Data sharing method of multi-agent system, multi-agent system and electronic equipment

Abstract

The invention relates to a data sharing method of a multi-agent system, the multi-agent system and electronic equipment. The data sharing method of the multi-agent system comprises the following steps: S0, establishing a data index table; the data index table records the content and the possible source of each data type; s1, receiving a data request sent by a first service agent; s2, searching the data index table according to the content of the data request to obtain a target data type, and determining a possible source of the target data type; s3, determining a second business agent according to the possible source of the target data type; s4, searching a database of a second business agent according to the content of the data request to obtain target data conforming to the data request; and S5, forwarding the target data to the first service agent. According to the method, the accuracy rate, the recall rate and the efficiency of data sharing among multiple agents are considered, and the intelligent level of a multi-agent system is ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a data sharing method for a multi-agent system, a multi-agent system and an electronic device. Background Art

[0002] A multi-agent system (MAS), also known as a multi-agent system or a self-organized system, is a computing system or group system composed of multiple interacting agents. Each agent in a multi-agent system can make decisions and perform actions independently, and multiple agents can complete more complex tasks together through communication and collaboration. Currently, multi-agent systems have been applied in the fields of smart terminals, robots, traffic scheduling, distributed computing, etc.

[0003] For example, the smart assistant on a smartphone is a multi-agent system. Assume that the smart assistant includes multiple agents such as a language recognition agent, a language generation agent, a camera agent, an image processing agent, and a memory agent. When a user says to the smart assistant, "It's a great day today, can you take a photo for me?", the multiple agents of the smart assistant respond to the user through the following process: S01, the language recognition agent converts the user's input voice into text, identifies the user's intention by analyzing the text content, and then conveys the intention to the language generation agent and the camera agent. S02, after understanding the user's intention, the language generation agent generates a reply: "Of course! I will help you open the camera application on your phone right away, so that you can easily capture this beautiful weather!" S03, the camera agent calls the camera application on the phone to take a photo for the user, and then further transmits the photo to the image processing agent. S04, the image processing agent adds the user's preferred filter to the photo based on the user's historical preferences memorized by the memory agent, and then transmits the processed photo to the language generation agent. S05: The language generation agent sends the processed photo to the user and generates a reply: “I hope this photo can capture this beauty! If there is anything else I can help you with, please let me know!” The above response process involves the problem of data sharing among multiple agents. For example, the language recognition agent needs to share the text recognition content with the language generation agent and the camera agent, the camera agent needs to share the original photo with the image processing agent, and the memory agent needs to share the memory related to the user's historical preferences with the image processing agent.

[0004] For the data sharing problem among multiple agents, the existing data sharing methods of multi-agent systems include: 1. All agents share a single largest database. In the above example of taking a photo, the language recognition agent uploads the text recognition content to a shared database accessible to all agents. Except for the language generation agent and the photo taking agent, any other agent can obtain the text recognition content. Although this solution can guarantee the recall rate of data sharing (any agent in need will not miss the text recognition content data), since not every agent is related to the text recognition content, this method will cause irrelevant agents to spend computing power to process a large amount of redundant information, which is costly and inefficient.

[0005] 2. Let each agent decide what data to share with other agents, and let the data be transmitted point-to-point between agents. In the above example of the photo response, the language generation agent determines through thinking that the text recognition content is related to the photo-taking agent and the language generation agent, so the text recognition content is shared point-to-point with the photo-taking agent and the language generation agent, and other unrelated agents cannot obtain the text recognition content. Although this solution avoids unrelated agents spending computing power to process redundant information, each agent only judges whether the data needs to be shared from an individual perspective, without a global perspective, and omissions often occur, resulting in a decrease in the intelligence level of the multi-agent system. Summary of the invention

[0006] Based on this, the object of the present invention is to provide a data sharing method for a multi-agent system, a multi-agent system and an electronic device.

[0007] The present invention provides a data sharing method for a multi-agent system, including the steps of: S0, establishing a data index table; the data index table records the content and possible sources of each data type; S1, receiving a data request sent by a first business agent; S2, searching the data index table according to the content of the data request to obtain a target data type, and determining the possible sources of the target data type; S3, determining a second business agent according to the possible sources of the target data type; S4, searching the database of the second business agent according to the content of the data request to obtain target data that meets the data request; S5, forwarding the target data to the first business agent.

[0008] The data index table of the present invention records all the data types that each business agent may master. By searching the data index table, all second business agents related to the data request can be accurately and comprehensively determined. Compared with the solution that all agents share the same database, the present invention can ensure that the second business agent searched is related to the content of the data request, avoiding irrelevant agents from expending computing power to process redundant information, and has higher accuracy and efficiency. Compared with the point-to-point data sharing solution, the present invention can search all possible sources of target data from a global perspective, ensuring that no target data that meets the data request is missed, and has a higher recall rate. Therefore, the present invention takes into account the accuracy, recall rate and efficiency of data sharing, and ensures the intelligence level of the multi-agent system.

[0009] Furthermore, in step S0, the step of establishing a data index table is specifically: by communicating with each business agent in the multi-agent system, the data type mastered by each business agent is obtained, thereby establishing a data index table.

[0010] Furthermore, in step S0, when the data type mastered by the business agent changes, the data index table is updated.

[0011] Furthermore, in step S5, before forwarding the target data to the first business agent, it is determined whether the first business agent has the authority to access the target data; only after confirming that the first business agent has the authority to access the target data, the target data is forwarded to the first business agent.

[0012] Furthermore, in step S5, after confirming that the first business agent has the authority to access the target data, the target data is encrypted using an encryption algorithm, and then the encrypted target data is forwarded to the first business agent.

[0013] Furthermore, in step S4, when searching the database of the second business agent according to the content of the data request, potential data is also foreseen; and when the potential data meets the data request, a business action is executed, thereby converting the potential data into target data output.

[0014] Further, in step S3, when the number of possible sources of the target data type is multiple, multiple second business intelligent entities are determined; then, the databases of the multiple second business intelligent entities are searched according to the content of the data request to obtain multiple target data; finally, the multiple target data are aggregated and the aggregated target data are forwarded to the first business intelligent entity.

[0015] Furthermore, the method further includes the steps of: establishing a cache area and saving recent high-frequency access data in the cache area; before step S2, the method further includes the steps of: first searching the cache area according to the content of the data request; if the target data that meets the data request exists in the cache area, directly forwarding the target data to the first business intelligent entity; if the target data that meets the data request does not exist in the cache area, continuing to execute step S2.

[0016] Based on the same inventive concept, the present invention also provides a multi-agent system, which includes: an administrator agent, a first business agent and a second business agent; the administrator agent, the first business agent and the second business agent realize data sharing through the following process: S0, the administrator agent establishes a data index table; the data index table records the content and possible sources of each data type; S1, the first business agent sends a data request to the administrator agent; S2, the administrator agent searches the data index table according to the content of the data request, obtains the target data type, and determines the possible sources of the target data type; S3, the administrator agent determines the second business agent according to the possible sources of the target data type, and forwards the data request to the second business agent; S4, the second business agent searches its own database according to the content of the data request, obtains the target data that meets the data request, and sends the target data to the administrator agent; S5, the administrator agent forwards the target data to the first business agent.

[0017] Based on the same inventive concept, the present invention also provides an electronic device, which includes: a processor; a memory for storing a computer program executed by the processor; wherein, when the processor executes the computer program, it implements any of the above-mentioned data sharing methods for a multi-agent system.

[0018] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a module schematic diagram of the multi-agent system of the present invention; Figure 2 It is a timing flow chart of the data sharing method of the multi-agent system of the present invention. DETAILED DESCRIPTION

[0020] It should be clear that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of the present application.

[0021] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. The singular forms of "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0022] In addition, in the description of this application, unless otherwise specified, "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0023] It should be understood that the embodiments of the present application are not limited to the precise structures described above and shown in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the appended claims.

[0024] The present invention manages the data sharing among multiple agents in a unified manner through a centralized administrator agent. The administrator agent has a global perspective, can obtain the data types mastered by each agent, and can judge all possible sources of a certain data type through thinking, thereby guiding the data to be purposefully shared among agents in need, taking into account the accuracy, recall rate and efficiency of data sharing, and ensuring the intelligence level of the multi-agent system.

[0025] See also Figure 1 , Figure 1 It is a module diagram of the multi-agent system of the present invention. The multi-agent system of the present invention includes: an administrator agent and multiple business agents. The business agent refers to an agent in the multi-agent system, other than the administrator agent, that is responsible for a specific business. The administrator agent establishes a communication channel with each business agent to uniformly manage the data sharing between business agents. In the example of the photo response described in the background technology, the language recognition agent, the language generation agent, the photo agent and the image processing agent are all business agents, which are respectively responsible for the specific businesses of language recognition, language generation, photo taking and image processing.

[0026] See also Figure 2 , Figure 2 The multi-agent system of the present invention realizes data sharing through the following data sharing method: S0, the administrator agent establishes a data index table; the data index table records the content and possible sources of each data type.

[0027] Specifically, the administrator agent counts all business agents in the multi-agent system, assigns a unique identifier to each business agent, and establishes a communication channel with each business agent. The administrator agent communicates with each business agent, obtains the type of data mastered by each business agent, and generates a data index table; the data index table records the content and possible sources of each data type. A data type may have multiple sources. For example, the camera agent and the image processing agent may both master photo-type data. At this time, the data index table will record all possible sources of this type of data. When the type of data mastered by the business agent changes, the business agent sends a notification message to the administrator agent; the administrator agent updates the data index table based on the notification message.

[0028] S1, a first business agent sends a data request to an administrator agent, wherein the content of the data request is the content of the data to be shared required by the first business agent.

[0029] S2, the administrator agent searches the data index table according to the content of the data request, obtains the target data type related to the data request, and determines the possible source of the target data type. The administrator agent searches the data index table in an intelligent way based on its thinking ability, rather than the precise matching based on key-value pairs used by traditional database management tools to query the database. Even if the content of the data request does not completely match the content stored in the data index table, the administrator agent can still derive the most likely source of the target data through thinking.

[0030] S3, the administrator agent determines a second business agent according to the possible sources of the target data type, and forwards the data request to the second business agent. When there are multiple possible sources of the target data type, the number of the second business agents is multiple, and the administrator agent forwards the data request to multiple second business agents.

[0031] S4, the second business agent searches its own database according to the content of the data request, obtains target data that meets the data request, and sends the target data to the administrator agent. When there are multiple second business agents, multiple second business agents all search their own databases according to the content of the data request, obtain multiple target data that meet the data request, and send the multiple target data to the administrator agent. The way the second business agent searches its own database is also an intelligent search based on thinking ability.

[0032] S5, the administrator agent forwards the target data to the first business agent. When there are multiple target data, the multiple target data may or may not be repeated. At this time, the administrator agent summarizes the multiple target data, merges the repeated target data, and finally forwards the summarized target data to the first business agent, thereby achieving the effect of not missing target data and minimizing redundant data.

[0033] Furthermore, since the target data may involve user privacy data, and the business agent may be connected to an untrusted external application, in order to protect user privacy, in step S5, before forwarding the target data to the first business agent, the administrator agent also determines whether the first business agent has the authority to access the target data; after confirming that the first business agent has the authority to access the target data, the administrator agent forwards the target data to the first business agent.

[0034] Furthermore, in order to prevent the data transmission process from being eavesdropped and causing the user's privacy to be leaked. In step S5, after confirming that the first business agent has the authority to access the target data, the administrator agent also encrypts the target data through an encryption algorithm, and then forwards the encrypted target data to the first business agent. After receiving the encrypted target data, the first business agent restores the original target data through a preset decryption algorithm.

[0035] Furthermore, when the second business agent searches its own database according to the content of the data request, it can also foresee potential data and execute business actions when the potential data meets the data request, thereby converting the potential data into real target data output. The potential data refers to data that does not exist at present, but the agent can foresee its future existence through its thinking ability.

[0036] For example, the image processing agent can process an existing color image to obtain a black-and-white image; before the image processing agent performs the image processing action, the black-and-white image does not exist temporarily. When the image processing agent receives a data request with the content of "black-and-white image", the image processing agent can foresee the possibility of the future existence of black-and-white images through its thinking ability, and regard the currently non-existent black-and-white images as potential data for search; when it is determined that the potential data meets the data request, the image processing agent performs the image processing action, converts the potential black-and-white image into a real black-and-white image, and outputs the real black-and-white image as the target data to the administrator agent.

[0037] Furthermore, the business agent may repeatedly request access to data of the same content in a short period of time. In order to improve the efficiency of data sharing, the administrator agent also establishes a cache area to save recent high-frequency access data in the cache area. After receiving a data request, the administrator agent first searches the cache area according to the content of the data request; if there is target data that meets the data request in the cache area, the target data is directly forwarded to the first business agent; if there is no target data that meets the data request in the cache area, step S2 is continued to search the data index table according to the content of the data request.

[0038] Example 1 In this embodiment, the multi-agent system is an intelligent assistant system installed on an intelligent terminal, which includes: an administrator agent, a memory agent, and a music agent. The memory agent and the music agent are both business agents.

[0039] The memory agent stores the "music type data that the user has recently preferred". The music agent has music recommendation and music playback functions, but does not have a memory function. When the user asks the intelligent assistant to recommend music to him, the music agent needs to obtain the "music type data that the user has recently preferred" from the memory agent in order to complete the music recommendation task based on the user's preference. Therefore, when the music agent performs the music recommendation task, the music agent is the first business agent that issues a data request, and the memory agent is the second business agent that is a possible source of the target data type. In the data index table established by the administrator agent, the possible sources of "user preference data" are pre-recorded, including the memory agent.

[0040] The administrator agent, music agent and memory agent share data through the following process: S1, the music agent sends a data request to the administrator agent, and the content of the data request is "the user's recent preferred music type data".

[0041] S2, the administrator agent searches the data index table according to the content of the data request, determines that the target data type is "user preference data", and determines that the possible sources of "user preference data" include memory agents.

[0042] S3, the administrator agent establishes communication with the memory agent, and forwards the data request with the content of "the user's recent preferred music type data" to the memory agent.

[0043] S4, the memory agent searches its own database according to the content of the data request, obtains the target data that meets the data request, and the content is "the most frequently played music types by users in the past 7 days are classical music and jazz", and then sends the target data to the administrator agent.

[0044] S5, the administrator agent forwards the target data to the music agent, so that the music agent can recommend a personalized playlist of mainly classical music and jazz to the user based on the target data.

[0045] The present invention has the following technical effects: 1. The data index table maintained by the administrator agent of the present invention records all data types that each business agent may master. Therefore, the administrator agent has a unique global perspective and can accurately and comprehensively determine all second business agents related to the data request through intelligent search. Compared with the solution that all agents share the same database, the present invention can ensure that the second business agent searched is related to the content of the data request, avoiding irrelevant agents from expending computing power to process redundant information, and has higher accuracy and efficiency. Compared with the point-to-point data sharing solution, the present invention can search for all possible sources of target data from a global perspective, ensuring that no target data that meets the data request is missed, and has a higher recall rate. In summary, the present invention takes into account the accuracy, recall rate and efficiency of data sharing, and ensures the intelligence level of the multi-agent system.

[0046] 2. Since the agents involved in the present invention all have intelligent thinking capabilities, the process of the administrator agent determining the second business agent, and the process of the second business agent searching for target data, are both intelligent search processes based on thinking capabilities. Although traditional database management tools can also record data and query data, traditional database management tools do not have thinking capabilities. When querying data, they can only perform precise matching based on key-value pairs stored in the database, and cannot identify potential data that does not exist temporarily. The administrator agent / second business agent of the present invention can flexibly perform intelligent searches and identify potential data through thinking capabilities. Even if the content of the data request does not completely match the content stored in the data index table / database, the administrator agent / second business agent can still obtain the closest search results through thinking, and the level of intelligence is relatively high.

[0047] Based on the same inventive concept, the present application also provides an electronic device, which may be a terminal device such as a server, a desktop computing device or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). The device includes one or more processors and a memory, wherein the processor is used to execute a program to implement the data sharing method of the multi-agent system of the embodiment of the present invention; and the memory is used to store a computer program executable by the processor.

[0048] Based on the same inventive concept, the present application also provides a computer-readable storage medium, corresponding to the embodiment of the aforementioned multi-agent system data sharing method, wherein the computer-readable storage medium stores a computer program thereon, and when the program is executed by the processor, the steps of the multi-agent system data sharing method recorded in any of the aforementioned embodiments are implemented.

[0049] The present application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-usable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0050] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, and the present invention is also intended to include these modifications and modifications.

Claims

1. A data sharing method for a multi-agent system, characterized in that: Includes steps: S0, establishing a data index table; the data index table records the content and possible sources of each data type; S1, receiving a data request sent by a first business agent; S2, searching the data index table according to the content of the data request, obtaining the target data type, and determining the possible source of the target data type; S3, determining a second business agent according to possible sources of the target data type; S4, searching the database of the second business agent according to the content of the data request to obtain target data that meets the data request; S5: forward the target data to the first business agent.

2. The data sharing method of a multi-agent system according to claim 1, characterized in that: In step S0, the step of establishing a data index table is specifically: by communicating with each business agent in the multi-agent system, the data type mastered by each business agent is obtained, thereby establishing a data index table.

3. The data sharing method of a multi-agent system according to claim 2, characterized in that: In step S0, when the data type mastered by the business agent changes, the data index table is updated.

4. The data sharing method of a multi-agent system according to claim 1, characterized in that: In step S5, before forwarding the target data to the first business agent, it is determined whether the first business agent has the authority to access the target data; after confirming that the first business agent has the authority to access the target data, the target data is forwarded to the first business agent.

5. The data sharing method of a multi-agent system according to claim 4, characterized in that: In step S5, after confirming that the first business agent has the authority to access the target data, the target data is encrypted using an encryption algorithm, and then the encrypted target data is forwarded to the first business agent.

6. The data sharing method of a multi-agent system according to claim 1, characterized in that: In step S4, when searching the database of the second business agent according to the content of the data request, potential data is also foreseen; and when the potential data meets the data request, a business action is executed, thereby converting the potential data into target data output.

7. The data sharing method of a multi-agent system according to claim 1, characterized in that: In step S3, when there are multiple possible sources of the target data type, multiple second business agents are determined; Then, the databases of multiple second business intelligence entities are searched according to the content of the data request to obtain multiple target data; finally, the multiple target data are summarized and the summarized target data are forwarded to the first business intelligence entity.

8. The data sharing method of a multi-agent system according to claim 1, characterized in that: The steps also include: establishing a cache area, and storing recent high-frequency access data in the cache area; Before step S2, the method further includes the following steps: first searching the cache area according to the content of the data request; if target data that meets the data request exists in the cache area, directly forwarding the target data to the first business intelligent body; If the target data that meets the data request does not exist in the cache area, then continue to execute step S2.

9. A multi-agent system, characterized in that: include: An administrator agent, a first business agent, and a second business agent; the administrator agent, the first business agent, and the second business agent share data through the following process: S0, the administrator agent establishes a data index table; the data index table records the content and possible sources of each data type; S1, the first business agent sends a data request to the administrator agent; S2, the administrator agent searches the data index table according to the content of the data request, obtains the target data type, and determines the possible source of the target data type; S3, the administrator agent determines a second business agent according to possible sources of the target data type, and forwards the data request to the second business agent; S4, the second business agent searches its own database according to the content of the data request, obtains the target data that meets the data request, and sends the target data to the administrator agent; S5, the administrator agent forwards the target data to the first business agent.

10. An electronic device, characterized in that: include: processor; a memory for storing a computer program executed by the processor; Wherein, when the processor executes the computer program, it implements the data sharing method of any multi-agent system described in claims 1-8.

Images