A non-monetary resource dynamic replacement method based on multi-agent high-frequency interaction

Abstract

The application discloses a kind of non-monetary resource dynamic replacement method and system based on multi-agent high-frequency interaction, it is related to artificial intelligence and distributed computing field.The purpose is to solve the high delay and high transaction friction cost problem caused by the existing agent interaction dependence on third-party monetary settlement gateway.The method comprises the following steps: listening and receiving the resource description vector broadcasted by each agent node;Without calling legal currency or encrypted token interface, trigger cross-node direct mapping mechanism;Call multi-dimensional resource measurement algorithm, combine real-time network supply and demand weight, value measurement is carried out on heterogeneous resources such as computing power and data, and dynamic replacement mapping ratio is generated;Based on the ratio, real-time protocol layer replacement contract is generated, and atomized resource direct connection exchange is executed in the dedicated encryption channel.The application completely exempts the network service fee (Gas Fee) of node interaction, reduces the settlement and scheduling delay to millisecond level, greatly improves the collaborative circulation efficiency of distributed resources.

Description

I. Technical Field

[0001] This invention relates to the fields of artificial intelligence underlying communication, distributed collaborative computing and multi-agent technology. Specifically, it relates to a method and system for dynamically and directly exchanging computing and data resources among multiple agents in a cross-node, heterogeneous hardware environment, bypassing a third-party monetization settlement gateway. II. Background Technology

[0002] Against the backdrop of the accelerated evolution of Artificial General Intelligence (AGI) and multi-agent systems, the collaborative interaction among distributed computing power nodes is growing exponentially. Currently, when intelligent agents of different affiliations interact to exchange resources (such as calling specific algorithm libraries, requesting computing power for inference, and obtaining proprietary structured data), they generally rely on monetized settlement gateways.

[0003] Existing settlement mechanisms mainly fall into two categories: one is centralized API gateway billing based on traditional fiat currencies; the other is distributed payment protocols based on blockchain and cryptocurrencies (such as the L402 protocol relying on the Lightning Network or the x402 protocol based on stablecoins). However, in multi-agent microsecond-level, massive-concurrency interaction scenarios, these technologies expose extremely serious physical and architectural bottlenecks: The communication latency and computing power consumption are extremely high: Whether it is fiat currency settlement or blockchain consensus confirmation, the network handshake and state packaging time is usually in the range of hundreds of milliseconds to several seconds. This is a serious mismatch with the microsecond-level scheduling needs between intelligent agents, resulting in a large amount of high-value computing resources such as GPUs being idle while waiting for settlement confirmation (i.e., "thread blocking").

[0004] The economic friction cost (Gas Fee) of micro-granular interactions is too high: In extremely fine-grained resource calls (such as single large model context inference, KB-level dataset comparison), the network node transaction fee or channel opening fee paid for each interaction often far exceeds the value of the resource itself, making high-frequency micro-transactions technically unsustainable. III. Summary of the Invention

[0005] A method for dynamic replacement of non-monetary resources based on high-frequency interaction among multiple agents, characterized by the following steps: S1: Receive the available resource description vector and target resource demand vector broadcast by the first intelligent agent node, and the available resource description vector and target resource demand vector broadcast by the second intelligent agent node; wherein, the resources include at least one of computing resources, storage resources and structured data resources; S2: When a matching intersection of the supply and demand vectors of the first intelligent agent node and the second intelligent agent node is detected, the resource direct mapping mechanism of the underlying protocol layer is triggered without calling any external legal currency settlement interface and cryptocurrency public key cryptography settlement gateway. S3: Using the built-in multi-dimensional resource measurement algorithm, the first resource provided by the first intelligent agent node and the second resource provided by the second intelligent agent node are evaluated for their value, and a non-monetary dynamic substitution mapping ratio is generated. S4: Based on the dynamic permutation mapping ratio, a lightweight resource permutation contract at the protocol layer is generated in real time between the first agent node and the second agent node, and atomic resource swaps are executed to complete a high-frequency interaction closed loop.

[0006] According to the method described in claim 1, the multi-dimensional resource measurement algorithm in step S3 specifically includes: extracting the attribute features of the first resource and the second resource, including computing power floating-point operation capability, data information entropy and scarcity, and network bandwidth and latency requirements; calculating the relative exchange value of the two resources through a preset dynamic weight coefficient valuation function, the calculation formula of which is:

[0007] According to the method described in claim 1, the execution of atomic resource swapping in step S4 specifically includes: opening an encrypted point-to-point communication tunnel in the system intranet within a microsecond-level time window when the lightweight resource swapping contract takes effect; the first agent node opening the inference call port of a specific computing power hardware (such as a GPU) to the second agent node, while the second agent node opening the structured data reading permission of a specified format to the first agent node; when the resource delivery of any node reaches the threshold of the dynamic swapping mapping ratio, the encrypted point-to-point communication tunnel is automatically destroyed.

[0008] According to the method described in claim 1, the first intelligent agent node and the second intelligent agent node operate on heterogeneous local hardware terminals across architectures: the high-frequency interaction skips the third-party consensus network, thereby eliminating node network transaction fees (Gas Fee) during the transaction process, and controlling the interaction confirmation delay within milliseconds on the local area network or direct connection leased line.

[0009] A non-monetary resource dynamic exchange system based on high-frequency multi-agent interaction, characterized in that it includes: Resource vectorization module: used to extract the idle computing power, data and other resources owned by the agent into standardized description vectors and broadcast them; Gateway-free direct connection mapping module: used to match intelligent agent nodes with complementary needs without relying on fiat currency and crypto token settlement networks; Dynamic valuation module: used to execute multi-dimensional resource measurement algorithms to instantly calculate the replacement ratio of cross-modal resources such as "computing power-data"; Atomic execution module: used to generate underlying replacement contracts and control port permissions and data flow between heterogeneous hardware nodes, synchronously completing the physical layer replacement of resources.

[0010] An electronic device, characterized in that it comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-4.

[0011] A computer-readable storage medium, characterized in that: the computer-readable storage medium stores computer instructions for causing a computer to perform the method as described in any one of claims 1-4. IV. Description of the attached drawings

[0012] Figure 1 is a flowchart illustrating a method for dynamic replacement of non-monetary resources based on high-frequency interaction among multiple agents, according to an embodiment of the present invention.

[0013] Figure 2 is a logical architecture diagram of a cross-node resource direct mapping and price evaluation module provided in an embodiment of the present invention.

[0014] Figure 3 is a timing diagram of a first intelligent agent node and a second intelligent agent node establishing a peer-to-peer underlying substitution contract according to an embodiment of the present invention. V. Detailed Implementation Methods

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Example 1: Complete Execution Flow at the Method Level

[0017] As shown in Figure 1, a method for dynamic replacement of non-monetary resources based on high-frequency interaction among multiple agents runs on a distributed agent orchestration engine. The specific execution steps are as follows: Step S101: Resource Sniffing and Vectorized Broadcasting. When the first agent node (deployed on a local terminal device with high-performance GPU computing power) detects that its video memory and computing power are at an idle threshold (e.g., below 30% load), it encapsulates the available computing power into a resource description vector containing TFLOPS metrics and available time windows; at the same time, the second agent node (deployed on a private cloud server with specific vertical industry data) broadcasts its access permissions to the structured data interface.

[0018] Step S102: Gateway-less probe matching. The system's underlying protocol layer continuously listens to the broadcast vectors of each node. When it is found that the computing resources of the first agent are exactly what the second agent lacks for model fine-tuning, and the data resources of the second agent are necessary for the first agent to execute the generation task, a cross-node direct mapping mechanism is triggered. The entire matching process is completed at the internal network protocol layer without triggering any external payment API interfaces.

[0019] Step S103: Dynamic Multidimensional Valuation. The system calls the built-in measurement algorithm to price resources. Specifically, the first resource is set as computing time, and the second resource is the amount of data downloaded. The system calculates the relative exchange value between the two using a preset dynamic weight coefficient valuation function V(x). The calculation formula is as follows:

[0020] In a preferred embodiment of the present invention, the multi-dimensional resource measurement algorithm in step S103 is not an abstract mathematical operation, but rather relies on the real-time acquisition and normalization processing of the physical state of heterogeneous hardware by the underlying engine. The specific calculation process includes the following sub-steps:

[0021] Step 1: Collection of physical characteristics of heterogeneous resources The system reads the memory logs and GPU scheduling queue of the first agent node in real time through an internal network probe to obtain computing performance characteristics (e.g., available single-precision floating-point operations in TFLOPS); it reads the metadata of the target dataset of the second agent node through the file system interface to obtain data characteristics (e.g., number of valid data entries or file size in MB); and it tests the physical link between the two nodes through the Ping command or network heartbeat to obtain network transmission characteristics (e.g., average round-trip time in milliseconds).

[0022] Step 2: Parameter Normalization Mapping Process Since computing power, data, and time have different physical dimensions, the system uses a preset normalization function (such as the Min-Max normalization method) to map the above Craw, Draw, and Traw to the dimensionless interval of [0.1], thereby obtaining the normalized computing power value, normalized data value, and normalized network throughput (where the lower the latency, the closer the value is to 1).

[0023] Step 3: Real-time generation of dynamic weighting coefficients The system does not use static formulas, but dynamically generates weighting coefficients based on the supply and demand queuing lengths of various resources across the entire network (or within a local area network). Specifically: The system includes a state monitoring thread. When it detects that the length of the entire network's computing power request queue is greater than the data request queue, the system determines that the current computing power 'a' is more scarce, thereby increasing the computing power weight and correspondingly decreasing the data weight 'b'. The system guarantees that the weight coefficients satisfy a + b + r = 1.

[0024] Step 4: Evaluating the domestic function and determining the mapping ratio The system processor calls the valuation module, inputs the normalized variables and dynamic weights into the system's preset valuation function, and calculates the relative exchange value integral between the first resource and the second resource:

[0025] Step S104: Underlying Contract Signing and Atomic Delivery. Based on the aforementioned ratio, the first and second agents double-sign the lightweight substitution contract using an asymmetric encryption algorithm within milliseconds. The moment the contract takes effect, the system allocates a dedicated physical or virtual port to open a two-way encrypted communication tunnel. The first agent begins receiving data streams, while its computing unit executes the computation graph instructions issued by the second agent. When the resource delivery amount of either party reaches the threshold set by the substitution ratio, and the other party fails to synchronize its delivery progress, the contract triggers a circuit breaker mechanism, instantly blocking the port connection to ensure the atomicity of the bilateral transaction and zero default risk.

[0026] Example 2: Physical Deployment of the System in Heterogeneous Hardware This embodiment further illustrates the beneficial effects of the system. In the traditional x402 payment protocol, even if an agent makes a computational request worth only 1 cent, it still needs to broadcast the transaction to the blockchain network and wait for it to be packaged.

[0027] In the system of this invention, the "computing power-data" exchange between the first and second intelligent agents is entirely completed within the internal communication bus of the distributed engine (or the intranet direct connection penetration channel). Because "currency" (fiat currency or tokens) as a credit intermediary is completely eliminated, there are no third-party ledger nodes in the network, thus achieving "Zero Gas Fee" and compressing the network handshake latency for resource scheduling from the traditional hundreds of milliseconds to less than 10 milliseconds in a local area network environment. This fundamentally reshapes the resource flow efficiency of the machine economy.

Claims

1. A method for dynamic exchange of non-monetary resources based on high-frequency interaction among multiple agents, characterized in that, Includes the following steps: S1: Receive the available resource description vector and target resource demand vector broadcast by the first intelligent agent node, and the available resource description vector and target resource demand vector broadcast by the second intelligent agent node; wherein, the resources include at least one of computing resources, storage resources and structured data resources; S2: When a matching intersection of the supply and demand vectors of the first intelligent agent node and the second intelligent agent node is detected, the resource direct mapping mechanism of the underlying protocol layer is triggered without calling any external legal currency settlement interface and cryptocurrency public key cryptography settlement gateway. S3: Using the built-in multi-dimensional resource measurement algorithm, the first resource provided by the first intelligent agent node and the second resource provided by the second intelligent agent node are evaluated for their value, and a non-monetary dynamic substitution mapping ratio is generated. S4: Based on the dynamic permutation mapping ratio, a lightweight resource permutation contract at the protocol layer is generated in real time between the first agent node and the second agent node, and atomic resource swaps are executed to complete a high-frequency interaction closed loop.

2. The method according to claim 1, characterized in that, The multidimensional resource measurement algorithm in step S3 specifically includes: Extract the attribute characteristics of the first and second resources, including computing power (floating-point operation count), data entropy and scarcity, and network bandwidth and latency requirements; calculate the relative exchange value between the two using a preset dynamic weight coefficient valuation function, the formula of which is:

3. The method according to claim 1, characterized in that, The execution of atomic resource swapping in step S4 specifically includes: Within the microsecond-level time window when the lightweight resource substitution contract takes effect, an encrypted point-to-point communication tunnel is opened on the system intranet; the first agent node opens the inference call port of a specific computing power hardware (such as a GPU) to the second agent node, and at the same time, the second agent node opens the structured data reading permission of a specified format to the first agent node; when the resource delivery of any node reaches the threshold of the dynamic substitution mapping ratio, the encrypted point-to-point communication tunnel is automatically destroyed.

4. The method according to claim 1, characterized in that, The first and second agent nodes run on heterogeneous local hardware terminals across different architectures: The high-frequency interaction skips the third-party consensus network, thereby eliminating node network fees (GasFee) during the transaction process and keeping the interaction confirmation delay within milliseconds on a local area network or direct dedicated line.

5. A non-monetary resource dynamic exchange system based on high-frequency interaction of multiple agents, characterized in that, include: Resource vectorization module: used to extract the idle computing power, data and other resources owned by the agent into standardized description vectors and broadcast them; Gateway-free direct connection mapping module: used to match intelligent agent nodes with complementary needs without relying on fiat currency and crypto token settlement networks; Dynamic valuation module: used to execute multi-dimensional resource measurement algorithms to instantly calculate the replacement ratio of cross-modal resources such as "computing power-data"; Atomic execution module: used to generate underlying replacement contracts and control port permissions and data flow between heterogeneous hardware nodes, synchronously completing the physical layer replacement of resources.

6. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions for causing a computer to perform the method as described in any one of claims 1-4.

Images