figure Description

[0019] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of a method for evidence storage and trusted auditing of multi-agent collaborative behavior.

[0020] Figure 2 This is a flowchart of the overall process for evidence storage and trusted auditing of multi-agent collaborative behavior. Detailed Implementation

[0021] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0022] Example 1 The present invention provides a method and system for evidence storage and trusted auditing of multi-agent collaborative behavior, comprising: establishing a unique trusted identifier for each agent through an on-chain digital identity mechanism; converting complex user tasks into standardized collaborative processes defined by smart contracts; recording key agent behaviors through on-chain evidence storage; and realizing automatic advancement of the collaborative process and dispute resolution based on the state machine logic of smart contracts.

[0023] The method for evidence storage and trusted auditing of multi-agent collaborative behavior, such as Figure 1 As shown, it includes: Step 101: Create a unique on-chain digital identity for each intelligent agent, and bind and store the digital identity with its historical interaction records and reputation evaluation data on the blockchain; Specifically, step 101 includes: when each AI Agent joins the collaboration network for the first time, it generates a unique on-chain digital identity using decentralized identity (DID) technology or a blockchain wallet address, such as an Ethereum address or Polygon DID. This identity is bound to its historical interaction records and reputation score, forming an immutable "behavioral history." The identity information is registered to the blockchain via a smart contract as a basic trust credential for subsequent collaboration; wherein, the historical interaction records include the number of successful task completions, average response time, and historical collaboration success rate; the reputation score is generated and updated by other Agents or users based on historical collaboration results.

[0024] Step 102: Receive the target task submitted by the user, decompose the target task into multiple target sub-tasks, and encode the task flow of each target sub-task, the intelligent agent capability requirements of the target sub-task, the cooperation rules between target sub-tasks, and the reward and punishment mechanism into a smart contract. Specifically, step 102 includes: Complex tasks submitted by users, such as "researching competitor A's latest market dynamics and generating an analysis report," are parsed into multi-stage processes by the task decomposition module and encoded into the main smart contract. This contract explicitly defines: Task flow: Clarify the execution order of each collaboration stage and the type of Agent capabilities required for each stage; for example, stage 1 requires a data collection Agent, and stage 2 requires a data analysis Agent. Collaboration rules: Define the handover conditions between agents in adjacent stages. For example, the data collection agent needs to submit the IPFS storage address and hash value of the data, and the subsequent analysis can only be started after the data analysis agent verifies it. Reward and penalty mechanism: pre-allocated reward amount for each stage, such as 10 tokens for successful data collection and 20 tokens for completed analysis; staking rules, such as deducting 5 staked tokens and failure penalty clauses if the target is not met.

[0025] Step 103: Select the appropriate intelligent agent based on the intelligent agent capability requirements required for the target sub-task and the corresponding reputation evaluation data of the intelligent agent. During the execution of the target sub-task by the selected intelligent agent, submit the key behavioral data of the selected intelligent agent that meets the preset requirements to the blockchain in the form of a transaction for evidence storage. Specifically, step 103 includes: When executing a task, each agent must submit its key operational steps to the blockchain for notarization in the form of transactions, specifically including: After completing the task, the data collection agent submits the storage address of the collected data (such as IPFS hash), the data digest (such as SHA-256 hash value), and the collection timestamp to the smart contract. When the data analysis agent calls an external API, it records the API name, request parameters, response time, and key characteristics of the returned result (such as digest hash). When the Agent generates the final report, it includes the hash value of the report content and intermediate parameters from the generation process.

[0026] All evidence storage data is completed through the smart contract's write interface, ensuring that it is tamper-proof and can be audited by other agents and users.

[0027] Step 104: Based on the collaboration rules between target sub-tasks, automatically advance the collaboration to the next stage according to the evidence verification results of the previous stage, and trigger the preset arbitration logic to handle abnormal situations when disputes occur.

[0028] Specifically, step 104 includes: As a state machine, a smart contract internally maintains a state indicator for the current collaboration stage, such as "waiting for data collection to complete," "data analysis in progress," or "report generation complete." Only when the goal of the previous stage is achieved, such as when the hash value submitted by the data collection agent is verified as valid, will the contract automatically trigger a state transition to the next stage and release the corresponding agent's reward tokens. If a dispute arises, such as a downstream agent deeming the upstream data unqualified, the contract's built-in decentralized arbitration logic will be triggered. Invoke preset verification rules, such as data hash comparison and API call parameter compliance check; If verification fails, the responsible party's deposit will be deducted and the task will be reassigned. If a dispute cannot be resolved automatically, the final status can be determined by a vote of users or the community.

[0029] In this embodiment, the blockchain is a distributed ledger that supports smart contracts (such as Ethereum or Fabric), and each AI Agent interacts with the blockchain nodes through a decentralized network to achieve behavior storage and state synchronization.

[0030] The present invention also provides a multi-agent collaborative behavior evidence storage and trusted auditing system. The multi-agent collaborative behavior evidence storage and trusted auditing system can be implemented by executing the process steps of the multi-agent collaborative behavior evidence storage and trusted auditing method. That is, those skilled in the art can understand the multi-agent collaborative behavior evidence storage and trusted auditing method as a preferred implementation of the multi-agent collaborative behavior evidence storage and trusted auditing system.

[0031] This invention addresses the technical problems of lack of trust, unauditable behavior, and difficulty in value distribution in existing technologies by encoding complex task intentions into smart contracts and utilizing the immutability of blockchain to achieve on-chain evidence storage of agent behavior and automatic advancement of collaborative status. This results in the construction of a trustworthy, traceable, and auditable multi-agent collaboration framework.

[0032] Example 2 Example 2 is a preferred example of Example 1. A method for evidence storage and trusted auditing of multi-agent cooperative behavior provided by the present invention includes: Step 201: Agent on-chain digital identity registration and binding; Step 202: Smart contract encoding of task intent; Step 203: On-chain evidence storage of key actions; Step 204: Autonomous Collaboration and Dispute Resolution.

[0033] Specifically, such as Figure 2 The diagram shown is an overall flowchart for multi-agent collaboration.

[0034] In one implementation, step 203 specifically includes: Step 2031: After the data collection agent completes the task, it extracts the storage address of the collected data, such as the IPFS CID and content hash value; Step 2032: Package the hash value, collection timestamp, and data source identifier, such as API name or webpage URL, into a notarized transaction and submit it to the smart contract; Step 2033: The smart contract verifies the transaction signature and writes the evidence data to the blockchain to ensure it comes from a legitimate agent. Step 2034: When the data analysis agent receives data, it queries the hash value of the upstream data through the contract and recalculates and compares it with the hash of the locally received data; Step 2035: If the comparison is successful, the status will be transferred to the "data analysis stage"; otherwise, an exception will be recorded and the process will be paused.

[0035] In one implementation, step 204 specifically includes: Step 2041: The smart contract monitors the target conditions of the current stage, such as data hash verification passing and API call parameters being compliant; Step 2042: If the conditions are met, automatically release the next stage reward tokens and update the status to "Next Agent is Executable"; Step 2043: If the downstream Agent submits a dispute, such as declaring that the upstream data is incomplete, the contract's built-in verification script is triggered, such as recalculating the hash or checking the API logs; Step 2044: If verification fails, deduct the responsible party's stake, such as 5 tokens of the data collection agent, and redistribute or terminate the task; if the dispute is complex, initiate a community voting mechanism to let users decide the final result.

[0036] The specific embodiments described in this invention are for illustrative purposes only. In actual deployment, they can be adapted to different blockchain platforms (such as Ethereum and Fabric) and smart contract languages ​​(such as Solidity and Go), and support the expansion of more agent types (such as hardware control agents and financial analysis agents) and collaborative scenarios (such as supply chain collaboration and scientific research experiment automation). Those skilled in the art can adjust the module details according to their needs without departing from the core ideas of this invention.

[0037] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0038] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for evidence storage and trusted auditing of multi-agent collaborative behavior, characterized in that, include Step S1: Create a unique on-chain digital identity for each intelligent agent, and bind and store the digital identity with its historical interaction records and reputation evaluation data on the blockchain; Step S2: Receive the target task submitted by the user, decompose the target task into multiple target sub-tasks, and encode the task flow of each target sub-task, the intelligent agent capability requirements of the target sub-task, the cooperation rules between target sub-tasks, and the reward and punishment mechanism into a smart contract. Step S3: Select the appropriate intelligent agent based on the intelligent agent's capability requirements for the target sub-task and the corresponding reputation evaluation data of the intelligent agent. During the execution of the target sub-task by the selected intelligent agent, submit the key behavioral data of the selected intelligent agent that meets the preset requirements to the blockchain in the form of a transaction for evidence storage. Step S4: Based on the collaboration rules between target sub-tasks, automatically advance the collaboration to the next stage according to the evidence verification results of the previous stage, and trigger the preset arbitration logic to handle abnormal situations when disputes occur.

2. The method for evidence storage and trusted auditing of multi-agent cooperative behavior according to claim 1, characterized in that, Step S1 includes: Step S1.1: Generate the on-chain digital identity of the smart agent using a decentralized identity identifier (DID) or a blockchain wallet address; Step S1.2: Bind and store the on-chain digital identity with its historical interaction records and reputation evaluation data on the blockchain; The historical interaction records include: the number of times the agent successfully completed tasks, the average response time, and the historical collaboration success rate. The reputation evaluation data is generated and updated by other intelligent agents or users based on historical interaction records.

3. The method for evidence storage and trusted auditing of multi-agent cooperative behavior according to claim 1, characterized in that, The key behavioral data that meet the preset requirements include: The data acquisition agent submits the data storage address and data content hash value; The key feature hashes of the external API name, request parameters, response timestamp, and returned results called by the data analysis agent meet the preset requirements; The result is the hash value of the final report submitted by the generating agent and a summary of the intermediate parameters of the generation process; The key behavioral data that meets the preset requirements is recorded in the blockchain ledger in an immutable transaction form through the smart contract's write interface.

4. The method for evidence storage and trusted auditing of multi-agent cooperative behavior according to claim 1, characterized in that, Step S4 includes: Step S4.1: Maintain the state of the agent in the current collaboration phase, including: waiting for data collection to complete, data analysis in progress, and report generation completed; Step S4.2: When the handover conditions between agents in adjacent stages are met, the state transition to the next stage is triggered, and the corresponding reward tokens are released according to the reward and punishment mechanism; Step S4.3: When a downstream agent raises an objection to the upstream action, a preset verification script is invoked for automatic arbitration. If the verification fails, the responsible party's deposit is deducted according to the reward and punishment mechanism and the task is redistributed. If the preset verification script cannot complete the automatic arbitration, the current dispute is considered to be a complex situation, and a community voting mechanism is initiated for users to decide the final state. The preset verification scripts include: data content hash consistency verification, API call log integrity check, and collaboration rule parameter range verification.

5. The method for evidence storage and trusted auditing of multi-agent cooperative behavior according to claim 1, characterized in that, The method further includes: the blockchain is a distributed ledger that supports smart contracts, and each smart agent interacts with the blockchain node through a decentralized network to achieve behavior storage and state synchronization.

6. A multi-agent collaborative behavior evidence storage and trusted auditing system, characterized in that, include Module M1: Creates a unique on-chain digital identity for each intelligent agent and binds and stores the digital identity with its historical interaction records and reputation evaluation data on the blockchain; Module M2: Receives the target task submitted by the user, decomposes the target task into multiple target sub-tasks, and encodes the task flow of each target sub-task, the intelligent agent capability requirements of the target sub-task, the cooperation rules between target sub-tasks, and the reward and punishment mechanism into a smart contract. Module M3: Based on the agent capability requirements of the target sub-task and the corresponding reputation evaluation data of the agent, select the appropriate agent. During the execution of the target sub-task by the selected agent, submit the key behavioral data of the agent that meets the preset requirements to the blockchain in the form of transactions for storage. Module M4: Based on the collaboration rules between target subtasks, it automatically advances the collaboration to the next stage according to the evidence verification results of the previous stage, and triggers preset arbitration logic to handle abnormal situations when disputes occur.

7. The multi-agent cooperative behavior evidence storage and trusted auditing system according to claim 6, characterized in that, The module M1 includes: Module M1.1: Generates the on-chain digital identity of the smart agent through decentralized identity identifier DID or blockchain wallet address; Module M1.2: Binds and stores the on-chain digital identity with its historical interaction records and reputation evaluation data on the blockchain; The historical interaction records include: the number of times the agent successfully completed tasks, the average response time, and the historical collaboration success rate. The reputation evaluation data is generated and updated by other intelligent agents or users based on historical interaction records.

8. The multi-agent cooperative behavior storage and trusted auditing system according to claim 6, characterized in that, The key behavioral data that meet the preset requirements include: The data acquisition agent submits the data storage address and data content hash value; The key feature hashes of the external API name, request parameters, response timestamp, and returned results called by the data analysis agent meet the preset requirements; The result is the hash value of the final report submitted by the generating agent and a summary of the intermediate parameters of the generation process; The key behavioral data that meets the preset requirements is recorded in the blockchain ledger in an immutable transaction form through the smart contract's write interface.

9. The multi-agent cooperative behavior storage and trusted auditing system according to claim 6, characterized in that, The module M4 includes: Module M4.1: Maintains the state of the agents in the current collaboration phase, including: waiting for data collection to complete, data analysis in progress, and report generation completed; Module M4.2: When the handover conditions between agents in adjacent stages are met, the state transition to the next stage is triggered, and the corresponding reward tokens are released according to the reward and punishment mechanism; Module M4.3: When a downstream agent raises an objection to the upstream action, a preset verification script is invoked for automatic arbitration. If the verification fails, the responsible party's deposit is deducted according to the reward and punishment mechanism and the task is reassigned. If the preset verification script cannot complete the automatic arbitration, the current dispute is considered to be a complex situation, and a community voting mechanism is initiated for users to decide the final state. The preset verification scripts include: data content hash consistency verification, API call log integrity check, and collaboration rule parameter range verification.

10. The multi-agent cooperative behavior storage and trusted auditing system according to claim 6, characterized in that, The method further includes: the blockchain is a distributed ledger that supports smart contracts, and each smart agent interacts with the blockchain node through a decentralized network to achieve behavior storage and state synchronization.

Images