Diversified interactive agent construction method and system based on offline learning

Abstract

The invention discloses a diversified interaction agent construction method and system based on off-line learning, the system comprises an off-line circulation system and an agent interaction system, the off-line circulation system comprises a process splitting unit, a self-evolution unit, a storage processing unit and a model calling unit, the intelligent agent interaction system comprises an information sharing unit and a cooperation unit. The offline circulating system processes basic functions of the intelligent agent under the independent condition of the single intelligent agent by using an offline strengthening algorithm, and the basic functions comprise a thinking function, a sensing function and an execution function; the intelligent agent interaction system uses an interaction adjustment learning algorithm to enhance information in an offline circulation system, uses a distance contraction adjustment algorithm to improve the cooperation efficiency of the intelligent agent interaction system, and performs information interaction with other single intelligent agents to effectively complete complex tasks. According to the method, efficient learning of the intelligent agent in the offline state is realized, and the data processing efficiency of the diversified intelligent agents for completing complex tasks is improved.

Description

Technical Field

[0001] The present invention relates to the field of multi-agent interaction technology, and in particular to a method and system for constructing diversified interactive agents based on offline learning. Background Art

[0002] Agent is a core concept in the field of artificial intelligence. It refers to an entity that can perceive the environment and make decisions based on the perception results to achieve specific goals. Agents can be in the form of software, such as an automated trading system, or physical entities, such as an autonomously navigating robot. Offline agent, also known as batch reinforcement learning, is a variant of reinforcement learning. It mainly studies how to use large-scale static data sets collected in advance to train reinforcement learning agents without any form of online interaction and exploration. A multi-agent system is a computing system composed of multiple agents that interact, communicate, collaborate or compete in a shared environment to complete complex tasks that are difficult for a single agent to complete.

[0003] Single agent and multi-agent have their own advantages and are suitable for different vertical fields. The reinforcement learning principle of single agent is based on Markov decision making. In simple terms, it can be divided into state set S, action set A, and reward R. The state and reward at the next moment are only related to the action at the previous moment, and have nothing to do with the state before. The principle of its model is to let the agent learn by trial and error. If a certain strategy can be rewarded, the strategy of the agent to produce this behavior will be strengthened. Its purpose is to act in a single environment and get the maximum reward as much as possible. The application field is also relatively wide at present. For example, the training of continuous actions in racing games: controlling the steering wheel, accelerator, brake and other actions can be decided by DDPG, A3C, PPO algorithms. The training of some discrete actions, such as the Go agent AlphaGo, can be decided by algorithms such as Q-Learning. The decision of multi-agents is not only related to their own actions, but also to the actions of other agents in the system. There will be more than two agents in a multi-agent system, and they generally have a cooperative or competitive relationship. Such a model is called a Markov game, and its state transitions conform to Markov decisions and relationships conform to games. In a multi-agent model, the goal of each agent is to find the optimal strategy to maximize the long-term cumulative reward in any state. Due to the complexity of the model and the large number of interference factors, there are currently few commercial products of multi-agent models.

[0004] The intelligent agent system disclosed in Chinese invention patent CN117789729A actively determines the topic on the system side, identifies the interlocutor and conducts a conversation with him, so that the conversation of the passengers in the vehicle becomes smooth and a pleasant space is created. The microphone installed in the vehicle cabin is used to collect the voice information of the passengers. The control unit enables the analysis unit to analyze the speech content contained in the collected voice information, determine the listener, retrieve the recent event from the mobile terminal of the passenger and determine the topic, and output the voice information from the speaker. In this way, the system side actively determines the listener, determines the topic based on the recent event and outputs it. This invention is mainly used in the field of vehicle voice interaction, but the idea can also be used in other intelligent agent fields. However, the idea of ​​this invention is limited to offline intelligent agents, which is actually a transplantation of basic artificial intelligence and cannot be widely used in most complex scenarios.

[0005] The health intelligent all-in-one machine disclosed in Chinese invention patent CN109859843B includes a health detection module, a main controller, a power module, a face recognition module, a voice module, a data storage module and a display module; the health detection module is electrically connected to the main controller; the main controller is electrically connected to each module, analyzes and processes the user's health data, and controls the signal transmission between each module; the power module is electrically connected to the main controller and each module; the face recognition module includes a camera, an identity recognition unit and an emotion recognition unit, and the camera is electrically connected to the identity recognition unit and the emotion recognition unit; the voice module includes a voice input module, a voiceprint recognition unit and a voice recognition unit; the voice input module is electrically connected to the voiceprint recognition unit, the voice recognition unit and the voice broadcast unit; the data storage module is used to store user data, and the display module is used to display the interface. This invention is a relatively simple human-computer interaction device system for individual user health examinations by Yuyue Medical Equipment Company. This invention mainly solves the problem of human-computer friendliness, and there is no further optimization for the actual function expansion and data judgment accuracy.

[0006] In the existing invention patents or actual products, only a single product or a single function in the industry or field is described to construct an intelligent body. In fact, there are many more complex problems in the actual application of intelligent bodies, which require a comprehensive intelligent body performance improvement system that includes offline and online processing of diverse problems, such as learning evaluation, medical health rating, agricultural product futures grading, etc. Therefore, the existing intelligent body processing system is not ideal for handling complex problems. Summary of the invention

[0007] The present invention provides a method and system for constructing a diversified interactive intelligent agent based on offline learning, so as to solve the technical problems that the existing intelligent agents rely on online construction during learning, cannot perform diversified interactions with multiple preset functions, and have weak interaction functions between intelligent agents, and low efficiency and accuracy.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: The present invention provides a method and system for constructing a diversified interactive intelligent agent based on offline learning, comprising: An offline cycle system and an intelligent agent interaction system, wherein the offline cycle system comprises a process splitting unit, a self-evolution unit, a storage processing unit, and a model calling unit, and the intelligent agent interaction system comprises an information sharing unit and a collaboration unit; The offline loop system uses an offline reinforcement algorithm to process the basic functions of a single agent under independent conditions, wherein the basic functions of the agent include thinking function, perception function, and execution function; The intelligent agent interaction system uses an interactive adjustment learning algorithm to enhance the information stored in the processing unit in the offline circulation system, and uses a distance contraction adjustment algorithm to improve the collaborative efficiency of the intelligent agent interaction system. It can exchange information with other single intelligent agents and effectively complete complex tasks.

[0009] Furthermore, the offline reinforcement algorithm regards the basic functions of a single agent under independent conditions as an offline function trajectory; Model offline functional trajectories in a serialized manner and mark the agent action feedback of each time series; The offline function trajectory is predicted using the agent action feedback through an offline MARL data-driven training paradigm to obtain a regressed offline function trajectory; Track through the offline function After serialization modeling, the regression offline function trajectory is obtained Realize the basic functions of the intelligent agent under the independent condition of the single intelligent agent.

[0010] Furthermore, the interactive regulation learning algorithm enhances the stability of the agent through unified training and distributed implementation of the basic functions of the agent; The interactive adjustment learning algorithm comprises the steps of: S1 normalizes and reconstructs the regression offline function trajectory; S2 uses hidden state analysis to build a self-consistent model; S3 improves the collaboration efficiency of the intelligent agent interaction system through a distance contraction adjustment algorithm.

[0011] Furthermore, when interacting with other single agents, the agent error is reduced by introducing a spontaneous pre-training conversion model; The spontaneous pre-training conversion model fits the normalized regression offline function trajectory G of multiple agents through the MAPPO algorithm to improve the accuracy of completing complex tasks.

[0012] The beneficial effects brought about by the technical solution provided by the present invention include at least: The offline loop system uses an offline reinforcement algorithm to process the basic functions of a single agent under independent conditions, including thinking function, perception function, and execution function, thus achieving efficient learning of the agent in an offline state.

[0013] The offline reinforcement algorithm is used to treat the basic functions of a single agent under independent conditions as an offline function trajectory, and the offline function trajectory is serialized and modeled, marking the agent action feedback of each time series. The standardized description of a single agent is completed, which facilitates the data processing of diverse agents in the subsequent improvement steps.

[0014] An interactive adjustment learning algorithm is used to enhance the information stored in the processing unit in the offline cycle system, enhance the agent data in the offline state, and improve the processing efficiency of diversified agent data.

[0015] The distance contraction adjustment algorithm is used to improve the collaborative efficiency of the intelligent agent interaction system, and information is exchanged with other single intelligent agents, thereby improving the efficiency of completing complex tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 Construct a system architecture diagram for diverse interactive agents based on offline learning; Figure 2 A flow chart of the interactive adjustment learning algorithm method; Figure 3 It is a schematic diagram of the distance contraction adjustment algorithm method; Figure 4 Schematic diagram of the diversified intelligent agent structure for medical teaching. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Example

[0019] This embodiment provides a method and system for constructing a diversified interactive intelligent agent based on offline learning.

[0020] Please refer to Figure 1The illustrated embodiment of the present invention provides a schematic diagram of the structure of a diversified interactive intelligent agent construction system based on offline learning.

[0021] The system comprises an offline cycle system and an intelligent agent interaction system, wherein the offline cycle system comprises a process splitting unit, a self-evolution unit, a storage processing unit, and a model calling unit, and the intelligent agent interaction system comprises an information sharing unit and a collaboration unit; The offline loop system uses an offline reinforcement algorithm to process the basic functions of a single agent under independent conditions, wherein the basic functions of the agent include thinking function, perception function, and execution function; The intelligent agent interaction system uses an interactive adjustment learning algorithm to enhance the information stored in the processing unit in the offline circulation system, and uses a distance contraction adjustment algorithm to improve the collaborative efficiency of the intelligent agent interaction system. It can exchange information with other single intelligent agents and effectively complete complex tasks.

[0022] Specifically, the offline loop system is a diversified unit set that enables an agent to realize preset functions without information interaction with other agents. The diversified unit set is composed of different types of unit sets, including process splitting units, self-evolution units, storage processing units, and model calling units. The unit set can independently complete the preset function, and the diversified unit set can complete multiple preset functions; The process splitting unit decomposes a complex function into a set of simple functions; The self-evolution unit improves the effect of achieving the preset function of the unit set by evaluating the unit set to achieve the preset function; The storage processing unit uses the accumulated existing unit set to guide the implementation of the preset function; The model calling unit can call an external model to enhance the unit set function, and the external model is an existing database and plug-in library that assists in implementing the current unit set.

[0023] Specifically, the agent interaction system improves the performance of agents through the collaboration of different agents; The information sharing unit can share the diverse unit set in the offline cycle system with other agents; The collaboration unit enhances the ability of multiple agents to complete preset functions through collaboration of multiple agents.

[0024] Specifically, the thinking function is that the agent can split complex tasks into simpler subtasks and continuously analyze the results to achieve improvements and thinking capabilities; The perception function realizes external input through text input, image capture, and voice recognition; The execution function can call an external model to assist the intelligent agent in realizing the preset function.

[0025] Specifically, the offline reinforcement algorithm considers the basic functions of a single agent under independent conditions as an offline function trajectory , expressed as: ; in, Indicates the observation state, Represents the global state, Indicates instructions, is the reinforcement value, T represents the time series; Model offline functional trajectories in a serialized manner and mark the agent action feedback of each time series It is expressed as: ; Where T represents the time series, e represents the natural constant, Represented as the agent action feedback of the previous time series; Through the offline MARL data-driven training paradigm, the agent action feedback is used Predict the offline function trajectory to obtain the regression offline function trajectory , expressed as: ; in, is the observed state of the agent in time series t, is the global state of the agent in time series t; Track through the offline function After serialization modeling, the regression offline function trajectory is obtained Realize the basic functions of the intelligent agent under the independent condition of the single intelligent agent.

[0026] Specifically, the interactive regulation learning algorithm enhances the stability of the agent through unified training and distributed implementation of the basic functions of the agent; Please refer to Figure 2 The flowchart of the interactive adjustment learning algorithm method is shown.

[0027] The interactive adjustment learning algorithm comprises the steps of: S1 normalizes and reconstructs the regression offline function trajectory; S2 uses hidden state analysis to build a self-consistent model; S3 improves the collaboration efficiency of the intelligent agent interaction system through a distance contraction adjustment algorithm.

[0028] Specifically, the agent action feedback of each time series , expressed as: ; in, Represents the normalized agent action feedback of the previous time series; The normalization strategy C is expressed as: ; Among them, K represents the sampling degree of time series; For the normalized regression offline function trajectory G after normalization and reconstruction, it is expressed as: ;

[0029] Complete offline function trace After serialization modeling, the regression offline function trajectory is obtained The process is normalized to obtain the normalized regression offline function trajectory.

[0030] Specifically, the hidden state analysis method first samples other agents n to obtain the sampling trajectories of other agents , expressed as: ; Among them, v represents the sampling degree, Represents the sampling state of agent n in time series t.

[0031] Specifically, the sampling trajectories of other agents are analyzed through the nearest neighbor algorithm. Probability distribution extraction is performed, and the contraction adjustment algorithm is used when the probability distribution is extracted. The contraction adjustment algorithm is used to calculate the sampling trajectory of other intelligent agents. Compare with the normalized regression offline function trajectory G to obtain the normalized regression offline function shrinkage trajectory of the time series with the minimum difference , expressed as: ; in, represents the normalized regression offline feature trajectory vector, Represents all normalized regression offline function trajectory vectors obtained, Indicates the minimum value; Among them, The exponential moving average method is used to improve efficiency when traversing all normalized regression offline function trajectory vectors.

[0032] Please refer to Figure 3 Schematic diagram of the distance contraction adjustment algorithm method shown.

[0033] Specifically, when interacting with other single agents, the agent error is reduced by introducing a spontaneous pre-training conversion model; The spontaneous pre-training conversion model fits the normalized regression offline function trajectory G of multiple agents through the MAPPO algorithm to improve the accuracy of completing complex tasks.

[0034] Second implementation This embodiment provides an offline circulation system and an intelligent agent interaction system, which is an application in the field of intelligent agent medical teaching.

[0035] Please refer to Figure 4 The schematic diagram of the diversified intelligent agent structure for medical teaching is shown.

[0036] In this embodiment, the diverse agents include medical teaching information management agents, collaborative interaction agents, student assistant agents, and medical teaching exploration agents.

[0037] In the field of medical education, medical teaching information management is the core link, which covers the sharing, acquisition and composition of medical knowledge between teachers and students.

[0038] It should be noted that medical school teachers often find it difficult to accurately grasp the details of each medical student's mastery of medical knowledge. Although medical students have a certain degree of analysis and evaluation of their own learning situation, they are affected by their experience and knowledge reserves and it is difficult for them to accurately grasp the real situation of medical learning. As an educational platform, a diversified interactive intelligent agent system with medical knowledge and rich teaching resources can significantly promote the management and dissemination of knowledge.

[0039] The specific application forms of the medical teaching information management system in medical teaching include medical knowledge agents, medical analysis agents, medical examination paper agents and medical operation agents.

[0040] The medical knowledge agent can automatically collect medical students' questions in the medical field, obtain answers from the medical data set, and classify the questions into question sets, and continuously update the data set and question set; the medical analysis agent automatically analyzes relevant medical materials based on the status of medical students; the medical examination paper agent selects questions from the data set and automatically forms medical examination papers according to the goals set in advance by medical teachers, and forms examination papers for students in a personalized way; the medical homework agent automatically generates homework and gives reminders based on the learning progress of the target medical students, the goals set by the teachers, the learning progress of the students, the test scores, etc. By making rational use of the database, the medical teaching information management agent can effectively improve the learning ability and efficiency of medical students.

[0041] It should be noted that the medical teaching information management agent, as a core agent, can be used offline. Offline does not mean disconnection from the network, but means temporarily not interacting with other agents. 1. Medical teaching information management agent

[0042]

[0043] The specific application forms of collaborative interactive agents in medical teaching include medical team agents and medical evaluation agents.

[0044] Medical group agents can automatically group students according to multiple factors such as the learning progress of medical students, teacher grouping, medical students' learning progress, test scores, and test score composition, and group students with similar learning progress and targeted teachers together to achieve efficient teaching and efficient learning. Medical evaluation agents can conduct targeted and comprehensive evaluations based on the learning situation of medical students, and can also provide suggestions to teachers and students based on their learning situation, and collect feedback on students' later learning progress and results to improve the capabilities of students and teachers.

[0045] 2. Collaborative Interactive Agents

[0046] The specific application forms of student assistant agents in medical teaching include medical self-service testing agents, medical auxiliary reminder agents, and medical teaching management and control agents.

[0047] Medical self-service testing agents can help medical students generate self-tests during self-study and help them quickly grasp their learning status and progress. Medical assistance agents help medical students develop learning plans and assist teachers in developing learning progress plans and managing students. Medical teaching management agents control account permissions, manage learning materials, examination materials, data verification, and other functions. 3. Student Assistant Agent

[0048]

[0049] The specific application forms of medical teaching exploration agents in medical teaching include medical courseware collection agents and medical teaching agents.

[0050] The medical courseware collection agent can make full use of the learning situation of existing medical students and the teaching situation of teachers to collect and automatically integrate the publicly available teaching resources on the Internet to form a teaching resource library, which is provided to teachers and students after review. The medical teaching agent can compare and integrate the acquired teaching resource library with existing data and teacher resource data to generate new artificial intelligence courseware as an additional learning aid for students or as a reference for teachers.

[0051] IV. Medical Teaching Exploration Agent

[0052] The function realized by any intelligent agent in the medical teaching information management system is regarded as an offline function trajectory, and serialized modeling is performed. The intelligent agent action feedback of each time series is marked, and the offline function trajectory is predicted using the intelligent agent action feedback through the offline MARL data-driven training paradigm to obtain the regressed offline function trajectory. The regressed offline function trajectory is obtained through serialized modeling to realize the basic functions of the intelligent agent under the condition of a single independent agent.

[0053] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0054] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems) and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0055] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0056] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0057] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A method and system for constructing a diversified interactive intelligent agent based on offline learning, characterized in that: include: An offline cycle system and an intelligent agent interaction system, wherein the offline cycle system comprises a process splitting unit, a self-evolution unit, a storage processing unit, and a model calling unit, and the intelligent agent interaction system comprises an information sharing unit and a collaboration unit; The offline loop system uses an offline reinforcement algorithm to process the basic functions of a single agent under independent conditions, wherein the basic functions of the agent include thinking function, perception function, and execution function; The intelligent agent interaction system uses an interactive adjustment learning algorithm to enhance the information stored in the processing unit in the offline circulation system, and uses a distance contraction adjustment algorithm to improve the collaborative efficiency of the intelligent agent interaction system. It can exchange information with other single intelligent agents and effectively complete complex tasks.

2. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The offline cycle system includes a process splitting unit, a self-evolution unit, a storage processing unit, and a model calling unit; wherein, The offline cycle system is a diversified unit set that realizes preset functions without information interaction with other intelligent agents. The diversified unit set is composed of different types of unit sets, including process splitting units, self-evolution units, storage processing units, and model calling units. The unit set can independently complete the preset function, and the diversified unit set can complete multiple preset functions; The process splitting unit can decompose complex functions into a collection of simple functions; The self-evolution unit improves the effect of achieving the preset function of the unit set by evaluating the unit set to achieve the preset function; The storage processing unit uses the accumulated existing unit set to guide the implementation of the preset function; The model calling unit can call an external model to enhance the unit set function, and the external model is an existing database and plug-in library that assists in implementing the current unit set.

3. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The agent interaction system comprises an information sharing unit and a collaboration unit; wherein: The agent interaction system improves the performance of agents through the collaboration of different agents; The information sharing unit can share the diverse unit set in the offline cycle system with other agents; The collaboration unit enhances the ability of multiple agents to complete preset functions through collaboration of multiple agents.

4. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The basic functions of the intelligent agent include thinking function, perception function and execution function; among which: The thinking function is that the intelligent agent can split complex tasks into simpler subtasks and continuously analyze the results to achieve improvement and thinking capabilities; The perception function realizes external input through text input, image capture, and voice recognition; The execution function can call an external model to assist the intelligent agent in realizing the preset function.

5. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The offline cycle system uses an offline reinforcement algorithm to process the basic functions of the agent under the condition of a single agent being independent; wherein: The offline reinforcement algorithm considers the basic functions of a single agent under independent conditions as an offline function trajectory. , expressed as: ； in, Indicates the observation state, Represents the global state, Indicates instructions, is the reinforcement value, T represents the time series; Model offline functional trajectories in a serialized manner and mark the agent action feedback of each time series , expressed as: ； Where T represents the time series, e represents the natural constant, Represented as the agent action feedback of the previous time series; Through the offline MARL data-driven training paradigm, the agent action feedback is used Predict the offline function trajectory to obtain the regression offline function trajectory , expressed as: ； in, is the observed state of the agent in time series t, is the global state of the agent in time series t; Track through the offline function After serialization modeling, the regression offline function trajectory is obtained Realize the basic functions of the single agent under independent conditions.

6. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The agent interaction system uses an interactive adjustment learning algorithm to enhance the information stored in the processing unit in the offline circulation system, and uses a distance contraction adjustment algorithm to improve the collaboration efficiency of the agent interaction system; wherein: The interactive adjustment learning algorithm enhances the stability of the agent through unified training and distributed implementation of the basic functions of the agent; The interactive adjustment learning algorithm comprises the steps of: S1 normalizes and reconstructs the regression offline function trajectory; S2 uses hidden state analysis to build a self-consistent model; S3 improves the collaboration efficiency of the intelligent agent interaction system through a distance contraction adjustment algorithm.

7. The S1 of claim 6 normalizes and reconstructs the regression offline function trajectory, characterized in that: Agent action feedback for each time series , expressed as: ； in, Represents the normalized agent action feedback of the previous time series; The normalization strategy C is expressed as: ； Among them, K represents the sampling degree of time series; For the normalized regression offline function trajectory G after normalization and reconstruction, it is expressed as: ； Complete offline function trace After serialization modeling, the regression offline function trajectory is obtained The process is normalized to obtain the normalized regression offline function trajectory.

8. The S2 of claim 6 uses hidden state analysis to establish a self-consistent model, characterized in that: The hidden state analysis method first samples other agents n to obtain sampling trajectories of other agents , expressed as: ； Among them, v represents the sampling degree, Represents the sampling state of agent n in time series t.

9. The S3 of claim 6 improves the collaboration efficiency of the agent interaction system by using a distance contraction adjustment algorithm, characterized in that: Sampling trajectories of other agents through the nearest neighbor algorithm Probability distribution extraction is performed, and the contraction adjustment algorithm is used when the probability distribution is extracted. The contraction adjustment algorithm is used to calculate the sampling trajectory of other intelligent agents. Compare with the normalized regression offline function trajectory G to obtain the normalized regression offline function shrinkage trajectory of the time series with the minimum difference , expressed as: ； in, represents the normalized regression offline feature trajectory vector, Represents all normalized regression offline function trajectory vectors obtained, Indicates the minimum value; Among them, The exponential moving average method is used to improve efficiency when traversing all normalized regression offline function trajectory vectors.

10. The method and system for constructing a diversified interactive intelligent agent based on offline learning as claimed in claim 1, characterized in that: The said can exchange information with other single intelligent agents to effectively complete complex tasks; wherein: The agent error is reduced by introducing a spontaneous pre-training conversion model when interacting with other single agents; The spontaneous pre-training conversion model fits the normalized regression offline function trajectory G of multiple agents through the MAPPO algorithm to improve the accuracy of completing complex tasks.

Images