An ethical behavior demonstration device and system based on a humanoid robot

Abstract

The present application relates to the technical field of robot control, and especially relates to a moral behavior demonstration device and system based on a humanoid robot, which comprises a robot main body, a robot head, a camera, a robot arm, a robot leg, a robot mouth, a receiving microphone, a moral behavior demonstration device and system based on a humanoid robot, and the like.The system generates moral behavior samples in accordance with the cognitive level of children in real time by using a moral generation adversarial network through a decision center module, judges the personality type of children according to child data through an execution network module, and implements an adaptive moral behavior demonstration scheme for different children according to the personality type of children, so that the effect of moral behavior demonstration for the individual differences of different children is achieved, and the data properties of behavior decision and behavior decision parameters are output through an evolution cloud brain module, the privacy data of different children are stored, and the privacy of personal data is ensured.

Description

Technical Field

[0001] This invention relates to the field of robotic arm control technology, and in particular to a moral behavior demonstration device and system based on a humanoid robot. Background Technology

[0002] Childhood is a critical period for the formation of moral cognition and behavioral habits. Psychological research shows that children's moral development at this stage follows a process from external discipline to autonomy, requiring the internalization of moral principles through concrete behavioral demonstrations, interactive experiences, and emotional resonance. Traditional moral education for children mainly relies on families, schools, and media. However, these educational methods have significant limitations: in the family environment, parents' educational levels vary, and some parents lack scientific methods for moral guidance; school education is limited by teaching resources and class size, making it difficult to achieve personalized moral education; and media content tends to be entertainment-oriented, making it difficult to systematically and deeply convey moral norms.

[0003] Chinese Patent Publication No. CN114239847A discloses an artificial intelligence method and robot based on human morality, including: a step of acquiring behavioral and moral data; a supervised training step of the artificial intelligence moral judgment model; a step of acquiring candidate behaviors; and a step of judging the ethical behavior of artificial intelligence. The above method, system, and robot use artificial intelligence to judge whether its own behavior conforms to human morality, and then select behaviors that conform to human morality to execute, thereby avoiding the ethical risks of artificial intelligence. However, the moral behavior demonstration robot of this scheme only uses the universal judgment of "general human morality," without adapting to the cognitive patterns of children as a special group, and cannot demonstrate moral behavior according to the individual differences of different children. Summary of the Invention

[0004] To address this, the present invention provides a moral behavior demonstration device and system based on a humanoid robot, which overcomes the problem of low adaptability of moral behavior demonstration robots in the prior art, which only make universal judgments based on "universal human morality" without adapting to the cognitive patterns of children as a special group, and cannot demonstrate moral behavior according to the individual differences of different children.

[0005] To achieve the above objectives, in one aspect, the present invention provides a moral behavior demonstration device based on a humanoid robot, comprising:

[0006] The robot body is connected to the robot head, robotic arms, robotic legs, and a system of humanoid robot-based moral behavior demonstration devices.

[0007] The robot's head is connected to the robot's body, a camera, and a system for demonstrating ethical behavior based on a humanoid robot.

[0008] A camera, which is connected to the robot's head and a system of humanoid robot-based demonstration devices for ethical behavior;

[0009] A robotic arm connected to a system of robot bodies and humanoid robot-based demonstration devices for ethical behavior;

[0010] The robotic legs are connected to the robot body and the system of a humanoid robot-based demonstration device for ethical behavior;

[0011] A robotic mouth, which is connected to a robot head and a system of humanoid robot-based moral behavior demonstration devices;

[0012] A microphone, which is connected to the main body of the robot;

[0013] A system based on a humanoid robot for demonstrating ethical behavior, which is connected to the robot's head, camera, robotic arm, robotic legs, robot body, and robotic mouth.

[0014] On the other hand, the present invention also provides a system for a moral behavior demonstration device based on a humanoid robot, comprising:

[0015] The sensory center module is used to acquire data about children;

[0016] The decision-making central module is used to acquire behavioral decisions and decision parameters based on children's data using the moral generative adversarial network, and then send the behavioral decisions and decision parameters to the execution network module to adjust the behavioral decision parameters based on the children's data;

[0017] The execution network module is used to acquire execution plans based on behavioral decisions and decision parameters, control the humanoid robot according to the execution plan, calibrate decision parameters based on children's data, and revise the calibration process of decision parameters.

[0018] The Evolutionary Cloud Brain module is used to judge the deviation of emotional fluctuations based on children's data, obtain the judgment result, update the execution reference library based on the judgment result, and also correct the adjustment of behavioral decisions based on the judgment result.

[0019] Furthermore, the decision center module inputs the children's behavior and speech data from the children's data into the moral generative adversarial network to obtain the behavioral decisions and decision parameters output by the moral generative adversarial network. The decision center module then sends the behavioral decisions and decision parameters to the execution network module. The decision parameters include language intensity weight α and physical contact weight β, with α+β=1.

[0020] Furthermore, the moral behavior demonstration device and system based on humanoid robots is characterized in that, when the decision-making central module adjusts the behavioral decision parameters according to the children's data, it inputs the geographical coordinates from the children's data into the cultural analysis model, outputs the physical contact resistance γ from the cultural analysis model, and adjusts the physical contact weight β according to the physical contact resistance γ to obtain the adjusted physical contact weight β1. β1 is set to (1-γ)×β, and the language intensity weight α is adjusted according to the adjusted physical contact weight β1 to obtain the adjusted language intensity weight α1. α1 is set to 1-β1, and the adjusted language intensity weight α1 and the adjusted physical contact weight β1 are used as the adjusted decision parameters, and the decision parameters are replaced with the adjusted decision parameters.

[0021] Furthermore, when the execution network module obtains the execution plan based on the behavioral decisions and decision parameters, it inputs the behavioral decisions and decision parameters into the execution comparison library, compares the behavioral decisions and decision parameters with the indexes of the behavioral decisions and decision parameters, and outputs the execution plan based on the comparison results, wherein:

[0022] When the behavior decision and decision parameters match the behavior decision and decision parameter index, the execution plan corresponding to the behavior decision and decision parameter index will be output.

[0023] When the behavioral decision and decision parameters are inconsistent with the behavioral decision and decision parameter index, the behavioral decision and decision parameters are uploaded to the Evolutionary Cloud Brain module, which then outputs the execution plan based on the behavioral decision and decision parameters.

[0024] Furthermore, the execution network module controls the humanoid robot according to the execution plan.

[0025] Furthermore, when the execution network module calibrates the decision parameters based on the child data, it inputs the child's visual data and audio data after execution into the emotion perception model to obtain the child's emotional fluctuation P output by the emotion perception model. The module then compares the child's emotional fluctuation P with preset minimum emotional fluctuation Pmin and preset maximum child emotional fluctuation Pmax. Based on the comparison results, it judges the child's emotional fluctuation status and calibrates the decision parameters accordingly.

[0026] When P≤Pmin, the execution network module determines that the child's emotional fluctuation is low and does not calibrate the decision parameters;

[0027] When Pmin < P ≤ Pmax, the execution network module determines the child's emotional fluctuation as moderate, calibrates the decision parameters, and calibrates the physical contact weight β using the first fluctuation coefficient b1, setting b1 = 0.85 + 0.1 × e -(P-Pmin)Let e ​​be the base of the natural logarithm. We obtain the first proofread limb contact weight β2 and set β2 = b1 × β. Based on the first proofread limb contact weight β2, we proofread the language intensity weight α to obtain the first proofread language intensity weight α2. We set α2 = 1 - β2. We replace the limb contact weight β and language intensity weight α with the first proofread limb contact weight β2 and the first proofread language intensity weight α2, and output the execution plan based on the first proofread limb contact weight β2 and the first proofread language intensity weight α2.

[0028] When P > Pmax, the execution network module determines that the child's emotional fluctuation is high, adjusts the decision parameters, and adjusts the physical contact weight β using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e -(P-Pmax) Let e ​​be the base of the natural logarithm. The second proofread limb contact weight β3 is obtained. Let β3 = b2 × β. The language intensity weight α is proofread according to the second proofread limb contact weight β3 to obtain the second proofread language intensity weight α3. Let α3 = 1 - β3. Replace the limb contact weight β and the language intensity weight α with the second proofread limb contact weight β3 and the second proofread language intensity weight α3. Output the execution plan according to the second proofread limb contact weight β3 and the second proofread language intensity weight α3.

[0029] Furthermore, when revising the decision parameter calibration process, the execution network module calculates the emotional response time Δt1 based on the decision execution time T0 and the first emotional change time T1, setting Δt1 = T1 - T0. It then compares the emotional response time Δt1 with the preset emotional response time Δt10, determines the child's characteristic type based on the comparison result, and revises the decision parameter calibration process based on the determination result. Specifically:

[0030] When △t1≤△t10, the execution network module determines that the child's characteristic type is rapid response type, and calculates the peak emotion change time △t2 based on the first emotion change time T1 and the peak emotion change time T2, setting △t2=T2-T1, and comparing the peak emotion change time △t2 with the preset peak emotion change time △t20, wherein:

[0031] If △t2≤△t20, the execution network module determines that the child's characteristic type is the rapid resonance type in the rapid response type, and does not revise the decision parameter calibration process;

[0032] If △t2>△t20, the execution network module determines that the child's characteristic type is the rational observation type in the rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α through the first fluctuation coefficient b1, setting b1=0.85+0.1×e". -(P-Pmin) Let e ​​be the base of the natural logarithm. The third-corrected language intensity weight α4 is obtained, and α4 is set to b1 × α. The physical contact weight β is corrected based on the third-corrected language intensity weight α4, resulting in the third-corrected physical contact weight β4, which is set to 1 - α4. The physical contact weight β and language intensity weight α are replaced with the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4, and the execution plan is output based on the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "Correcting the language intensity weight α using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e." -(P-Pmax) e is the base of the natural logarithm. The fourth proofreading language intensity weight α5 is obtained. α5 is set to b2×α. The limb contact weight β is proofread according to the fourth proofreading language intensity weight α5 to obtain the fourth proofreading limb contact weight β5. β5 is set to 1-α5. The limb contact weight β and language intensity weight α are replaced with the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5. The execution plan is output according to the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5.

[0033] When △t1 > △t10, the execution network module determines that the child's characteristic type is non-rapid response, and calculates the total emotional change duration △t3 based on the emotional return to normalcy time T3 and the decision execution time T0, setting △t3 = T3 - T0, and comparing the total emotional change duration △t3 with the preset total emotional change duration △t30, wherein:

[0034] If △t3≤△t30, the execution network module determines that the child's characteristic type is low-participation in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the calibration of the decision parameters is canceled, and the physical contact weight β is revised by the first revision coefficient dz, which is set to dz=1.76-0.21×e -(P-Pmax)The first revised physical contact weight β8 is obtained, β8 = β × dz. The language intensity weight α is then revised based on this first revised weight β8, resulting in the first revised language intensity weight α8. α8 is set to 1 - β8. The physical contact weight β and language intensity weight α are replaced with the first revised weights β8 and α8, respectively. The execution plan is then output based on these weights. When the child's emotional fluctuation is high, the calibration of the decision parameters is canceled. The physical contact weight β is revised using the second revision coefficient dq, set to dq = 1.96 - 0.21 × e. -(P-Pmax) The second revised physical contact weight β9 is obtained, β9 = β × dq. The language intensity weight α is corrected according to the second revised physical contact weight β9 to obtain the second revised language intensity weight α9. α9 = 1 - β9 is set. The physical contact weight β and the language intensity weight α are replaced with the second revised physical contact weight β9 and the second revised language intensity weight α9. The execution plan is output according to the second revised physical contact weight β9 and the second revised language intensity weight α9.

[0035] If △t3 > △t30, the execution network module determines that the child's characteristic type is the delayed response type in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α through the third fluctuation coefficient b3, setting b3 = 1.24 - 0.20 × e -(P-Pmin) Let e ​​be the base of the natural logarithm. The fifth-corrected language intensity weight α6 is obtained, and α6 is set to b3 × α. The physical contact weight β is corrected based on the fifth-corrected language intensity weight α6, resulting in the fifth-corrected physical contact weight β6, which is set to 1 - α6. The physical contact weight β and language intensity weight α are replaced with the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6, and the execution plan is output based on the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "Correcting the language intensity weight α using the fourth fluctuation coefficient b4, setting b4 = 1.56 - 0.33 × e." -(P-Pmax)Let e ​​be the base of the natural logarithm. We obtain the language intensity weight α7 after the sixth proofreading. We set α7 = b4 × α. Based on the language intensity weight α7 after the sixth proofreading, we calibrate the limb contact weight β to obtain the limb contact weight β7 after the sixth proofreading. We set β7 = 1 - α7. We replace the limb contact weight β and language intensity weight α with the limb contact weight β7 and language intensity weight α7 after the sixth proofreading. We then output the execution plan based on the limb contact weight β7 and language intensity weight α7 after the sixth proofreading.

[0036] Furthermore, when the evolutionary cloud brain module outputs the data attributes of behavioral decisions and behavioral decision parameters, it calculates the emotional fluctuation deviation ΔP based on the child's emotional fluctuation P and the average emotional fluctuation Py of children in the cloud database, setting ΔP=(P-Py) / Py. It then compares the emotional fluctuation deviation ΔP with a preset emotional fluctuation deviation ΔP0, judges the emotional fluctuation deviation based on the comparison result, and outputs the data attributes of behavioral decisions and behavioral decision parameters based on the judgment result. It also corrects the adjustment of behavioral decisions based on the judgment result, wherein:

[0037] When △P≤△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is small, and outputs non-private data as the data attributes of behavioral decision and behavioral decision parameters, without correcting the adjustment of behavioral decision.

[0038] When △P>△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is large, and outputs private data as the data attribute of behavioral decision and the data attribute of behavioral decision parameters to correct the adjustment of behavioral decision. The correction method is to cancel the adjustment of behavioral decision parameters by the decision center module.

[0039] Furthermore, when the evolutionary cloud brain module determines that the emotional fluctuation deviation is appropriate and outputs non-private data as the data attributes of the behavioral decision and the behavioral decision parameters, the evolutionary cloud brain module uploads the behavioral decision and the behavioral decision parameters to the cloud, analyzes the behavioral decision and the behavioral decision parameters through the cloud, obtains the cloud analysis execution plan, and uses the cloud analysis execution plan as the first update dataset to update the execution comparison library.

[0040] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a multi-dimensional perception system through a camera and a microphone, which can capture children's behavior and voice data in real time and accurately. The intelligent analysis model built into the system uses advanced algorithms to deeply mine the children's needs and behavioral characteristics behind the data, thereby quickly generating accurate and comprehensive execution decisions, covering anthropomorphic body movements and context-appropriate language expressions. The robot, through the coordinated operation of its head, arms, legs, and mouth, vividly and realistically demonstrates positive moral behavior, not only achieving dynamic monitoring and scientific guidance of children's behavior, but also giving moral education intuitiveness and fun through the concrete expression of human-computer interaction, significantly improving the efficiency of children's moral cognition and behavior shaping, and providing intelligent and personalized moral education solutions tailored for children's growth.

[0041] In particular, the system utilizes a moral generative adversarial network through a decision-making central module to generate moral behavior samples that conform to children's cognitive levels in real time. The system also uses an execution network module to determine children's personality types based on their data and implements appropriate moral behavior demonstration programs for different children based on their personality types, so as to achieve the effect of demonstrating moral behavior in response to the individual differences of different children. Furthermore, the system uses an evolutionary cloud brain module to output the data attributes of behavioral decisions and behavioral decision parameters, and stores the privacy data of different children to ensure the privacy of personal data. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the structure of the moral behavior demonstration device based on a humanoid robot in this embodiment;

[0043] Figure 2 This is a schematic diagram of the system structure of the moral behavior demonstration device based on a humanoid robot in this embodiment. Detailed Implementation

[0044] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0045] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0046] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0047] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0048] Please see Figure 1 The diagram shown is a structural schematic of a moral behavior demonstration device based on a humanoid robot, as described in this embodiment. The device includes:

[0049] The robot body 1 is connected to the robot head 2, the robotic arm 4, the robotic legs 5 and the system of the humanoid robot-based moral behavior demonstration device, and serves as the main frame of the robot.

[0050] The robot head 2 is connected to the robot body 1, the camera 3, and the system 8 of the humanoid robot-based moral behavior demonstration device, and is used to execute the limb action plan in the execution decision;

[0051] Camera 3, which is connected to the robot head 2 and the system 8 of the humanoid robot-based moral behavior demonstration device, includes a first camera 301 and a second camera 302. Both the first camera 301 and the second camera 302 are located on the robot head 2 and are used to collect children's behavior and visual data of children after execution.

[0052] The robotic arm 4, connected to the robot body 1 and the system 8 of the humanoid robot-based moral behavior demonstration device, includes a first robotic arm joint 401, a second robotic arm joint 402, a third robotic arm joint 403, and a fourth robotic arm joint 404. The first robotic arm joint 401 and the second robotic arm joint 402 are connected via a movable connector, and the third robotic arm joint 403 and the fourth robotic arm joint 404 are connected via a movable connector. It is used to execute limb movement schemes in the execution decision.

[0053] The robotic leg 5, connected to the robot body 1 and the system 8 of the humanoid robot-based moral behavior demonstration device, includes a first robotic leg joint 501, a second robotic leg joint 502, a third robotic leg joint 503, and a fourth robotic leg joint 504. The first robotic leg joint 501 and the second robotic leg joint 502 are connected via a movable connector, and the third robotic leg joint 503 and the fourth robotic leg joint 504 are connected via a movable connector. It is used to execute the limb movement plan in the execution decision.

[0054] The robotic mouth 6, which is connected to the robotic head 2 and the humanoid robot-based moral behavior demonstration device, is used to execute language schemes in the decision-making process.

[0055] The microphone 7 is connected to the robot body 1 and is used to collect children's voice and audio data of children after execution;

[0056] The system 8, which is a moral behavior demonstration device based on a humanoid robot, is connected to the robot head 2, camera 3, robotic arm 4, robotic leg 5, robot body 1, and robotic mouth 6. It is used to analyze children's behavior, children's speech, children's visual data after execution, and children's audio data after execution to obtain execution decisions, and to control the robot head 2, robotic arm 4, robotic leg 5, and robotic mouth 5 according to the execution decisions.

[0057] Specifically, it is understood that this embodiment does not limit the specific model of the active connector, and those skilled in the art can set it according to the actual situation, such as a reversible planetary roller screw.

[0058] Specifically, the humanoid robot-based moral behavior demonstration device is used for children's moral behavior education. The device constructs a multi-dimensional perception system using a camera and microphone, enabling real-time and accurate capture of children's behavior and voice data. The system's built-in intelligent analysis model uses advanced algorithms to deeply mine the children's needs and behavioral characteristics behind the data, thereby quickly generating accurate and comprehensive execution decisions. This includes anthropomorphic body movements and context-appropriate language expressions. Through the coordinated operation of its head, arms, legs, and mouth, the robot vividly and realistically demonstrates positive moral behavior, not only achieving dynamic monitoring and scientific guidance of children's behavior but also, through concrete human-computer interaction, giving moral education intuitiveness and fun. This significantly improves the efficiency of children's moral cognition and behavior shaping, providing a tailored, intelligent, and personalized moral education solution for children's growth.

[0059] Please see Figure 2 The diagram shown is a structural schematic of the system of the moral behavior demonstration device based on a humanoid robot in this embodiment. The system includes:

[0060] The sensory center module is used to acquire data about children;

[0061] The decision center module is used to acquire behavioral decisions and decision parameters based on children's data using a moral generative adversarial network, and send the behavioral decisions and decision parameters to the execution network module to adjust the behavioral decision parameters based on the children's data. The decision center module is connected to the perception center module.

[0062] The execution network module is used to acquire the execution plan based on behavioral decisions and decision parameters, control the humanoid robot according to the execution plan, calibrate the decision parameters based on children's data, and revise the calibration process of the decision parameters. The execution network module is connected to the decision center module.

[0063] The Evolutionary Cloud Brain module is used to output data attributes of behavioral decisions and behavioral decision parameters based on children's data. The Evolutionary Cloud Brain module is connected to the Execution Network module.

[0064] Specifically, the system of the humanoid robot-based moral behavior demonstration device is applied in a humanoid robot-based moral behavior demonstration device. This system not only compensates for the shortcomings of traditional education but also provides children with safe, scientific, and warm moral guidance in the era of artificial intelligence, becoming a beneficial extension of family and school education. It fundamentally promotes the systematic improvement of children's moral qualities. The system acquires children's data through a perception center module to facilitate subsequent moral decisions for children. The system uses a moral generative adversarial network (GAN) through a decision center module to generate moral behavior samples that match children's cognitive levels in real time. The system also uses an execution network module to determine children's personality types based on the data and implements appropriate moral behavior demonstration schemes for different children based on their personality types, achieving the effect of demonstrating moral behavior to individual differences among children. Furthermore, the system uses an evolutionary cloud brain module to output data attributes of behavioral decisions and behavioral decision parameters, storing the privacy data of different children to ensure the confidentiality of personal data.

[0065] Specifically, the perception center module acquires child data, which includes child behavior, child speech, post-execution child visual data, and post-execution child audio data. The child behavior and post-execution child visual data are acquired through a camera, while the child speech and post-execution child audio data are acquired through a microphone. The child behavior refers to the child's actions, such as clapping or grabbing. The child speech refers to the sounds the child makes while performing the actions, such as crying or swearing. The post-execution child visual data refers to the behavioral image data of the child after the humanoid robot performs the moral behavior demonstration. The post-execution child audio data refers to the child's speech after the humanoid robot performs the moral behavior demonstration.

[0066] Specifically, the perception center module acquires children's data to make subsequent moral decisions about the children.

[0067] Specifically, the decision center module inputs children's behavior and speech data from children's data into the moral generative adversarial network to obtain the behavioral decisions and decision parameters output by the moral generative adversarial network. The decision center module then sends the behavioral decisions and decision parameters to the execution network module. The decision parameters include language intensity weight α and physical contact weight β, with α+β=1.

[0068] Specifically, the aforementioned moral generative adversarial network (GAN) refers to a GAN that takes children's behavior and speech as input and outputs behavioral decisions and decision parameters. The decision-making central module constructs the moral GAN ​​using a moral GAN ​​construction method. This embodiment does not limit the specific method of constructing the moral GAN; those skilled in the art can set it according to actual needs. For example, the GAN can be trained using an adversarial dataset to obtain the moral GAN. The adversarial dataset refers to the training dataset used to construct the moral GAN, and includes historically acquired children's behavior, historically acquired children's speech, and historical... The acquired child's voice corresponds to behavioral decisions and decision parameters. The behavioral decisions refer to the moral demonstration decisions that the robot needs to make. The decision parameters refer to the weight parameters used to make behavioral choices for the humanoid robot's behavioral decisions. The decision parameters correspond to the behavioral decisions. The language intensity weight refers to the language output weight coefficient in the humanoid robot's moral demonstration. The limb contact weight refers to the output weight coefficient of the physical behavior in the humanoid robot's moral demonstration. This embodiment does not limit the specific method by which the decision center module sends the behavioral decision parameters to the execution network module. Those skilled in the art can set it according to actual needs, such as sending the behavioral decision parameters to the execution network module via wireless signals.

[0069] Specifically, the decision-making central module converts children's behavior into corresponding behavioral decisions and behavioral decision parameters, so that the humanoid robot's moral behavior demonstration device can be controlled according to the behavioral decisions and behavioral decision parameters, thereby achieving the purpose of demonstrating moral behavior corresponding to the scenarios in which children's behavior and speech are generated.

[0070] Specifically, when the decision-making central module adjusts the behavioral decision parameters based on the children's data, it inputs the geographical coordinates from the children's data into the cultural analysis model. The cultural analysis model outputs the physical contact resistance γ, and adjusts the physical contact weight β based on the physical contact resistance γ to obtain the adjusted physical contact weight β1. β1 is set to (1-γ)×β. The language intensity weight α is then adjusted based on the adjusted physical contact weight β1 to obtain the adjusted language intensity weight α1. α1 is set to 1-β1. The adjusted language intensity weight α1 and the adjusted physical contact weight β1 are used as the adjusted decision parameters, and the decision parameters are replaced with the adjusted decision parameters.

[0071] Specifically, the geographic coordinates refer to the coordinates of the current geographical location of the humanoid robot. This embodiment does not limit the method of obtaining geographic coordinates; those skilled in the art can set it according to actual needs, such as obtaining geographic coordinates through a locator. The cultural analysis model refers to a recurrent neural network model that takes geographic coordinates as input and physical contact resistance as output. The decision center module constructs the cultural analysis model using a cultural analysis model construction method. This embodiment does not limit the specific method of constructing the cultural analysis model; those skilled in the art can set it according to actual needs, such as training the recurrent neural network model using a cultural analysis dataset to obtain the cultural analysis model. The cultural analysis dataset refers to the training dataset used to construct the cultural analysis model. The cultural analysis dataset includes historically collected geographic coordinates and the physical contact resistance corresponding to the historically collected geographic coordinates, where γ≤1. The physical contact resistance corresponding to the historically collected geographic coordinates is the physical contact resistance judged by experts based on their experience after cultural analysis of the geographic coordinates.

[0072] Specifically, the decision-making central module obtains the degree of physical contact resistance corresponding to the geographical coordinates by analyzing the geographical coordinates, avoiding incorrect moral behavior demonstrations caused by local cultural resistance to physical contact, increasing the regional adaptability of the humanoid robot, and adapting to the moral behavior demonstrations of children in multiple regions.

[0073] Specifically, when the execution network module obtains an execution plan based on behavioral decisions and decision parameters, it inputs the behavioral decisions and decision parameters into an execution comparison database, compares the behavioral decisions and decision parameters with their indexes, and outputs the execution plan based on the comparison results, wherein:

[0074] When the behavior decision and decision parameters match the behavior decision and decision parameter index, the execution plan corresponding to the behavior decision and decision parameter index will be output.

[0075] When the behavioral decision and decision parameters are inconsistent with the behavioral decision and decision parameter index, the behavioral decision and decision parameters are uploaded to the Evolutionary Cloud Brain module, which then outputs the execution plan based on the behavioral decision and decision parameters.

[0076] Specifically, the execution reference library refers to an index library used to compare and output execution schemes corresponding to behavioral decisions and decision parameters. The behavioral decision and decision parameter index refers to a guided reference used to retrieve execution schemes. The execution schemes include language schemes and body movement schemes. This embodiment does not limit the specific construction method of the execution reference library. Those skilled in the art can set it according to the actual situation. For example, taking behavioral decision-comfort as an example, a comfort execution scheme table is set up. When the language intensity weight α=0, there is no language scheme. When the language intensity weight 0<α≤0.3, the humanoid robot outputs the language "Don't cry." as the language scheme. When the language intensity weight 0.3<α≤0.6, the humanoid robot outputs the language "Don't be sad." as the language scheme. When the language intensity weight 0.6<α≤0. 8. Using the humanoid robot's output language "I know you're sad, I'll be with you!" as the language scheme, when the language intensity weight is 0.8 < α ≤ 1.0, the humanoid robot's output language "Seeing you cry really breaks my heart, can I hug you?" is used as the language scheme. When the limb contact weight β = 0, there is no limb action scheme. When the limb contact weight is 0 < β ≤ 0.1, the humanoid robot's output action "nod slightly" is used as the limb action scheme. When the limb contact weight is 0.1 < β ≤ 0.4, the humanoid robot's output action "pat on the shoulder" is used as the limb action scheme. When the limb contact weight is 0.4 < β ≤ 0.7, the humanoid robot's output action "shake hands" is used as the limb action scheme. When the limb contact weight is 0.7 < β ≤ 1.0, the humanoid robot's output action "hug" is used as the limb action scheme.

[0077] Specifically, the execution network module outputs behavioral decisions and execution schemes corresponding to decision parameters from the execution reference library, so that the humanoid robot can subsequently be controlled to demonstrate moral behavior to children based on children's behavior and speech.

[0078] Specifically, the execution network module controls the humanoid robot according to the execution plan.

[0079] Specifically, this embodiment does not limit the specific control methods for controlling the humanoid robot according to the execution plan. Those skilled in the art can set them according to the actual situation, such as using a PID control algorithm to control the humanoid robot according to the execution plan.

[0080] Specifically, the execution network module controls the humanoid robot according to the execution plan in order to demonstrate moral behavior to children.

[0081] Specifically, when the execution network module calibrates the decision parameters based on the child data, it inputs the child's post-execution visual data and post-execution audio data into the emotion perception model to obtain the child's emotional fluctuation P output by the emotion perception model. The module then compares the child's emotional fluctuation P with preset minimum emotional fluctuation Pmin and preset maximum child emotional fluctuation Pmax. Based on the comparison results, it judges the child's emotional fluctuation status and calibrates the decision parameters accordingly.

[0082] When P≤Pmin, the execution network module determines that the child's emotional fluctuation is low and does not calibrate the decision parameters;

[0083] When Pmin < P ≤ Pmax, the execution network module determines the child's emotional fluctuation as moderate, calibrates the decision parameters, and calibrates the physical contact weight β using the first fluctuation coefficient b1, setting b1 = 0.85 + 0.1 × e -(P-Pmin) Let e ​​be the base of the natural logarithm. We obtain the first proofread limb contact weight β2 and set β2 = b1 × β. Based on the first proofread limb contact weight β2, we proofread the language intensity weight α to obtain the first proofread language intensity weight α2. We set α2 = 1 - β2. We replace the limb contact weight β and language intensity weight α with the first proofread limb contact weight β2 and the first proofread language intensity weight α2, and output the execution plan based on the first proofread limb contact weight β2 and the first proofread language intensity weight α2.

[0084] When P > Pmax, the execution network module determines that the child's emotional fluctuation is high, adjusts the decision parameters, and adjusts the physical contact weight β using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e -(P-Pmax)Let e ​​be the base of the natural logarithm. The second proofread limb contact weight β3 is obtained. Let β3 = b2 × β. The language intensity weight α is proofread according to the second proofread limb contact weight β3 to obtain the second proofread language intensity weight α3. Let α3 = 1 - β3. Replace the limb contact weight β and the language intensity weight α with the second proofread limb contact weight β3 and the second proofread language intensity weight α3. Output the execution plan according to the second proofread limb contact weight β3 and the second proofread language intensity weight α3.

[0085] Specifically, the preset minimum emotional fluctuation refers to the minimum preset value used to judge the emotional fluctuation of children, and the preset maximum emotional fluctuation refers to the maximum preset value used to judge the emotional fluctuation of children. This embodiment does not limit the specific values ​​of the preset minimum and preset maximum emotional fluctuations. Those skilled in the art can set them according to actual needs, but they must meet the needs of judging the emotional fluctuation of children. For example, the specific values ​​of the preset minimum and preset maximum emotional fluctuations can be set according to the judgment experience of child behavior experts. The emotional fluctuation of children refers to the emotional fluctuation of children after the humanoid robot demonstrates moral behavior according to the execution decision, as judged by the child's emotional fluctuation and the preset minimum and preset maximum emotional fluctuations. The emotional fluctuation of children includes low emotional fluctuation, medium emotional fluctuation, and high emotional fluctuation.

[0086] Specifically, the execution network module judges the child's emotional fluctuations and adjusts the decision parameters based on the judgment results to reduce the child's emotional fluctuations. The first fluctuation coefficient is determined when the child's emotional fluctuation is moderate, indicating that the child understands and is thinking about the meaning of the moral behavior demonstration, but cannot apply it independently. By decreasing the value of the first fluctuation coefficient as the child's emotional fluctuation increases, the weight of physical contact is reduced, allowing the child to understand the moral behavior through words. The second fluctuation coefficient is determined when the child's emotional fluctuation is high, indicating that the child clearly recognizes the moral behavior demonstration but cannot apply it independently. In this case, by significantly decreasing the value of the second fluctuation coefficient as the child's emotional fluctuation increases, the weight of physical contact is significantly reduced, allowing the child to gradually understand the purpose of the behavioral decision. This achieves the goal of demonstrating moral behavior in a way that is tailored to the individual differences of different children.

[0087] Specifically, when revising the decision parameter calibration process, the execution network module calculates the emotional response time Δt1 based on the decision execution time T0 and the first emotional change time T1, setting Δt1 = T1 - T0. It then compares the emotional response time Δt1 with a preset emotional response time Δt10, determines the child's characteristic type based on the comparison result, and revises the decision parameter calibration process accordingly.

[0088] When △t1≤△t10, the execution network module determines that the child's characteristic type is rapid response type, and calculates the peak emotion change time △t2 based on the first emotion change time T1 and the peak emotion change time T2, setting △t2=T2-T1, and comparing the peak emotion change time △t2 with the preset peak emotion change time △t20, wherein:

[0089] If △t2≤△t20, the execution network module determines that the child's characteristic type is the rapid resonance type in the rapid response type, and does not revise the decision parameter calibration process;

[0090] If △t2>△t20, the execution network module determines that the child's characteristic type is the rational observation type in the rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α through the first fluctuation coefficient b1, setting b1=0.85+0.1×e". -(P-Pmin) Let e ​​be the base of the natural logarithm. The third-corrected language intensity weight α4 is obtained, and α4 is set to b1 × α. The physical contact weight β is corrected based on the third-corrected language intensity weight α4, resulting in the third-corrected physical contact weight β4, which is set to 1 - α4. The physical contact weight β and language intensity weight α are replaced with the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4, and the execution plan is output based on the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "Correcting the language intensity weight α using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e." -(P-Pmax) e is the base of the natural logarithm. The fourth proofreading language intensity weight α5 is obtained. α5 is set to b2×α. The limb contact weight β is proofread according to the fourth proofreading language intensity weight α5 to obtain the fourth proofreading limb contact weight β5. β5 is set to 1-α5. The limb contact weight β and language intensity weight α are replaced with the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5. The execution plan is output according to the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5.

[0091] When △t1 > △t10, the execution network module determines that the child's characteristic type is non-rapid response, and calculates the total emotional change duration △t3 based on the emotional return to normalcy time T3 and the decision execution time T0, setting △t3 = T3 - T0, and comparing the total emotional change duration △t3 with the preset total emotional change duration △t30, wherein:

[0092] If △t3≤△t30, the execution network module determines that the child's characteristic type is low-participation in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the calibration of the decision parameters is canceled, and the physical contact weight β is revised by the first revision coefficient dz, which is set to dz=1.76-0.21×e -(P-Pmax) The first revised physical contact weight β8 is obtained, β8 = β × dz. The language intensity weight α is then revised based on this first revised weight β8, resulting in the first revised language intensity weight α8. α8 is set to 1 - β8. The physical contact weight β and language intensity weight α are replaced with the first revised weights β8 and α8, respectively. The execution plan is then output based on these weights. When the child's emotional fluctuation is high, the calibration of the decision parameters is canceled. The physical contact weight β is revised using the second revision coefficient dq, set to dq = 1.96 - 0.21 × e. -(P-Pmax) The second revised physical contact weight β9 is obtained, β9 = β × dq. The language intensity weight α is corrected according to the second revised physical contact weight β9 to obtain the second revised language intensity weight α9. α9 = 1 - β9 is set. The physical contact weight β and the language intensity weight α are replaced with the second revised physical contact weight β9 and the second revised language intensity weight α9. The execution plan is output according to the second revised physical contact weight β9 and the second revised language intensity weight α9.

[0093] If △t3 > △t30, the execution network module determines that the child's characteristic type is the delayed response type in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α through the third fluctuation coefficient b3, setting b3 = 1.24 - 0.20 × e -(P-Pmin)Let e ​​be the base of the natural logarithm. The fifth-corrected language intensity weight α6 is obtained, and α6 is set to b3 × α. The physical contact weight β is corrected based on the fifth-corrected language intensity weight α6, resulting in the fifth-corrected physical contact weight β6, which is set to 1 - α6. The physical contact weight β and language intensity weight α are replaced with the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6, and the execution plan is output based on the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "Correcting the language intensity weight α using the fourth fluctuation coefficient b4, setting b4 = 1.56 - 0.33 × e." -(P-Pmax) Let e ​​be the base of the natural logarithm. We obtain the language intensity weight α7 after the sixth proofreading. We set α7 = b4 × α. Based on the language intensity weight α7 after the sixth proofreading, we calibrate the limb contact weight β to obtain the limb contact weight β7 after the sixth proofreading. We set β7 = 1 - α7. We replace the limb contact weight β and language intensity weight α with the limb contact weight β7 and language intensity weight α7 after the sixth proofreading. We then output the execution plan based on the limb contact weight β7 and language intensity weight α7 after the sixth proofreading.

[0094] Specifically, the decision execution time refers to the time when the humanoid robot demonstrates moral behavior based on the executed decision; the first emotional change time refers to the time when the child's emotions begin to change; the peak emotional change time refers to the time when the child's emotions are at their greatest; the emotional return to normalcy time refers to the time when the child's emotions no longer change; the preset emotional response time, the preset peak emotional change time, and the preset total emotional change duration are preset values ​​used to determine the child's characteristic type. This embodiment does not limit the specific values ​​of the preset emotional response time, the preset peak emotional change time, and the preset total emotional change duration. Those skilled in the art can set them according to actual conditions, but they must meet the requirements for determining the child's characteristic type. For example, the preset emotional response time and the preset peak emotional change time can be adjusted based on the experience of child behavior experts. The specific values ​​for the time of emotional change and the preset total duration of emotional change are set. The child characteristic type refers to the child's personality characteristics judged based on the time of emotional change. The child characteristic type includes the child characteristic type of rapid response and non-rapid response. The rapid response type includes rapid resonance type and rational observation type. The non-rapid response type includes low participation type and delayed reaction type. The rapid resonance type means that the child can quickly resonate with the humanoid robot's behavioral and moral demonstration behavior. The rational response type means that the child can rationally deal with the humanoid robot's behavioral and moral demonstration behavior, but the weight of physical actions needs to be increased. The delayed reaction type means that the child reacts slowly to the humanoid robot's behavioral and moral demonstration behavior, and the weight of language intensity needs to be increased. The low participation type means that the child has a weak sense of participation in the humanoid robot's behavioral and moral demonstration behavior, and the weight of physical actions needs to be significantly increased.

[0095] Specifically, the execution network module adjusts the moral demonstration behavior scheme based on the individual differences of different children by judging the child's characteristic type. This makes the humanoid robot's moral demonstration behavior more suitable for children with different personality traits. When the child is a rational observer, they can rationally deal with the humanoid robot's moral demonstration behavior. By slightly increasing the weight of physical actions, the rational observer's understanding of the robot's moral behavior demonstration is enhanced. When the child is a delayed-response type, they react slowly to the humanoid robot's moral demonstration behavior. The weight of language intensity needs to be increased. This is achieved by increasing the value of the language intensity weight through the third fluctuation coefficient, which increases the delayed-response child's understanding of the moral behavior. Since the third fluctuation coefficient changes with emotional fluctuations... The weight of language intensity increases with the increase of the emotional fluctuation value. When the emotional fluctuation value is greater, the weight of language intensity increases. Through language education, children with delayed response can understand moral behavior. When children are low-participation, they have a weak sense of participation in the moral demonstration behavior of the humanoid robot. It is necessary to significantly increase the weight of physical actions. By adjusting the weight of physical actions, the weight of physical contact is adjusted through the first and second revision coefficients. When the first and second revision coefficients increase with the increase of emotional fluctuation, the weight of physical contact will also increase. This achieves the goal of increasing the weight of physical contact when the emotional fluctuation is too high, thereby increasing the participation of such children and enabling the humanoid robot to adjust to a moral behavior demonstration scheme suitable for low-participation children.

[0096] Specifically, when the evolutionary cloud brain module outputs the data attributes of behavioral decisions and behavioral decision parameters, it calculates the emotional fluctuation deviation ΔP based on the child's emotional fluctuation P and the average emotional fluctuation Py of children in the cloud database, setting ΔP=(P-Py) / Py. It then compares the emotional fluctuation deviation ΔP with a preset emotional fluctuation deviation ΔP0, judges the emotional fluctuation deviation based on the comparison result, and outputs the data attributes of behavioral decisions and behavioral decision parameters based on the judgment result. Furthermore, it corrects the adjustment of behavioral decisions based on the judgment result, wherein:

[0097] When △P≤△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is small, and outputs non-private data as the data attributes of behavioral decision and behavioral decision parameters, without correcting the adjustment of behavioral decision.

[0098] When △P>△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is large, and outputs private data as the data attribute of behavioral decision and the data attribute of behavioral decision parameters to correct the adjustment of behavioral decision. The correction method is to cancel the adjustment of behavioral decision parameters by the decision center module.

[0099] Specifically, the average emotional fluctuation of children in the cloud database refers to the average emotional fluctuation of children after executing the decision in the cloud database. The preset emotional fluctuation deviation refers to a preset value for judging the emotional fluctuation deviation. This embodiment does not limit the specific value of the preset emotional fluctuation deviation. Those skilled in the art can set it according to the actual situation, but it must meet the requirements for judging the emotional fluctuation deviation. For example, it can be set according to the experience of child psychologists on children's emotional changes. The emotional fluctuation deviation refers to the magnitude of the deviation between the child's emotional fluctuation judged by the emotional fluctuation deviation and the average emotional fluctuation of children in the cloud database. The emotional fluctuation deviation includes the emotional fluctuation deviation being small and the emotional fluctuation deviation being large.

[0100] Specifically, the evolutionary cloud brain module judges the universality of a child's emotional fluctuation deviation by analyzing the deviation. If the child's emotional deviation is too large, it proves that the child is different from ordinary children. The child's personality and habits are personal privacy data and will not be disclosed to protect the user's privacy. If the child's emotional fluctuation is greater than the average emotional fluctuation, the child may not be adapted to the local culture or may not be a native of the area. In this case, the adjustment of behavioral decision parameters based on geographical coordinates is canceled to achieve the effect of adapting to the child and improve the adaptability of the humanoid robot.

[0101] Specifically, when the evolutionary cloud brain module determines that the emotional fluctuation deviation is appropriate and outputs non-private data as the data attributes of the behavioral decision and the data attributes of the behavioral decision parameters, the evolutionary cloud brain module uploads the behavioral decision and the behavioral decision parameters to the cloud, analyzes the behavioral decision and the behavioral decision parameters through the cloud, obtains the cloud analysis execution plan, and uses the cloud analysis execution plan as the first update dataset to update the execution comparison library.

[0102] Specifically, the cloud refers to a virtual computing environment consisting of a remote server cluster and related services connected via a network. This embodiment does not limit the specific analysis method for analyzing behavioral decisions and behavioral decision parameters through the cloud. Those skilled in the art can set it according to the actual situation. For example, child behavior experts in the cloud can analyze behavioral decisions and behavioral decision parameters based on their experience. This embodiment does not limit the specific method for uploading the behavioral decisions and behavioral decision parameters to the cloud. Those skilled in the art can set it according to the actual situation. For example, the behavioral decisions and behavioral decision parameters can be uploaded to the cloud via a wireless network.

[0103] Specifically, the evolutionary cloud brain module transmits the behavioral decisions and behavioral decision parameters that are determined to be non-private data to the cloud, where experts analyze them so that the humanoid robot can fill the gaps in its own database and deal with more scenarios that require demonstration of ethical behavior.

[0104] Specifically, when the Evolutionary Cloud Brain module determines that the emotional fluctuation deviation is inappropriate and outputs private data as the data attribute and data attribute of the behavioral decision, the Evolutionary Cloud Brain module uploads the behavioral decision and behavioral decision parameters to the edge node and sends the behavioral decision and behavioral decision parameters to the child's family. The child's family analyzes the behavioral decision and behavioral decision parameters to obtain the edge execution plan, and uses the edge execution plan as the second update dataset to update the execution comparison library.

[0105] Specifically, the edge node refers to a computing node deployed at the network edge close to the data source or user terminal. This embodiment does not limit the specific analysis method for analyzing behavioral decisions and behavioral decision parameters by the child's family. Those skilled in the art can set it according to the actual situation, such as having the child's family analyze the behavioral decisions and behavioral decision parameters based on the actual situation and past experience. This embodiment does not limit the specific method for uploading the behavioral decisions and behavioral decision parameters to the edge node. Those skilled in the art can set it according to the actual situation, such as uploading the behavioral decisions and behavioral decision parameters to the edge node via a wireless network.

[0106] Specifically, the evolutionary cloud brain module sends the behavioral decisions and behavioral decision parameters that are determined to be private data to the child's family members, who then analyze them. This allows the humanoid robot to fill the gaps in its own database, respond to more scenarios that require moral behavior demonstrations, and at the same time, ensure the privacy of the child's personal data.

[0107] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A system for demonstrating moral behavior using a humanoid robot, characterized in that, The humanoid robot-based moral behavior demonstration device includes: The robot body is connected to the robot head, robotic arms, robotic legs, and a system of humanoid robot-based moral behavior demonstration devices. The robot's head is connected to the robot's body, a camera, and a system for demonstrating ethical behavior based on a humanoid robot. A camera is connected to a system that includes a robot head and a humanoid robot-based demonstration device for ethical behavior. A robotic arm connected to a system of robot bodies and humanoid robot-based demonstration devices for ethical behavior; The robotic legs are connected to the robot body and the system of a humanoid robot-based demonstration device for ethical behavior; A robotic mouth, which is connected to a robot head and a system of humanoid robot-based moral behavior demonstration devices; A microphone, which is connected to the main body of the robot; A system based on a humanoid robot for demonstrating ethical behavior, which is connected to the robot's head, camera, robotic arm, robotic legs, robot body, and robotic mouth. The system of the humanoid robot-based moral behavior demonstration device includes: The sensory center module is used to acquire data about children; The decision-making central module is used to acquire behavioral decisions and decision parameters based on children's data using the moral generative adversarial network, and then send the behavioral decisions and decision parameters to the execution network module to adjust the behavioral decision parameters based on the children's data; The execution network module is used to acquire execution plans based on behavioral decisions and decision parameters, control the humanoid robot according to the execution plan, calibrate decision parameters based on children's data, and revise the calibration process of decision parameters. The Evolutionary Cloud Brain module is used to judge the deviation of emotional fluctuations based on children's data, obtain the judgment result, update the execution control library based on the judgment result, and also correct the adjustment of behavioral decisions based on the judgment result; The decision center module inputs children's behavior and speech data from children's data into the moral generative adversarial network to obtain the behavioral decisions and decision parameters output by the moral generative adversarial network. The decision center module then sends the behavioral decisions and decision parameters to the execution network module. The decision parameters include language intensity weight α and physical contact weight β, and α+β=1 is set. When the decision-making central module adjusts the behavioral decision parameters based on the children's data, it inputs the geographical coordinates from the children's data into the cultural analysis model. The cultural analysis model outputs the physical contact resistance γ, and adjusts the physical contact weight β based on the physical contact resistance γ to obtain the adjusted physical contact weight β1. β1 is set to (1-γ)×β. The language intensity weight α is then adjusted based on the adjusted physical contact weight β1 to obtain the adjusted language intensity weight α1. α1 is set to 1-β1. The adjusted language intensity weight α1 and the adjusted physical contact weight β1 are used as the adjusted decision parameters, and the decision parameters are replaced with the adjusted decision parameters. When the execution network module obtains an execution plan based on behavioral decisions and decision parameters, it inputs the behavioral decisions and decision parameters into an execution comparison database, compares the behavioral decisions and decision parameters with their indexes, and outputs the execution plan based on the comparison results. When the behavior decision and decision parameters match the behavior decision and decision parameter index, the execution plan corresponding to the behavior decision and decision parameter index will be output. When the behavior decision and decision parameters are inconsistent with the behavior decision and decision parameter index, the behavior decision and decision parameters are uploaded to the Evolutionary Cloud Brain module, which then outputs the execution plan based on the behavior decision and decision parameters. When the execution network module calibrates the decision parameters based on the child data, it inputs the child's post-execution visual data and post-execution audio data into the emotion perception model to obtain the child's emotion fluctuation P output by the emotion perception model. The module then compares the child's emotion fluctuation P with preset minimum emotion fluctuation Pmin and preset maximum child emotion fluctuation Pmax. Based on the comparison results, it judges the child's emotion fluctuation status and calibrates the decision parameters accordingly. When P≤Pmin, the execution network module determines that the child's emotional fluctuation is low and does not calibrate the decision parameters; When Pmin < P ≤ Pmax, the execution network module determines the child's emotional fluctuation as moderate, calibrates the decision parameters, and calibrates the physical contact weight β using the first fluctuation coefficient b1, setting b1 = 0.85 + 0.1 × e -(P-Pmin) Let e ​​be the base of the natural logarithm. We obtain the first proofread limb contact weight β2 and set β2 = b1 × β. Based on the first proofread limb contact weight β2, we proofread the language intensity weight α to obtain the first proofread language intensity weight α2. We set α2 = 1 - β2. We replace the limb contact weight β and language intensity weight α with the first proofread limb contact weight β2 and the first proofread language intensity weight α2, and output the execution plan based on the first proofread limb contact weight β2 and the first proofread language intensity weight α2. When P > Pmax, the execution network module determines that the child's emotional fluctuation is high, adjusts the decision parameters, and adjusts the physical contact weight β using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e -(P-Pmax) Let e ​​be the base of the natural logarithm. The second proofread limb contact weight β3 is obtained. Let β3 = b2 × β. The language intensity weight α is proofread according to the second proofread limb contact weight β3 to obtain the second proofread language intensity weight α3. Let α3 = 1 - β3. Replace the limb contact weight β and the language intensity weight α with the second proofread limb contact weight β3 and the second proofread language intensity weight α3. Output the execution plan according to the second proofread limb contact weight β3 and the second proofread language intensity weight α3.

2. The system of the moral behavior demonstration device based on a humanoid robot according to claim 1, characterized in that, The execution network module controls the humanoid robot according to the execution plan.

3. The system of the moral behavior demonstration device based on a humanoid robot according to claim 1, characterized in that, When revising the decision parameter calibration process, the execution network module calculates the emotional response time Δt1 based on the decision execution time T0 and the first emotional change time T1, setting Δt1 = T1 - T0. It then compares the emotional response time Δt1 with the preset emotional response time Δt10, determines the child's characteristic type based on the comparison result, and revises the decision parameter calibration process accordingly. When △t1≤△t10, the execution network module determines that the child's characteristic type is rapid response type, and calculates the peak emotion change time △t2 based on the first emotion change time T1 and the peak emotion change time T2, setting △t2=T2-T1, and comparing the peak emotion change time △t2 with the preset peak emotion change time △t20, wherein: If △t2≤△t20, the execution network module determines that the child's characteristic type is the rapid resonance type in the rapid response type, and does not revise the decision parameter calibration process; If △t2 > △t20, the execution network module determines that the child's characteristic type is the rational observation type within the rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α using the first fluctuation coefficient b1, setting b1 = 0.85 + 0.1 × e -(P-Pmin) Let e ​​be the base of the natural logarithm. The third-corrected language intensity weight α4 is obtained, and α4 is set to b1 × α. The physical contact weight β is corrected based on the third-corrected language intensity weight α4, resulting in the third-corrected physical contact weight β4, which is set to 1 - α4. The physical contact weight β and language intensity weight α are replaced with the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4, and the execution plan is output based on the third-corrected physical contact weight β4 and the third-corrected language intensity weight α4. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "Correcting the language intensity weight α using the second fluctuation coefficient b2, setting b2 = 0.65 + 0.2 × e." -(P-Pmax) e is the base of the natural logarithm. The fourth proofreading language intensity weight α5 is obtained. α5 is set to b2×α. The limb contact weight β is proofread according to the fourth proofreading language intensity weight α5 to obtain the fourth proofreading limb contact weight β5. β5 is set to 1-α5. The limb contact weight β and language intensity weight α are replaced with the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5. The execution plan is output according to the fourth proofreading limb contact weight β5 and the fourth proofreading language intensity weight α5. When △t1 > △t10, the execution network module determines that the child's characteristic type is non-rapid response, and calculates the total emotional change duration △t3 based on the emotional return to normalcy time T3 and the decision execution time T0, setting △t3 = T3 - T0, and comparing the total emotional change duration △t3 with the preset total emotional change duration △t30, wherein: If △t3≤△t30, the execution network module determines that the child's characteristic type is low-participation in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the calibration of the decision parameters is canceled, and the physical contact weight β is revised by the first revision coefficient dz, which is set to dz=1.76-0.21×e -(P-Pmax) The first revised physical contact weight β8 is obtained, β8 = β × dz. The language intensity weight α is then revised based on this first revised weight β8, resulting in the first revised language intensity weight α8. α8 is set to 1 - β8. The physical contact weight β and language intensity weight α are replaced with the first revised weights β8 and α8, respectively. The execution plan is then output based on these weights. When the child's emotional fluctuation is high, the calibration of the decision parameters is canceled. The physical contact weight β is revised using the second revision coefficient dq, set to dq = 1.96 - 0.21 × e. -(P-Pmax) The second revised physical contact weight β9 is obtained, β9 = β × dq. The language intensity weight α is corrected according to the second revised physical contact weight β9 to obtain the second revised language intensity weight α9. α9 = 1 - β9 is set. The physical contact weight β and the language intensity weight α are replaced with the second revised physical contact weight β9 and the second revised language intensity weight α9. The execution plan is output according to the second revised physical contact weight β9 and the second revised language intensity weight α9. If △t3 > △t30, the execution network module determines that the child's characteristic type is the delayed response type in the non-rapid response type, and revises the decision parameter calibration process. When the child's emotional fluctuation is moderate, the decision parameter calibration process is revised to: "Calibrate the language intensity weight α through the third fluctuation coefficient b3, setting b3 = 1.24 - 0.20 × e -(P-Pmin) Let e ​​be the base of the natural logarithm. The fifth-corrected language intensity weight α6 is obtained, and α6 is set to b3 × α. The physical contact weight β is corrected based on the fifth-corrected language intensity weight α6, resulting in the fifth-corrected physical contact weight β6, which is set to 1 - α6. The physical contact weight β and language intensity weight α are replaced with the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6, and the execution plan is output based on the fifth-corrected physical contact weight β6 and the fifth-corrected language intensity weight α6. When the child's emotional fluctuation is high, the process of correcting the decision parameters is revised to: "The language intensity weight α is corrected using the fourth fluctuation coefficient b4, and b4 is set to 1.56 - 0.33 × e." -(P-Pmax) Let e ​​be the base of the natural logarithm. We obtain the language intensity weight α7 after the sixth proofreading. We set α7 = b4 × α. Based on the language intensity weight α7 after the sixth proofreading, we calibrate the limb contact weight β to obtain the limb contact weight β7 after the sixth proofreading. We set β7 = 1 - α7. We replace the limb contact weight β and language intensity weight α with the limb contact weight β7 and language intensity weight α7 after the sixth proofreading. We then output the execution plan based on the limb contact weight β7 and language intensity weight α7 after the sixth proofreading.

4. The system of the moral behavior demonstration device based on a humanoid robot according to claim 3, characterized in that, When the evolutionary cloud brain module outputs the data attributes of behavioral decisions and behavioral decision parameters, it calculates the emotional fluctuation deviation ΔP based on the child's emotional fluctuation P and the average emotional fluctuation Py of children in the cloud database, setting ΔP=(P-Py) / Py. It then compares the emotional fluctuation deviation ΔP with a preset emotional fluctuation deviation ΔP0, judges the emotional fluctuation deviation based on the comparison result, and outputs the data attributes of behavioral decisions and behavioral decision parameters based on the judgment result. Furthermore, it corrects the adjustment of behavioral decisions based on the judgment result, wherein: When △P≤△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is small, and outputs non-private data as the data attributes of behavioral decision and behavioral decision parameters, without correcting the adjustment of behavioral decision. When △P>△P0, the evolutionary cloud brain module determines that the emotional fluctuation deviation is large, and outputs private data as the data attribute of behavioral decision and the data attribute of behavioral decision parameters to correct the adjustment of behavioral decision. The correction method is to cancel the adjustment of behavioral decision parameters by the decision center module.

5. The system of the moral behavior demonstration device based on a humanoid robot according to claim 4, characterized in that, When the evolutionary cloud brain module determines that the emotional fluctuation deviation is appropriate and outputs non-private data as the data attributes of the behavioral decision and the behavioral decision parameters, the evolutionary cloud brain module uploads the behavioral decision and the behavioral decision parameters to the cloud, analyzes the behavioral decision and the behavioral decision parameters through the cloud, obtains the cloud analysis execution plan, and uses the cloud analysis execution plan as the first update dataset to update the execution comparison library.

Images