Method and device for constructing decentralized intelligent agent network by using computing power of intelligent computing center cloud platform

Abstract

The application provides a method and device for building a decentralized intelligent agent network by using computing power of an intelligent computing center cloud platform, and relates to the technical fields of intelligent computing centers, intelligent computing centers and computing power infrastructure. The method is executed by a first intelligent agent and includes the following steps: S1: sending broadcast information and receiving broadcast information sent by other intelligent agents; the broadcast information contains the capability information of the intelligent agent sending the broadcast information, a routing table maintained by the intelligent agent sending the broadcast information and an intelligent agent capability table; S2: updating the maintained routing table and intelligent agent capability table according to the received broadcast information; S3: receiving a user task sent by an edge intelligent agent, decomposing the user task into multiple subtasks, selecting a second intelligent agent for executing the subtasks according to the maintained intelligent agent capability table and routing table, and distributing the subtasks to the second intelligent agent; and S4: receiving the subtask execution results of the second intelligent agent, generating a task execution result and sending the task execution result to the edge intelligent agent.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent computing centers, wisdom computing centers and computing power infrastructure, in particular to a method and device for building a decentralized agent network by computing power of an intelligent computing center cloud platform. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "wisdom computing centers" have emerged.

[0003] An "intelligent computing center" refers to a facility that provides required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model inference scenarios) by using large-scale heterogeneous computing power resources, including general computing power and intelligent computing power. An intelligent computing center covers facilities, hardware and software, and can provide full-stack capabilities from bottom computing power to top application enablement.

[0004] An "intelligent computing center" includes but is not limited to a "wisdom computing center".

[0005] A "wisdom computing center", also known as an artificial intelligence computing center, is a type of computing power infrastructure based on artificial intelligence theory and using artificial intelligence computing architecture to provide computing power services, data services and algorithm services required by artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "wisdom computing centers". It is the ability of computer equipment or computing / data centers to process information. It is the ability of computer hardware and software to work together to perform certain computing needs. It is the computing power that achieves the output of target results by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity and data storage power, and mainly provides services to society through computing power infrastructure.

[0007] Agents based on large language models or multi-modal large models have become the core components of the intelligent computing center service ecosystem. Intelligent computing centers provide strong computing power support and resource scheduling capabilities for agents. Current agent technology is evolving from single agent to multi-agent system to cope with complex task scenarios and distributed task processing needs. In existing multi-agent systems, although a multi-agent collaboration mechanism is introduced, a centralized collaboration mechanism is generally used to manage the behavior and resource allocation of all agents. This multi-agent system requires the topology of the agent to be determined at the system design stage and known to all participating agents. When the number of agents increases significantly and needs to adapt to complex tasks, the existing solution causes the multi-agent system to lack further dynamic expansion capability, easily forms a performance bottleneck and increases the risk of failure, limiting the flexibility and robustness of the multi-agent system. SUMMARY

[0008] The application provides a method and device for building a decentralized intelligent agent network by using computing power of an intelligent computing center cloud platform, and aims to solve the problem that the multi-agent system adopting a centralized cooperation mechanism lacks further dynamic expansion capability, is prone to performance bottleneck and increased failure risk, and limits flexibility and robustness of the multi-agent system.

[0009] To solve the above technical problems, the application is implemented as follows:

[0010] In a first aspect, the application provides a method for building a decentralized intelligent agent network by using computing power of an intelligent computing center cloud platform, comprising the following steps:

[0011] Step S1: a first intelligent agent sends broadcast information and receives broadcast information sent by other intelligent agents, wherein the broadcast information contains capability information of the intelligent agent sending the broadcast information, a routing table maintained by the intelligent agent sending the broadcast information, and an intelligent agent capability table;

[0012] Step S2: the first intelligent agent updates the routing table and the intelligent agent capability table maintained by itself according to the received broadcast information;

[0013] Step S3: when the first intelligent agent receives a user task sent by an edge intelligent agent, the first intelligent agent decomposes the user task into multiple subtasks, selects a second intelligent agent for executing each subtask according to the intelligent agent capability table maintained by itself, and distributes the subtasks to the second intelligent agent according to the routing table maintained by itself;

[0014] Step S4: the first intelligent agent receives subtask execution results returned by the second intelligent agent, generates a task execution result based on all the subtask execution results, and sends the task execution result to the edge intelligent agent;

[0015] In the routing table, a third intelligent agent reachable by the intelligent agent maintaining the routing table and a forwarding path between the third intelligent agent and the intelligent agent are contained, and in the intelligent agent capability table, capability information of the third intelligent agent is contained.

[0016] Optionally, the first intelligent agent comprises at least one of the following:

[0017] An intelligent agent deployed inside the intelligent computing center;

[0018] An intelligent agent deployed outside the intelligent computing center;

[0019] And / or, the other intelligent agents comprise at least one of the following:

[0020] An intelligent agent deployed in a first intelligent computing center, wherein the first intelligent computing center is an intelligent computing center to which the first intelligent agent belongs;

[0021] an agent deployed in the second intelligent computing center;

[0022] an agent deployed outside the intelligent computing center.

[0023] Optionally, the step S3 comprises:

[0024] Step S31: when the agent capability table maintained by the first agent contains multiple candidate agents with the same or similar capabilities, obtaining a comprehensive value score of the multiple candidate agents, and selecting a second agent from the multiple candidate agents to execute the subtask according to the comprehensive value score of the multiple candidate agents.

[0025] The comprehensive value score is determined based on at least one of the following target indicators: computing node performance, network performance indicator, historical performance indicator, load state indicator, and cost benefit indicator.

[0026] Optionally, the comprehensive value score is a weighted sum of at least two of the target indicators.

[0027] Optionally, the step S3 further comprises:

[0028] Step S32: the first agent collects the target indicators of each agent in the agent capability table maintained by the first agent, and calculates the comprehensive value score of each agent in the agent capability table maintained by the first agent according to the collected target indicators.

[0029] Alternatively, step S33: the first agent receives the comprehensive value score reported by other agents.

[0030] Optionally, before the step S1, further comprising:

[0031] Step S0: the first agent determines capability information; when the first agent is an agent in a first intelligent computing center for executing task decomposition and distribution, the capability information determined by the first agent includes the capability information of all agents that can be aggregated by the first intelligent computing center.

[0032] In a second aspect, the present application provides a device for building a decentralized agent network by computing power of an intelligent computing center cloud platform, comprising:

[0033] a transceiving module, configured to send broadcast information and receive broadcast information sent by other agents, wherein the broadcast information contains the capability information of the agent sending the broadcast information, the routing table and the agent capability table maintained by the agent sending the broadcast information;

[0034] An updating module is configured to update a routing table and an agent capability table maintained by the intelligent computing center cloud platform according to the received broadcast information.

[0035] A scheduling module is configured to, in a case where a user task sent by an edge agent is received, decompose the user task into a plurality of subtasks, select a second agent for executing each of the subtasks according to the agent capability table maintained by the intelligent computing center cloud platform, and distribute the subtasks to the second agent according to the routing table maintained by the intelligent computing center cloud platform.

[0036] A feedback module is configured to receive a subtask execution result returned by the second agent, generate a task execution result based on all the subtask execution results, and send the task execution result to the edge agent.

[0037] The routing table includes a third agent reachable by the agent maintaining the routing table and a forwarding path between the agent and the third agent, and the agent capability table includes capability information of the third agent.

[0038] In a third aspect, the present application provides an electronic device, comprising a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform according to the first aspect.

[0039] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the steps of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform according to the first aspect.

[0040] In a fifth aspect, the present application provides a computer program product comprising computer instructions, wherein the computer instructions, when executed by a processor, implement the steps of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform according to the first aspect.

[0041] In the present application, the dynamic routing mechanism is used to complete the discovery and connection between agents, thereby dynamically building a multi-agent system with a decentralized network. The agents can join or exit the network at any time without knowing the global topology structure in advance, so that the multi-agent system can have a dynamic expansion capability, reduce performance bottlenecks and failure risks, and greatly improve the flexibility and robustness of the multi-agent system. Meanwhile, through the self-organizing cooperation mode between agents, the decomposition of complex tasks and the allocation of computing power resources are realized, thereby efficiently completing the processing of complex tasks. BRIEF DESCRIPTION OF DRAWINGS

[0042] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments with reference to the drawings. The drawings are for purposes of illustration only and are not intended to limit the present application thereto. Identical reference numerals have been used throughout the drawings to denote identical elements. In the drawings:

[0043] Figure 1 A flowchart of a method for constructing a decentralized agent network by using the computing power of the intelligent computing center cloud platform of the present application;

[0044] Figure 2 A structural diagram of the multi-agent system of the present application;

[0045] Figure 3 A structural diagram of the device for constructing a decentralized agent network by using the computing power of the intelligent computing center cloud platform of the present application;

[0046] Figure 4 A structural diagram of the electronic device of the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0048] The technical terms involved in the present application will be described simply below.

[0049] The "computing power" described in the present application refers to the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to perform certain computing requirements, the computing power to achieve target result output by processing information data, and a new type of productivity integrating information computing power, network carrying capacity, and data storage capacity, which mainly provides services to society through computing power infrastructure.

[0050] The "Computational Power (CP)" described in the present application refers to: the ability of a data center server to process data and achieve result output, which is a comprehensive index for measuring the computing power of a data center, including general computing power, supercomputing power and intelligent computing power. The commonly used unit of measurement is the number of floating point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS), and the larger the value represents the stronger the comprehensive computing power. According to estimates, 1 EFLOPS is about 5 Tianhe 2A or 500,000 mainstream server CPUs or 2 million mainstream notebook computers. The calculation formula is: CP = CP 通用 + CP 智能 + CP 超级 .

[0051] The "Network Power (NP)" described in the present application refers to: the performance of the computing power facility data transmission capability, including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, and other comprehensive capabilities, involving data center internal and data center network transmission, which is a comprehensive index for measuring network transmission scheduling capability.

[0052] The "Storage Power (SP)" described in the present application refers to: the comprehensive capability of data center in data storage capacity, performance, safety and reliability, and green low carbon, which is a comprehensive index for measuring the data storage capability of data center, including storage array and other external storage devices and server built-in storage devices. The commonly used unit of measurement for storage capacity is exabyte (EB, 1 EB = 2^60 bytes), the commonly used unit of measurement for performance is the number of read and write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB), and the disaster recovery ratio is an important performance of safety and reliability.

[0053] The "Computing Power Infrastructure" described in the present application refers to: a new type of information infrastructure integrating information computing power, network carrying capacity and data storage power, which can realize centralized computing, storage, transmission and application of information.

[0054] The "new type of information infrastructure" described in the present application refers to: mainly including 5G network, fiber broadband network, backbone network, international communication network, satellite internet and other network infrastructure, data center, general computing power center, intelligent computing center, supercomputing center and other computing power infrastructure, artificial intelligence, blockchain, quantum computing and other new technology facilities.

[0055] The "computing power" described in the present application includes: general computing power, intelligent computing power and supercomputing power.

[0056] The "general computing power" described in the application refers to the computing power provided by a CPU (Central Processing Unit) chip-based server, which is used to support basic general computing such as cloud computing and edge computing.

[0057] The "intelligent computing power" described in the application refers to the large-scale deployment of computing platforms based on GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit) and other special chips for various artificial intelligence innovation applications, such as natural language processing, machine vision, etc.

[0058] The "super computing power" described in the application refers to the computing power provided by supercomputers and other high-performance computing clusters. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for computing in cutting-edge scientific fields, such as planet simulation, drug molecule design, gene analysis, etc.

[0059] The "intelligent computing center" described in the application refers to a facility that provides the required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model inference scenarios) by using large-scale heterogeneous computing resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware and software, and can provide full-stack capabilities from bottom computing power to top application enablement.

[0060] The "intelligent computing center cloud platform" described in the application refers to a cloud computing platform based on the hardware resources and software resources of the intelligent computing center.

[0061] The "intelligent computing center" described in the application includes but is not limited to "intelligent computing center".

[0062] The "intelligent computing center" described in the application, i.e. artificial intelligence computing center, is based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services and algorithm services required by artificial intelligence applications.

[0063] The "computing center" described in the application refers to a facility that uses infrastructure such as wind, fire, water and electricity and IT hardware and software equipment as the main components, with computing power, carrying capacity and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0064] The "supercomputing center" described in the present application refers to a supercomputing data center based on a supercomputer or a large-scale computing cluster, which can provide large-scale computing, storage and network services, and is widely used in aerospace, national defense, oil exploration, climate modeling and genome sequencing.

[0065] The "computing power resource" described in the present application refers to the technology and facilities required for the development of digital society, including but not limited to CPU, GPU, etc. computing resources, switches, routers, etc. network resources, storage arrays, distributed storage, etc. storage resources, firewalls, intrusion detection systems, etc. security resources, and power, water, electricity, etc. support and protection resources.

[0066] The "computing power node" described in the present application refers to the computing resources of a server / container that can handle computing tasks.

[0067] The "large language model" described in the present application refers to a large language model (LLM), which is a language model with a large number of parameters, designed to understand and generate human language, trained on a large amount of text data, and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0068] The "multimodal large model" described in the present application refers to a model that jointly trains text, image, video, audio and other multimodal information, including but not limited to multimodal large language models.

[0069] The "agent" described in the present application refers to an agent that can perceive the environment and take actions to achieve specific goals, and is an important part of the intelligent computing center cloud platform. It can be software, hardware or a system, with autonomy, adaptability and interaction ability. The agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions based on its own learned knowledge and algorithms, and then executes actions to influence the environment or achieve predetermined goals. Agents are widely used in the field of artificial intelligence, commonly found in automated systems, robots, virtual assistants and game characters, and their core is to autonomously learn and continuously evolve to better complete tasks and adapt to complex environments.

[0070] To solve the problem of the existing multi-agent system using a centralized collaboration mechanism that lacks further dynamic expansion capability, is prone to performance bottlenecks and increases the risk of failure, and limits the flexibility and robustness of the system, please refer to Figure 1 The present application provides a method for building a decentralized agent network of an intelligent computing center cloud platform, which comprises:

[0071] Step S1: the first intelligent agent sends broadcast information and receives broadcast information sent by other intelligent agents, the broadcast information containing capability information of the intelligent agent sending the broadcast information, a routing table and an intelligent agent capability table maintained by the intelligent agent;

[0072] In the present application, the first intelligent agent comprises at least one of the following: an intelligent agent deployed inside an intelligent computing center; and an intelligent agent deployed outside the intelligent computing center. That is, the first intelligent agent can be an intelligent agent deployed inside the intelligent computing center or an intelligent agent deployed outside the intelligent center.

[0073] The first intelligent agent sends broadcast information to other intelligent agents, the broadcast information sent by the first intelligent agent containing capability information of the first intelligent agent, a routing table and an intelligent agent capability table maintained by the first intelligent agent. The routing table sent by the first intelligent agent contains a third intelligent agent reachable by the first intelligent agent and a forwarding path between the first intelligent agent and the third intelligent agent, and the intelligent agent capability table sent by the first intelligent agent contains capability information of the third intelligent agent.

[0074] In the present application, the capability information of the intelligent agent is stored in a preset data table in the form of a structured description statement. The capability range of the intelligent agent covers two levels of basic function capability and business application capability. The basic function capability includes but is not limited to: document parsing processing capability, code generation and analysis capability, image recognition and processing capability, and multi-modal information processing capability. The business application capability includes but is not limited to: specialized processing capability of specific business scenarios such as life scenarios and financial data scenarios.

[0075] In the present application, the third intelligent agent refers to a destination node in the routing table, and the forwarding path between the first intelligent agent and the third intelligent agent refers to a forwarding path of information sent by the first intelligent agent to the third intelligent agent.

[0076] Optionally, the first intelligent agent periodically sends broadcast information. The sending period can be set as required.

[0077] The first intelligent agent can also receive broadcast information sent by other intelligent agents, the broadcast information sent by the other intelligent agents containing capability information of the other intelligent agents, a routing table and an intelligent agent capability table maintained by the other intelligent agents. The routing table sent by the other intelligent agents contains a third intelligent agent reachable by the other intelligent agents and a forwarding path between the other intelligent agents and the third intelligent agent, and the intelligent agent capability table sent by the other intelligent agents contains capability information of the third intelligent agent.

[0078] In the present application, the third agent refers to a destination node in a routing table, and the forwarding path between the other agents and the third agent refers to a forwarding path of information sent by the other agents to the third agent.

[0079] In the present application, the first agent receives broadcast information sent by adjacent other agents. It should be noted that the adjacent agents in the present application can be connected by wired mode or by wireless mode.

[0080] Optionally, the other agents periodically send broadcast information. The sending period can be set according to needs. Further optionally, the sending period of each agent (including the first agent and the other agents) can be the same.

[0081] In the present application, optionally, the other agents include at least one of the following:

[0082] The agent deployed in the first intelligent computing center, wherein the first intelligent computing center is an intelligent computing center to which the first agent belongs;

[0083] The agent deployed in the second intelligent computing center;

[0084] The agent deployed outside the intelligent computing center.

[0085] The routing table in the present application can refer to Table 1, which is a routing table maintained by the agent A.

[0086] Table 1

[0087]

[0088] The agent capability table in the present application can refer to Table 2, which is an agent capability table maintained by the agent A.

[0089] Table 2

[0090]

[0091] The agent A can be the first agent or the other agent.

[0092] In the present application, as shown in Table 3, the routing table and the agent capability table can also be integrated into one table.

[0093] Table 3

[0094]

[0095] It should be noted that the format of the above routing table and agent capability table is only an example, and other formats can also be used.

[0096] Step S2: the first agent updates the routing table and the agent capability table maintained by itself according to the received broadcast information;

[0097] In the present application, after the first agent receives the broadcast information sent by other agents, it needs to query whether the identity of the other agent is contained in the routing table and the agent capability table maintained by itself. If not, the identity of the other agent needs to be added to the routing table and the agent capability table, and the forwarding path between the first agent and the other agent is generated in the routing table, and the capability information of the other agent is recorded in the agent capability table. In addition, the agent that can be reached by the first agent through the relay of the other agent and the forwarding path between the first agent and the agent that can be reached through the relay of the other agent also need to be determined according to the routing table of all the other agents, and are updated to the routing table maintained by the first agent. At the same time, the capability information of the agent that can be reached through the relay of the other agent is updated to the agent capability table maintained by the first agent.

[0098] If the identity of the other agent is contained in the routing table and the agent capability table maintained by itself, the first agent also needs to update the forwarding path between the other agent and the agent that can be reached through the relay of the other agent in the routing table maintained by itself, and update the capability information of the other agent and the agent that can be reached through the relay of the other agent recorded in the agent capability table maintained by itself.

[0099] So far, through the above dynamic routing mechanism, the mutual discovery and connection between agents can be completed, thereby dynamically constructing a multi-agent system with a decentralized network.

[0100] Please refer to Figure 2 , Figure 2 is a structural schematic diagram of the multi-agent system of the present application, Figure 2 In the present application, each agent (A, B, C,...) maintains its own routing table and agent capability table, and periodically sends broadcast information. The agent receiving the broadcast information needs to update the routing table and the agent capability table maintained by itself based on the received broadcast information, to complete the mutual discovery and connection between agents. The first agent can be any agent in the present application. Figure 2 In the present application, each agent (A, B, C,...) maintains its own routing table and agent capability table, and periodically sends broadcast information. The agent receiving the broadcast information needs to update the routing table and the agent capability table maintained by itself based on the received broadcast information, to complete the mutual discovery and connection between agents. The first agent can be any agent in the present application.

[0101] Figure 2In the present application, the intelligent computing center 1 and the intelligent computing center 2 each include a plurality of internal intelligent agents, and can also include aggregatable external intelligent agents, wherein the intelligent agent K is an aggregatable external intelligent agent of the intelligent computing center 1, and the intelligent agent I and the intelligent agent J are aggregatable external intelligent agents of the intelligent computing center 2. The "aggregatable external intelligent agent" in the present application refers to an intelligent agent deployed outside the intelligent computing center and capable of cooperating with the intelligent agents inside the intelligent computing center to complete a task.

[0102] The above intelligent agent includes three main modules, one of which is a routing module for sending and receiving broadcast information and maintaining a routing table, a capability information module for determining self-capability information and maintaining an intelligent agent capability table, and a task execution module for executing a task. In addition, some intelligent agents can also have a task decomposition module for executing task decomposition.

[0103] Step S3: In the case that the first intelligent agent receives a user task sent by the edge intelligent agent, the user task is decomposed into a plurality of subtasks, and a second intelligent agent for executing each subtask is selected according to the intelligent agent capability table maintained by the first intelligent agent, and the subtask is distributed to the second intelligent agent according to the routing table maintained by the first intelligent agent.

[0104] It should be noted that the first intelligent agent in the present application is an intelligent agent with task decomposition and computing resource allocation capabilities, which can be any intelligent agent in the above-mentioned decentralized network. Preferably, based on security and computing power considerations, it can be an intelligent agent inside the intelligent computing center. Further preferably, it is an intelligent agent inside the intelligent computing center that serves as a connection with external intelligent agents, such as the intelligent agent A and the intelligent agent F in the above-mentioned intelligent computing center. Figure 2 In the present application, optionally, one intelligent agent for task decomposition and computing resource allocation can be provided for each intelligent computing center.

[0105] In the present application, the edge intelligent agent can be deployed at the network edge or at the user side, and is responsible for the access and preliminary processing of user tasks.

[0106] In an optional manner, when the edge intelligent agent receives a user task sent by a user, the user task can be directly forwarded to the first intelligent agent. In this case, the edge intelligent agent can select the first intelligent agent in various ways, for example, the intelligent agent inside the intelligent computing center closest to it can be selected as the first intelligent agent, or the intelligent agent inside the intelligent computing center with the lowest cost can be selected as the first intelligent agent, etc.

[0107] In another alternative, when the edge agent receives a user task sent by a user, the user task can be preprocessed, for example, the required capabilities of the user task are evaluated, so as to select a first agent that is suitable for the capabilities of the user task, or select a smart computing center that is suitable for the capabilities of the user task. Alternatively, the user task can also be preliminarily decomposed.

[0108] In the present application, the edge agent needs to query the routing table maintained by itself, and send the user task to the first agent according to the forwarding path between the selected first agent and the edge agent queried from the routing table.

[0109] In the present application, the first agent can decompose the user task into a plurality of sub-tasks that can be executed independently, which can include a plurality of sub-tasks that can be executed simultaneously, or a sub-task that needs to wait for the execution of other sub-tasks. The first agent determines the required agent capabilities of each sub-task, and selects a second agent for executing each sub-task from the maintained agent capability table. Each second agent can execute one sub-task, or a plurality of sub-tasks. The first agent distributes the sub-tasks to the corresponding second agents according to the maintained routing table.

[0110] Step S4: The first agent receives the sub-task execution results returned by the second agent, generates a task execution result based on all the sub-task execution results, and sends the task execution result to the edge agent.

[0111] The first agent can combine the sub-task execution results returned by a plurality of second agents to obtain the task execution result, or analyze the sub-task execution results returned by a plurality of second agents to obtain the task execution result, according to specific circumstances.

[0112] The first agent sends the task execution result to the edge agent according to the query routing table and the forwarding path between the first agent and the edge agent in the routing table.

[0113] The routing table (including the routing table of the above-mentioned first agent and the routing table of other agents) contains a third agent that can be reached by the agent that maintains the routing table, and a forwarding path between the third agent and the agent, and the agent capability table contains the capability information of the third agent.

[0114] In the application, through the dynamic routing mechanism, the mutual discovery and connection between agents are completed, thereby dynamically constructing a multi-agent system with a decentralized network, and the agent can join or exit the network at any time without predicting the global topology structure, so that the multi-agent system can have dynamic expansion capability, reduce performance bottleneck and failure risk, and greatly improve the flexibility and robustness of the multi-agent system. At the same time, through the self-organizing cooperation mode between agents, the decomposition of complex tasks and the allocation of computing power resources are realized, thereby efficiently completing the processing of complex tasks.

[0115] It should be noted that the agent in the application is an agent based on a large language model or a multi-modal large model. The "based on a large language model or a multi-modal large model" means that the large language model or the multi-modal large model is called to perform tasks such as reasoning and thinking, design, dialogue with users, scheme design generation and / or code generation by organizing input information and designing prompt words.

[0116] In the application, optionally, the step S3 comprises:

[0117] Step S31: When the agent capability table maintained by the first agent contains multiple candidate agents with the same or similar capabilities, obtaining a comprehensive value score of the multiple candidate agents, and selecting a second agent for executing the subtask from the multiple candidate agents according to the comprehensive value score of the multiple candidate agents; wherein the comprehensive value score is determined based on at least one of the following target indicators: computing power node performance, network performance indicator, historical performance indicator, load state indicator, cost benefit indicator.

[0118] The computing power node performance can include at least one of the following: CPU performance, GPU performance, memory capacity, storage speed and other hardware indicators.

[0119] The network performance indicator can include at least one of the following: access delay, bandwidth capacity, network stability and other connection quality indicators.

[0120] The historical performance indicator can include at least one of the following: historical task completion success rate, historical task processing quality and the like.

[0121] The load state indicator can include at least one of the following: load condition, available resource amount and the like.

[0122] The cost benefit indicator can include at least one of the following: use cost, cost performance ratio of expected income and the like.

[0123] In the application, optionally, the comprehensive value score is the weighted sum of at least two of the target indicators, that is, the comprehensive value of the agent is evaluated from multiple dimensions.

[0124] Optionally, in the application, the intelligent agent with the highest comprehensive value score is selected as the second intelligent agent.

[0125] Optionally, in the application, the process of selecting the second intelligent agent can be performed for a certain subtask. For example, for a certain subtask, a plurality of candidate intelligent agents with the same or similar capabilities are queried from the intelligent agent capability table. At this time, the intelligent agent with the highest comprehensive value score is selected as the second intelligent agent.

[0126] The process of selecting the second intelligent agent can also be performed for a plurality of subtasks at the same time. For example, a plurality of subtasks need the same or similar intelligent agents to execute, and a plurality of candidate intelligent agents with the same or similar capabilities are queried from the intelligent agent capability table. The number of candidate intelligent agents is greater than the number of subtasks. At this time, the intelligent agents with the highest comprehensive value scores are selected from the queried candidate intelligent agents as the second intelligent agents, which are used to process the plurality of subtasks, respectively.

[0127] Optionally, in the application, the step S3 further comprises:

[0128] Step S32: The first intelligent agent collects the target indicators of each intelligent agent in the intelligent agent capability table maintained by the first intelligent agent, and calculates the comprehensive value scores of each intelligent agent in the maintained intelligent agent capability table according to the collected target indicators.

[0129] Optionally, in the application, the first intelligent agent can periodically collect the target indicators of each intelligent agent in the intelligent agent capability table maintained by the first intelligent agent, and calculate the comprehensive value scores of each intelligent agent in the maintained intelligent agent capability table. The comprehensive value scores of each intelligent agent are recorded in the maintained intelligent agent capability table, thereby facilitating the query.

[0130] The first intelligent agent can also collect the target indicators of each intelligent agent in the intelligent agent capability table maintained by the first intelligent agent in real time when the user task needs to be executed, and calculate the comprehensive value scores of each intelligent agent in the maintained intelligent agent capability table, thereby making the collected target indicators more real-time.

[0131] Optionally, in the application, the step S3 further comprises:

[0132] Step S33: The first intelligent agent receives the comprehensive value scores reported by other intelligent agents.

[0133] In the application, each target indicator of the other intelligent agent can be collected by the other intelligent agent itself, and the comprehensive value score of the other intelligent agent can be calculated based on the target indicators.

[0134] In the present application, each intelligent agent can periodically send its own comprehensive value score, for example, by sending the comprehensive value score through broadcast information. Each intelligent agent can also report its own comprehensive value score based on the reporting request sent by the first intelligent agent.

[0135] In the present application, optionally, before the step S1, the method further comprises:

[0136] Step S0: the first intelligent agent determines the capability information; wherein, when the first intelligent agent is an intelligent agent in a first intelligent computing center for performing task decomposition and distribution, the capability information determined by the first intelligent agent includes the capability information of all intelligent agents that can be aggregated by the first intelligent computing center.

[0137] For example, please refer to Figure 2 , assuming that the first intelligent agent is intelligent agent A, the capability information determined by intelligent agent A includes the capability information of all intelligent agents that can be aggregated by intelligent computing center 1, i.e. the capability information of intelligent computing center 1 as a whole, including the capability information of intelligent agents A, B, C and K. Assuming that the first intelligent agent is intelligent agent F, the capability information determined by intelligent agent F includes the capability information of all intelligent agents that can be aggregated by intelligent computing center 2, i.e. the capability information of intelligent computing center 2 as a whole, including the capability information of intelligent agents F, G, H, I and J.

[0138] And, the capability determined by other intelligent agents other than the intelligent agents for performing task decomposition and distribution is the capability of itself, such as Figure 2 the capability information determined by intelligent agent B in

[0139] Please refer to Figure 3 The present application also provides an apparatus 10 for building a decentralized intelligent agent network by computing power of an intelligent computing center cloud platform, comprising:

[0140] A transceiving module 11 for sending broadcast information and receiving broadcast information sent by other intelligent agents, wherein the broadcast information contains the capability information of the intelligent agent sending the broadcast information, the routing table and the intelligent agent capability table maintained by the intelligent agent sending the broadcast information;

[0141] An updating module 12 for updating the routing table and the intelligent agent capability table maintained by itself according to the received broadcast information;

[0142] A scheduling module 13 for, in the case of receiving a user task sent by an edge intelligent agent, decomposing the user task into a plurality of sub-tasks, selecting a second intelligent agent for performing each of the sub-tasks according to the intelligent agent capability table maintained by itself, and distributing the sub-tasks to the second intelligent agent according to the routing table maintained by itself;

[0143] The feedback module 14 is configured to receive the sub-task execution result returned by the second agent, generate a task execution result based on all the sub-task execution results, and send the task execution result to the edge agent.

[0144] The routing table includes a third agent reachable by the agent maintaining the routing table and a forwarding path between the agent and the third agent, and the agent capability table includes capability information of the third agent.

[0145] In the present application, the dynamic routing mechanism is used to complete the discovery and connection between agents, thereby dynamically constructing a multi-agent system with a decentralized network. The agents can join or exit the network at any time without knowing the global topology structure in advance, so that the multi-agent system can have dynamic expansion capability, reduce performance bottlenecks and failure risks, and greatly improve the flexibility and robustness of the multi-agent system. At the same time, through the self-organizing cooperation mode between agents, the decomposition of complex tasks and the allocation of computing resources are realized, thereby efficiently completing the processing of complex tasks.

[0146] Optionally, the first agent includes at least one of:

[0147] An agent deployed inside an intelligent computing center;

[0148] An agent deployed outside an intelligent computing center;

[0149] And / or, the other agent includes at least one of:

[0150] An agent deployed in a first intelligent computing center, which is the intelligent computing center to which the first agent belongs;

[0151] An agent deployed in a second intelligent computing center;

[0152] An agent deployed outside an intelligent computing center.

[0153] Optionally, the scheduling module 13 includes:

[0154] The selection sub-module is configured to, when the agent capability table maintained by the first agent includes multiple candidate agents with the same or similar capabilities, obtain a comprehensive value score of the multiple candidate agents, and select a second agent from the multiple candidate agents to execute the sub-task according to the comprehensive value score of the multiple candidate agents.

[0155] The comprehensive value score is determined based on at least one of the following target indicators: computing node performance, network performance indicators, historical performance indicators, load state indicators, and cost-effectiveness indicators.

[0156] Optionally, the comprehensive value score is a weighted sum of at least two of the target indicators.

[0157] Optionally, the scheduling module 13 comprises:

[0158] a calculation sub-module, configured to collect the target indicators of each agent in the agent capability table maintained by the calculation sub-module, and calculate a comprehensive value score of each agent in the agent capability table according to the collected target indicators;

[0159] Optionally, the receiving sub-module is configured to receive the comprehensive value score reported by other agents.

[0160] Optionally, the apparatus further comprises:

[0161] a determination module, configured to determine capability information; when the first agent is an agent in a first intelligent computing center for performing task decomposition and distribution, the capability information determined by the first agent comprises capability information of all agents that can be aggregated by the first intelligent computing center to which the first agent belongs.

[0162] For details, please refer to Figure 4 The application further provides an electronic device 20 comprising a processor 21, a memory 22, and a computer program stored in the memory 22 and executable on the processor 21, wherein the computer program is executed by the processor 21 to implement each process of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform and achieve the same technical effects. To avoid repetition, details are not repeated here.

[0163] The application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement each process of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform and achieve the same technical effects. To avoid repetition, details are not repeated here. The computer readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0164] The application further provides a computer program product comprising computer instructions, wherein the computer instructions are executed by a processor to implement each process of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform and achieve the same technical effects. To avoid repetition, details are not repeated here. Figure 1 The application further provides a computer program product comprising computer instructions, wherein the computer instructions are executed by a processor to implement each process of the method for building a decentralized agent network by computing power of the intelligent computing center cloud platform and achieve the same technical effects. To avoid repetition, details are not repeated here.

[0165] It should be noted that, in this document, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0166] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, and the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a plurality of instructions to make a terminal (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) execute the method described in various embodiments of the present application.

[0167] The embodiments of the present application are described above in conjunction with the drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative, not restrictive. Those skilled in the art can make many forms under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims, which all belong to the protection of the present application.

Claims

1. A method for constructing a decentralized intelligent agent network based on computing power of an intelligent computing center cloud platform, characterized in that, Comprising: Step S1: the first intelligent agent sends broadcast information and receives broadcast information sent by other intelligent agents, the broadcast information containing the capability information of the intelligent agent sending the broadcast information, the routing table and the intelligent agent capability table maintained by the intelligent agent sending the broadcast information; Step S2: the first intelligent agent updates the routing table and the intelligent agent capability table maintained by itself according to the received broadcast information; Step S3: in the case that the first intelligent agent receives a user task sent by an edge intelligent agent, the first intelligent agent decomposes the user task into multiple sub-tasks, selects a second intelligent agent to execute each sub-task according to the intelligent agent capability table maintained by itself, and distributes the sub-tasks to the second intelligent agent according to the routing table maintained by itself; Step S4: the first intelligent agent receives the sub-task execution results returned by the second intelligent agent, generates a task execution result based on all the sub-task execution results, and sends the task execution result to the edge intelligent agent; Wherein, the routing table contains a third intelligent agent reachable by the intelligent agent maintaining the routing table and a forwarding path between the third intelligent agent and the intelligent agent, and the intelligent agent capability table contains the capability information of the third intelligent agent; The first intelligent agent is an intelligent agent in an intelligent computing center that plays a role of connecting with external intelligent agents; The other intelligent agents include at least one of the following: Intelligent agents deployed in a first intelligent computing center, the first intelligent computing center being an intelligent computing center to which the first intelligent agent belongs; Intelligent agents deployed in a second intelligent computing center; Intelligent agents deployed outside the intelligent computing center; Wherein, the edge intelligent agent selects the first intelligent agent in one of the following ways: selecting an intelligent agent inside the intelligent computing center closest to it as the first intelligent agent, selecting an intelligent agent inside the intelligent computing center with the lowest cost as the first intelligent agent, selecting a first intelligent agent with a capability suitable for the user task, and selecting an intelligent agent in an intelligent computing center with a capability suitable for the user task as the first intelligent agent; Wherein, the step S3 includes: Step S31: when the intelligent agent capability table maintained by the first intelligent agent contains multiple candidate intelligent agents with the same or similar capabilities, obtaining a comprehensive value score of the multiple candidate intelligent agents, and selecting a second intelligent agent to execute the sub-task from the multiple candidate intelligent agents according to the comprehensive value score of the multiple candidate intelligent agents; Wherein, the comprehensive value score is determined based on at least one of the following target indicators: computing node performance indicator, network performance indicator, historical performance indicator, load state indicator, and cost-effectiveness indicator.

2. The method of claim 1, wherein, The comprehensive value score is the weighted sum of at least two of the target indicators.

3. The method according to any of claims 1 or 2, characterized in that, The step S3 further includes: Step S32: the first intelligent agent collects the target indicators of each intelligent agent in the intelligent agent capability table maintained by it, and calculates the comprehensive value score of each intelligent agent in the intelligent agent capability table maintained by it according to the collected target indicators; Or, step S33: the first intelligent agent receives the comprehensive value score reported by other intelligent agents.

4. The method of claim 1, wherein, Before the step S1, it further includes: Step S0: the first intelligent agent determines capability information; wherein, when the first intelligent agent is an intelligent agent in a first intelligent computing center for performing task decomposition and distribution, the capability information determined by the first intelligent agent includes capability information of all intelligent agents that can be aggregated by the first intelligent computing center to which the first intelligent agent belongs.

5. An apparatus for constructing a decentralized intelligent agent network based on computing power of an intelligent computing center cloud platform, characterized in that, The device is applied to a first intelligent agent, and the device comprises: a transceiving module configured to send broadcast information and receive broadcast information sent by other intelligent agents, the broadcast information containing capability information of an intelligent agent sending the broadcast information, a routing table maintained by the intelligent agent sending the broadcast information, and an intelligent agent capability table; an updating module configured to update the routing table and the intelligent agent capability table maintained by itself according to the received broadcast information; a scheduling module configured to, when a user task sent by an edge intelligent agent is received, decompose the user task into a plurality of subtasks, select a second intelligent agent for performing each of the subtasks according to the intelligent agent capability table maintained by itself, and distribute the subtasks to the second intelligent agent according to the routing table maintained by itself; a feedback module configured to receive subtask execution results returned by the second intelligent agent, generate a task execution result based on all the subtask execution results, and send the task execution result to the edge intelligent agent; wherein the routing table contains a third intelligent agent reachable by the intelligent agent maintaining the routing table and a forwarding path between the intelligent agent and the third intelligent agent, and the intelligent agent capability table contains capability information of the third intelligent agent; the first intelligent agent is an intelligent agent in an intelligent computing center that serves as a connection with external intelligent agents; the other intelligent agents include at least one of: an intelligent agent deployed in a first intelligent computing center, the first intelligent computing center being an intelligent computing center to which the first intelligent agent belongs; an intelligent agent deployed in a second intelligent computing center; an intelligent agent deployed outside an intelligent computing center; wherein the edge intelligent agent selects the first intelligent agent in one of the following ways: selecting an intelligent agent inside a nearest intelligent computing center as the first intelligent agent, selecting an intelligent agent inside an intelligent computing center with the lowest cost as the first intelligent agent, selecting a first intelligent agent that is suitable for a user task in terms of capability, and selecting an intelligent agent in an intelligent computing center that is suitable for a user task in terms of capability as the first intelligent agent; wherein the scheduling module comprises: a selection submodule configured to, when the intelligent agent capability table maintained by the first intelligent agent contains a plurality of candidate intelligent agents with the same or similar capabilities, obtain a comprehensive value score of the plurality of candidate intelligent agents, and select a second intelligent agent for performing the subtask from the plurality of candidate intelligent agents according to the comprehensive value scores of the plurality of candidate intelligent agents; wherein the comprehensive value score is determined based on at least one of the following target indicators: a computing node performance indicator, a network performance indicator, a historical performance indicator, a load state indicator, and a cost-benefit indicator.

6. An electronic device, comprising: comprises: A processor, a memory, and a program stored on the memory and executable on the processor, the program, when executed by the processor, implements the steps of the method for building a decentralized intelligent agent network based on the computing power of the intelligent computing center cloud platform according to any one of claims 1 to 4.

7. A computer readable storage medium characterized in that, A computer program is stored on the computer readable storage medium, and the computer program, when executed by the processor, implements the steps of the method for building a decentralized intelligent agent network based on the computing power of the intelligent computing center cloud platform according to any one of claims 1 to 4.

8. A computer program product, characterised in that, Computer instructions are included, and the computer instructions, when executed by the processor, implement the steps of the method for building a decentralized intelligent agent network based on the computing power of the intelligent computing center cloud platform according to any one of claims 1 to 4.

Images