Social interaction cognition method and system based on value function

Abstract

The invention provides a social interaction cognition method and system based on a value function. The method comprises the following steps: constructing a cognition framework based on the value function; and inputting the current scene into the cognitive framework to obtain an interaction decision. The method for inputting the current scene into the cognitive framework to obtain the interactive decision comprises the following steps of: inputting the current scene into a perception module to generate perception data; the belief module stores the sensing data and converts the sensing data into belief at the current moment; the reverse deduction module deduces beliefs, intentions and values of other people according to the belief at the current moment; the value module, the belief module, the inference module and the plug-in module output the current intention of the intelligent agent through the intention module; the planning module combines the current intention with the state of the current scene and plans an action; and the action module executes actions to obtain interaction decisions. According to the method, the value of human to various states involved in the social interaction process can be learned, and decisions are made based on the value, so that human-like social interaction is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a social interaction cognition method and system based on value function. Background Art

[0002] At present, the methods for artificial intelligence systems to achieve social interaction mainly include the following aspects: Natural Language Processing (NLP): Enables AI to understand and generate human language through speech recognition and generation techniques. For example, chatbots use NLP to understand user input and generate appropriate responses.

[0003] Affective computing: AI systems use technologies such as facial expression recognition and speech emotion analysis to identify and respond to the user's emotional state.

[0004] Computer Vision: Used to recognize and understand visual information. This is particularly important in social robots, helping them recognize users and the environment.

[0005] Multimodal interaction: Combining vision, hearing, and language processing to enable more natural interactions. For example, smart assistants provide more accurate responses by combining voice and visual information.

[0006] These methods combine different technologies to enable AI to interact with humans more effectively. Interactive robots built based on the above methods either detect and identify human social interaction cues and respond accordingly through pre-defined rule sets, or perform dynamic modeling of one or several social interactions in simple interaction scenarios to simulate social interaction scenarios. None of them achieve human-like interaction. When interacting with people, the people interacting will feel obviously stiff. How to achieve more human-like interaction with AI is a technology urgently needed in fields such as intelligent chatbots. Summary of the invention

[0007] One of the purposes of the present invention is to provide a social interaction cognition method and system based on value function, which can learn the value of various states involved in the social interaction process by humans, make decisions based on this, and realize human-like social interaction.

[0008] An embodiment of the present invention provides a social interaction recognition method based on a value function, comprising: Construct a cognitive framework based on value functions; Input the current situation into the cognitive framework to obtain interactive decisions.

[0009] Preferably, the cognitive framework includes: a perception module, a belief module, a reverse inference module, a value module, an intention module, a planning module, an action module and a plug-in module.

[0010] Preferably, the current scenario is input into the cognitive framework to obtain an interactive decision, including: Input the current scenario into the perception module to generate perception data; The belief module stores the perception data and converts it into the belief of the current moment; The reverse inference module infers other people’s beliefs, intentions, and values ​​based on current beliefs; The value module, together with the belief module, inference module, and plug-in module, outputs the current intention of the agent through the intention module; The planning module combines the current intention with the state of the current situation to plan the action; The action module executes actions to obtain interactive decisions.

[0011] Preferably, the plug-in module includes: a cause-effect library and a common sense library.

[0012] Preferably, the value module, belief module, inference module, and plug-in module together output the current intention of the agent through the intention module, including: Construct a belief graph based on the current belief of the belief module and the beliefs of others in the inference module; Parse the belief graph to obtain multiple goals; Apply the value model to analyze multiple goals and determine current intent.

[0013] Preferably, the application value module analyzes multiple goals to determine the current intent, including: Conduct value analysis on each goal and calculate the value of each goal based on the value function; The highest value among all goals is selected as the current intention. The value decision function is as follows: In the formula, Representing an actor intention; For the actor The present moment Time span to the future All reachable states of For the actor value.

[0014] The value function includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters that represent the value of different states in order to achieve one's current intention, and parameters that represent the value of different states in the social field; The present invention also provides a social interaction cognitive system based on value function, including: a building module and a decision module, wherein the building module builds a cognitive framework based on value function; the decision module inputs the current situation into the cognitive framework to obtain an interactive decision.

[0015] Preferably, the cognitive framework includes: a perception module, a belief module, a reverse inference module, a value module, an intention module, a planning module, an action module and a plug-in module.

[0016] Preferably, the decision module performs the following operations: Input the current scenario into the perception module to generate perception data; The belief module stores the perception data and converts it into the belief of the current moment; The reverse inference module infers other people’s beliefs, intentions, and values ​​based on current beliefs; The value module, together with the belief module, inference module, and plug-in module, outputs the current intention of the agent through the intention module; The planning module combines the current intention with the state of the current situation to plan the action; The action module executes actions to obtain interactive decisions.

[0017] The present invention has the following advantages: Based on the traditional belief-desire-intention framework, the original desire is expanded into a broader value, expanding the scope of application of the cognitive framework.

[0018] From the collected human data, we learn the value of actions, intentions, states, and other objects in human cognition when making decisions.

[0019] By combining the data of social interactions between people with previously learned values, we learn the weights that different human personalities attach to various values ​​when they interact socially, and thus make comprehensive decisions.

[0020] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0021] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1A schematic diagram of a social interaction cognition method based on a value function in an embodiment of the present invention; Figure 2 Schematic diagram of a social interaction cognitive system based on value function in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0024] The embodiment of the present invention provides a social interaction cognition method based on value function, such as Figure 1 As shown, including: Step 1: Construct a cognitive framework based on value function; The cognitive framework includes: perception module, belief module, reverse inference module, value module, intention module, planning module, action module and plug-in module; Among them, the value module mainly includes the value function, which includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters that represent the value of different states in order to complete one's current intention, and parameters that represent the value of different states in the social field; the definition formula of the value function is as follows: ; In the formula, The value of a specific action, e.g. when the agent is tired, it will not walk as much; the value of one's own goals; the value to other people’s goals; , , They are preset coefficients to describe the agent's emphasis on different values, which will be affected by the values; Step 2: Input the current situation into the cognitive framework to obtain interactive decision.

[0025] Among them, the current scenario is input into the cognitive framework to obtain interactive decisions, including: inputting the current scenario into the perception module to generate perception data; the belief module stores the perception data and converts it into the belief at the current moment; the reverse inference module infers the beliefs, intentions and values ​​of other people based on the beliefs at the current moment; the value module, together with the belief module, the inference module and the plug-in module, outputs the current intention of the intelligent agent through the intention module; the planning module combines the current intention with the state of the current scenario to plan the action; the action module executes the action to obtain interactive decisions. Among them, the plug-in module includes: causal library and common sense library. That is, the intelligent agent first needs to perceive the current situation through the perception module to obtain some observable data, which is called perception data. The specific perception means can be implemented through a pre-trained perception model; for example: the perception data can be the existence of person A in the environment, the object B held by task A, the direction of person A's hand movement, the direction of person A's footsteps, etc.; the belief module generates the belief at the current moment based on the perception data, that is, it infers the facts existing in the environment, and infers the belief through a pre-configured neural network model for inferring beliefs, so as to infer the beliefs about various objects; for example: the belief of moving forward with the footsteps, the belief of raising the hand, the belief in object B, etc.; inference The modules include: belief inference submodule, intention inference submodule and value inference submodule; these submodules all use different pre-trained converged neural network models to infer other people's intentions, beliefs and values; the decision of the intelligent agent is a comprehensive decision-making process, that is, the value module, belief module, inference module and plug-in module together output the current intention of the intelligent agent through the intention module, including: constructing a belief graph based on the belief module's current beliefs and other people's beliefs in the inference module; parsing the belief graph to obtain multiple goals (for example: the goal of following footsteps to move forward, the purpose of raising the hand, the purpose of moving object B, etc.); applying the value module to analyze multiple goals and determine the current intention. Among them, applying the value module to analyze multiple goals and determine the current intention includes: performing value analysis on each goal and calculating the value of each goal based on the value function; The highest value among all goals is selected as the current intention. The value decision function is as follows: ; In the formula, Representing an actor intention; For the actor The present moment Time span to the future All reachable states of For the actor When Approaching zero is a reflective intention; approaching the preset time c is a planned intention; approaching , for pure desire; Among them, value is modeled as a series of parameters, and belief is modeled as an AND / OR graph. In belief, it includes both the belief about the world at the current moment and the belief about the psychological state of other intelligent agents. Value and belief will work together to generate the goals of the intelligent agent, including both long-term goals and short-term goals. These goals will be unfolded layer by layer, with some sub-goals in the middle layer, and finally unfolded into a series of actions. The actions are executed in the environment. After the actions are executed, the environment will generate new observations. The intelligent agent updates the belief based on the new observations, and so on. In general, beliefs can be processed by a pre-trained neural network model to generate multiple goals. Each goal finds the corresponding parameters in the value, and then participates in the screening of the value decision function to obtain the current intention. After the current intention is obtained, it is planned and output through the planning module and the action module; the planning module and the action module are both pre-configured neural network models that have been trained and converged, so as to guide the formation of human-like output.

[0026] Generally speaking, the present invention is a further optimization based on the traditional BDI framework to achieve more human-like interactive decision-making of intelligent agents; among them, the BDI framework (Belief-Desire-Intention) specifically refers to: describing the behavior and decision-making process of intelligent agents. It is based on three core concepts: beliefs, desires, and intentions. Beliefs represent the intelligent agent's cognition of the world; desires are the goals that the intelligent agent hopes to achieve; intentions are the specific goals and plans that the intelligent agent chooses to achieve. Through the interaction of these three components, the BDI framework helps design intelligent agents that can simulate the human decision-making process.

[0027] In one embodiment, a social interaction cognition method based on a value function further includes: obtaining a preset number of intentions before a current intention and feedback of a current scenario after the output of the corresponding intention as correction analysis data; Comprehensively analyze and correct the analysis data and correct the value function; When making decisions, the revised value function is used.

[0028] Among them, comprehensive analysis and correction of analysis data are performed to correct the value function, including: Sort the intents from the last to the last time, and match the feedback to the intents. The feedback refers to the changes in the current scene after the intent is output, i.e., the changes in the data perceived by the perception module. The intention and feedback are reminded in sequence, and standard quantification is performed to form a row of data parameters in the analysis data set; each data parameter represents a different intention and a different feedback; the standard quantification rules are configured and stored in advance; Match the analysis data set with the retrieved data set corresponding to each correction coefficient set in the pre-configured correction library; extract the corresponding correction coefficient set; the correction coefficient set includes the correction coefficients corresponding to each parameter in the value function; the correction library is pre-configured, and in the correction library, the correction coefficient set is associated with the retrieved data set one by one.

[0029] This embodiment comprehensively analyzes previous interactions and makes decision corrections to ensure the accuracy of the output of the current intention, thereby providing a more intelligent interactive experience.

[0030] For the cognitive framework of the present invention, in order to enable it to achieve more human-like interactive cognition of the intelligent body, the construction of the value function when constructing the cognitive framework is particularly important; how to achieve the construction of an accurate value function depends on whether the required behavioral analysis data can be collected comprehensively and accurately; in one embodiment, the data used to construct the cognitive framework based on the value function has the following collection methods: the first is to obtain the required data by screening the existing data on the big data platform, for example: screening and classifying the data of various social interactions according to pre-configured screening and classification rules; each classification of the screening and classification corresponds to each pre-configured scenario, and in the scenario to which the value function to be constructed is applicable, extract the data corresponding to the classification corresponding to the scenario; when screening and classifying the existing data on the big data platform, the specific steps are as follows: segment each data to obtain multiple keywords; construct a keyword set based on multiple keywords; compare the keyword set with each keyword in the pre-configured classification library The standard word set corresponding to the classification is matched; the classification corresponding to the matched standard word set is used as the classification of the data; the second is to collect social interaction data by publishing scenario data; the data on the big data platform is existing and the acquisition speed is fast; while the data obtained by publishing scenario data is slow. The present invention combines the two collection methods, and the specific steps are as follows: extract the data corresponding to the classification of the big data platform, and determine the statistical parameters of the data; when the statistical parameters do not meet the preset conditions; publish the scenario data and obtain the collected data; update the statistical parameters based on the collected data, and end the collection when the preset conditions are met; wherein the statistical parameters include: the total number of data, the number of data of each different interactive decision data, the number of interactive decision data, etc.; the preset conditions include: the data threshold corresponding to the total number of data, the data threshold corresponding to each different interactive decision data, the number threshold corresponding to the number of interactive decision data, etc.; In addition, when collecting data, the validity of the data needs to be verified. The specific verification method can be to analyze the user who provides the data, so as to indirectly verify the data. The specific analysis process is as follows: obtain the user's authority information, and determine the first score value based on the authority information and a pre-configured first score table; obtain the user's provided data corresponding to the historical collection in the collection module, calculate the similarity between the historical collection provided data and the currently provided data, and determine the second score value based on the similarity and a pre-configured second score table; construct an analysis vector based on the time interval of the user's provided data for a preset number of times before the collection module, and select the pre-configured third score library according to the analysis vector. The corresponding third score value is retrieved from the database; when the sum of the first score value, the second score value and the third score value is greater than the preset score threshold, the verification is passed; wherein, in the second score table, the second score value corresponds to the similarity one by one, and the specific correspondence can be configured with an intermediate value (for example, a similarity of 0.8 corresponds to a score of 100), and starting from the similarity of the intermediate value, as the similarity increases or decreases, the second score value gradually decreases; the third score value in the third score library is associated with the analysis vector one by one, and the third score library is configured in advance; when configuring, the following rules are followed: the shorter the average interval time, the lower the third score value; for the same average time interval, the smaller the item in the interval time, the lower the third score value; The user's permission information is determined dynamically based on the initial configuration value and then the user's behavior is analyzed to ensure the accuracy of the determination; the permission value corresponding to the specific permission information is determined by counting the number of times the data provided by the user is judged to be accurate and the number of times it is judged to be inaccurate; when the number of times it is judged to be inaccurate is greater than the preset threshold, the permission value is set to zero; otherwise, it is the sum of the initial configuration value, the number of times it is judged to be accurate multiplied by the value of the first adjustment value, and the number of times it is judged to be accurate multiplied by the value of the second adjustment value; wherein the first adjustment value is positive, the second adjustment value is negative, and the absolute value of the first adjustment value is less than the absolute value of the second adjustment value. Through the difference between the first adjustment value and the second adjustment value, it is possible to adapt to the punitive reduction of the permission value for malicious provision; to avoid malicious provision of data next time.

[0031] In actual use, different value functions are called according to the current scenario to make targeted decisions, which improves the level of humanization; and different scenario configurations correspond to different value functions of different personalities to adapt to the interactive experience of different personalities.

[0032] In order to realize the automatic learning of intelligent agents, the social interaction cognition method based on value function also includes: Obtain feedback data after the action is output; analyze the feedback data, and adjust the value parameters of the decision corresponding to the action module based on the analysis data; The analysis is performed based on the standard feedback library configured for the action. The feedback data is matched with each standard feedback data in the standard feedback library, and the adjustment strategy corresponding to the matching standard feedback data is extracted. The value parameter is adjusted based on the adjustment strategy. In order to avoid over-adjustment, an adjustable interval of the value parameter is configured, and the adjustable interval has an upper limit and a lower limit. When the adjusted value is greater than the upper limit, the upper limit is used as the adjusted value. When the adjusted value parameter is less than the lower limit, the lower limit is used as the adjusted value. By configuring the standard feedback library in advance, the automatic optimization of the value function in advance is facilitated for subsequent interactions. Since decision-making through decision functions uses the largest value as the guide for the final action, there is causal stability and it cannot reflect the subjectivity of human interaction. Therefore, in one embodiment, the social interaction cognition method based on the value function also includes: making random decisions by configuring a random module; the random module is a triggering module, and is only triggered when the value used to make the decision is calculated, and there is a difference between the value corresponding to at least one target and the maximum value within a preset random trigger threshold. The random pool is configured according to the random module to perform random triggering. The configuration of the random pool: according to the difference, the corresponding pre-configured matching table is retrieved, and the two target placement numbers specified in the matching table are placed in the random pool; then the target is randomly extracted from the random pool.

[0033] The present invention also provides a social interaction cognitive system based on value function, such as Figure 2 As shown, it includes: a construction module 1 and a decision module 2, wherein the construction module 1 constructs a cognitive framework based on the value function; the decision module 2 inputs the current scenario into the cognitive framework to obtain an interactive decision.

[0034] Among them, the cognitive framework includes: perception module, belief module, reverse inference module, value module, intention module, planning module, action module and plug-in module.

[0035] The decision module 2 performs the following operations: Input the current scenario into the perception module to generate perception data; The belief module stores the perception data and converts it into the belief of the current moment; The reverse inference module infers other people’s beliefs, intentions, and values ​​based on current beliefs; The value module, together with the belief module, inference module, and plug-in module, outputs the current intention of the agent through the intention module; The planning module combines the current intention with the state of the current situation to plan the action; The action module executes actions to obtain interactive decisions.

[0036] Plug-in modules include: causal library and common sense library.

[0037] Among them, the value module, belief module, inference module, and plug-in module together output the current intention of the agent through the intention module, including: Construct a belief graph based on the current belief of the belief module and the beliefs of others in the inference module; Parse the belief graph to obtain multiple goals; Apply the value model to analyze multiple goals and determine current intent.

[0038] The application value module analyzes multiple goals to determine the current intent, including: Conduct value analysis on each goal and calculate the value of each goal based on the value function; The highest value among all goals is selected as the current intention. The value decision function is as follows: ; In the formula, Representing an actor intention; For the actor The present moment Time span to the future All reachable states of For the actor value.

[0039] Among them, the value function includes parameters that represent the preference for different types of actions caused by an individual's personality, parameters that represent the value of different states in order to achieve one's current intention, and parameters that represent the value of different states in the social field.

[0040] In one embodiment, a social interaction cognition system based on value function further includes: a correction module; The correction module performs the following operations: Obtaining a preset number of intentions before the current intention and feedback of the current scenario after the corresponding intention is output as correction analysis data; Comprehensively analyze and correct the analysis data and correct the value function; When making decisions, the revised value function is used.

[0041] Among them, comprehensive analysis and correction of analysis data are performed to correct the value function, including: Sort the intents from the last to the last time, and match the feedback to the intents. The feedback refers to the changes in the current scene after the intent is output, i.e., the changes in the data perceived by the perception module. The intention and feedback are reminded in sequence, and standard quantification is performed to form a row of data parameters in the analysis data set; each data parameter represents a different intention and a different feedback; the standard quantification rules are configured and stored in advance; Match the analysis data set with the retrieved data set corresponding to each correction coefficient set in the pre-configured correction library; extract the corresponding correction coefficient set; the correction coefficient set includes the correction coefficients corresponding to each parameter in the value function; the correction library is pre-configured, and in the correction library, the correction coefficient set is associated with the retrieved data set one by one.

[0042] This embodiment comprehensively analyzes previous interactions and makes decision corrections to ensure the accuracy of the output of the current intention, thereby providing a more intelligent interactive experience.

[0043] The system of the present invention integrates two collection methods to configure a data acquisition module, and the specific steps of the data acquisition module to acquire data are as follows: extracting data corresponding to the classification of the big data platform and determining the statistical parameters of the data; when the statistical parameters do not meet the preset conditions; publishing the scenario data and acquiring the collected data; updating the statistical parameters based on the collected data, and ending the collection when the preset conditions are met; wherein the statistical parameters include: one or more combinations of the total number of data, the number of data of each different interactive decision data, the number of interactive decision data, etc.; the preset conditions include: one or more combinations of the data threshold corresponding to the total number of data, the data threshold corresponding to each different interactive decision data, the number threshold corresponding to the number of interactive decision data, etc.; In addition, when collecting data, the validity of the data needs to be verified. The system configuration verification module of the present invention can be verified by analyzing the user who provides the data, so as to indirectly verify the data. The specific analysis process is as follows: obtaining the user's authority information, and determining a first score value based on the authority information and a pre-configured first score table; obtaining the user's provided data corresponding to the historical collection in the collection module, calculating the similarity between the historical collection provided data and the currently provided data, and determining a second score value based on the similarity and a pre-configured second score table; constructing an analysis vector based on the time interval of the user's provided data for a preset number of times before the collection module, and calculating the second score value based on the analysis vector from the pre-configured second score table. The corresponding third score value is retrieved from the configured third score library; when the sum of the first score value, the second score value and the third score value is greater than the preset score threshold, the verification is passed; wherein, the second score value in the second score table corresponds to the similarity one by one, and the specific correspondence can be configured with an intermediate value (for example, a similarity of 0.8 corresponds to a score of 100), and starting from the similarity of the intermediate value, as the similarity increases or decreases, the second score value gradually decreases; the third score value in the third score library is associated with the analysis vector one by one, and the third score library is configured in advance; when configuring, the following rules are followed: the shorter the average interval time, the lower the third score value; for the same average time interval, the smaller the item in the interval time, the lower the third score value; The user's permission information is determined dynamically based on the initial configuration value and then the user's behavior is analyzed to ensure the accuracy of the determination; the permission value corresponding to the specific permission information is determined by counting the number of times the data provided by the user is judged to be accurate and the number of times it is judged to be inaccurate; when the number of times it is judged to be inaccurate is greater than the preset threshold, the permission value is set to zero; otherwise, it is the sum of the initial configuration value, the number of times it is judged to be accurate multiplied by the value of the first adjustment value, and the number of times it is judged to be accurate multiplied by the value of the second adjustment value; wherein the first adjustment value is positive, the second adjustment value is negative, and the absolute value of the first adjustment value is less than the absolute value of the second adjustment value. Through the difference between the first adjustment value and the second adjustment value, it is possible to adapt to the punitive reduction of the permission value for malicious provision; to avoid malicious provision of data next time.

[0044] In order to realize the automatic learning of the intelligent agent, the social interaction cognitive system based on the value function further includes: a feedback module; the feedback module performs the following operations: Obtain feedback data after the action is output; analyze the feedback data, and adjust the value parameters of the decision corresponding to the action module based on the analysis data; The analysis is performed based on the standard feedback library configured for the action. The feedback data is matched with each standard feedback data in the standard feedback library, and the adjustment strategy corresponding to the matching standard feedback data is extracted. The value parameter is adjusted based on the adjustment strategy. In order to avoid over-adjustment, an adjustable interval of the value parameter is configured, and the adjustable interval has an upper limit and a lower limit. When the adjusted value is greater than the upper limit, the upper limit is used as the adjusted value. When the adjusted value parameter is less than the lower limit, the lower limit is used as the adjusted value. By configuring the standard feedback library in advance, the automatic optimization of the value function in advance is facilitated for subsequent interactions. Since the decision function is used to make a decision, the maximum value is used as the guide for the final action, which has causal stability and cannot reflect the subjectivity of human interaction. Therefore, in one embodiment, the social interaction cognition system based on the value function further includes: a random decision module, which performs the following operations: Make random decisions by configuring the random module; the random module is a trigger module, which is triggered only when the value used to make the decision is calculated and the difference between the value corresponding to at least one target and the maximum value is within the preset random trigger threshold; configure the random pool according to the random module and perform random triggering. The configuration of the random pool: call the corresponding pre-configured ratio table according to the difference, and place the two target placement numbers specified in the ratio table into the random pool; then randomly extract the target from the random pool.

[0045] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A social interaction cognition method based on value function, characterized in that: include: Construct a cognitive framework based on value functions; Input the current situation into the cognitive framework to obtain interactive decisions.

2. The social interaction cognition method based on value function as claimed in claim 1, characterized in that: The cognitive framework includes: perception module, belief module, reverse inference module, value module, intention module, planning module, action module and plug-in module.

3. The social interaction cognition method based on value function as claimed in claim 2, characterized in that: Input the current situation into the cognitive framework to obtain interactive decisions, including: Input the current scenario into the perception module to generate perception data; The belief module stores the perception data and converts it into the belief of the current moment; The reverse inference module infers other people’s beliefs, intentions, and values ​​based on current beliefs; The value module, together with the belief module, inference module, and plug-in module, outputs the current intention of the agent through the intention module; The planning module combines the current intention with the state of the current situation to plan the action; The action module executes actions to obtain interactive decisions.

4. The social interaction cognition method based on value function as claimed in claim 3, characterized in that: Plug-in modules include: causal library and common sense library.

5. The social interaction cognition method based on value function as claimed in claim 3, characterized in that: The value module, belief module, inference module, and plug-in module together output the current intention of the agent through the intention module, including: Construct a belief graph based on the current belief of the belief module and the beliefs of others in the inference module; Parse the belief graph to obtain multiple goals; Apply the value model to analyze multiple goals and determine current intent.

6. The social interaction cognition method based on value function as claimed in claim 5, characterized in that: Apply the value model to multiple goal analysis to determine current intent, including: Conduct value analysis on each goal and calculate the value of each goal based on the value function; Select the highest value among all goals as the current intention.

7. The social interaction cognition method based on value function as claimed in claim 1, characterized in that: The value function includes parameters that represent the preference for different types of actions caused by an individual's personality, parameters that represent the value of different states in order to achieve one's current intention, and parameters that represent the value of different states in the social field.

8. A social interaction cognitive system based on value function, characterized in that: include: The building module and the decision module, wherein the building module constructs a cognitive framework based on the value function; the decision module inputs the current scenario into the cognitive framework to obtain an interactive decision.

9. The social interaction cognitive system based on value function as claimed in claim 8, characterized in that: The cognitive framework includes: perception module, belief module, reverse inference module, value module, intention module, planning module, action module and plug-in module.

10. The social interaction cognitive system based on value function as claimed in claim 9, characterized in that: The decision module performs the following operations: Input the current scenario into the perception module to generate perception data; The belief module stores the perception data and converts it into the belief of the current moment; The reverse inference module infers other people’s beliefs, intentions, and values ​​based on current beliefs; The value module, together with the belief module, inference module, and plug-in module, outputs the current intention of the agent through the intention module; The planning module combines the current intention with the state of the current situation to plan the action; The action module executes actions to obtain interactive decisions.

Images