Value belief simulation agent decision-making method and system

Abstract

According to the value belief simulation agent decision-making method and system, through a big data search engine tool, various human decision-making data are collected, invalid decisions and small correlation decisions are screened and removed, and various types of values in cognition during human decision-making are automatically learned from the various human decision-making data; based on the current belief and the multi-type value, multi-type value weights during social interaction are autonomously learned, a long-term intention and a short-term intention are generated, value weight state selection is carried out, and a corresponding value item intention is obtained; performing long-term intention series actions and short-term intention series actions on the long-term intention and the short-term intention respectively to form action sets; according to social interaction process data and corresponding value item intentions, multi-type value decision guidance corresponding to diversified human personalities is counted, intelligent action decisions corresponding to the human personalities are selected from a series of action sets to be selected, and a simulation agent action decision system is constructed to simulate a human decision process and continuously learn progressive optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of statistical analysis information processing of intelligent agent training data, and more specifically, to a value belief simulation intelligent agent decision-making system and method. Background Art

[0002] At present, the BDI framework model (Belief-Desire-Intention) is usually used to describe 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; and 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. The Schwartz Theory of Basic Values ​​summarizes ten basic values ​​that reflect the basic motivations or goals of individuals in life choices. The ten basic values ​​are as follows: (1) Power: the pursuit of social status, prestige, and control over others and resources, with the goal of achieving personal dominance. (2) Achievement: the pursuit of personal success, showing one's abilities and talents, and winning the respect and recognition of others. (3) Hedonism: the pursuit of happiness and enjoyment of life, and the satisfaction of sensory needs. (4) Stimulation: desire for a fresh, stimulating and varied life, avoiding boredom and repetition. (5) Self-direction: value independence and creativity, pursue freedom to express oneself and make choices. (6) Universalism: focus on equality, justice and social welfare, respect nature and humans, and pursue universal happiness of mankind. (7) Benevolence: value friendship, help and care for others, especially within close groups. (8) Tradition: respect cultural and religious traditions, maintain and defend traditional customs and beliefs. (9) Conformity: follow social norms, avoid causing trouble or harm to others, and obey rules and expectations. (10) Security: pursue security, harmony and stability, including security at the personal, family and social levels; methods for intelligent agents to achieve social interaction include computer vision: used to recognize and understand visual information; this is particularly important in social robots, helping them to identify users and environments; multimodal interaction: combining vision, hearing and language processing to achieve a more natural way of interaction. For example, intelligent assistants provide more accurate responses by combining voice and visual information; how to improve the decision-making ability of intelligent agents and the operating efficiency of decision-making models by combining different technologies so that intelligent agents can interact with humans more effectively and other issues remain to be resolved; therefore, it is necessary to propose a value belief simulation intelligent agent decision-making system and method to at least partially solve the problems existing in the existing technology. Summary of the invention

[0003] A series of simplified concepts are introduced in the content of the invention, which will be further described in detail in the specific implementation method section; the content of the invention of the present invention does not mean to attempt to limit the key features and essential technical features of the technical solution claimed for protection, nor does it mean to attempt to determine the scope of protection of the technical solution claimed for protection.

[0004] To at least partially solve the above problems, the present invention provides a value belief simulation intelligent agent decision-making method, comprising: S10, through big data search engine tools, collects a variety of human decision-making data and screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision-making data; S20, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; S30, respectively expanding the long-term intention series of actions and the short-term intention series of actions to form an action set based on the long-term intention and the short-term intention, and obtaining a series of actions to be selected; S40, based on the social interaction process data and the corresponding value items, count the various value decision orientations corresponding to the diverse human personalities, select the intelligent action decisions corresponding to the human personalities from the series of actions to be selected, build a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

[0005] Preferably, S10 includes: S101, collects a variety of human decision-making data through big data search engine tools; S102, preprocessing the human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data; S103, automatically learning multiple types of values ​​in human cognition when making decisions from multiple human decision data based on the pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0006] Preferably, S20 includes: S201, based on current beliefs and multiple types of values, autonomously learn the weights of multiple types of values ​​in social interaction, generate long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; S202, selecting the state with the highest value in the accessible space according to multiple time scales and multiple types of values; S203, at the current time scale, takes the state with the highest value as the corresponding value item intention; uses the value function of the agent's purpose from the present moment to the future time span, the agent's purpose and the agent's value, the value function includes parameters that represent the individual's personality leading to preferences for different types of actions, parameters for the value of different states in order to complete one's current intentions, and parameters for the value of different states in the social field, to calculate and obtain the agent's corresponding value item intention.

[0007] Preferably, S30 includes: S301, respectively expand the long-term intention and the short-term intention to obtain a series of actions of the long-term intention and a series of actions of the short-term intention; S302, forming an action set by combining the long-term intention series actions and the short-term intention series actions, and obtaining a series action set to be selected.

[0008] Preferably, S40 includes: S401, based on the social interaction process data and the corresponding value items, statistically analyze the various value decision orientations corresponding to the diverse human personalities; S402, based on the candidate series of action sets and multiple types of value decision orientations, and targeting diverse human personalities, select from the candidate series of action sets, select intelligent action decisions corresponding to human personalities, construct a simulated intelligent agent to simulate the human decision-making process, and continuously learn and gradually optimize; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

[0009] The present invention provides a value belief simulation intelligent agent decision-making system, comprising: The decision data automatic learning subsystem collects a variety of human decision data through big data search engine tools, screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision data; The value weight intention generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; The intention expansion action set subsystem expands the long-term intention and the short-term intention into a long-term intention series of actions and a short-term intention series of actions to form an action set, and obtains a series of actions to be selected; The personality-oriented value intelligent decision-making subsystem, based on the social interaction process data and the corresponding value item intentions, counts the various types of value decision orientations corresponding to diverse human personalities, selects intelligent action decisions corresponding to human personalities from the series of actions to be selected, and constructs a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

[0010] Preferably, the decision data automatic learning subsystem includes: The decision data collection subsystem collects various human decision data through big data search engine tools; The preprocessing screening and removal subsystem preprocesses the human decision data, screens and removes invalid decisions and decisions with small correlation, and obtains the preprocessed human decision data; The cognitive value automatic learning subsystem automatically learns various types of values ​​in human cognition when making decisions from a variety of human decision data based on pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0011] Preferably, the value weight intention generating subsystem includes: The value weight generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; The value state selection subsystem selects the state with the highest value in the accessible space based on multiple time scales and multiple types of values; The value item intention mapping subsystem takes the highest value state as the corresponding value item intention at the current time scale; uses the purpose of the agent from the present moment to the future, the purpose of the agent and the value function of the agent, the value function includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete one's current intentions, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the agent.

[0012] Preferably, the intended action set subsystem includes: The long-term and short-term intention expansion subsystem expands the long-term intention and the short-term intention respectively to obtain the long-term intention series of actions and the short-term intention series of actions; The series action aggregation subsystem forms an action set by combining the long-term intention series actions and the short-term intention series actions, and obtains a set of candidate series actions.

[0013] Preferably, the personality-oriented value intelligent decision-making subsystem includes: The personality-corresponding decision-making orientation subsystem counts various types of value decision orientations corresponding to diverse human personalities based on the social interaction process data and the corresponding value item intentions; The value-oriented intelligent action decision subsystem selects from the series of actions to be selected and various value decision orientations, targeting diverse human personalities, selects intelligent action decisions corresponding to human personalities, builds a simulated intelligent agent to simulate the human decision-making process, and continuously learns and optimizes gradually; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

[0014] Compared with the prior art, the present invention has at least the following beneficial effects: The present invention provides a value belief simulation intelligent agent decision-making method and system. Through a big data search engine tool, a variety of human decision-making data are collected and invalid decisions and decisions with small correlation are screened out, and a variety of values ​​in human cognition when making decisions are automatically learned from a variety of human decision-making data; based on current beliefs and a variety of values, a variety of value weights in social interaction are autonomously learned, long-term intentions and short-term intentions are generated, and value weight state selection is performed to obtain corresponding value item intentions; long-term intentions and short-term intentions are respectively expanded into a long-term intention series of actions and a short-term intention series of actions to form an action set, and a series of actions to be selected are obtained; according to the social interaction process data and the corresponding value item intentions, a variety of value decision orientations corresponding to diverse human personalities are statistically analyzed, and intelligent action decisions corresponding to human personalities are selected from the series of actions to be selected, and a simulated intelligent agent action decision system is constructed to simulate the human decision-making process and continuously learn progressive optimization; it can provide Improve the decision-making ability of intelligent agents and the computing efficiency of decision-making models; can more efficiently collect a variety of human decision-making data and screen out invalid decisions and decisions with low correlation, greatly improve the processing efficiency of complex human decision-making data, automatically learn a variety of values ​​in human decision-making cognition, and significantly improve the value learning speed; autonomous learning of multiple value weights improves the accuracy of value weight selection and social interaction intention generation; respectively expand long-term intention series actions and short-term intention series actions to form action sets, which significantly optimizes the time and space selection of the series of actions to be selected; the accuracy of statistical analysis of multiple value decision-making orientations corresponding to diverse human personalities is significantly improved, and the selection of series actions in intelligent action decisions corresponding to human personalities is more comprehensive, reducing the probability of missing selection; significantly improve the comprehensiveness of the simulated intelligent agent action decision-making system and the accuracy of the simulated human decision-making process, and continuous learning and gradual optimization continuously improve the action decision-making ability of intelligent agents, which is of great significance and significant effect.

[0015] The value belief simulation intelligent agent decision-making method and system described in the present invention, other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will also be understood by technical personnel in the field through research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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 1 This is a diagram of an embodiment of a value belief simulation intelligent agent decision-making system described in the present invention.

[0017] Figure 2 This is a diagram showing an implementation example of a value belief simulation intelligent agent decision-making method and system operation framework described in the present invention.

[0018] Figure 3This is another embodiment diagram of the value belief simulation intelligent agent decision-making method and system framework described in the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, so that those skilled in the art can implement it according to the instructions; as shown in the figure, the present invention provides a value belief simulation intelligent agent decision-making method, including: S10, through big data search engine tools, collects a variety of human decision-making data and screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision-making data; S20, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; S30, respectively expanding the long-term intention series of actions and the short-term intention series of actions to form an action set based on the long-term intention and the short-term intention, and obtaining a series of actions to be selected; S40, based on the social interaction process data and the corresponding value items, count the various value decision orientations corresponding to the diverse human personalities, select the intelligent action decisions corresponding to the human personalities from the series of actions to be selected, build a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

[0020] The principle and effect of the above technical solution are as follows: the present invention provides a value belief simulation intelligent agent decision-making method, including: through a big data search engine tool, collecting a variety of human decision-making data and screening out invalid decisions and decisions with small correlation, automatically learning a variety of values ​​in human cognition when making decisions from a variety of human decision-making data; based on current beliefs and a variety of values, autonomously learning a variety of value weights during social interaction, generating long-term intentions and short-term intentions, and performing value weight state selection to obtain corresponding value item intentions; respectively expanding a long-term intention series of actions and a short-term intention series of actions for long-term intentions and short-term intentions to form an action set, and obtaining a series of actions to be selected; according to the social interaction process data and the corresponding value item intentions, statistically analyzing the diverse human The multi-type value decision-making orientation corresponding to the personality selects the intelligent action decision corresponding to the human personality from the series of actions to be selected, constructs a simulated intelligent action decision system to simulate the human decision-making process and continuously learns and gradually optimizes; the intelligent agent decision system simulates the intelligent agent action decision system including: obtaining a variety of value set data in the value module, including personality data and decision mode data; a variety of value set data is combined with the belief network architecture data and input into the long-term purpose intention unit and short-term purpose intention unit of the target module to generate long-term intention and short-term intention. The relationship between long-term intention and short-term intention includes: coexistence, conflict or non-conflict; long-term intention and short-term intention are input into the intelligent activation module to execute environmental scenario decisions; and value weight state selection is performed ; Values ​​are modeled as a series of parameters, while beliefs are modeled as an AND / OR graph; beliefs include both the current beliefs about the world and the beliefs about the psychological states of other intelligent agents; values ​​and beliefs work together to generate the goals of the intelligent agent, which include both long-term and short-term goals; these goals are expanded layer by layer, with some sub-goals in the middle layer, and finally expanded into a series of actions, which are executed in the environment. After the actions are executed, the environment will generate new observation decisions, and the intelligent agent will update its beliefs based on the new observation decisions, and so on; it can improve the decision-making ability of intelligent agents and the computing efficiency of decision-making models; it can more efficiently collect a variety of human decision data and filter out invalid decisions and decisions with small correlation, which is very useful for complex human The efficiency of decision-making data processing has been greatly improved, and the multiple types of values ​​in human cognition during decision-making are automatically learned, and the value learning speed is significantly improved; the autonomous learning of multiple types of value weights improves the accuracy of value weight selection and social interaction intention generation; the long-term intention series of actions and short-term intention series of actions are separately formed to form action sets, which significantly optimizes the time and space selection of the selected series of actions; the accuracy of statistical analysis of multiple types of value decision-making orientations corresponding to diverse human personalities is significantly improved, and the selection of series of actions in intelligent action decisions corresponding to human personalities is more comprehensive, reducing the probability of missing selection; the comprehensiveness of the simulated intelligent agent action decision-making system and the accuracy of the simulated human decision-making process are significantly improved, and continuous learning and gradual optimization continuously improve the intelligent agent action decision-making ability, which is of great significance and significant effect.

[0021] In one embodiment, S10 includes: S101, collects a variety of human decision-making data through big data search engine tools; S102, preprocessing the human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data; S103, automatically learning multiple types of values ​​in human cognition when making decisions from multiple human decision data based on the pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0022] The principle and effect of the above technical solution are as follows: collect a variety of human decision data through big data search engine tools; pre-process the human decision data, filter out invalid decisions and decisions with low correlation, and obtain pre-processed human decision data; automatically learn a variety of values ​​in human cognition when making decisions from a variety of human decision data based on the pre-processed human decision data; the multiple values ​​include: action value, intention value, state value; pre-process the human decision data, filter out invalid decisions and decisions with low correlation, and obtain the pre-processed human decision data including: extract the current environmental information sequence of current belief cognition from the human decision data; according to The current belief cognition conducts statistical analysis on human intention plans, and statistics the target achievement probability of human intention plans under the current environmental information sequence, and summarizes them to form a total target achievement probability set; the minimum probability of the total target achievement probability set and the first target achievement probability set that is less than the set screening reference probability are screened out; the minimum probability of the total target achievement probability set corresponds to an invalid decision, and the first target achievement probability set that is less than the set screening reference probability corresponds to a small correlation decision; the human decision data corresponding to the second target achievement probability set of the total target achievement probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0023] In one embodiment, S20 includes: S201, based on current beliefs and multiple types of values, autonomously learn the weights of multiple types of values ​​in social interaction, generate long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; S202, selecting the state with the highest value in the accessible space according to multiple time scales and multiple types of values; S203, at the current time scale, takes the state with the highest value as the corresponding value item intention; uses the value function of the agent's purpose from the present moment to the future time span, the agent's purpose and the agent's value, the value function includes parameters that represent the individual's personality leading to preferences for different types of actions, parameters for the value of different states in order to complete one's current intentions, and parameters for the value of different states in the social field, to calculate and obtain the agent's corresponding value item intention.

[0024] The principle and effect of the above technical solution are as follows: based on current beliefs and multiple types of values, autonomously learn the weights of multiple types of values ​​during social interaction, generate long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; select the state with the highest value in the reachable space according to multiple time scales and multiple types of values; take the state with the highest value as the corresponding value item intention at the current time scale; use the value function of the purpose of the intelligent agent from the present moment to the future time span, the purpose of the intelligent agent and the value of the intelligent agent, the value function includes parameters indicating the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete their current intentions, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the intelligent agent; The corresponding value item generation formula is as follows: in, Representing an actor The corresponding value item plan; For the actor From now on The time span between the future time t1 All reachable states of ; For the actor the value of ->0, indicating the reaction purpose at the present moment; ->C, represents the planning purpose of time span C; ->∞, indicating the whole purpose; the value function includes parameters indicating the preference for different types of actions due to one’s personality, parameters indicating the value of different states in order to complete one’s current intention, and parameters indicating the value of different states in the social field; the reachable state includes the state that the intelligent actor can reach through a finite planning step reachable space starting from the starting state.

[0025] In one embodiment, S30 includes: S301, respectively expand the long-term intention and the short-term intention to obtain a series of actions of the long-term intention and a series of actions of the short-term intention; S302, forming an action set by combining the long-term intention series actions and the short-term intention series actions, and obtaining a series action set to be selected.

[0026] The principle and effect of the above technical solution are: expand the long-term intention and short-term intention respectively to obtain the long-term intention series of actions and the short-term intention series of actions; form the long-term intention series of actions and the short-term intention series of actions into an action set to obtain a series of actions to be selected.

[0027] In one embodiment, S40 includes: S401, based on the social interaction process data and the corresponding value items, statistically analyze the various value decision orientations corresponding to the diverse human personalities; S402, based on the candidate series of action sets and multiple types of value decision orientations, and targeting diverse human personalities, select from the candidate series of action sets, select intelligent action decisions corresponding to human personalities, construct a simulated intelligent agent to simulate the human decision-making process, and continuously learn and gradually optimize; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

[0028] The principle and effect of the above technical solution are as follows: according to the social interaction process data and the corresponding value item intentions, the various types of value decision orientations corresponding to the diverse human personalities are statistically analyzed; according to the series of actions to be selected and the various types of value decision orientations, for the diverse human personalities, the intelligent action decisions corresponding to the human personalities are selected from the series of actions to be selected, and the simulated intelligent agent is constructed to simulate the human decision-making process and continuously learn and gradually optimize; according to the series of actions to be selected and the various types of value decision orientations, for the diverse human personalities, the intelligent action decisions corresponding to the human personalities are selected from the series of actions to be selected, and the simulated intelligent agent is constructed to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of actions to be selected and the various types of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed, and the simulated intelligent agent decision-oriented value function includes parameters indicating the preference for different types of actions caused by the individual's personality, parameters indicating the value of different states in order to complete their current intentions, and parameters indicating the value of different states in the social field; describing the degree to which the intelligent agent will be affected by different values, and constructing a simulated intelligent agent to simulate human decision-making by calculating the simulated intelligent agent decision-oriented value. The decision-making process is continuously learned and gradually optimized; according to the social interaction process data and the corresponding value items, the various types of value decision-making orientations corresponding to the diverse human personalities are statistically analyzed, including: the first type is the value of preference for different types of actions caused by personal personality, including: whether they are lively and active or whether they like socializing; the second type is to complete their current intentions, corresponding to the current intention state value; the third type of value of different states in the social field includes: when the action decision system wants to help others, the state value that enables others to complete their intentions more efficiently is higher; according to the series of actions to be selected and the various types of value decision-making orientations, for the diverse human personalities, the series of actions to be selected are selected, and the intelligent action decisions corresponding to the human personality are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of actions to be selected and the various types of value decision-making orientations, the decision-oriented value function of the simulated intelligent agent is constructed, and the decision-oriented value function of the simulated intelligent agent includes parameters indicating the preference for different types of actions caused by personal personality, parameters indicating the value of different states in order to complete their current intentions, and parameters indicating the value of different states in the social field; the formula of the decision-oriented value function of the simulated intelligent agent is as follows: ; in, To simulate the decision-making value of intelligent agents, The value of a particular action, the value of one's own goals; the value to other people’s goals; , , are preset weight coefficients respectively; for example: when the intelligent agent is in a state of cognitive fatigue, its intention is more likely to be a resting action rather than a movement action, the value of the resting action is higher than the value of the movement action, and the intelligent decision is to rest; when the intelligent agent is in a state of cognitive fatigue and there is danger in the environment, its intention is more likely to be a movement action to avoid the danger rather than a resting action, and the movement action to avoid the danger has a greater impact on the value of the self-goal and the value of other people's goals; in response to diverse human personalities, select from the set of candidate series of actions, and select the intelligent action decision corresponding to the human personality; describe the degree to which the intelligent agent will be affected by different values, and simulate the decision-oriented value of the intelligent agent through calculation, construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and optimize gradually.

[0029] The present invention provides a value belief simulation intelligent agent decision-making system, comprising: The decision data automatic learning subsystem collects a variety of human decision data through big data search engine tools, screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision data; The value weight intention generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; The intention expansion action set subsystem expands the long-term intention and the short-term intention into a long-term intention series of actions and a short-term intention series of actions to form an action set, and obtains a series of actions to be selected; The personality-oriented value intelligent decision-making subsystem, based on the social interaction process data and the corresponding value item intentions, counts the various types of value decision orientations corresponding to diverse human personalities, selects intelligent action decisions corresponding to human personalities from the series of actions to be selected, and constructs a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

[0030] The principles and effects of the above technical solution are as follows: The present invention provides a value belief simulation intelligent agent decision-making system, including: a decision data automatic learning subsystem, which collects a variety of human decision data and screens out invalid decisions and decisions with low correlation through a big data search engine tool, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision data; a value weight intention generation subsystem, which autonomously learns a variety of value weights during social interaction based on current beliefs and a variety of values, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; an intention expansion action set subsystem, which expands long-term intentions and short-term intentions into a series of long-term intention actions and a series of short-term intention actions to form actions. The personality-oriented value intelligent decision-making subsystem, based on the social interaction process data and the corresponding value item intentions, counts the various types of value decision orientations corresponding to the diverse human personalities, selects the intelligent action decisions corresponding to the human personality from the series of actions to be selected, and constructs a simulated intelligent agent action decision system to simulate the human decision-making process and continuously learn and gradually optimize; the intelligent agent decision system simulates the intelligent agent action decision system, including: obtaining a variety of value set data in the value module, including personality data and decision mode data; a variety of value set data is combined with the belief network architecture data and input into the long-term purpose intention unit and the short-term purpose intention unit of the target module to generate long-term intentions and short-term intentions, and the long-term intentions and short-term intentions are combined. Intentional relationships include: coexistence, conflict or non-conflict; long-term intentions and short-term intentions are input into the intelligent activation module for environmental scenario decision execution; and value weight state selection is performed; values ​​are modeled as a series of parameters, and beliefs are modeled as an AND / OR graph; beliefs include both the current belief about the world and the belief about the psychological state of other intelligent agents; values ​​and beliefs work together to generate the goals of the intelligent agent, including both long-term goals and short-term goals; these goals are expanded layer by layer, with some sub-goals in the middle layer, and finally expanded into a series of actions, which are executed in the environment. After the actions are executed, the environment will generate new observation decisions, and the intelligent agent will update its beliefs based on the new observation decisions, and so on; It can improve the decision-making ability of intelligent agents and the computing efficiency of decision-making models; it can more efficiently collect a variety of human decision-making data and filter out invalid decisions and decisions with low correlation, greatly improve the processing efficiency of complex human decision-making data, automatically learn a variety of values ​​in human cognition when making decisions, and significantly improve the value learning speed; autonomous learning of multiple value weights improves the accuracy of value weight selection and social interaction intention generation; separately expand long-term intention series actions and short-term intention series actions to form action sets, and significantly optimize the time and space selection of the series of actions to be selected; the accuracy of statistical analysis of multiple value decision-making orientations corresponding to diverse human personalities is significantly improved, and the selection of series actions in intelligent action decisions corresponding to human personalities is more comprehensive, reducing the probability of missed selection;Significantly improve the comprehensiveness of the simulated intelligent action decision-making system and the accuracy of the simulated human decision-making process. Continuous learning and gradual optimization will continuously improve the intelligent action decision-making ability, which has great significance and significant effects. ;

[0031] In one embodiment, the decision data automatic learning subsystem includes: The decision data collection subsystem collects various human decision data through big data search engine tools; The preprocessing screening and removal subsystem preprocesses the human decision data, screens and removes invalid decisions and decisions with small correlation, and obtains the preprocessed human decision data; The cognitive value automatic learning subsystem automatically learns various types of values ​​in human cognition when making decisions from a variety of human decision data based on pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0032] The principle and effect of the above technical solution are as follows: the decision data automatic learning subsystem includes: the decision data collection subsystem, which collects a variety of human decision data through big data search engine tools; the preprocessing screening and removal subsystem, which preprocesses the human decision data, screens out invalid decisions and decisions with low correlation, and obtains the preprocessed human decision data; the cognitive value automatic learning subsystem, based on the preprocessed human decision data, automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision data; the multiple values ​​include: action value, intention value, state value; preprocessing the human decision data, screening out invalid decisions and decisions with low correlation, and obtaining the preprocessed human decision data includes: converting human decision From the data, the current environmental information sequence of the current belief cognition is extracted; according to the current belief cognition, statistical analysis of human intention planning is performed, and the target achievement probability of the human intention plan based on the current environmental information sequence is statistically analyzed to form a total target achievement probability set; the minimum probability of the total target achievement probability set and the first target achievement probability set that is less than the set screening reference probability are screened and removed; the minimum probability of the total target achievement probability set corresponds to an invalid decision, and the first target achievement probability set that is less than the set screening reference probability corresponds to a small correlation decision, and the human decision data corresponding to the second target achievement probability set of the total target achievement probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

[0033] In one embodiment, the value weight intention generation subsystem includes: The value weight generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; The value state selection subsystem selects the state with the highest value in the accessible space based on multiple time scales and multiple types of values; The value item intention mapping subsystem takes the highest value state as the corresponding value item intention at the current time scale; uses the purpose of the agent from the present moment to the future, the purpose of the agent and the value function of the agent, the value function includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete one's current intentions, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the agent.

[0034] The principle and effect of the above technical solution are as follows: a value weight intention generation subsystem, including: a value weight generation subsystem, based on current beliefs and multiple types of values, autonomously learning multiple types of value weights during social interaction, generating long-term intentions and short-term intentions, the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; a value state selection subsystem, selecting the highest value state in the reachable space according to multiple time scales and multiple types of values; a value item intention mapping subsystem, taking the highest value state as the corresponding value item intention at the current time scale; using the purpose of the intelligent agent from the present moment to the future time span, the purpose of the intelligent agent and the value function of the intelligent agent, the value function includes parameters indicating the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete their current intentions, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the intelligent agent; The corresponding value item generation formula is as follows: in, Representing an actor The corresponding value item plan; For the actor From now on The time span between the future time t1 All reachable states of ; For the actor the value of ->0, indicating the reaction purpose at the present moment; ->C, represents the planning purpose of time span C; ->∞, indicating the whole purpose; the value function includes parameters indicating the preference for different types of actions due to one’s personality, parameters indicating the value of different states in order to complete one’s current intention, and parameters indicating the value of different states in the social field; the reachable state includes the state that the intelligent actor can reach through a finite planning step reachable space starting from the starting state.

[0035] In one embodiment, the intention to expand the action set subsystem includes: The long-term and short-term intention expansion subsystem expands the long-term intention and the short-term intention respectively to obtain the long-term intention series of actions and the short-term intention series of actions; The series action aggregation subsystem forms an action set by combining the long-term intention series actions and the short-term intention series actions, and obtains a set of candidate series actions.

[0036] The principle and effect of the above technical solution are: the intention expansion action set subsystem, including: the long-term and short-term intention expansion subsystem, which expands the long-term intention and the short-term intention respectively to obtain the long-term intention series of actions and the short-term intention series of actions; the series action collection subsystem, which forms an action set with the long-term intention series of actions and the short-term intention series of actions to obtain a series of actions to be selected.

[0037] In one embodiment, the personality-oriented value intelligent decision-making subsystem includes: The personality-corresponding decision-making orientation subsystem counts various types of value decision orientations corresponding to diverse human personalities based on the social interaction process data and the corresponding value item intentions; The value-oriented intelligent action decision subsystem selects from the series of actions to be selected and various value decision orientations, targeting diverse human personalities, selects intelligent action decisions corresponding to human personalities, builds a simulated intelligent agent to simulate the human decision-making process, and continuously learns and optimizes gradually; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

[0038] The principle and effect of the above technical solution are as follows: the personality-oriented value intelligent decision-making subsystem includes: the personality-corresponding decision-making subsystem, which counts various types of value decision-making orientations corresponding to diverse human personalities according to the social interaction process data and the corresponding value items; the value-oriented intelligent action decision-making subsystem, which selects from the series of actions to be selected according to the series of actions to be selected and various types of value decision-making orientations, targeting diverse human personalities, selects intelligent action decisions corresponding to human personalities, constructs a simulated intelligent body to simulate the human decision-making process and continuously learns and optimizes gradually; according to the series of actions to be selected and various types of value decision-making orientations, targeting diverse human personalities, Personality, select from the series of actions to be selected, select the intelligent action decision corresponding to the human personality, build a simulated intelligent agent to simulate the human decision-making process and continue to learn and optimize gradually, including: according to the series of actions to be selected and a variety of value decision-making orientations, build a simulated intelligent agent decision-oriented value function, the simulated intelligent agent decision-oriented 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 complete their current intentions, and parameters that represent the value of different states in the social field; describe the degree to which the intelligent agent will be affected by different values, and simulate through calculation The value of intelligent agent decision-making orientation is to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize; according to the social interaction process data and the corresponding value item intention, the various types of value decision orientations corresponding to the diverse human personality are statistically analyzed, including: the first type is the value of preference for different types of actions caused by personal personality, including: whether they are lively and active or whether they like socializing; the second type is to complete their current intentions, corresponding to the current intention state value; the third type of value of different states in the social field includes: when the action decision system wants to help others, the state that enables others to complete their intentions more efficiently has a higher value; according to the series of actions to be selected and the various types of value decision orientations, for the diverse human personalities, select from the series of actions to be selected, select the intelligent action decision corresponding to the human personality, construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of actions to be selected and the various types of value decision orientations, construct a simulated intelligent agent decision-making orientation value function, the simulated intelligent agent decision-making orientation value function includes parameters indicating the preference for different types of actions caused by personal personality, parameters indicating the value of different states in order to complete their current intentions, and parameters indicating the value of different states in the social field; the formula of the simulated intelligent agent decision-making orientation value function is as follows: ; in, To simulate the decision-making value of intelligent agents, The value of a particular action, the value of one's own goals; the value to other people’s goals; , , are preset weight coefficients respectively; for example: when the intelligent agent is in a state of cognitive fatigue, its intention is more likely to be a resting action rather than a movement action, the value of the resting action is higher than the value of the movement action, and the intelligent decision is to rest; when the intelligent agent is in a state of cognitive fatigue and there is danger in the environment, its intention is more likely to be a movement action to avoid the danger rather than a resting action, and the movement action to avoid the danger has a greater impact on the value of the self-goal and the value of other people's goals; in response to diverse human personalities, select from the set of candidate series of actions, and select the intelligent action decision corresponding to the human personality; describe the degree to which the intelligent agent will be affected by different values, and simulate the decision-oriented value of the intelligent agent through calculation, construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and optimize gradually.

[0039] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A value belief simulation agent decision-making method, characterized in that: include: S10, through big data search engine tools, collects a variety of human decision-making data and screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision-making data; S20, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; S30, respectively expanding the long-term intention series of actions and the short-term intention series of actions to form an action set based on the long-term intention and the short-term intention, and obtaining a series of actions to be selected; S40, based on the social interaction process data and the corresponding value items, count the various value decision orientations corresponding to the diverse human personalities, select the intelligent action decisions corresponding to the human personalities from the series of actions to be selected, build a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

2. A value belief simulation agent decision-making method according to claim 1, characterized in that: S10 includes: S101, collects a variety of human decision-making data through big data search engine tools; S102, preprocessing the human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data; S103, automatically learning multiple types of values ​​in human cognition when making decisions from multiple human decision data based on the pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

3. A value belief simulation agent decision-making method according to claim 1, characterized in that: The S20 includes: S201, based on current beliefs and multiple types of values, autonomously learn the weights of multiple types of values ​​in social interaction, generate long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; S202, selecting the state with the highest value in the accessible space according to multiple time scales and multiple types of values; S203, at the current time scale, takes the state with the highest value as the corresponding value item intention; uses the purpose of the agent from the present moment to the future time span, the purpose of the agent and the value function of the agent, the value function includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete one's current intention, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the agent.

4. A value belief simulation agent decision-making method according to claim 1, characterized in that: S30 includes: S301, respectively expand the long-term intention and the short-term intention to obtain a series of actions of the long-term intention and a series of actions of the short-term intention; S302, forming an action set by combining the long-term intention series actions and the short-term intention series actions, and obtaining a series action set to be selected.

5. A value belief simulation agent decision-making method according to claim 1, characterized in that: S40 includes: S401, based on the social interaction process data and the corresponding value items, statistically analyze the various value decision orientations corresponding to the diverse human personalities; S402, based on the candidate series of action sets and multiple types of value decision orientations, and targeting diverse human personalities, select from the candidate series of action sets, select intelligent action decisions corresponding to human personalities, construct a simulated intelligent agent to simulate the human decision-making process, and continuously learn and gradually optimize; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

6. A value belief simulation intelligent agent decision-making system, characterized in that: include: The decision data automatic learning subsystem collects a variety of human decision data through big data search engine tools, screens out invalid decisions and decisions with low correlation, and automatically learns a variety of values ​​in human cognition when making decisions from a variety of human decision data; The value weight intention generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and selects value weight states to obtain corresponding value item intentions; The intention expansion action set subsystem expands the long-term intention and the short-term intention into a long-term intention series of actions and a short-term intention series of actions to form an action set, and obtains a series of actions to be selected; The personality-oriented value intelligent decision-making subsystem, based on the social interaction process data and the corresponding value item intentions, counts the various types of value decision orientations corresponding to diverse human personalities, selects intelligent action decisions corresponding to human personalities from the series of actions to be selected, and constructs a simulated intelligent action decision system to simulate the human decision-making process and continuously learn and gradually optimize.

7. A value belief simulation agent decision system according to claim 6, characterized in that: Decision data automatic learning subsystem, including: The decision data collection subsystem collects various human decision data through big data search engine tools; The preprocessing screening and removal subsystem preprocesses the human decision data, screens and removes invalid decisions and decisions with small correlation, and obtains the preprocessed human decision data; The cognitive value automatic learning subsystem automatically learns various types of values ​​in human cognition when making decisions from a variety of human decision data based on pre-processed human decision data; The multiple types of values ​​include: action value, intention value, and state value; preprocessing human decision data, screening out invalid decisions and decisions with small correlation, and obtaining preprocessed human decision data includes: extracting the current environmental information sequence of the current belief cognition from the human decision data; performing statistical analysis of human intention plans based on the current belief cognition, and statistically analyzing the target realization probability of human intention plans under the current environmental information sequence, and summarizing them to form a total target realization probability set; screening out the minimum probability of the total target realization probability set and the first target realization probability set that is less than the set screening reference probability; the minimum probability of the total target realization probability set corresponds to an invalid decision, and the first target realization probability set that is less than the set screening reference probability corresponds to a decision with small correlation, and the human decision data corresponding to the second target realization probability set of the total target realization probability set that is not less than the set screening reference probability is selected as the effective preprocessed human decision data, and the preprocessed human decision data is obtained.

8. A value belief simulation agent decision system according to claim 6, characterized in that: The value weight intention generation subsystem includes: The value weight generation subsystem, based on current beliefs and multiple types of values, autonomously learns multiple types of value weights during social interaction, generates long-term intentions and short-term intentions, and the relationship between long-term intentions and short-term intentions includes: coexistence, conflict or non-conflict; The value state selection subsystem selects the state with the highest value in the accessible space based on multiple time scales and multiple types of values; The value item intention mapping subsystem takes the highest value state as the corresponding value item intention at the current time scale; it uses the purpose of the agent from the present moment to the future, the purpose of the agent and the value function of the agent, the value function includes parameters that represent the preference for different types of actions caused by the individual's personality, parameters for the value of different states in order to complete one's current intentions, and parameters for the value of different states in the social field, to calculate and obtain the corresponding value item intention of the agent.

9. A value belief simulation agent decision system according to claim 6, characterized in that: The intention expansion action set subsystem includes: The long-term and short-term intention expansion subsystem expands the long-term intention and the short-term intention respectively to obtain the long-term intention series of actions and the short-term intention series of actions; The series action aggregation subsystem forms an action set by combining the long-term intention series actions and the short-term intention series actions, and obtains a set of candidate series actions.

10. A value belief simulation agent decision system according to claim 6, characterized in that: Personality-oriented value intelligent decision-making subsystem, including: The personality-corresponding decision-making orientation subsystem counts various types of value decision orientations corresponding to diverse human personalities based on the social interaction process data and the corresponding value item intentions; The value-oriented intelligent action decision subsystem selects from the series of actions to be selected and various value decision orientations, targeting diverse human personalities, selects intelligent action decisions corresponding to human personalities, builds a simulated intelligent agent to simulate the human decision-making process, and continuously learns and optimizes gradually; According to the series of action sets to be selected and a variety of value decision orientations, in view of diverse human personalities, selection is made from the series of action sets to be selected, and intelligent action decisions corresponding to human personalities are selected to construct a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize, including: according to the series of action sets to be selected and a variety of value decision orientations, a simulated intelligent agent decision-oriented value function is constructed. The simulated intelligent agent decision-oriented 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 complete one's current intentions, and parameters that represent the value of different states in the social field; it describes the degree to which the intelligent agent will be affected by different values, and constructs a simulated intelligent agent to simulate the human decision-making process and continuously learn and gradually optimize by calculating the decision-oriented value of the simulated intelligent agent.

Images