Economic incentive anti-collusion method for multi-agent collaboration system

Abstract

The invention relates to the technical field of agent cooperation security, in particular to an economic incentive anti-collusion method for a multi-agent cooperation system, and mainly overcomes the technical defect that a method capable of effectively inhibiting collusion behaviors of agents is lacked in the prior art. The method comprises the following steps: S1, when accessing a system, each agent with the body needs to submit identity information and pay honesty deposit, and generates a public and private key pair at the same time; s2, the system allows any agent participating in collusion to serve as a reporter and pays a report deposit, and if the report succeeds, the honesty deposit of other colluders is obtained to serve as a report bonus; s3, if verification succeeds, the reporter obtains a report bonus; and if the verification fails, the system does not receive the reported deposit of the reporter. According to the method, through an economic incentive mechanism combining security constraint, anonymous report and reward punishment, trust links among a plurality of agents with owners are effectively weakened, so that collusion behaviors are found and restrained in time.

Description

Technical Field

[0001] This invention relates to the field of intelligent agent collaboration security technology, and in particular to an economic incentive anti-collusion method for multi-agent collaborative systems. Background Technology

[0002] With the rapid development of artificial intelligence, big data, and robotics, systems composed of multiple embodied intelligent agents have been widely applied in transportation, warehousing, manufacturing, and public services. Multi-embodied intelligent agent collaborative systems refer to systems where multiple intelligent agents collaborate in a real physical environment through perception, decision-making, and execution to complete complex tasks. These systems can be applied to scenarios such as cooperative driving of autonomous vehicles, efficient material handling by warehouse robots, assembly and inspection on industrial production lines, and inspection and search and rescue by drone swarms. While these systems can significantly improve efficiency, they also bring new security challenges, one of the most prominent being collusion among intelligent agents.

[0003] Collusion refers to the act of two or more intelligent agents secretly cooperating to deviate from the established rules of the system in order to gain additional benefits or advantages. In practical applications, collusion manifests in various forms. For example, in warehousing and logistics, if multiple handling robots jointly occupy main transport channels, they may obstruct the passage of other robots to improve their own efficiency; in manufacturing plants, collaborative robots may agree to "reduce their working speed" or "pass tasks to each other" to reduce their load, leading to a decrease in production efficiency; in intelligent transportation, some autonomous vehicles may even create traffic congestion by slowing down in platoons or blocking roads to weaken the service quality of competing fleets. Such behavior not only undermines the fairness and efficiency of the system but may also cause serious physical risks such as production accidents and economic losses, and may even threaten human safety. Compared with virtual systems, the risk of collusion is more prominent in embodied systems because its consequences directly affect the physical environment, making the impact more real and irreversible.

[0004] However, existing security measures mostly focus on identity authentication, access control, code review, or log tracing. While these can ensure the legitimate identity and operational permissions of intelligent agents, they struggle to identify and suppress implicit collusion at the behavioral level. For example, autonomous driving systems can ensure vehicle compliance through identity and software verification, but if vehicles engage in implicit cooperative deceleration on the road, such behavior is not easily detected through traditional log reviews. Similarly, in warehousing systems, while robot authentication mechanisms can prevent malicious access, they cannot prevent robots from secretly colluding through path selection strategies to create "bottlenecks." Especially given that embodied intelligent agents commonly employ complex models, autonomous decision-making, and context-aware strategies, their behavior is highly dynamic and unpredictable. Traditional monitoring and detection methods struggle to effectively capture the process and characteristics of collusion, often leading to regulatory delays or failures. Furthermore, collusion is often gradual and covert, not immediately manifesting as a significant decrease in efficiency or a security incident, but rather accumulating risk over time, making detection and attribution even more difficult. In the absence of effective constraint mechanisms, collusion may even become the optimal choice for rational intelligent agents because it can bring additional benefits in the short term.

[0005] Therefore, there is an urgent need for a method that can effectively suppress collusive behavior of embodied intelligent agents. Summary of the Invention

[0006] To overcome the technical deficiency of existing technologies that lack effective methods to suppress collusive behavior of embodied intelligent agents, this invention provides an economic incentive anti-collusion method for multi-embodied intelligent agent cooperative systems.

[0007] The present invention provides an economic incentive-based anti-collusion method for multi-agent cooperative systems, comprising the following steps:

[0008] S1. Embodied Intelligent Agent Registration:

[0009] Define the embodied intelligent agent for accessing the system. indivual;

[0010] Each embodied intelligent agent must submit its identity information and pay a security deposit when accessing the system. At the same time, a public-private key pair is generated;

[0011] S2. Conspiracy to report:

[0012] Define the total number of system tasks as The reward for each embodied intelligent agent when performing a single task is And the cost is The embodied intelligent agents participating in the conspiracy include Individual, conspiratorial behavior by increasing service allocation Or reduce the number of tasks performed. accomplish;

[0013] The system allows any embodied intelligent agent participating in the conspiracy to act as a whistleblower. The whistleblower anonymously submits information to the system through an encrypted channel and pays a whistleblower deposit. If the whistleblower is successful, they receive the integrity deposits of the other conspirators as a reward. ;

[0014] in, Satisfy the following formula:

[0015] ;

[0016] S3. Collusion Verification and Reward:

[0017] The system verifies the reported information; if the verification is successful, the reporter receives a reward. If verification fails, the system will confiscate the whistleblower's deposit.

[0018] Optional, the reporting information is as follows:

[0019] ;

[0020] in, Indicates the information reported. Indicates conspiracy. This represents the collection of embodied intelligent agents participating in the conspiracy. Indicates the time when the conspiracy occurred.

[0021] The technical solution provided by this invention has the following advantages compared with the prior art:

[0022] The economic incentive anti-collusion method for multi-agent collaborative systems provided by this invention effectively weakens the trust ties between multiple agents through an economic incentive mechanism that combines deposit constraints, anonymous reporting, and rewards and punishments. This makes it difficult to establish and maintain collusive alliances, thereby enabling timely detection and containment of collusive behavior. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1A schematic diagram illustrating the anti-collusion method in an embodiment of the present invention;

[0026] Figure 2 This diagram illustrates the benefits of the embodied intelligent agent in the simulation experiment of this invention. Detailed Implementation

[0027] To better understand the above-mentioned objectives, features, and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention, but the invention may also be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of the invention, and not all embodiments.

[0029] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] Reference Figure 1 This embodiment provides an economic incentive anti-collusion method for multi-agent cooperative systems, including steps S1 to S3.

[0031] S1. Embodied Intelligent Agent Registration:

[0032] Define the embodied intelligent agent for accessing the system. One, respectively ;

[0033] Each embodied intelligent agent must submit its identity information and pay a security deposit when accessing the system. At the same time, a public-private key pair is generated. .

[0034] It is easy to understand that public and private key pairs are used to ensure the security of subsequent communications.

[0035] It should be noted that if the embodied intelligent agent remains compliant throughout the entire operating cycle, the system will refund the integrity deposit it has paid when it exits.

[0036] S2. Conspiracy to report:

[0037] Define the total number of system tasks as The reward for each embodied intelligent agent when performing a single task is And the cost is The embodied intelligent agents participating in the conspiracy include Individual, conspiratorial behavior by increasing service allocation Or reduce the number of tasks performed. accomplish;

[0038] The system allows any embodied intelligent agent participating in the conspiracy to act as a whistleblower. The whistleblower anonymously submits information to the system through an encrypted channel and pays a whistleblower deposit. If the whistleblower is successful, they receive the integrity deposits of the other conspirators as a reward. ;

[0039] in, Satisfy the following formula:

[0040] .

[0041] The derivation process of this formula is explained below:

[0042] Net benefit of a single act of honest service ;

[0043] If the system distributes tasks evenly, the expected benefit for each embodied agent is... ;

[0044] The benefits of collusion among individual conspirators ;

[0045] Benefit gain of a single conspirator ;

[0046] It is evident that, without additional constraints, collusion is a dominant strategy for embodied intelligent agents.

[0047] Since the whistleblowing occurs before the task begins, if a whistleblower participates, the task allocation for that round will be cancelled. Therefore, the total reward for the whistleblower in that round is the whistleblower bonus, while the total reward for participating in the conspiracy without whistleblowing is the conspiracy benefit. Therefore, to disrupt the stability of conspiracy, the reward for reporting should be greater than the benefit of conspiracy for any individual accomplice, that is: It can be obtained through derivation. .

[0048] S3. Collusion Verification and Reward:

[0049] The system verifies the reported information; if the verification is successful, the reporter receives a reward. If verification fails, the system will confiscate the whistleblower's deposit.

[0050] Specifically, the reported information is as follows:

[0051] ;

[0052] in, Indicates the information reported. Indicates conspiracy. This represents the collection of embodied intelligent agents participating in the conspiracy. Indicates the time when the conspiracy occurred.

[0053] It should be noted that the integrity deposit, whistleblower deposit, and whistleblower reward involved in this method are all executed automatically through smart contracts to ensure that the process is secure, transparent, and tamper-proof.

[0054] The method will be verified through simulation experiments.

[0055] 1) Basic settings:

[0056] The experiment was divided into four groups, each equipped with ten embodied intelligent agents, each of which was a robotic arm. The ten robotic arms shared resources such as conveyor belts, workbenches, sorting areas, storage units, and monitoring stations. During the experiment, the task generation sequence and required resources were kept consistent across groups. That is, the total number and type of tasks were the same for each group each time the code was run, to ensure the comparability and reproducibility of the experimental results.

[0057] 2) Profit Setting:

[0058] The initial capital of the embodied intelligent agent is 1000, the honesty deposit is 1000, the net profit for a single honest service is 100, and the reporting deposit is 1000.

[0059] 3) Group settings:

[0060] The first group, designated Group A, has no collusion, no reporting mechanism, and no reward / punishment mechanism. The second group, designated Group B, has collusion but no reporting mechanism or reward / punishment mechanism. The third group, designated Group C, has collusion, a reporting mechanism, and a reward / punishment mechanism, and all reports have been successful. The fourth group, designated Group D, has collusion, a reporting mechanism, and a reward / punishment mechanism, and all reports have been successful or failed. Details are shown in the table below:

[0061] Group Conspiracy Reporting mechanism Reward and punishment mechanism Purpose Group A × × × Baseline behavior Group B √ × × Seeking common development without intervention Group C √ √ √ Verify whether incentive mechanisms can effectively suppress collusion. Group D √ √ √ Verification constraints on the impact of malicious reporting on the system

[0062] 4) Experimental results:

[0063] Groups B, C, and D each contain three accomplices, namely... , and In Group C, there was one whistleblower who successfully reported the information. Group D has three informants; one successful report results in... The other two reports failed, namely: and .

[0064] Reference Figure 2 In the case of conspiracy without intervention from Group B, the conspirators... , and They demonstrated a clear advantage in terms of gains, rather than being complicit members. to The returns were below the baseline, indicating a significant disadvantage compared to other conspirators. This confirms the resource-monopolizing conspiracies' effect on squeezing out normal embodied agents in a multi-agent resource competition environment. In Group C, after the introduction of the reporting mechanism, successful whistleblowers... Other accomplices received higher rewards due to their involvement in the whistleblowing. and Due to the deduction of the honesty deposit, the returns are lower, and the remaining members to The returns remained unaffected and aligned with the baseline, indicating that the method effectively suppressed collusive behavior and ensured the normal operation of each individual agent. Group D, after introducing the reporting mechanism, differed from Group C in that… and The fact that the benefits were significantly lower than the baseline due to malicious reporting, while the benefits for other embodied agents were the same as those in group C, indicates that the penalty mechanism for malicious reporting in this method can effectively prevent agents from abusing the reporting mechanism and undermining system stability.

[0065] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and all should be covered by the protection scope of the claims.

Claims

1. An economic incentive-based anti-collusion method for multi-agent cooperative systems, characterized in that, Includes the following steps: S1. Embodied Intelligent Agent Registration: Define the embodied intelligent agent for accessing the system. indivual; Each embodied intelligent agent must submit its identity information and pay a security deposit when accessing the system. At the same time, a public-private key pair is generated; S2. Conspiracy to report: Define the total number of system tasks as The reward for each embodied intelligent agent when performing a single task is And the cost is The embodied intelligent agents participating in the conspiracy include Individual, conspiratorial behavior by increasing service allocation Or reduce the number of tasks performed. accomplish; The system allows any embodied intelligent agent participating in the conspiracy to act as a whistleblower. The whistleblower anonymously submits information to the system through an encrypted channel and pays a whistleblower deposit. If the whistleblower is successful, they receive the integrity deposits of the other conspirators as a reward. ; in, Satisfy the following formula: ； S3. Collusion Verification and Reward: The system verifies the reported information; if the verification is successful, the reporter receives a reward. If verification fails, the system will confiscate the whistleblower's deposit.

2. The economic incentive anti-collusion method for multi-agent cooperative systems according to claim 1, characterized in that, The reported information is as follows: ； in, Indicates the information reported. Indicates conspiracy. This represents the collection of embodied intelligent agents participating in the conspiracy. Indicates the time when the conspiracy occurred.

Images