Intelligent agent construction method and device, task processing method and device, equipment and medium

Abstract

The invention discloses an agent construction method, a task processing method and device, equipment and a medium. The intelligent agent construction method comprises the following steps: performing the following operations through an initial intelligent agent to construct and obtain a target intelligent agent: obtaining target information based on perception information obtained by perception of the initial intelligent agent; determining corresponding task processing rule data based on the first information type of the target information; performing decision processing on the task processing rule data based on the target information to obtain a task decision result; and performing task processing based on the task decision result. According to the method, the construction efficiency and effect of the intelligent agent can be improved, and the task processing capacity of the intelligent agent is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent agent technology, and in particular to an intelligent agent construction method, task processing method, device, equipment and medium. Background Art

[0002] As a virtual object or virtual model, an intelligent agent can perceive the surrounding environment information and perform corresponding actions like humans. By building an intelligent agent, humans can handle some tasks. The most important part of an intelligent agent is its brain, which has the ability to reason and make decisions. The development of intelligent agents has gone through many stages. For example, the early intelligent agent brains included multimodal expert systems based on machine learning, and recently developed into multimodal large models as the brains of intelligent agents for reasoning and decision-making. Among them, the large model has excellent performance in language understanding and generation, but as the brain of an intelligent agent, it still lacks common sense reasoning and understanding, requires huge computing resources, the model is not interpretable, and the model may be biased. Summary of the invention

[0003] The embodiments of the present application are intended to solve at least one of the technical problems in the related art to a certain extent. To this end, the purpose of the embodiments of the present application is to propose a method and device for constructing an intelligent agent, a method and device for processing tasks performed by an intelligent agent, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] An embodiment of the present application provides a method for constructing an intelligent agent, which includes constructing a target intelligent agent by performing the following operations by an initial intelligent agent: obtaining target information based on perception information perceived by the initial intelligent agent; determining corresponding task processing rule data based on a first information type of the target information; performing decision processing by the task processing rule data based on the target information to obtain a task decision result; and performing task processing based on the task decision result.

[0005] Exemplarily, obtaining target information based on the perception information perceived by the initial intelligent agent includes: performing cognitive processing on the perception information to obtain cognitive information; and performing information analysis processing on at least one of the perception information and the cognitive information to obtain the target information.

[0006] Exemplarily, the initial intelligent agent includes an instinctive brain, and the cognitive information includes instinctive cognitive information; the cognitive processing of the perceptual information to obtain cognitive information includes: determining a first weight corresponding to the instinctive brain based on a second information type of the perceptual information; and the instinctive brain cognitively processes the perceptual information based on the first weight to obtain the instinctive cognitive information.

[0007] Exemplarily, the initial intelligent agent includes a rational brain, and the cognitive information includes rational cognitive information; the cognitive processing of the perceptual information to obtain cognitive information includes: determining a second weight corresponding to the rational brain based on a second information type of the perceptual information; and the rational brain cognitively processes the perceptual information based on the second weight to obtain the rational cognitive information.

[0008] Exemplarily, the initial intelligent agent includes an emotional brain, and the cognitive information includes emotional cognitive information; the cognitive processing of the perceptual information to obtain cognitive information includes: determining a third weight corresponding to the emotional brain based on a second information type of the perceptual information; and the emotional brain cognitively processes the perceptual information based on the third weight to obtain the emotional cognitive information.

[0009] Exemplarily, the second information type includes at least one of the following: a physical signal type of the external environment, a physical state signal type of the initial intelligent agent.

[0010] Exemplarily, determining the corresponding task processing rule data based on the first information type of the target information includes: determining the corresponding task type based on the first information type; and determining the task processing rule data corresponding to the task type.

[0011] Exemplarily, the first information type includes at least one of the following: an environmental information type, a stimulus information type, a state information type of the initial agent, and an interest information type.

[0012] Exemplarily, determining the corresponding task type based on the first information type includes at least one of the following: when the first information type is the environmental information type, determining the corresponding task type is an attention task type; when the first information type is the stimulus information type, determining the corresponding task type is a reflexive task type; when the first information type is the state information type of the agent, determining the corresponding task type is an unconscious task type; when the first information type is the interest information type, determining the corresponding task type is a memory task type.

[0013] Exemplarily, the task processing rule data performs decision processing based on the target information to obtain a task decision result, including at least one of the following: the task processing rule data corresponding to the attention task type performs an attention decision on the target information to obtain a cognitive task decision result and an action task decision result; the task processing rule data corresponding to the reflexive task type performs a trigger decision on the target information to obtain an action task decision result; the task processing rule data corresponding to the unconscious task type performs a subconscious decision on the target information to obtain at least one of a cognitive task decision result and an action task decision result; the task processing rule data corresponding to the memory task type performs a memory search decision on the target information to obtain at least one of a cognitive task decision result and an action task decision result.

[0014] Exemplarily, the task processing based on the task decision result includes at least one of the following: when the task decision result includes multiple first actions to be executed, executing the multiple first actions to be executed based on the priorities of the multiple first actions to be executed; when the task decision result includes multiple second actions to be executed, selecting at least one second action to be executed from the multiple second actions to be executed based on a preset selection rule, and executing the at least one second action to be executed.

[0015] Exemplarily, the multiple first actions to be executed include at least one of the following: a first sub-action to be executed of an emergency type obtained based on the perception information, a second sub-action to be executed of a non-emergency type obtained based on the perception information, and a third sub-action to be executed based on the cognitive information, wherein the priority of the first sub-action to be executed is higher than the priority of the second sub-action to be executed and the priority of the third sub-action to be executed.

[0016] Exemplarily, the method for constructing the intelligent agent also includes: converting the information format of at least one of the perception information and the cognitive information to obtain information in a knowledge graph format; and constructing a knowledge graph based on the information in the knowledge graph format to obtain a constructed current knowledge graph.

[0017] Exemplarily, the task processing rule data performs decision processing based on the target information to obtain a task decision result, including: the task processing rule data uses the information in the knowledge graph format as a query condition, performs information query in the current knowledge graph, and obtains the task decision result.

[0018] Exemplarily, the information in the knowledge graph format includes multiple subject information and relationship information, and the relationship information represents the relationship between the multiple subject information; the task processing rule data uses the information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain the task decision result, including: using at least one of the multiple subject information and / or the relationship information as a query condition to perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the multiple subject information and / or other subject information associated with the relationship information; and obtaining the task decision result based on the other relationship information and at least one of the other subject information.

[0019] Exemplarily, the method for constructing the intelligent agent further includes: when the current knowledge graph is constructed, associating the current knowledge graph with the historical knowledge graph.

[0020] Exemplarily, the task processing rule data uses information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain the task decision result, including: the task processing rule data uses information in the knowledge graph format as a first query condition to perform information query in the current knowledge graph to obtain a first query result; when the first query result is a preset query result, the task processing rule data uses information in the knowledge graph format and at least one of the first query result as a second query condition to perform information query in the historical knowledge graph to obtain a second query result; and the task decision result is obtained based on at least one of the first query result and the second query result.

[0021] Exemplarily, the initial intelligent agent includes a perception layer, a cognitive layer, a decision-making layer and an execution layer; the cognitive layer includes a reptilian brain, a rational brain and an emotional brain; the decision-making layer includes a fast system and a slow system; the instinctive brain includes the reptilian brain and the fast system.

[0022] Another embodiment of the present application provides a task processing method performed by an intelligent agent, the task processing method including the following operations performed by a constructed intelligent agent: obtaining target information based on perception information perceived by the constructed intelligent agent; determining corresponding task processing rule data based on a first information type of the target information; performing decision processing by the task processing rule data based on the target information to obtain a task decision result; and performing task processing based on the task decision result.

[0023] Another embodiment of the present application provides a device for constructing an intelligent body, which includes: a first acquisition module, used to obtain target information based on perception information perceived by an initial intelligent body; a first determination module, used to determine corresponding task processing rule data based on a first information type of the target information; a first decision module, used to perform decision processing based on the target information by the task processing rule data to obtain a task decision result; and a first processing module, used to perform task processing based on the task decision result.

[0024] Another embodiment of the present application provides a task processing device, which includes: a second acquisition module, used to obtain target information based on perception information perceived by a constructed intelligent agent; a second determination module, used to determine corresponding task processing rule data based on a first information type of the target information; a second decision module, used to perform decision processing based on the target information by the task processing rule data to obtain a task decision result; and a second processing module, used to perform task processing based on the task decision result.

[0025] Another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in any one of the above-mentioned intelligent agent construction method and task processing method are implemented.

[0026] Another embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the above-mentioned intelligent agent construction method and task processing method.

[0027] Another embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed by a processor of a computer device, the computer device can perform the steps of the method described in any of the above embodiments.

[0028] In the embodiment of the present application, the initial agent performs the following operations to construct the target agent: based on the perception information perceived by the initial agent, the target information is obtained; based on the first information type of the target information, the corresponding task processing rule data is determined; the task processing rule data performs decision processing based on the target information to obtain a task decision result; and the task is processed based on the task decision result. The method of the present application can improve the efficiency and effect of the construction of the agent and improve the task processing capability of the agent. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of a method for training an intelligent agent provided in an embodiment of the present application;

[0030] Figure 2 A schematic diagram of a method for implementing humanoid general artificial intelligence provided by an embodiment of the present application;

[0031] Figure 3 A schematic diagram of a triple brain provided for an embodiment of the present application;

[0032] Figure 4 A schematic diagram of a fast system and a slow system in a decision layer provided for an embodiment of the present application;

[0033] Figure 5 A schematic diagram of an intelligent agent working framework provided for an implementation method of the present application;

[0034] Figure 6 A flowchart of a method for constructing an intelligent agent provided in an embodiment of the present application;

[0035] Figure 7 A flowchart for obtaining target information provided by an implementation method of the present application;

[0036] Figure 8 A flow chart for obtaining instinctive cognitive information provided by an embodiment of the present application;

[0037] Fig. 9 A flow chart for obtaining rational cognitive information provided for an implementation method of the present application;

[0038] Fig.10 A flowchart of obtaining emotion recognition information provided by an embodiment of the present application;

[0039] Fig.11 A flowchart for determining task processing rule data provided for an implementation method of the present application;

[0040] Fig.12 A schematic diagram of the workflow of the instinctive brain classified by task provided in the implementation mode of the present application;

[0041] Fig.13 A schematic diagram of the instinctive brain working mode provided for the implementation mode of the present application;

[0042] Fig.14 A flowchart for constructing the current knowledge graph provided for the implementation method of this application;

[0043] Fig.15 A flowchart for obtaining a task decision result provided for an implementation method of the present application;

[0044] Fig.16 A flowchart for obtaining a task decision result provided for an implementation method of the present application;

[0045] Fig.17A flowchart of a task processing method performed by an agent provided in an embodiment of the present application;

[0046] Fig.18 A schematic diagram of a device for constructing an intelligent entity provided in an embodiment of the present application;

[0047] Fig.19 A schematic diagram of a task processing device provided in an embodiment of the present application;

[0048] Fig. 20 A block diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0049] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0050] The development of intelligent agents has gone through many stages. The mainstream intelligent agent learning includes multimodal expert systems based on machine learning, and has recently developed into multimodal large models as the brain of the intelligent agent for reasoning and decision-making. Among them, the large model has excellent performance in language understanding and generation, but as the brain of the intelligent agent, it still lacks common sense reasoning and understanding, requires huge computing resources, the model is not interpretable, and the model may be biased.

[0051] Figure 1 A schematic diagram of a method for training an intelligent agent provided in an embodiment of the present application.

[0052] like Figure 1 As shown, an embodiment of the present application proposes a method for training an intelligent agent, which trains the policy network of the intelligent agent so that it can make the best decision based on the current state observation value to achieve precise control of the virtual object and maximize the reward return value. However, from the perspective of achieving general artificial intelligence, this method uses a supervised learning method to train the intelligent agent. This method has a low degree of versatility, a single objective function (maximizing reward), resulting in poor training results, and a lack of reasoning and cognitive abilities. The development of intelligent agents needs to place more emphasis on versatility, autonomous learning ability, and cognitive abilities.

[0053] Figure 2 A schematic diagram of a method for implementing humanoid general artificial intelligence provided in an embodiment of the present application.

[0054] like Figure 2As shown, the embodiments of the present application realize an artificial intelligence system with more cognitive and emotional capabilities by modeling human-like functions such as attention mechanism, self-awareness and empathy, so that the machine can process and understand various types of information more like humans. However, this method realizes the instinctive reaction of the machine through pre-made programs, which is not universal enough and difficult to handle unexpected situations; the reasoning model of this method includes complex technologies such as attention mechanism, multi-level feature extraction, chain associative activation, etc., which requires complex computing resources, such as a large amount of computing resources and storage resources to implement, resulting in the high cost of implementing this method and difficulty in expansion; the model of this method is not interpretable.

[0055] In view of the problems existing in the intelligent agent in the above-mentioned implementation mode, the present application proposes an optimized method for constructing an intelligent agent and a method for processing tasks performed by the intelligent agent. Figure 3-Figure 17 The method for constructing an intelligent agent and the method for processing tasks performed by the intelligent agent in the embodiment of the present application are described in detail.

[0056] Figure 3 Schematic diagram of a triple brain provided for an embodiment of the present application.

[0057] like Figure 3 As shown, the reptilian brain in the human brain is also called the primitive brain, including the brain stem and cerebellum. It is the first brain component to appear and controls desires and instincts. The limbic system in the human brain is located in the middle part of the brain, including the hypothalamus, hippocampus and amygdala, which controls human emotions. The reptilian brain and limbic system control 95% of human behavior. The neocortex is located above the hemispheres of the human brain, occupying 94% of the entire surface of the cerebral cortex of an adult, and controls rational thinking. This application splits the intelligent body from a cognitive perspective, splitting it into three brains with cognitive abilities, namely the reptilian brain, emotional brain and rational brain. The reptilian brain controls instinctive thoughts and physiological consciousness, the emotional brain is used to generate emotions, and the rational brain is used to process information in terms of logic, language, creativity and thinking.

[0058] Figure 4 A schematic diagram of a fast system and a slow system in a decision layer provided for an implementation manner of the present application.

[0059] like Figure 4 As shown, the present application also splits the intelligent agent from the perspective of decision-making into two systems: a fast system and a slow system. The fast system is a commonly used, intuitive, unconscious decision-making system. The slow system is a conscious decision-making system that requires active control.

[0060] Figure 5 A schematic diagram of the intelligent agent working framework provided for an implementation manner of the present application.

[0061] like Figure 5As shown, the initial agent includes an original virtual object, which is composed of some object definition information, data processing rules, logic codes, etc. The original virtual object may be an object that has not yet performed task processing. Alternatively, the initial agent may also be an agent that has some learning capabilities by performing some task processing on the original virtual object.

[0062] In one example, the initial intelligent agent includes a perception layer, a cognitive layer, a decision layer, and an execution layer; the cognitive layer includes a reptilian brain, a rational brain, and an emotional brain; the decision layer includes a fast system and a slow system. Among them, the instinctive brain in the embodiment of the present application may include a reptilian brain and a fast system.

[0063] Specifically, the perception layer perceives external information and its own information, obtains perception information, and directly transmits the perception information to the cognitive layer or transmits it to the cognitive layer after preprocessing the perception information. The cognitive layer includes the reptilian brain, the rational brain, and the emotional brain. After the perception information reaches the cognitive layer, the cognitive layer determines the cognitive processing weight of the triple brain according to the type of perception information. The triple brain performs cognitive processing based on the weight to obtain cognitive layer information. All cognitive layer information is summarized into cognitive information and handed over to the decision layer. The decision layer includes a fast system and a slow system. The decision layer first comprehensively analyzes the current state of the intelligent body based on the perception information and cognitive information, and dispatches attention between the fast system and the slow system according to the state. Finally, the fast and slow systems generate the actions that need to be executed at the moment according to the summarized information through rule processing (executed by the fast system) or logical inference (executed by the slow system). The generated action instructions are handed over to the execution layer, and the execution layer executes the corresponding actions after receiving the action instructions.

[0064] To be more specific, the "instinctive brain" with basic cognitive and decision-making functions includes the reptilian brain used to process instinctive thoughts and physiological consciousness, as well as the commonly used, intuitive, unconscious fast system.

[0065] It can be seen that the present application splits the cognitive and decision-making brains of the intelligent body according to functional logic, mainly into an "instinctive brain" with basic cognitive and decision-making functions, an "emotional brain" with the ability to generate emotions, and a "reasoning brain" with advanced thinking, cognitive and decision-making functions. Among them, the "instinctive brain" is responsible for processing cognitive and decision-making tasks that do not require complex reasoning, reducing the computing resources required for the intelligent body to perform tasks and improving the interpretability of the intelligent body. Part of the cognitive processing and part of the decision-making processing in the construction method and task processing method of the intelligent body in the embodiments below mainly rely on the "instinctive brain" to achieve, of course, another part of the cognitive processing can also be achieved by the emotional brain and the reasoning brain, and another part of the decision-making processing can also be achieved by the slow system.

[0066] Figure 6 It is a flowchart of a method for constructing an intelligent body according to one embodiment of the present application.

[0067] like Figure 6 As shown, the method for constructing an intelligent agent includes performing the following operations S1-S4 by the initial intelligent agent to construct a target intelligent agent.

[0068] S1, based on the perception information perceived by the initial agent, obtain the target information.

[0069] As an example, the initial intelligent agent can obtain visual, tactile and other sensory information, for example, the sensory information can be obtained through various types of sensors, for example, visual sensory information can be obtained through a camera. The sensory information can be sensory information from the outside world, or sensory information from the initial intelligent agent itself. For example, the sensory information is obtained by the sensory layer of the initial intelligent agent, and then the sensory information is processed by at least one of the cognitive layer and the decision layer to obtain the target information.

[0070] S2: Determine corresponding task processing rule data based on the first information type of the target information.

[0071] S3, the task processing rule data performs decision processing based on the target information to obtain the task decision result.

[0072] Specifically, different types of target information need to be processed by different task processing rule data. The first type of target information characterizes the task type, and the task processing rule data applicable to processing the task type can be determined based on the first information type of the target information. The task processing rules can be pre-set, or updated or changed during the process of the agent performing task processing according to the improvement of the task processing capability.

[0073] As an example, after determining the task processing rule data corresponding to the target information, the task processing rule data performs decision processing based on the target information, and the obtained task decision results can be multiple types of results, including, for example, cognitive task decision results and action task decision results. The cognitive task decision results usually indicate that the initial intelligent agent does not need to perform actions, but only needs to perform cognitive operations. The action task decision results indicate that the intelligent agent needs to perform corresponding actions.

[0074] S4, perform task processing based on the task decision results.

[0075] After the decision layer determines the task decision result, it transmits the result to the execution layer. The execution layer performs task processing according to the task decision result. The task processing can be a cognitive processing or the execution of corresponding actions.

[0076] It can be understood that the above-mentioned method for constructing an intelligent agent uses the perception information perceived by the initial intelligent agent, and determines the task processing rule data corresponding to the perception information, and the task processing rule data performs decision processing based on the target information, obtains the task decision result and performs task processing. The intelligent agent is constructed in this way, which improves the efficiency and effect of constructing the intelligent agent, improves the task processing ability of the intelligent agent, and improves the interpretability of the intelligent agent. Different from the method of large model training and learning, the method for constructing an intelligent agent of the present application can better construct the target intelligent agent, so that the constructed target intelligent agent can efficiently handle general tasks.

[0077] In one embodiment of the present application, Figure 7 As shown, based on the perception information perceived by the initial intelligent agent, the target information is obtained, including steps S11-S12.

[0078] S11, cognitive processing is performed on the perception information to obtain cognitive information.

[0079] Exemplarily, the initial intelligent agent obtains the perception information by perceiving the external environment information or its own state information through the perception layer. The cognitive layer of the initial intelligent agent cognitively processes the perception information to obtain cognitive information. The perception information includes different types of perception information, such as perception information including a second information type, and cognitive processing is performed on the perception information in different ways according to different second information types. The second information type includes, for example, a physical signal type of the external environment, a physical state signal type of the initial intelligent agent, and the like.

[0080] S12, performing information analysis processing on at least one of the perception information and the cognitive information to obtain target information.

[0081] Then, the perception layer can send the perception information directly to the decision layer, and the cognitive layer can send the cognitive information to the decision layer. The decision layer performs information analysis on at least one of the perception information and the cognitive information to obtain the target information. The information analysis process may include parsing and splitting at least one of the perception information and the cognitive information to obtain different types of target information, such as obtaining the target information of the first information type, and cognitively processing the perception information in different ways according to the second information type. The second information type includes, for example, environmental information type, stimulating information type, initial intelligent agent state information type, information type of interest, and the like. Environmental information includes the environmental scene in which the intelligent agent is located, stimulating information includes dangerous information, the initial intelligent agent state information includes the intelligent agent's own state information, and the information of interest includes environmental information that attracts the attention of the intelligent agent.

[0082] The initial intelligent agent may include the instinctive brain (composed of the reptilian brain and the fast system), the rational brain, and the emotional brain. The instinctive brain, the rational brain, and the emotional brain process the perceived information differently, and the cognitive information obtained is also different. After the perceived information reaches the cognitive layer, the cognitive processing weights of the instinctive brain, the rational brain, and the emotional brain are determined according to the type of perceived information, and cognitive processing of the perceived information is performed based on the weights.

[0083] In one embodiment of the present application, Figure 8 As shown, the initial intelligent agent includes an instinctive brain, and the cognitive information includes instinctive cognitive information. The cognitive processing of the perception information to obtain cognitive information includes steps S111-S112.

[0084] S111, determining a first weight corresponding to the instinctive brain based on the second information type of the perceived information.

[0085] S112, the instinctive brain performs cognitive processing on the perception information based on the first weight to obtain instinctive cognitive information.

[0086] Specifically, the instinctive brain includes the reptilian brain and the fast system, which are responsible for processing cognitive and decision-making tasks that do not require complex reasoning. When the instinctive brain performs cognitive processing on the perceptual information, the initial intelligent agent determines the first weight corresponding to the instinctive brain based on the second information type of the perceptual information, and the instinctive brain performs cognitive processing on the perceptual information based on the first weight, and the cognitive information obtained is the instinctive cognitive information.

[0087] As an example, the second information type includes at least one of the following: a physical signal of the external environment, a physical state signal of the initial agent.

[0088] Specifically, the second information type is the information type of perception information, and the perception information may be a physical signal of the external environment or a physical state signal of the initial intelligent agent. For example, the physical signal of the external environment includes the perception information of the five senses of the intelligent agent and external information, and the physical state signal of the initial intelligent agent includes the fatigue level of the initial intelligent agent. The first weight corresponding to the instinctive brain may be different for different second information types. For example, when the physical signal of the external environment includes information that does not require complex reasoning, and the physical state signal of the initial intelligent agent is information that requires complex reasoning, the first weight corresponding to the physical signal of the external environment is larger, and the first weight corresponding to the physical state signal of the initial intelligent agent is smaller.

[0089] In one embodiment of the present application, Fig. 9 As shown, the initial intelligent agent includes a rational brain, and the cognitive information includes rational cognitive information; cognitive processing is performed on the perception information to obtain cognitive information, including steps S113-S114.

[0090] S113, determining a second weight corresponding to the rational brain based on the second information type of the perceived information.

[0091] S114, the rational brain performs cognitive processing on the perception information based on the second weight to obtain rational cognitive information.

[0092] Specifically, when the rational brain cognitively processes the perceived information, the initial intelligent agent determines the second weight corresponding to the rational brain based on the second information type of the perceived information, and the rational brain cognitively processes the perceived information based on the second weight, and the obtained cognitive information is rational cognitive information.

[0093] Similar to the instinctive brain, the rational brain may have different corresponding second weights for different types of second information. For example, when the physical signal of the external environment includes information that does not require rational reasoning, and the body state signal of the initial intelligent agent is information that requires rational reasoning, the second weight corresponding to the physical signal of the external environment is smaller, and the second weight corresponding to the body state signal of the initial intelligent agent is larger.

[0094] In one embodiment of the present application, Fig.10 As shown, the initial intelligent agent includes an emotional brain, and the cognitive information includes emotional cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including S115-S116.

[0095] S115, determining a third weight corresponding to the emotional brain based on the second information type of the perceived information.

[0096] S116, the emotional brain performs cognitive processing on the perception information based on the third weight to obtain emotional cognitive information.

[0097] Specifically, the initial intelligent agent determines the third weight corresponding to the emotional brain based on the second information type of the perceived information, and the emotional brain cognitively processes the perceived information based on the third weight, and the obtained cognitive information is emotional cognitive information.

[0098] Similar to the instinctive brain or the rational brain, the third weights corresponding to the emotional brain may be different for different types of second information. For example, when the physical signal of the external environment includes information with a low degree of impact on emotions, and the physical state signal of the initial intelligent agent is information with a high degree of impact on emotions, the third weight corresponding to the physical signal of the external environment is smaller, and the third weight corresponding to the physical state signal of the initial intelligent agent is larger.

[0099] It can be understood that the above-mentioned method of constructing an intelligent agent splits the intelligent agent into three brains: the instinctive brain, the rational brain, and the emotional brain. The three brains have different processing capabilities and tendencies when processing data. Therefore, different weights are configured for the three brains, making the three brains more targeted when cognitively processing perceptual information, and having higher processing effects and efficiency. Among them, the "instinctive brain" is mainly responsible for cognitive and decision-making tasks that do not require complex reasoning, which reduces the computing resources required for the intelligent agent to perform tasks and improves the interpretability of the intelligent agent. Different from the large model training and learning method, this method of constructing an intelligent agent can better construct the target intelligent agent. The target intelligent agent can process general tasks with low power consumption based on the instinctive brain, thereby improving task processing efficiency.

[0100] On the other hand, by splitting the cognitive layer into three types of brains: the reptilian brain, the emotional brain, and the rational brain, the three different brains process the perceptual information according to the corresponding weights. This allows for more efficient and quick cognitive processing of perceptual information. For example, when performing complex logical reasoning on perceptual information, the weight of the rational brain will be greater, which is more conducive to the classification and processing of perceptual information.

[0101] As an example, the perception layer is responsible for perceiving external information and its own state information, that is, obtaining the physical signals of the external environment and the physical state signals of the initial intelligent body (such as the fatigue state of the initial intelligent body) to obtain the perception information, and the perception information is converted into a unified format after interpretation and handed over to the cognitive layer. The cognitive layer receives the perception information and determines the second information type of the perception information. The initial intelligent body performs triple brain cognitive inference based on the second information type of the perception information. The instinctive brain generates desire information and memory retrieval based on rules, the emotional brain generates emotional information, and the rational brain generates ideas based on the perception information and self-awareness. Different types of brains process different types of perception information based on different weights to obtain cognitive information. Specifically, after the perception information reaches the cognitive layer, the cognitive processing weight of the triple brain is determined according to the second information type of the perception information, and cognitive processing is performed based on the weight. After the triple brain cognitive processing, the instinctive cognitive information, rational cognitive information, and emotional cognitive information processed by the triple brain are summarized into cognitive information, and the perception information and cognitive information are sent to the decision layer, and then the decision layer performs information parsing processing on at least one of the perception information and cognitive information to obtain the target information. Next, the decision layer of the initial intelligent agent determines the corresponding task processing rule data based on the first information type of the target information.

[0102] In one embodiment of the present application, Fig.11 As shown, based on the first information type of the target information, the corresponding task processing rule data is determined, including S21-S22.

[0103] S21: Determine a corresponding task type based on the first information type.

[0104] S22, determining task processing rule data corresponding to the task type.

[0105] Specifically, the present application divides the types of processing tasks, and different types of target information correspond to different task types and also correspond to different task processing rule data.

[0106] As an example, the first information type includes at least one of the following: environmental information type, stimulating information type, initial agent state information type, and interest information type. According to the first information type of the target information, a corresponding task type is determined.

[0107] As an example, determining the corresponding task type based on the first information type includes at least one of the following:

[0108] In the case where the first information type is an environmental information type, it is determined that the corresponding task type is an attention task type.

[0109] In the case where the first information type is a stimulating information type, it is determined that the corresponding task type is a reflective task type.

[0110] In the case where the first information type is the state information type of the agent, it is determined that the corresponding task type is an unconscious task type.

[0111] In the case where the first information type is the information type of interest, it is determined that the corresponding task type is the memory task type.

[0112] Specifically, the present application divides task types into four types, namely, attention task type, reflective task type, unconscious task type and memory task type.

[0113] Attention tasks are tasks that require the instinctive brain to obtain a lot of attention and dispatch the fast and slow systems to execute. Attention tasks include, for example: tasks that attract attention such as moving objects, unusual patterns, sounds, and strong lights; tasks that require "looking around" in unfamiliar environments; tasks that require the agent to pay attention to and follow people or new objects that appear in the room; and tasks that require observation and gradual adaptation to the layout of the environment when entering a new environment.

[0114] Reflexive tasks do not require any attention scheduling and are executed immediately once triggered. Examples include: hearing a horn on a noisy street, which requires the agent to pay attention to look for potential dangers; touching a hot object, which requires the agent to quickly withdraw its hand to avoid injury; and detecting an impending collision and automatically avoiding it to protect itself.

[0115] Unconscious tasks are subconscious tasks that do not require attention. For example, they include: small movements made to relieve the physical discomfort of the intelligent agent, such as turning the head and blinking; tasks that generate ideas based on physiological needs.

[0116] Memory tasks are tasks that do not require complex reasoning, memory retrieval, or attention scheduling. For example, when seeing an acquaintance, the agent's brain needs to automatically search for the acquaintance's information, quickly identify and understand the task; when meeting an acquaintance, the agent needs to greet the acquaintance.

[0117] The initial agent needs to determine the corresponding task type based on the first information type of the target information. When the target information is environmental information, the corresponding task is an attention task. When the target information is stimulating information, the corresponding task is determined to be a reflex task. When the target information is the state information of the agent, the corresponding task is determined to be an unconscious task. When the target information is information of interest, the corresponding task is determined to be a memory task. According to different types of tasks, the corresponding task processing rule data is determined, and the corresponding task processing rule data is processed to obtain a decision result.

[0118] In one embodiment of the present application, the task processing rule data performs decision processing based on the target information to obtain a task decision result, including at least one of the following:

[0119] The task processing rule data corresponding to the attention task type is used to make attention decisions on the target information, and cognitive task decision results and action task decision results are obtained.

[0120] For example, attention tasks include: tasks that attract attention such as moving objects, unusual patterns, sounds, and strong lights; tasks that require "looking around" in unfamiliar environments; tasks that require the agent to pay attention to and follow people or new objects that appear in the room; and tasks that require the agent to observe and gradually adapt to the layout of the environment when entering a new environment. The task decision results corresponding to these tasks include cognitive task decision results and action task decision results, that is, these tasks all require cognition and the execution of corresponding actions, such as recognizing the unfamiliar environment and performing the action of "looking around" in an unfamiliar environment.

[0121] The task processing rule data corresponding to the reflective task type is used to trigger the target information and obtain the action task decision result.

[0122] For example, reflexive tasks include: tasks that require the agent to pay attention to potential dangers when hearing a horn on a noisy street; tasks that require the agent to quickly withdraw its hand to avoid injury when touching a hot object; and tasks that automatically avoid collisions to protect itself when detecting an impending collision. The task decision results corresponding to these tasks include action-type task decision results, that is, these tasks require the execution of corresponding actions, such as the action of "quickly withdrawing your hand" when touching a hot object.

[0123] A subconscious decision is made on the target information based on the task processing rule data corresponding to the unconscious task type to obtain at least one of a cognitive task decision result and an action task decision result.

[0124] Unconscious tasks include: small movements made to relieve the physical discomfort of the intelligent agent, such as turning the head and blinking. The task decision results corresponding to this task include cognitive task decision results and action task decision results, that is, to perform this task, it is necessary to recognize the physical discomfort and make "turning the head and blinking" movements; tasks that generate ideas based on physiological needs. The task decision results corresponding to this task include cognitive task decision results, that is, to perform this task, it is necessary to generate ideas based on physiological needs without performing actions.

[0125] A memory search decision is performed on the target information according to the task processing rule data corresponding to the memory task type to obtain at least one of a cognitive task decision result and an action task decision result.

[0126] Memory tasks include: seeing an acquaintance, which requires the intelligent agent's brain to automatically search for acquaintance information, quickly identify and understand the task. The task decision results corresponding to this task include cognitive task decision results, that is, executing this task requires identifying input but no action is required; the decision results corresponding to the task of greeting an acquaintance include cognitive task decision results and action task decision results, that is, executing this task requires retrieving cognitive input from memory and making a "greeting" action.

[0127] The initial agent determines the corresponding task type (attention task type, reflex task type, unconscious task type, memory task type) according to the first information type of the target information (environmental information type, stimulus information type, state information type of the initial agent, and information type of interest), and determines the task processing rule data corresponding to the task type based on the task type, and the task processing rule data performs decision processing based on the target information to obtain the task decision result (cognitive task decision result and action task decision result). Finally, the task processing is performed according to the task decision result.

[0128] As an example, the workflow of instinctive brain task classification is as follows Fig.12As shown, the instinctive brain decision module receives the original perceptual information and cognitive information (cognitive information includes rational cognitive information, instinctive cognitive information, and emotional cognitive information). First, it is divided into four different target information (environmental information, stimulating information, initial intelligent agent state information, and interest information) according to the instinctive brain task classification. Then determine the different task types corresponding to different target information (four task types: attention task, reflex task, unconscious task, and memory task). Environmental information corresponds to attention tasks, stimulating information corresponds to reflex tasks, i.e., emergency response, the state information of the initial intelligent agent corresponds to unconscious tasks, and the interest information, i.e., focus information (focus information is information that needs attention), corresponds to memory tasks. In one example, it is necessary to perform memory retrieval on the focus information to obtain memory retrieval results, and the memory retrieval results can be used as supplementary cognitive information for decision making. Based on the decision rules corresponding to different task types, the target information is processed separately to obtain decision results, and the decision results include actions or cognition. Then, different tasks are triggered according to the decision results, such as triggering attention tasks, emergency responses, and small movements. The task decision results include cognitive task decision results and action task decision results. The tasks corresponding to cognitive task decision results need to directly update cognitive information, and the tasks corresponding to action task decision results need to perform corresponding actions. In one example, the action task decision results can correspond to multiple actions, and multiple actions can be performed, or the action management module can select a suitable action to perform.

[0129] It should be noted that action management includes determining the execution priority of actions, or selecting some actions to execute from a variety of actions.

[0130] For example, when the task decision result includes multiple first actions to be performed, the multiple first actions to be performed are performed based on the priorities of the multiple first actions to be performed.

[0131] For example, when the task decision result includes multiple second actions to be executed, at least one second action to be executed is selected from the multiple second actions to be executed based on a preset selection rule, and the at least one second action to be executed is executed.

[0132] Specifically, if the task decision result includes multiple first actions to be executed, among all the first actions to be executed, the multiple first actions to be executed can be executed in sequence according to the execution priority of the actions, or at least one first action to be executed can be selected from the multiple first actions to be executed according to a preset selection rule.

[0133] As an example, the multiple first actions to be executed include at least one of the following: a first sub-action to be executed of an emergency type obtained based on perception information, a second sub-action to be executed of a non-emergency type obtained based on perception information, and a third sub-action to be executed based on cognitive information, wherein the priority of the first sub-action to be executed is higher than the priority of the second sub-action to be executed and the priority of the third sub-action to be executed.

[0134] Specifically, the action management module can execute the sub-actions to be executed according to the priority of the sub-actions to be executed. The sub-actions to be executed can be a first sub-action to be executed of an emergency type, a second sub-action to be executed of a non-emergency type, and a third sub-action to be executed based on cognitive information. The first sub-action to be executed and the second sub-action to be executed are usually determined based on the perception information and are executed according to the emergency type. The first sub-action to be executed is more urgent than the second sub-action to be executed, and the priority of the first sub-action to be executed is higher than that of the second sub-action to be executed. The third sub-action to be executed is a cognitive action, and its priority is also lower than the priority of the first sub-action to be executed.

[0135] As an example, Fig.13 The working mode of the instinctive brain shown in the figure, the instinctive brain includes cognitive modules (reptile brain) and decision modules (fast system). The cognitive modules in the instinctive brain handle simple cognitive tasks, for example: the perception module first perceives the "hunger information" of the body, and the instinctive brain cognitive module generates the desire to "eat" based on the "hunger information"; the perception module perceives that an "acquaintance" appears in the field of vision, and the instinctive brain cognitive module automatically retrieves the "basic information about the acquaintance" in the memory to form a cognition of the "acquaintance".

[0136] The input to the decision-making module in the instinctive brain may be perceptual information that has not been processed by the cognitive module, or it may be cognitive information (instinctive cognitive information, rational cognitive information, emotional cognitive information) that has been processed by the cognitive module.

[0137] The decision-making module can execute multiple actions to be executed in sequence according to the execution priority of the actions. For example, multiple first actions to be executed include: a first sub-action to be executed of an emergency type obtained based on the perception information (based on the perception of danger based on the perception information, a decision to take an evasive action), a second sub-action to be executed of a non-emergency type obtained based on the perception information (based on the perception of hunger based on the perception information, a decision to eat), and a third sub-action to be executed based on the cognitive information (for example, a person or object of interest appears in rational cognition, and a basic action such as "stare at it" is decided).

[0138] Based on the priority, if the decision module senses the presence of danger based on the perception information and decides to take an evasive action, the evasive action will be performed first.

[0139] If no emergency situation is perceived, the decision-making module can make a comprehensive decision based on instinctive cognitive information, rational cognitive information and emotional cognitive information at the same time. For example, a person or object of interest appears in rational cognition, and a basic action such as "staring at it" is decided; the desire information of "wanting to eat" appears in instinctive cognition or "basic information about acquaintances" in memory is automatically retrieved, and the action of "eating" is decided or the cognition of "acquaintances" is formed; specific emotions appear in emotional cognition, and a decision is made to respond to the specific emotion.

[0140] It should be noted that the priority of the first sub-action to be executed of the emergency type obtained based on the perception information is higher than the second sub-action to be executed of the non-emergency type obtained based on the perception information and the third sub-action to be executed based on the cognitive information. That is, the first sub-action to be executed of the emergency type obtained based on the perception information is executed first.

[0141] It can be understood that the embodiments of the present application execute actions based on the priority of the actions, so that the intelligent agent can process urgent tasks first, improve the effect of task processing, and make the intelligent agent's ability in task processing closer to that of humans.

[0142] In one embodiment of the present application, Fig.14 As shown, the method for constructing an intelligent agent also includes S41-S42.

[0143] S41, converting the information format of at least one of the perception information and the cognitive information to obtain information in a knowledge graph format.

[0144] S42, constructing a knowledge graph based on the information in the knowledge graph format to obtain the constructed current knowledge graph.

[0145] Specifically, at least one of the perception information and the cognitive information is converted into an information format to obtain information in a knowledge graph format. The information format conversion can be converted into an RDF (Resource Description Framework) format. It should be noted that the information format conversion is not limited to the RDF format, as long as it is the data format of the knowledge graph. Based on the information in the knowledge graph format, a knowledge graph is constructed to obtain a constructed current knowledge graph.

[0146] As an example, information in the knowledge graph format can be a type of information that includes a subject, a predicate, and an object. The subject, for example, corresponds to an entity, the predicate, for example, corresponds to a relationship, and the object, for example, corresponds to another entity or an attribute value. The knowledge graph format includes, for example: entity → another entity, or entity → attribute value, "→" represents a relationship, and the information format conversion process can be executed in a decision-making module or in a preprocessing module outside the instinctive brain. For example, if the perception information received by the intelligent agent is "I know person A, and person A is a star player", then the conversion format is "I → person A → star player", the first "→" represents the recognition relationship, and the second "→" represents the identity relationship.

[0147] As an example, the method for constructing an intelligent agent also includes: when the current knowledge graph is constructed, associating the current knowledge graph with the historical knowledge graph.

[0148] Specifically, the present application can also recursively construct a knowledge graph based on parts of speech, and can also add the currently constructed knowledge graph to the historical knowledge base for updating the knowledge base.

[0149] After building the current knowledge graph, you can perform information query in the current knowledge graph to obtain task decision results.

[0150] As an example, the task processing rule data performs decision processing based on the target information to obtain the task decision result, including: the task processing rule data uses the information in the knowledge graph format as the query condition, performs information query in the current knowledge graph, and obtains the task decision result.

[0151] Specifically, the instinctive brain can analyze perceptual information and cognitive information and split it into different task types, such as splitting input information into environmental information, stimulus information, initial agent status information, and information of interest. Different information types correspond to different task modules, and the corresponding task modules perform the function of querying the knowledge graph. With the information in the knowledge graph format as the query condition, information is queried in the current knowledge graph to obtain the task decision result, and the query result is the decision result.

[0152] For example, the knowledge graph can be queried based on any category of subject, predicate, and object. For example, if the agent currently sees "personnel A", the memory-based task module is called to query the graph based on "personnel A" to obtain the information "I know A, he is a star player" as the query result, and a decision result is generated based on the query result, such as rapid recognition and understanding (no action required).

[0153] In another embodiment of the present application, Fig.15As shown, the information in the knowledge graph format includes multiple subject information and relationship information, and the relationship information represents the relationship between multiple subject information; the task processing rule data uses the information in the knowledge graph format as the query condition, performs information query in the current knowledge graph, and obtains the task decision result, including steps S401-S402.

[0154] S401, using at least one of the multiple subject information and / or relationship information as query conditions, perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the multiple subject information and / or other subject information associated with the relationship information.

[0155] S402: Obtain a task decision result based on at least one of other relationship information and other subject information.

[0156] Taking "I→personnel A→star player; the first "→" indicates an acquaintance relationship, and the second "→" indicates an identity relationship" as an example, the information in the knowledge graph format includes multiple subject information (I, person A, star player) and relationship information (acquaintance relationship, identity relationship). The relationship information indicates the relationship between multiple subject information. The subject can be a subject or an object, and the relationship information can be a predicate. When performing information query in the current knowledge graph using the information in the knowledge graph format as the query condition, at least one of the multiple subject information (such as at least one of me, person A, and star player) and / or the relationship information (acquaintance relationship, identity relationship) can be used as the query condition to perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the multiple subject information and / or other subject information associated with the relationship information. For example, the agent currently sees person A, and uses the subject information of "person A" to perform information query in the current knowledge graph, and obtains other relationship information associated with the subject information and / or other subject information associated with the relationship information. For example, the other relationship information found includes, for example, unit relationship (unit relationship indicates the unit to which person A belongs), and other associated subject information includes, for example, team B (team B indicates the team to which person A belongs). Based on at least one of the other relationship information and other subject information, a task decision result is obtained, and the task decision result includes, for example, a cognitive task decision result, that is, the relevant information of person A is quickly recognized and understood (without action).

[0157] In another embodiment of the present application, as shown in the figure, the task processing rule data uses information in the knowledge graph format as a query condition, performs information query in the current knowledge graph, and obtains the task decision result, including S411-S413.

[0158] S411, using the task processing rule data in the format of a knowledge graph as the first query condition, performing information query in the current knowledge graph to obtain a first query result.

[0159] S412, when the first query result is a preset query result, the task processing rule data uses the information in the knowledge graph format and at least one of the first query results as the second query condition to perform information query in the historical knowledge graph to obtain a second query result.

[0160] S413: Obtain a task decision result based on at least one of the first query result and the second query result.

[0161] As an example, the task processing rule data uses the information in the knowledge graph format as the first query condition, performs information query in the current knowledge graph, obtains the first query result, and determines whether the first query result meets the preset query result. The preset query result is, for example, a result that does not meet the query purpose (i.e., the required result is not found). If the first query result does not meet the preset query result, it means that the query purpose has been achieved based on the current knowledge graph, and the first query result is used as the final task decision result. If the first query result is the preset query result, it means that the query purpose has not been achieved, and it is necessary to continue to query the historical knowledge graph for information query. The information in the historical knowledge graph is usually more than that in the current knowledge graph, and the information in the historical knowledge graph is more comprehensive, and the query consumes more time and resources. In the case where the first query result is the preset query result, the task processing rule data uses the information in the knowledge graph format and at least one of the first query result as the second query condition, performs information query in the historical knowledge graph, obtains the second query result, and uses the second query result as the task decision result.

[0162] The method for constructing an intelligent agent in the embodiment of the present application is based on human brain theory and working mode, and the cognitive and decision-making parts in the intelligent agent are split into different modules according to functional characteristics, each performing its own duties, so as to improve the interpretability of the intelligent agent. Among them, the instinctive brain can quickly perform the conversion of perception, cognition and decision-making process based on knowledge and rules, and improve the response speed to general situations and emergencies. Intelligent agents (virtual people, virtual objects) based on the instinctive brain can handle general tasks with low power consumption. In addition, the knowledge graph is used to recognize oneself and the world, reducing cognitive complexity to better explain the intelligent agent.

[0163] This application also proposes a task processing method performed by an intelligent agent.

[0164] In one embodiment of the present application, Fig.17 As shown, the task processing method performed by the intelligent agent includes: the constructed intelligent agent performs the following operations S5-S8.

[0165] S5, based on the perception information perceived by the constructed intelligent agent, obtain the target information.

[0166] S6: Determine corresponding task processing rule data based on the first information type of the target information.

[0167] S7, the task processing rule data performs decision processing based on the target information to obtain a task decision result.

[0168] S8, performing task processing based on the task decision result.

[0169] Specifically, after the intelligent agent is constructed according to the intelligent agent construction method of the above embodiment, the intelligent agent can be used to implement task execution, which is a task learning process. The constructed intelligent agent continuously performs the above task processing operations to improve the intelligent agent execution ability. The target intelligent agent includes the intelligent agent obtained by performing some task processing in the previous stage (which already has task processing and learning capabilities), and then the target intelligent agent can continue to perform task processing. The process of the target intelligent agent performing task processing is similar to the process of constructing the target intelligent agent above, and will not be repeated here.

[0170] The present application also proposes a device for constructing an intelligent agent.

[0171] In one embodiment of the present application, Fig.18 As shown, the constructing device 800 of the intelligent body includes:

[0172] The first acquisition module 810 is used to obtain target information based on the perception information perceived by the initial intelligent agent. The first determination module 820 is used to determine the corresponding task processing rule data based on the first information type of the target information. The first decision module 830 is used to perform decision processing based on the target information by the task processing rule data to obtain a task decision result. The first processing module 840 is used to perform task processing based on the task decision result.

[0173] As an example, the first obtaining module 810 is used to: perform cognitive processing on the perception information to obtain cognitive information; and perform information parsing processing on at least one of the perception information and the cognitive information to obtain target information.

[0174] As an example, the initial intelligent agent includes an instinctive brain, and the cognitive information includes instinctive cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: determining a first weight corresponding to the instinctive brain based on a second information type of the perceptual information; and the instinctive brain cognitively processes the perceptual information based on the first weight to obtain instinctive cognitive information.

[0175] As an example, the initial intelligent agent includes a rational brain, and the cognitive information includes rational cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: determining a second weight corresponding to the rational brain based on a second information type of the perceptual information; and the rational brain cognitively processes the perceptual information based on the second weight to obtain rational cognitive information.

[0176] As an example, the initial intelligent agent includes an emotional brain, and the cognitive information includes emotional cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: determining a third weight corresponding to the emotional brain based on a second information type of the perceptual information; and the emotional brain cognitively processes the perceptual information based on the third weight to obtain emotional cognitive information.

[0177] As an example, the second information type includes at least one of the following: a physical signal type of the external environment, a physical state signal type of the initial agent.

[0178] As an example, the first determination module 820 is used to: determine the corresponding task type based on the first information type; and determine the task processing rule data corresponding to the task type.

[0179] As an example, the first information type includes at least one of the following: an environmental information type, a stimulus information type, a state information type of an initial agent, and an interest information type.

[0180] As an example, based on the first information type, determining the corresponding task type includes at least one of the following: when the first information type is an environmental information type, determining the corresponding task type is an attention task type; when the first information type is a stimulus information type, determining the corresponding task type is a reflexive task type; when the first information type is an intelligent agent's state information type, determining the corresponding task type is an unconscious task type; when the first information type is an interest information type, determining the corresponding task type is a memory task type.

[0181] As an example, the first decision module 830 is used to perform at least one of the following: using the task processing rule data corresponding to the attention task type, an attention decision is made on the target information to obtain a cognitive task decision result and an action task decision result; using the task processing rule data corresponding to the reflective task type, a trigger decision is made on the target information to obtain an action task decision result; using the task processing rule data corresponding to the unconscious task type, a subconscious decision is made on the target information to obtain at least one of a cognitive task decision result and an action task decision result; using the task processing rule data corresponding to the memory task type, a memory search decision is made on the target information to obtain at least one of a cognitive task decision result and an action task decision result.

[0182] As an example, the first processing module 840 is used to perform at least one of the following: when the task decision result includes multiple first actions to be executed, execute the multiple first actions to be executed based on the priority of the multiple first actions to be executed; when the task decision result includes multiple second actions to be executed, select at least one second action to be executed from the multiple second actions to be executed based on a preset selection rule, and execute at least one second action to be executed.

[0183] As an example, the multiple first actions to be executed include at least one of the following: a first sub-action to be executed of an emergency type obtained based on perception information, a second sub-action to be executed of a non-emergency type obtained based on perception information, and a third sub-action to be executed based on cognitive information, wherein the priority of the first sub-action to be executed is higher than the priority of the second sub-action to be executed and the priority of the third sub-action to be executed.

[0184] As an example, the intelligent body construction device 800 also includes: a first conversion module, used to convert the information format of at least one of the perception information and the cognitive information to obtain information in the knowledge graph format; and a first construction module, used to construct a knowledge graph based on the information in the knowledge graph format to obtain a constructed current knowledge graph.

[0185] As an example, the first decision module 830 is used to: use the task processing rule data in the format of the knowledge graph as a query condition, perform information query in the current knowledge graph, and obtain a task decision result.

[0186] As an example, information in the knowledge graph format includes multiple subject information and relationship information, and the relationship information represents the relationship between the multiple subject information; the task processing rule data uses the information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain a task decision result, including: using at least one of the multiple subject information and / or relationship information as a query condition to perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the multiple subject information and / or other subject information associated with the relationship information; and obtaining a task decision result based on the other relationship information and at least one of the other subject information.

[0187] As an example, the construction device of the intelligent agent also includes: a first association module, which is used to associate the current knowledge graph with the historical knowledge graph when the current knowledge graph is constructed.

[0188] As an example, the task processing rule data uses information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain a task decision result, including: the task processing rule data uses information in the knowledge graph format as a first query condition to perform information query in the current knowledge graph to obtain a first query result; when the first query result is a preset query result, the task processing rule data uses information in the knowledge graph format and at least one of the first query results as a second query condition to perform information query in the historical knowledge graph to obtain a second query result; and based on at least one of the first query result and the second query result, obtain the task decision result.

[0189] As an example, the initial intelligent agent includes a perception layer, a cognitive layer, a decision-making layer and an execution layer; the cognitive layer includes a reptilian brain, a rational brain and an emotional brain; the decision-making layer includes a fast system and a slow system; and the instinctive brain includes a reptilian brain and a fast system.

[0190] The present application also proposes a task processing device.

[0191] In one embodiment of the present application, Fig.19 As shown, the task processing device 900 includes: a second acquisition module 910, which is used to obtain target information based on the perception information obtained by the constructed intelligent agent. A second determination module 920 is used to determine the corresponding task processing rule data based on the first information type of the target information. A second decision module 930 is used to perform decision processing based on the target information by the task processing rule data to obtain a task decision result. A second processing module 940 is used to perform task processing based on the task decision result.

[0192] As an example, the second obtaining module 910 is used to: perform cognitive processing on the perception information to obtain cognitive information; and perform information parsing processing on at least one of the perception information and the cognitive information to obtain target information.

[0193] As an example, the constructed intelligent agent includes an instinctive brain, and the cognitive information includes instinctive cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: determining a first weight corresponding to the instinctive brain based on a second information type of the perceptual information; and the instinctive brain cognitively processes the perceptual information based on the first weight to obtain instinctive cognitive information.

[0194] As an example, the constructed intelligent agent includes a rational brain, and the cognitive information includes rational cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: determining a second weight corresponding to the rational brain based on a second information type of the perceptual information; and the rational brain cognitively processes the perceptual information based on the second weight to obtain rational cognitive information.

[0195] As an example, the constructed intelligent agent includes an emotional brain, and cognitive information includes emotional cognitive information; cognitive processing is performed on the perceptual information to obtain cognitive information, including: based on the second information type of the perceptual information, determining a third weight corresponding to the emotional brain; and the emotional brain cognitively processes the perceptual information based on the third weight to obtain emotional cognitive information.

[0196] As an example, the second information type includes at least one of the following: a physical signal type of the external environment, a physical state signal type of the constructed intelligent agent.

[0197] As an example, the second determination module 920 is used to: determine the corresponding task type based on the first information type; and determine the task processing rule data corresponding to the task type.

[0198] As an example, the first information type includes at least one of the following: an environmental information type, a stimulus information type, a state information type of a constructed intelligent agent, and an interest information type.

[0199] As an example, based on the first information type, determining the corresponding task type includes at least one of the following: when the first information type is an environmental information type, determining the corresponding task type is an attention task type; when the first information type is a stimulus information type, determining the corresponding task type is a reflexive task type; when the first information type is an intelligent agent's state information type, determining the corresponding task type is an unconscious task type; when the first information type is an interest information type, determining the corresponding task type is a memory task type.

[0200] As an example, the second decision module 930 is used to perform at least one of the following: using the task processing rule data corresponding to the attention task type, an attention decision is made on the target information to obtain a cognitive task decision result and an action task decision result; using the task processing rule data corresponding to the reflective task type, a trigger decision is made on the target information to obtain an action task decision result; using the task processing rule data corresponding to the unconscious task type, a subconscious decision is made on the target information to obtain at least one of a cognitive task decision result and an action task decision result; using the task processing rule data corresponding to the memory task type, a memory search decision is made on the target information to obtain at least one of a cognitive task decision result and an action task decision result.

[0201] As an example, the second processing module 940 is used to perform at least one of the following: when the task decision result includes multiple first actions to be executed, execute the multiple first actions to be executed based on the priority of the multiple first actions to be executed; when the task decision result includes multiple second actions to be executed, select at least one second action to be executed from the multiple second actions to be executed based on a preset selection rule, and execute at least one second action to be executed.

[0202] As an example, the multiple first actions to be executed include at least one of the following: a first sub-action to be executed of an emergency type obtained based on perception information, a second sub-action to be executed of a non-emergency type obtained based on perception information, and a third sub-action to be executed based on cognitive information, wherein the priority of the first sub-action to be executed is higher than the priority of the second sub-action to be executed and the priority of the third sub-action to be executed.

[0203] As an example, the task processing device 900 also includes: a second conversion module, used to convert the information format of at least one of the perception information and the cognitive information to obtain information in the knowledge graph format; and a second construction module, used to construct a knowledge graph based on the information in the knowledge graph format to obtain a constructed current knowledge graph.

[0204] As an example, the second decision module 930 is used to: use the task processing rule data in the format of the knowledge graph as a query condition, perform information query in the current knowledge graph, and obtain a task decision result.

[0205] As an example, information in the knowledge graph format includes multiple subject information and relationship information, and the relationship information represents the relationship between the multiple subject information; the task processing rule data uses the information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain a task decision result, including: using at least one of the multiple subject information and / or relationship information as a query condition to perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the multiple subject information and / or other subject information associated with the relationship information; and obtaining a task decision result based on the other relationship information and at least one of the other subject information.

[0206] As an example, the task processing device 900 also includes: a second association module, which is used to associate the current knowledge graph with the historical knowledge graph when the current knowledge graph is constructed.

[0207] As an example, the task processing rule data uses information in the knowledge graph format as a query condition to perform information query in the current knowledge graph to obtain a task decision result, including: the task processing rule data uses information in the knowledge graph format as a first query condition to perform information query in the current knowledge graph to obtain a first query result; when the first query result is a preset query result, the task processing rule data uses information in the knowledge graph format and at least one of the first query results as a second query condition to perform information query in the historical knowledge graph to obtain a second query result; and based on at least one of the first query result and the second query result, obtain the task decision result.

[0208] As an example, a constructed intelligent entity includes a perception layer, a cognitive layer, a decision-making layer, and an execution layer; the cognitive layer includes the reptilian brain, the rational brain, and the emotional brain; the decision-making layer includes the fast system and the slow system; and the instinctive brain includes the reptilian brain and the fast system.

[0209] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the intelligent agent construction method in any of the above-mentioned embodiments are implemented.

[0210] The embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements a method for constructing an intelligent agent and a method for processing tasks executed by the intelligent agent when executing the computer program. Of course, the method for constructing an intelligent agent and the method for processing tasks executed by the intelligent agent can be executed by different electronic devices.

[0211] like Fig. 20 As shown, for ease of understanding, the embodiment of the present application shows a specific electronic device 1100.

[0212] The electronic device 1100 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0213] like Fig. 20As shown, the device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the electronic device 1100 can also be stored. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0214] Multiple components in the electronic device 1100 are connected to the I / O interface 1105, including: an input unit 1106, such as a keyboard, a mouse, etc.; an output unit 1107, such as various types of displays, speakers, etc.; a storage unit 1108, such as a disk, an optical disk, etc.; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1109 allows the electronic device 1100 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0215] The computing unit 1101 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1101 executes the various methods described above, such as at least one of the methods for constructing an agent and the task processing methods performed by the agent. For example, in some embodiments, any one or more of the methods for constructing an agent and the task processing methods performed by the agent may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, any one or more of the methods for constructing an agent and the task processing methods performed by the agent described above may be executed. Alternatively, in other embodiments, the computing unit 1101 may be configured as any one or more of the method for constructing an intelligent agent and the method for processing a task performed by the intelligent agent in any other appropriate manner (eg, by means of firmware).

[0216] This application is based on human brain theory and working mode, and splits the cognitive and decision-making parts of the intelligent agent into different modules according to functional characteristics, each performing its own duties, and improving the interpretability of the intelligent agent. Among them, the instinctive brain can quickly perform the conversion from perception to cognition and then to decision-making based on knowledge and rules, and improve the response speed to general situations and emergencies. Intelligent agents (virtual people, virtual objects) based on the instinctive brain can handle general tasks with low power consumption. And use knowledge graphs to recognize oneself and the world, reduce cognitive complexity, and better explain the intelligent agent.

[0217] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this application, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing in a suitable manner if necessary, and then stored in a computer memory.

[0218] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0219] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0220] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0221] In addition, the terms "first", "second", etc. used in the embodiments of the present application are only used for descriptive purposes and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in the present embodiment. Therefore, the features defined by the terms "first", "second", etc. in the embodiments of the present application can explicitly or implicitly indicate that at least one of the features is included in the embodiment. In the description of the present application, the word "multiple" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0222] In this application, unless otherwise clearly specified or limited in the embodiments, the terms "installed", "connected", "connected" and "fixed" etc. appearing in the embodiments should be understood in a broad sense. For example, the connection can be a fixed connection, a detachable connection, or an integrated connection. It can be understood that it can also be a mechanical connection, an electrical connection, etc.; of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal connection of two elements, or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to the specific implementation situation.

[0223] In the present application, unless otherwise clearly specified and limited, a first feature being “above” or “below” a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being “above”, “above”, and “above” a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being “below”, “below”, and “below” a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0224] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in the field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for constructing an intelligent agent, characterized in that: The method for constructing an intelligent agent includes constructing a target intelligent agent by performing the following operations on an initial intelligent agent: Based on the perception information perceived by the initial agent, obtaining target information; Determining corresponding task processing rule data based on the first information type of the target information; The task processing rule data performs decision processing based on the target information to obtain a task decision result; as well as Task processing is performed based on the task decision result.

2. The method for constructing an intelligent agent according to claim 1, characterized in that: The step of obtaining target information based on the perception information perceived by the initial agent includes: performing cognitive processing on the perception information to obtain cognitive information; and Perform information analysis on at least one of the perception information and the cognitive information to obtain the target information.

3. The method for constructing an intelligent agent according to claim 2, characterized in that: The initial intelligent agent includes an instinctive brain, and the cognitive information includes instinctive cognitive information; the cognitive processing of the perception information to obtain cognitive information includes: Determining a first weight corresponding to the instinctive brain based on a second information type of the perception information; and The instinctive brain performs cognitive processing on the perception information based on the first weight to obtain the instinctive cognitive information.

4. The method for constructing an intelligent agent according to claim 2, characterized in that: The initial intelligent agent includes a rational brain, and the cognitive information includes rational cognitive information; the cognitive processing of the perception information to obtain cognitive information includes: Determining a second weight corresponding to the rational brain based on a second information type of the perception information; and The rational brain cognitively processes the perception information based on the second weight to obtain the rational cognitive information.

5. The method for constructing an intelligent agent according to claim 2, characterized in that: The initial intelligent agent includes an emotional brain, and the cognitive information includes emotional cognitive information; the cognitive processing of the perception information to obtain cognitive information includes: Determining a third weight corresponding to the emotional brain based on the second information type of the perception information; and The emotional brain performs cognitive processing on the perception information based on the third weight to obtain the emotional cognitive information.

6. The method for constructing an intelligent agent according to any one of claims 3 to 5, characterized in that: The second information type includes at least one of the following: The physical signal type of the external environment and the physical state signal type of the initial intelligent agent.

7. The method for constructing an intelligent agent according to any one of claims 1 to 5, characterized in that: The determining corresponding task processing rule data based on the first information type of the target information includes: Based on the first information type, determining a corresponding task type; and Task processing rule data corresponding to the task type is determined.

8. The method for constructing an intelligent agent according to claim 7, characterized in that: The first information type includes at least one of the following: Environmental information type, stimulus information type, state information type of the initial intelligent agent, and interest information type.

9. The method for constructing an intelligent agent according to claim 8, characterized in that: The determining of the corresponding task type based on the first information type includes at least one of the following: In a case where the first information type is the environmental information type, determining that the corresponding task type is an attention task type; In the case where the first information type is the stimulus information type, determining that the corresponding task type is a reflective task type; In the case where the first information type is the state information type of the agent, determining that the corresponding task type is an unconscious task type; In the case where the first information type is the information type of interest, it is determined that the corresponding task type is a memory task type.

10. The method for constructing an intelligent agent according to claim 9, characterized in that: The task processing rule data performs decision processing based on the target information to obtain a task decision result, including at least one of the following: Performing attention decision on the target information according to the task processing rule data corresponding to the attention task type, and obtaining cognitive task decision results and action task decision results; Using the task processing rule data corresponding to the reflective task type, a trigger decision is made on the target information to obtain an action task decision result; Performing a subconscious decision on the target information according to the task processing rule data corresponding to the unconscious task type, and obtaining at least one of a cognitive task decision result and an action task decision result; The task processing rule data corresponding to the memory task type is used to perform a memory search decision on the target information to obtain at least one of a cognitive task decision result and an action task decision result.

11. The method for constructing an intelligent agent according to any one of claims 2 to 5, characterized in that: The performing task processing based on the task decision result includes at least one of the following: In a case where the task decision result includes a plurality of first actions to be performed, executing the plurality of first actions to be performed based on the priorities of the plurality of first actions to be performed; In the case that the task decision result includes a plurality of second actions to be performed, at least one second action to be performed is selected from the plurality of second actions to be performed based on a preset selection rule, and the at least one second action to be performed is performed.

12. The method for constructing an intelligent agent according to claim 11, characterized in that: The plurality of first actions to be performed include at least one of the following: A first sub-action to be performed of an emergency type obtained based on the perception information, a second sub-action to be performed of a non-emergency type obtained based on the perception information, and a third sub-action to be performed based on the cognitive information. The priority of the first sub-action to be executed is higher than the priority of the second sub-action to be executed and the priority of the third sub-action to be executed.

13. The method for constructing an intelligent agent according to any one of claims 2 to 5, characterized in that: The method for constructing the intelligent agent also includes: Performing information format conversion on at least one of the perception information and the cognitive information to obtain information in a knowledge graph format; and Based on the information in the knowledge graph format, a knowledge graph is constructed to obtain a constructed current knowledge graph.

14. The method for constructing an intelligent agent according to claim 13, characterized in that: The task processing rule data performs decision processing based on the target information to obtain a task decision result, including: The task processing rule data uses the information in the knowledge graph format as a query condition, performs information query in the current knowledge graph, and obtains the task decision result.

15. The method for constructing an intelligent agent according to claim 14, characterized in that: The information in the knowledge graph format includes multiple subject information and relationship information, and the relationship information represents the relationship between the multiple subject information; the task processing rule data uses the information in the knowledge graph format as a query condition, performs information query in the current knowledge graph, and obtains the task decision result, including: Using at least one of the plurality of subject information and / or the relationship information as query conditions, perform information query in the current knowledge graph to obtain other relationship information associated with at least one of the plurality of subject information and / or other subject information associated with the relationship information; and The task decision result is obtained based on at least one of the other relationship information and the other subject information.

16. The method for constructing an intelligent agent according to claim 14 or 15, characterized in that: The method for constructing the intelligent agent also includes: When the current knowledge graph is constructed, the current knowledge graph is associated with the historical knowledge graph.

17. The method for constructing an intelligent agent according to claim 16, characterized in that: The task processing rule data uses the information in the knowledge graph format as a query condition, performs information query in the current knowledge graph, and obtains the task decision result, including: The task processing rule data uses the information in the knowledge graph format as a first query condition, performs information query in the current knowledge graph, and obtains a first query result; In the case where the first query result is a preset query result, the task processing rule data uses the information in the knowledge graph format and at least one of the first query result as a second query condition to perform information query in the historical knowledge graph to obtain a second query result; and The task decision result is obtained based on at least one of the first query result and the second query result.

18. The method for constructing an intelligent agent according to claim 3, characterized in that: The initial intelligent agent includes a perception layer, a cognitive layer, a decision-making layer and an execution layer; the cognitive layer includes a reptilian brain, a rational brain and an emotional brain; the decision-making layer includes a fast system and a slow system; the instinctive brain includes the reptilian brain and the fast system.

19. A task processing method performed by an agent, characterized in that: The task processing method includes the following steps: Based on the perception information perceived by the constructed intelligent agent, obtaining target information; Determining corresponding task processing rule data based on the first information type of the target information; The task processing rule data performs decision processing based on the target information to obtain a task decision result; as well as Task processing is performed based on the task decision result.

20. A device for constructing an intelligent agent, characterized in that: The constructing device of the intelligent agent comprises: A first acquisition module, used for acquiring target information based on the perception information perceived by the initial intelligent agent; A first determination module, configured to determine corresponding task processing rule data based on a first information type of the target information; A first decision module, configured to perform decision processing based on the target information using the task processing rule data to obtain a task decision result; and The first processing module is used to perform task processing based on the task decision result.

21. A task processing device, characterized in that: The task processing device comprises: A second acquisition module is used to obtain target information based on the perception information perceived by the initial intelligent agent; A second determination module, configured to determine corresponding task processing rule data based on the first information type of the target information; A second decision module is used to perform decision processing based on the target information using the task processing rule data to obtain a task decision result; and The second processing module is used to perform task processing based on the task decision result.

22. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 19 are implemented.

23. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 19 are implemented.

Images