AI-driven full-perception simulation machine pet system

Abstract

The invention, which relates to the technical field of the simulation machine pet system, discloses an AI-driven full-perception simulation machine pet system comprising an AI training module, an olfactory perception module, an auditory perception module, a visual perception module, a navigation positioning module, a touch simulation module and a motion control module. The AI training module is used for generating a personalized behavior model through a machine learning algorithm based on the user behavior data and the environment data, and driving a mechanical pet to simulate the learning and growth process of a biological pet; the olfaction sensing module is used for collecting environment gas data through a gas sensor array, and identifying and feeding back a security threat signal to the main control system in combination with the behavior model output by the AI training module; the auditory perception module is used for receiving a sound wave signal through a microphone array and triggering an instruction response action generated by the AI training module after the sound wave signal is analyzed by a voice recognition algorithm; the visual perception module is used for capturing environment image data through a spherical camera.

Description

Technical Field

[0001] The present invention relates to the technical field of simulated robot pet systems, and in particular to an AI-driven fully perceptive simulated robot pet system. Background Art

[0002] The simulated robotic pet system is a high-tech product that leverages artificial intelligence (AI), sensor technology, and advanced mechanical design to simulate the behaviors and interactions of real pets. Designed to provide users with a companion that is both intelligent and emotionally comforting, this system is particularly suitable for family companionship, elderly care, and child education. It can not only recognize human voice commands and visual signals but also sense ambient odors and temperature changes, reacting accordingly. Therefore, utilizing advanced technologies to enhance the intelligence and safety of simulated robotic pet systems has become a pressing issue.

[0003] In the field of simulated robot pet systems, traditional robot pets often adopt preset behavioral patterns and lack the ability to adaptively learn based on user behavior and environmental changes. Existing robot pets only have one or a few sensory capabilities, making it difficult to fully understand the surrounding environment. At the same time, most robot pets can only provide limited emotional feedback and cannot simulate the warm touch and delicate reactions of real pets. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an AI-driven fully perceptual simulated robot pet system to solve the problem that traditional robot pets often adopt preset behavior patterns and lack the ability to adaptively learn according to user behavior and environmental changes. In addition, existing robot pets only have one or a few perceptual capabilities and find it difficult to fully understand the surrounding environment. At the same time, most robot pets can only provide limited emotional feedback and cannot simulate the warm touch and delicate reactions of real pets.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an AI-driven fully perceptive simulated robotic pet system, comprising:

[0008] AI training module, olfactory perception module, auditory perception module, visual perception module, navigation and positioning module, touch simulation module and motion control module;

[0009] The AI ​​training module is used to generate a personalized behavior model based on user behavior data and environmental data through a machine learning algorithm, driving the mechanical pet to simulate the learning and growth process of the biological pet;

[0010] The olfactory perception module is used to collect environmental gas data through the gas sensor array, and combined with the behavior model output by the AI ​​training module, identify and feed back security threat signals to the main control system;

[0011] The auditory perception module is used to receive sound wave signals through the microphone array, and after being analyzed by the speech recognition algorithm, trigger the command response action generated by the AI ​​training module;

[0012] The visual perception module is used to capture environmental image data through a spherical camera, identify target objects through an image processing algorithm, and plan movement paths based on the behavior model output by the AI ​​training module;

[0013] The navigation and positioning module is used to receive the environmental image data and external positioning signals output by the visual perception module, generate real-time position information through a multi-sensor fusion algorithm, and control the mechanical pet to perform obstacle avoidance or guidance tasks;

[0014] The touch simulation module is used to detect contact pressure through the pressure sensor in the silicone skin and simulate the body temperature and tactile feedback of the biological pet in combination with the temperature sensor data;

[0015] The motion control module is used to drive the ball-and-socket joint mechanical structure to perform multi-degree-of-freedom movements according to the behavioral instructions generated by the AI ​​training module, thereby achieving high-agility movement.

[0016] As a preferred solution of the AI-driven fully perceptive simulated robot pet system of the present invention, the system includes a user behavior acquisition unit, an environmental data acquisition unit, and a reinforcement learning algorithm unit, specifically:

[0017] The user behavior acquisition unit obtains the user's operation instruction sequence O=o1, o2, ..., o through a wearable device or a mobile terminal. n , interaction frequency f u and interactive method classification label L u ;

[0018] The environmental data acquisition unit receives data streams from the olfactory perception module, the auditory perception module, the visual perception module, and the tactile simulation module, and extracts the environmental state information vector, which is expressed as:

[0019] S env =[C gas ,A sound ,D vision ,T touch ];

[0020] Among them, C gas Indicates the gas concentration value, A sound Indicates the intensity of the sound source, D visionIndicates the target distance, T touch Indicates the comprehensive value of contact pressure and temperature;

[0021] The reinforcement learning algorithm unit adopts the Q-learning framework to construct the state space S=(S env ,L u ) and action space A = forward, backward, turn left, turn right, sit down, jump, respond to voice;

[0022] The reinforcement learning algorithm updates the policy based on the following reward function:

[0023] R(s,a)=α·U(L u )+β·E(S env );

[0024] Among them, U(L_u) is the user preference score, defined as:

[0025]

[0026] E(S env ) is the environmental adaptability score, defined as:

[0027]

[0028] Among them, α and β are weight coefficients respectively;

[0029] The algorithm uses Q table update, the expression is:

[0030]

[0031] Where η is the learning rate, γ is the discount factor, and s' represents the next state;

[0032] Output personalized behavior models to control the mechanical pet to exhibit differentiated response strategies in different scenarios.

[0033] As a preferred solution of the AI-driven fully perceptive simulated robot pet system of the present invention, the gas sensor array in the olfactory perception module includes multiple gas sensors corresponding to the detection channels of ammonia, hydrogen sulfide, methane and volatile organic compounds (VOCs), specifically:

[0034] Each gas sensor is made of metal oxide semiconductor material. When exposed to a specific gas, its resistance value R i Changes with gas concentration;

[0035] Output voltage signal V i Obtained through the voltage divider circuit:

[0036]

[0037] Among them, R0 is a fixed resistor, V ref is the reference voltage;

[0038] The ADC converter converts V i Convert to digital concentration value C i , the conversion formula is:

[0039] C i =k i ·(V i -b i );

[0040] Among them, k i is the calibration coefficient, b i is the offset;

[0041] The concentration value C of each channel i Input into the preset safety threshold judgment function:

[0042]

[0043] Among them, T j Indicates whether the jth gas component exceeds the safety threshold, C threshold,j Indicates the standard safety limit of the gas;

[0044] If any T j If it is 1, a threat signal A is generated. s =∑T j And send it to the main control system through serial communication;

[0045] The main control system is based on A s The size of the alarm triggers emergency avoidance actions or alarm prompts.

[0046] As a preferred solution of the AI-driven fully perceptive simulated robot pet system of the present invention, the microphone array in the auditory perception module is composed of four equidistantly distributed electret microphones, which are used to realize sound source localization and speech recognition functions, specifically including:

[0047] The four audio signals collected by the microphone array are represented as x1(t), x2(t), x3(t), and x4(t);

[0048] Perform short-time Fourier transform STFTx on each signal i (t)→X i (f,τ) obtains the frequency domain expression;

[0049] Compute the time difference using the cross-correlation function:

[0050]

[0051] Among them, τ ij represents the time delay between the i-th and j-th microphones, x j represents a conjugate complex number;

[0052] According to the time difference τ ij and the microphone spacing d, calculate the sound source direction angle θ, the expression is:

[0053]

[0054] Where c is the speed of sound in air;

[0055] At the same time, the audio signal is sent to the speech recognition engine, which uses the Hidden Markov Model (HMM) for semantic analysis and matches the preset command keyword library;

[0056] After a successful match, a corresponding command signal is generated and input into the AI ​​training module to trigger the corresponding behavior.

[0057] As a preferred solution of the AI-driven fully perceptive simulated robotic pet system of the present invention, the spherical camera in the visual perception module is composed of two wide-angle lenses, which are respectively installed on the left and right sides of the robotic pet's head to form a binocular stereo vision system, specifically including:

[0058] The two images collected by the spherical camera are I L (x,y) and I R (x,y);

[0059] Use SIFT algorithm to extract feature points and perform feature matching to obtain the corresponding pixel coordinate set (p L i,p R i);

[0060] Calculate the disparity map D(x,y)=x L -x R ;

[0061] Calculate the depth map using the camera intrinsic parameter matrix and baseline distance:

[0062]

[0063] Where Z represents the distance between the target point and the camera, f is the focal length, and b is the horizontal distance between the two cameras;

[0064] Fusion of depth map and RGB image to generate 3D point cloud data;

[0065] The point cloud data is processed using the YOLOv5 target detection algorithm to identify the target object category and location coordinates;

[0066] Output target location information to the navigation and positioning module for path planning.

[0067] As a preferred solution of the AI-driven fully perceptive simulated robot pet system of the present invention, the navigation and positioning module uses a multi-sensor fusion positioning method combining an inertial measurement unit (IMU), a GPS module, and a visual SLAM algorithm, specifically including:

[0068] The IMU module outputs three-axis acceleration a x ,a y ,a z and the three-axis angular velocity ω x ,ω y ,ω z ;

[0069] The GPS module provides longitude and latitude coordinates (lat, lon) and altitude h;

[0070] The visual SLAM algorithm is based on ORB feature extraction and BA optimization method to build a local map and estimate the real-time pose T v ision=(x,y,θ);

[0071] Multi-sensor fusion uses the extended Kalman filter EKF algorithm, and the state vector is defined as:

[0072] X=[x,y,θ,v x ,v y ] T ;

[0073] Among them, x, y are two-dimensional coordinates, θ is the heading angle, v x ,v y is the velocity component;

[0074] The Kalman gain matrix K is calculated as follows:

[0075] K=PH T (HPH T +R) -1 ;

[0076] Where P is the state covariance matrix, H is the observation matrix, and R is the observation noise covariance;

[0077] Output fused pose information X f used to control the mechanical pet to complete obstacle avoidance or guidance tasks.

[0078] As a preferred solution of the AI-driven fully perceptive simulated robot pet system of the present invention, the touch simulation module includes a plurality of pressure sensors and PTC heaters embedded in the silicone surface to simulate biological body temperature and contact feedback, specifically including:

[0079] Each pressure sensor outputs a voltage signal V p After being processed by the amplifying circuit, it is converted into contact force F = k·V p , where k is the calibration coefficient;

[0080] The temperature sensor outputs a resistance change ΔR, which is converted into a voltage signal V via a Wheatstone bridge circuit. T , and then convert it into temperature value T through table lookup method;

[0081] The control system adjusts the power of the PTC heater according to the difference ΔT between T and the target body temperature:

[0082]

[0083] Where P is the heating power, K p , K i , K d are the PID controller parameters;

[0084] When the pressure sensor detects that the contact force is greater than the set threshold F th When the electromagnetic vibrator is triggered, it generates feedback vibration, simulating the reaction of biological pets;

[0085] All tactile data is uploaded to the AI ​​training module to enhance the emotional interaction ability of the mechanical pet.

[0086] As a preferred solution of the AI-driven fully perceptive simulated robotic pet system of the present invention, the ball-and-socket joint in the motion control module is composed of a central ball head, a circular track, three sets of servo motors, and a harmonic reducer, and is used to achieve six-degree-of-freedom motion in three-dimensional space, specifically including:

[0087] The central ball head of the ball-and-socket joint is fixed to the end of the mechanical limb and embedded in the circular track, allowing free rotation in the x, y, and z directions;

[0088] Each rotational degree of freedom is driven by a servo motor through a harmonic reducer, forming three independent rotation axes:

[0089] The first axis is the pitch angle θ1;

[0090] The second axis is the yaw angle θ2;

[0091] The third axis is the roll angle θ3;

[0092] The output torque of each servo motor is calculated by the controller based on the inverse kinematics model:

[0093]

[0094] Among them, J(θ) is the Jacobian matrix, which represents the mapping relationship between the joint angle change and the end position, and Fext is the external force vector, which comes from the pressure sensor feedback of the touch simulation module;

[0095] The controller uses PID control strategy to adjust the servo motor angle in real time to achieve the target posture:

[0096]

[0097] Among them, u i (t) is the control input of the i-th motor, e i (t) represents the angle error, K p , K i , K d are the proportional, integral, and differential gain coefficients;

[0098] When performing complex actions, the system generates smooth time series instructions through trajectory planning algorithms:

[0099] θ(t)=θ0+(θ f -θ0)(1-cos(πt / T)) / 2;

[0100] Among them, θ0 is the initial angle, θ f is the target angle, T is the action duration;

[0101] The ball-and-socket joint, combined with the behavioral instructions output by the AI ​​training module, enables the mechanical pet to achieve highly agile movements.

[0102] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the AI-driven fully perceptive simulated robot pet system as described in the first aspect of the present invention is implemented.

[0103] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the AI-driven fully perceptive simulated robot pet system as described in the first aspect of the present invention.

[0104] The beneficial effects of the present invention are: by integrating AI training modules, olfactory perception modules, auditory perception modules, visual perception modules, navigation and positioning modules, tactile simulation modules and motion control modules, a bionic robot system with multimodal perception capabilities and intelligent interaction functions is constructed. The system can generate personalized behavior models based on user behavior and environmental data using reinforcement learning algorithms to simulate the learning and growth process of biological pets. At the same time, combined with hardware such as gas sensors, microphone arrays, spherical cameras, pressure and temperature sensors, it can realize real-time perception and response to multi-dimensional information such as environmental odors, sounds, images, and touch, and through the ball and socket joint structure and multi-degree-of-freedom motion control, it can complete high-agility physical movement performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0106] Figure 1 Schematic diagram of the AI-driven fully perceptual simulated robot pet system in Example 1. DETAILED DESCRIPTION

[0107] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0108] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0109] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0110] Example, see Figure 1 , is an embodiment of the present invention, which provides an AI-driven fully perceptive simulated robot pet system, comprising:

[0111] AI training module, olfactory perception module, auditory perception module, visual perception module, navigation and positioning module, touch simulation module and motion control module;

[0112] The AI ​​training module is used to generate personalized behavior models through machine learning algorithms based on user behavior data and environmental data, driving the learning and growth process of the mechanical pet to simulate the biological pet;

[0113] Furthermore, the AI ​​training module includes a user behavior collection unit, an environmental data collection unit, and a reinforcement learning algorithm unit, specifically:

[0114] The user behavior acquisition unit obtains the user's operation instruction sequence O=o1,o2,...,o through the wearable device or mobile terminal n , interaction frequency f u and interactive method classification label L u ;

[0115] The environmental data acquisition unit receives data streams from the olfactory perception module, auditory perception module, visual perception module, and tactile simulation module, and extracts the environmental state information vector, which is expressed as:

[0116] S env =[C gas ,A sound ,D vision ,T touch ];

[0117] Among them, C gas Indicates the gas concentration value, A sound Indicates the intensity of the sound source, D vision Indicates the target distance, T touch Indicates the comprehensive value of contact pressure and temperature;

[0118] The reinforcement learning algorithm unit uses the Q-learning framework to construct the state space S=(S env ,L u ) and action space A = forward, backward, turn left, turn right, sit down, jump, respond to voice;

[0119] The reinforcement learning algorithm updates the policy based on the following reward function:

[0120] R(s,a)=α·U(L u )+β·E(S env );

[0121] Among them, U(L_u) is the user preference score, defined as:

[0122]

[0123] E(S env ) is the environmental adaptability score, defined as:

[0124]

[0125] Among them, α and β are weight coefficients respectively;

[0126] The algorithm uses Q table update, the expression is:

[0127]

[0128] Where η is the learning rate, γ is the discount factor, and s' represents the next state;

[0129] Output personalized behavior models to control the mechanical pet to exhibit differentiated response strategies in different scenarios;

[0130] It should be noted that the AI ​​training module builds a data foundation for the learning and growth of mechanical pets through the synergy of the user behavior collection unit and the environmental data collection unit. The reinforcement learning algorithm unit adopts the Q-learning method, which enables the mechanical pet to continuously optimize its own behavior strategy according to user interaction methods and environmental feedback, thereby achieving personalized simulated response. The module not only improves the naturalness of human-computer interaction, but also provides a decision-making basis for the intelligent linkage of subsequent perception and execution modules.

[0131] The olfactory perception module is used to collect environmental gas data through a gas sensor array, combine it with the behavioral model output by the AI ​​training module, identify and feedback security threat signals to the main control system;

[0132] Furthermore, the gas sensor array in the olfactory perception module includes multiple gas-sensitive elements, corresponding to the detection channels of ammonia, hydrogen sulfide, methane and volatile organic compounds (VOCs), specifically:

[0133] Each gas sensor is made of metal oxide semiconductor material. When exposed to a specific gas, its resistance value R i Changes with gas concentration;

[0134] Output voltage signal V i Obtained through the voltage divider circuit:

[0135]

[0136] Among them, R0 is a fixed resistor, V ref is the reference voltage;

[0137] The ADC converter converts V i Convert to digital concentration value C i , the conversion formula is:

[0138] C i =k i ·(V i -b i );

[0139] Among them, k i is the calibration coefficient, b i is the offset;

[0140] The concentration value C of each channel i Input into the preset safety threshold judgment function:

[0141]

[0142] Among them, T j Indicates whether the jth gas component exceeds the safety threshold, C threshold,j Indicates the standard safety limit of the gas;

[0143] If any T j If it is 1, a threat signal A is generated. s =∑T j And send it to the main control system through serial communication;

[0144] The main control system is based on A s The size of the trigger emergency avoidance action or alarm prompt;

[0145] It should be noted that the gas sensor array in the olfactory perception module has the ability to detect a variety of harmful gases with high sensitivity. Combined with ADC conversion and safety threshold judgment mechanism, it realizes real-time monitoring of indoor air quality. The module can be used as part of the environmental safety early warning system in home or elderly care scenarios. When dangerous gas leaks are identified, it can trigger avoidance actions in time to ensure user health and living safety.

[0146] The auditory perception module is used to receive sound wave signals through the microphone array, analyze them through the speech recognition algorithm, and trigger the command response action generated by the AI ​​training module;

[0147] Furthermore, the microphone array in the auditory perception module consists of four equally spaced electret microphones, which are used to achieve sound source localization and speech recognition functions, including:

[0148] The four audio signals collected by the microphone array are represented as x1(t), x2(t), x3(t), and x4(t);

[0149] Perform short-time Fourier transform STFTx on each signal i (t)→X i (f,τ) obtains the frequency domain expression;

[0150] Compute the time difference using the cross-correlation function:

[0151]

[0152] Among them, τij represents the time delay between the i-th and j-th microphones, x j represents a conjugate complex number;

[0153] According to the time difference τ ij and the microphone spacing d, calculate the sound source direction angle θ, the expression is:

[0154]

[0155] Where c is the speed of sound in air;

[0156] At the same time, the audio signal is sent to the speech recognition engine, which uses the Hidden Markov Model (HMM) for semantic analysis and matches the preset command keyword library;

[0157] After a successful match, a corresponding command signal is generated and input into the AI ​​training module to trigger the corresponding behavior;

[0158] It should be noted that the auditory perception module adopts a four-microphone array structure, combined with short-time Fourier transform and cross-correlation function algorithm, to improve the accuracy of sound source positioning and speech recognition. Through the hidden Markov model HMM, the voice commands are semantically analyzed, so that the mechanical pet can respond to the user's voice control commands, enhancing the convenience and intelligence level of human-computer interaction.

[0159] The visual perception module is used to capture environmental image data through a spherical camera, identify target objects through image processing algorithms, and plan movement paths based on the behavior model output by the AI ​​training module;

[0160] Furthermore, the spherical camera in the visual perception module consists of two wide-angle lenses, installed on the left and right sides of the robot's head, forming a binocular stereo vision system, specifically including:

[0161] The two images collected by the spherical camera are I L (x,y) and I R (x,y);

[0162] Use SIFT algorithm to extract feature points and perform feature matching to obtain the corresponding pixel coordinate set (p L i,p R i);

[0163] Calculate the disparity map D(x,y)=x L -x R ;

[0164] Calculate the depth map using the camera intrinsic parameter matrix and baseline distance:

[0165]

[0166] Where Z represents the distance between the target point and the camera, f is the focal length, and b is the horizontal distance between the two cameras;

[0167] Fusion of depth map and RGB image to generate 3D point cloud data;

[0168] The point cloud data is processed using the YOLOv5 target detection algorithm to identify the target object category and location coordinates;

[0169] Output target location information to the navigation and positioning module for path planning;

[0170] It should be noted that the visual perception module realizes three-dimensional space modeling through a binocular stereo vision system, and combines SIFT feature matching and disparity calculation technology to effectively improve the accuracy of target recognition and distance estimation. The application of the YOLOv5 target detection algorithm enables the system to have rapid recognition and classification capabilities, providing reliable visual support for navigation path planning and enhancing the autonomous behavior capabilities of mechanical pets in complex environments.

[0171] The navigation and positioning module is used to receive the environmental image data and external positioning signals output by the visual perception module, generate real-time position information through a multi-sensor fusion algorithm, and control the robot pet to perform obstacle avoidance or guidance tasks;

[0172] Furthermore, the navigation and positioning module uses a combination of an inertial measurement unit (IMU), a GPS module, and a visual SLAM algorithm for multi-sensor fusion positioning, specifically including:

[0173] The IMU module outputs three-axis acceleration a x ,a y ,a z and the three-axis angular velocity ω x ,ω y ,ω z ;

[0174] The GPS module provides longitude and latitude coordinates (lat, lon) and altitude h;

[0175] The visual SLAM algorithm is based on ORB feature extraction and BA optimization method to build a local map and estimate the real-time pose T v ision=(x,y,θ);

[0176] Multi-sensor fusion uses the extended Kalman filter EKF algorithm, and the state vector is defined as:

[0177] X=[x,y,θ,v x ,v y ] T ;

[0178] Among them, x, y are two-dimensional coordinates, θ is the heading angle, vx ,v y is the velocity component;

[0179] The Kalman gain matrix K is calculated as follows:

[0180] K=PH T (HPH T +R) -1 ;

[0181] Where P is the state covariance matrix, H is the observation matrix, and R is the observation noise covariance;

[0182] Output fused pose information X f used, used to control the mechanical pet to complete obstacle avoidance or guidance tasks;

[0183] It should be noted that the navigation and positioning module integrates three positioning technologies: IMU, GPS and visual SLAM, and uses the extended Kalman filter algorithm to perform weighted fusion of multi-source information, which significantly improves the positioning stability and accuracy of the mechanical pet in different indoor and outdoor scenarios. The module provides key posture information support for functions such as obstacle avoidance, guidance and automatic cruising, and is an important component for achieving autonomous movement.

[0184] The touch simulation module is used to detect contact pressure through the pressure sensor in the silicone skin and combine it with the temperature sensor data to simulate the body temperature and tactile feedback of biological pets;

[0185] Furthermore, the silicone surface of the touch simulation module is embedded with multiple pressure sensors and PTC heaters to simulate biological body temperature and contact feedback, including:

[0186] Each pressure sensor outputs a voltage signal V p After being processed by the amplifying circuit, it is converted into contact force F = k·V p , where k is the calibration coefficient;

[0187] The temperature sensor outputs a resistance change ΔR, which is converted into a voltage signal V via a Wheatstone bridge circuit. T , and then convert it into temperature value T through table lookup method;

[0188] The control system adjusts the power of the PTC heater according to the difference ΔT between T and the target body temperature:

[0189]

[0190] Where P is the heating power, K p , K i , K d are the PID controller parameters;

[0191] When the pressure sensor detects that the contact force is greater than the set threshold Fth When the electromagnetic vibrator is triggered, it generates feedback vibration, simulating the reaction of biological pets;

[0192] All tactile data is uploaded to the AI ​​training module to enhance the emotional interaction ability of the mechanical pet;

[0193] It should be noted that the touch simulation module realizes the mechanical pet's perception of external contact and body temperature simulation functions through the coordinated work of the embedded pressure sensor and the PTC heating plate. The PID temperature control strategy ensures that the temperature is stable within the set range, and the electromagnetic vibrator is used to generate tactile feedback, further enhancing the mechanical pet's realistic expression and emotional interaction capabilities, and improving the realism of the user experience.

[0194] The motion control module is used to drive the ball-and-socket joint mechanical structure to perform multi-degree-of-freedom movements based on the behavioral instructions generated by the AI ​​training module, achieving high-agility movement;

[0195] Furthermore, the ball-and-socket joint in the motion control module consists of a central ball head, a circular track, three servo motors, and a harmonic reducer, which is used to achieve six degrees of freedom in three-dimensional space, including:

[0196] The central ball head of the ball-and-socket joint is fixed to the end of the mechanical limb and embedded in the circular track, allowing free rotation in the x, y, and z directions;

[0197] Each rotational degree of freedom is driven by a servo motor through a harmonic reducer, forming three independent rotation axes:

[0198] The first axis is the pitch angle θ1;

[0199] The second axis is the yaw angle θ2;

[0200] The third axis is the roll angle θ3;

[0201] The output torque of each servo motor is calculated by the controller based on the inverse kinematics model:

[0202]

[0203] Among them, J(θ) is the Jacobian matrix, which represents the mapping relationship between the joint angle change and the end position, and F ext is the external force vector, which comes from the pressure sensor feedback of the touch simulation module;

[0204] The controller uses PID control strategy to adjust the servo motor angle in real time to achieve the target posture:

[0205]

[0206] Among them, u i(t) is the control input of the i-th motor, e i (t) represents the angle error, K p , K i , K d are the proportional, integral, and differential gain coefficients;

[0207] When performing complex actions, the system generates smooth time series instructions through trajectory planning algorithms:

[0208] θ(t)=θ0+(θ f -θ0)(1-cos(πt / T)) / 2;

[0209] Among them, θ0 is the initial angle, θ f is the target angle, T is the action duration;

[0210] The ball-and-socket joint, combined with the behavioral instructions output by the AI ​​training module, enables the robotic pet to achieve highly agile movements;

[0211] It should be noted that the ball-and-socket joint structure design in the motion control module breaks through the freedom limitation of the traditional servo structure. Through the combined drive of three sets of servo motors and harmonic reducers, flexible motion control with six degrees of freedom is achieved. The inverse kinematics model based on the Jacobian matrix is ​​combined with the PID controller, making the mechanical pet more smooth and natural when performing complex movements such as jumping and turning, fully demonstrating its high agility and bionic characteristics.

[0212] This embodiment also provides a computer device suitable for the case of an AI-driven fully-perceptive simulated robot pet system, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the AI-driven fully-perceptive simulated robot pet system proposed in the above embodiment.

[0213] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0214] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the AI-driven fully perceptive simulated robot pet system proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0215] In summary, the present invention integrates AI training modules, olfactory perception modules, auditory perception modules, visual perception modules, navigation and positioning modules, tactile simulation modules and motion control modules to construct a bionic robot system with multimodal perception capabilities and intelligent interaction functions. The system can generate personalized behavior models based on user behavior and environmental data using reinforcement learning algorithms to simulate the learning and growth process of biological pets. At the same time, it combines hardware such as gas sensors, microphone arrays, spherical cameras, pressure and temperature sensors to achieve real-time perception and response to multi-dimensional information such as environmental odors, sounds, images, and touch, and completes high-agility physical movement performance through ball-and-socket joint structure and multi-degree-of-freedom motion control.

[0216] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An AI-driven fully perceptive simulated robotic pet system, characterized by: include: AI training module, olfactory perception module, auditory perception module, visual perception module, navigation and positioning module, touch simulation module and motion control module; The AI ​​training module is used to generate a personalized behavior model based on user behavior data and environmental data through a machine learning algorithm, driving the mechanical pet to simulate the learning and growth process of the biological pet; The olfactory perception module is used to collect environmental gas data through the gas sensor array, and combined with the behavior model output by the AI ​​training module, identify and feed back security threat signals to the main control system; The auditory perception module is used to receive sound wave signals through the microphone array, and after being analyzed by the speech recognition algorithm, trigger the command response action generated by the AI ​​training module; The visual perception module is used to capture environmental image data through a spherical camera, identify target objects through an image processing algorithm, and plan movement paths based on the behavior model output by the AI ​​training module; The navigation and positioning module is used to receive the environmental image data and external positioning signals output by the visual perception module, generate real-time position information through a multi-sensor fusion algorithm, and control the mechanical pet to perform obstacle avoidance or guidance tasks; The touch simulation module is used to detect contact pressure through the pressure sensor in the silicone skin and simulate the body temperature and tactile feedback of the biological pet in combination with the temperature sensor data; The motion control module is used to drive the ball-and-socket joint mechanical structure to perform multi-degree-of-freedom movements according to the behavioral instructions generated by the AI ​​training module, thereby achieving high-agility movement.

2. The AI-driven fully perceptive simulated robotic pet system according to claim 1, wherein: The AI ​​training module includes a user behavior acquisition unit, an environmental data acquisition unit, and a reinforcement learning algorithm unit, specifically: The user behavior acquisition unit obtains the user's operation instruction sequence O=o1, o2, ..., o through a wearable device or a mobile terminal. n , interaction frequency f u and interactive method classification label L u ; The environmental data acquisition unit receives data streams from the olfactory perception module, the auditory perception module, the visual perception module, and the tactile simulation module, and extracts the environmental state information vector, which is expressed as: S env =[C gas ,A sound ,D vision ,T touch ]; Among them, C gas Indicates the gas concentration value, A sound Indicates the intensity of the sound source, D vision Indicates the target distance, T touch Indicates the comprehensive value of contact pressure and temperature; The reinforcement learning algorithm unit adopts the Q-learning framework to construct the state space S=(S env ,L u ) and action space a = forward, backward, turn left, turn right, sit down, jump, respond to voice; The reinforcement learning algorithm updates the policy based on the following reward function: R(s,a)=α·U(L u )+β·E(S env ); Among them, U(L_u) is the user preference score, defined as: E(S env ) is the environmental adaptability score, defined as: Among them, α and β are weight coefficients respectively; The algorithm uses Q table update, the expression is: Where η is the learning rate, γ is the discount factor, and s' represents the next state; Output personalized behavior models to control the mechanical pet to exhibit differentiated response strategies in different scenarios.

3. The AI-driven fully perceptive simulated robotic pet system according to claim 2, wherein: The gas sensor array in the olfactory perception module includes multiple gas-sensitive elements, which correspond to the detection channels of ammonia, hydrogen sulfide, methane and volatile organic compounds (VOCs), specifically: Each gas sensor is made of metal oxide semiconductor material. When exposed to a specific gas, its resistance value R i Changes with gas concentration; Output voltage signal V i Obtained through the voltage divider circuit: Among them, R0 is a fixed resistor, V ref is the reference voltage; The ADC converter converts V i Convert to digital concentration value C i , the conversion formula is: C i =k i ·(V i -b i ); Among them, k i is the calibration coefficient, b i is the offset; The concentration value C of each channel i Input into the preset safety threshold judgment function: Among them, T j Indicates whether the jth gas component exceeds the safety threshold, C threshold,j Indicates the standard safety limit of the gas; If any T j If it is 1, a threat signal A is generated. s =∑T j And send it to the main control system through serial communication; The main control system is based on A s The size of the alarm triggers emergency avoidance actions or alarm prompts.

4. The AI-driven fully perceptive simulated robotic pet system according to claim 3, wherein: The microphone array in the auditory perception module consists of four equally spaced electret microphones, which are used to realize sound source localization and speech recognition functions, specifically including: The four audio signals collected by the microphone array are represented as x1(t), x2(t), x3(t), and x4(t); Perform short-time Fourier transform STFTx on each signal i (t)→X i (f,τ) obtains the frequency domain expression; Compute the time difference using the cross-correlation function: Among them, τ ij represents the time delay between the i-th and j-th microphones, x j represents a conjugate complex number; According to the time difference τ ij and the microphone spacing d, calculate the sound source direction angle θ, the expression is: Where c is the speed of sound in air; At the same time, the audio signal is sent to the speech recognition engine, which uses the Hidden Markov Model (HMM) for semantic analysis and matches the preset command keyword library; After a successful match, a corresponding command signal is generated and input into the AI ​​training module to trigger the corresponding behavior.

5. The AI-driven fully perceptive simulated robotic pet system according to claim 4, wherein: The spherical camera in the visual perception module consists of two wide-angle lenses, which are installed on the left and right sides of the mechanical pet's head to form a binocular stereo vision system. Specifically, it includes: The two images collected by the spherical camera are I L (x,y) and I R (x,y); Use SIFT algorithm to extract feature points and perform feature matching to obtain the corresponding pixel coordinate set (p L i,p R i); Calculate the disparity map D(x,y)=x L -x R ; Calculate the depth map using the camera intrinsic parameter matrix and baseline distance: Where Z represents the distance between the target point and the camera, f is the focal length, and b is the horizontal distance between the two cameras; Fusion of depth map and RGB image to generate 3D point cloud data; The point cloud data is processed using the YOLOv5 target detection algorithm to identify the target object category and location coordinates; Output target location information to the navigation and positioning module for path planning.

6. The AI-driven fully perceptive simulated robotic pet system according to claim 5, wherein: The navigation and positioning module uses a combination of an inertial measurement unit (IMU), a GPS module, and a visual SLAM algorithm to perform multi-sensor fusion positioning, specifically including: The IMU module outputs three-axis acceleration a x ,a y ,a z and the three-axis angular velocity ω x ,ω y ,ω z ; The GPS module provides longitude and latitude coordinates (lat, lon) and altitude h; The visual SLAM algorithm is based on ORB feature extraction and BA optimization method to build a local map and estimate the real-time pose T v ision=(x,y,θ); Multi-sensor fusion uses the extended Kalman filter EKF algorithm, and the state vector is defined as: X=[x,y,θ,v x ,v y ] T ; Among them, x, y are two-dimensional coordinates, θ is the heading angle, v x ,v y is the velocity component; The Kalman gain matrix K is calculated as follows: K=PH T (HPH T +R) -1 ; Where P is the state covariance matrix, H is the observation matrix, and R is the observation noise covariance; Output fused pose information X f used to control the mechanical pet to complete obstacle avoidance or guidance tasks.

7. The AI-driven fully perceptive simulated robotic pet system according to claim 6, wherein: The silicone surface of the touch simulation module is embedded with multiple pressure sensors and PTC heating sheets to simulate biological body temperature and contact feedback, specifically including: Each pressure sensor outputs a voltage signal V p After being processed by the amplifying circuit, it is converted into contact force F = k·V p , where k is the calibration coefficient; The temperature sensor outputs a resistance change ΔR, which is converted into a voltage signal V via a Wheatstone bridge circuit. T , and then convert it into temperature value T through table lookup method; The control system adjusts the power of the PTC heater according to the difference ΔT between T and the target body temperature: Where P is the heating power, K p , K i , K d are the PID controller parameters; When the pressure sensor detects that the contact force is greater than the set threshold F th When the electromagnetic vibrator is triggered, it generates feedback vibration, simulating the reaction of biological pets; All tactile data is uploaded to the AI ​​training module to enhance the emotional interaction ability of the mechanical pet.

8. The AI-driven fully perceptive simulated robotic pet system according to claim 7, wherein: The ball-and-socket joint in the motion control module consists of a central ball head, a circular track, three sets of servo motors and a harmonic reducer, and is used to achieve six degrees of freedom in three-dimensional space. Specifically, it includes: The central ball head of the ball-and-socket joint is fixed to the end of the mechanical limb and embedded in the circular track, allowing free rotation in the x, y, and z directions; Each rotational degree of freedom is driven by a servo motor through a harmonic reducer, forming three independent rotation axes: The first axis is the pitch angle θ1; The second axis is the yaw angle θ2; The third axis is the roll angle θ3; The output torque of each servo motor is calculated by the controller based on the inverse kinematics model: Among them, J(θ) is the Jacobian matrix, which represents the mapping relationship between the joint angle change and the end pose, and F ext is the external force vector, which comes from the pressure sensor feedback of the touch simulation module; The controller uses PID control strategy to adjust the servo motor angle in real time to achieve the target posture: Among them, u i (t) is the control input of the i-th motor, e i (t) represents the angle error, K p , K i , K d are the proportional, integral, and differential gain coefficients; When performing complex actions, the system generates smooth time series instructions through trajectory planning algorithms: θ(t)=θ0+(θ f -θ0)(1-cos(πt / T)) / 2; Among them, θ0 is the initial angle, θ f is the target angle, T is the action duration; The ball-and-socket joint, combined with the behavioral instructions output by the AI ​​training module, enables the mechanical pet to achieve highly agile movements.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the AI-driven fully perceptive simulated robot pet system according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the AI-driven fully perceptive simulated robot pet system according to any one of claims 1 to 8 are implemented.

Images