Exoskeleton control and interaction system and method for intelligent dolls

Abstract

This invention discloses an exoskeleton control and interaction system and method for intelligent dolls, relating to the field of artificial intelligence technology. The system includes a central control subsystem and an exoskeleton frame. The central control subsystem is responsible for AI interaction logic, motion planning, and multi-module collaboration. The exoskeleton frame is fixed to the intelligent doll via adjustable clamps, enabling motion actuation of the head, upper limbs, and torso. The central control subsystem possesses signal processing, instruction generation, and module collaboration functions, and consists of a processor, sensor module, drive module, communication module, power management module, and storage module. This invention effectively improves upon the static limitations of traditional BJD intelligent dolls and the excessive mechanization of existing interactive robots.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to an exoskeleton control and interaction system and method for intelligent androids. Background Technology

[0002] Ball-Jointed Dolls (BJDs) are widely used in collecting, photography, art exhibitions, and cultural creations due to their highly anthropomorphic appearance, adjustable ball joints, and replaceable head, hand, and clothing components. These intelligent dolls can be posed in different ways by adjusting the tightness of their joints and their angles; however, their movements are entirely dependent on manual adjustment, lacking automated capabilities and the ability to actively perform dynamic actions or create a natural, real-time interactive experience with the user. This limitation keeps BJDs largely at the level of "static display items," failing to meet current users' demands for immersive and intelligent experiences.

[0003] On the other hand, while existing interactive robots and intelligent dolls can achieve motion control through motors and servo motors, they generally suffer from the following problems: First, they are bulky in structure, usually using rigid mechanical frames and large drive units, resulting in a large overall size that is not suitable for combining with small and exquisite BJDs; second, their appearance has a strong mechanical feel, which cannot meet the high aesthetic requirements of BJD collectors and display users; and third, their interactive functions are limited, with most only able to execute preset actions or simple voice responses, lacking anthropomorphic emotional feedback and flexible behavior generation capabilities, and unable to deeply integrate with the dialogue, perception, and motion planning capabilities of modern artificial intelligence (AI) technology.

[0004] Furthermore, existing solutions typically require structural modifications to the internal structure of the intelligent doll, such as embedding drive mechanisms in the torso or joints. This not only undermines the original artistic and collectible value of BJDs but also increases costs and maintenance complexity, limiting their commercial applications. Especially in fields like exhibitions, educational companionship, and digital twin photography, where intelligent dolls must retain a realistic appearance and flexible design while simultaneously achieving dynamic interaction and AI-driven capabilities, no mature technology currently exists that can simultaneously meet these demands.

[0005] Therefore, there is an urgent need for an innovative system that does not require structural modifications to the BJD body, can be quickly assembled and disassembled, has a natural appearance, and can achieve dynamic movements and multimodal interactions through AI-driven design. This system would resolve the contradiction between the static limitations of traditional BJDs and the severe mechanization of existing interactive robots, and meet the market's demand for intelligent, immersive intelligent doll interaction. Summary of the Invention

[0006] The purpose of this invention is to provide an exoskeleton control and interaction system and method for intelligent androids, in order to solve the problems raised in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an exoskeleton control and interaction system for intelligent dolls, the system comprising a central control subsystem and an exoskeleton frame;

[0008] The central control subsystem is responsible for AI interaction logic, action planning, and multi-module collaboration.

[0009] The exoskeleton frame is fixed to the intelligent android by adjustable clamps, enabling motion drive of the head, upper limbs and torso.

[0010] The central control subsystem has signal processing, instruction generation, and module coordination functions, and consists of a processor, sensor module, driver module, communication module, power management module, and storage module.

[0011] The processor is selected from digital signal processors (DSPs) or ARM processors and uses an embedded system PC / 104. The processor is responsible for processing sensor data and performing motion planning and decision-making. Based on user input or AI-generated interactive logic, it coordinates and controls the movements of the exoskeleton.

[0012] The sensor module is used to collect the status information and external environment data of the intelligent puppet, providing a basis for control decisions; the sensor module includes a status sensing unit, a tactile sensing unit, an environmental sensing unit, and a data processing unit;

[0013] The state sensing unit consists of an accelerometer, a gyroscope, and a magnetometer;

[0014] The accelerometer, based on MEMS (Micro-Electro-Mechanical Systems) technology, detects acceleration using the principle of capacitance change. It measures the displacement of a mass block under inertial force and converts it into an electrical signal output, thus acquiring the linear acceleration data of the intelligent puppet. The gyroscope, based on the Coriolis effect, detects the vibration displacement of internal vibrating objects caused by the Coriolis force when the intelligent puppet rotates. The angular velocity is calculated by detecting this displacement. After fusing the two data sets, complementary filtering and Kalman filtering algorithms are used to eliminate noise interference and accurately calculate the intelligent puppet's attitude angles (pitch, roll, and yaw) and trajectory, achieving dynamic attitude stability control. The magnetometer determines the absolute orientation of the intelligent puppet by detecting the direction of the Earth's magnetic field. In complex electromagnetic environments, an adaptive Kalman filtering algorithm dynamically adjusts the weights to reduce magnetic field interference and ensure the accuracy of the intelligent puppet's orientation perception.

[0015] The tactile sensing unit includes a joint force sensor and a tactile sensor array; the joint force sensor is deployed at the exoskeleton joint and includes a variable-size sensor or a piezoresistive force sensor.

[0016] The strain gauge sensor is based on the piezoresistive effect. When the joint is subjected to force, the strain gauge attached to the elastic body deforms, causing a change in resistance. The change in resistance is converted into a voltage signal through a Wheatstone bridge circuit, thereby measuring the magnitude and direction of the force on the joint. The piezoresistive sensor utilizes the resistivity change characteristics of semiconductor materials under pressure to directly convert pressure into an electrical signal, realizing real-time monitoring of joint torque and contact force, and supporting the intelligent puppet to perform precise grasping, collision buffering and other motion control. The tactile sensor array is integrated on the surface of the intelligent puppet and uses piezoresistive materials and capacitive sensing technology. The capacitive tactile sensor consists of upper and lower electrodes and an elastic dielectric layer. When subjected to pressure, the change in the electrode spacing causes a change in capacitance. By scanning the capacitance change of each unit in the array, a pressure distribution image is generated, enabling the intelligent puppet to perceive the contact position, force and object contour, and realize human-like tactile feedback, such as simulating the force of a handshake and perceiving the texture of an object surface.

[0017] The environmental sensing unit includes a visual sensor and an environmental parameter sensor;

[0018] The visual sensors include RGB cameras, depth cameras, or LiDAR. The RGB cameras capture environmental images using a CMOS image sensor and utilize computer vision algorithms (such as YOLO and OpenCV libraries) for target recognition and scene classification. The depth cameras (such as ToF cameras) acquire 3D point cloud data of the environment by emitting infrared light and measuring the time-of-flight of reflected light, enabling obstacle detection, distance measurement, and intelligent android navigation and obstacle avoidance. The LiDAR emits a rotating laser beam and receives reflected signals to construct a high-precision environmental map, suitable for real-time perception in complex dynamic environments. The environmental parameter sensors include temperature and humidity sensors, gas sensors, and light sensors. The temperature and humidity sensors use digital temperature and humidity chips (such as DHT22) and sense environmental temperature and humidity through capacitive humidity sensing elements and thermistors. The gas sensors (such as the MQ series) utilize the conductivity changes of metal oxide semiconductors at different gas concentrations to detect harmful gases (such as CO2 and smoke). The light sensors sense ambient light intensity through photoresistors or silicon photodiodes to adjust the brightness of the intelligent android display or trigger specific scene responses (such as activating night mode).

[0019] Data collected by each sensor is transmitted via SPI and I 2 Transmitted to the processor via C or CAN bus.

[0020] The data processing unit employs multi-sensor data fusion technologies, such as distributed Kalman filtering and Bayesian networks, to synchronize heterogeneous data in time and fuse features, eliminating data conflicts and redundancy, and generating a unified environment and state perception model. Edge computing technology is used to complete data preprocessing (such as noise reduction and feature extraction) at sensor nodes, reducing data transmission volume and processor load, and improving the system's real-time response capability.

[0021] The drive module consists of a motor driver and a corresponding drive circuit; the drive module receives instructions from the processor to drive the joint movement of the exoskeleton frame and the movement of the micro-wheel chassis.

[0022] The communication module is responsible for data transmission between the smart doll and external devices. It uses wireless communication modules, such as Wi-Fi and Bluetooth, to connect the smart doll with the terminal device, making it convenient for users to remotely control the smart doll. It can also be used to download and update the motion library and obtain cloud AI services.

[0023] The power management module provides power to the central control subsystem and exoskeleton, and is responsible for power distribution, management and monitoring; lithium batteries or other materials can be used as power sources. The power management module must have overcharge and over-discharge protection functions to extend the power supply life and ensure the safe operation of the system.

[0024] The storage module is used to store pre-trained AI algorithm models, action libraries, user settings, and other data. For example, it stores different anthropomorphic action data so that it can be called according to different scenarios and interaction needs. It can also store the historical records of interactions between the intelligent puppet and the user to optimize the interactive experience.

[0025] The pre-trained AI algorithm model is a set of algorithm programs and parameters that support the three core functions of the exoskeleton control and interaction system for intelligent dolls: interaction logic parsing, action planning and decision-making, and multi-dimensional perception fusion. It needs to work in collaboration with the processor, sensor module, and driver module. Through the analysis and processing of intelligent doll state data, user interaction data, and external environment data, it generates action commands and interactive feedback that meet human-like requirements, and also has the ability to dynamically iterate and optimize based on interaction history.

[0026] The action library data in the storage module is stored according to a structured classification method to improve calling efficiency and adapt to different application scenarios; specifically, it can be divided into the following dimensions:

[0027] Basic motion library: contains basic body motion data such as nodding, waving, shaking head, and turning around. These motions serve as the basic units for the daily operation of the intelligent doll and can be retrieved through basic commands. For example, when a user is detected, the "waving" motion sequence can be invoked.

[0028] Emotional Expression Action Library: Combining the correspondence between facial expressions and body language in psychology, this library constructs action data that includes different emotions such as joy, anger, and sadness. For example, when the intelligent avatar recognizes a positive user review, it calls up the joyful action combination of "smiling + raising both hands"; when it perceives negative feedback, it presents a frustrated posture of "head down + arms hanging down".

[0029] Scene-specific action library: Preset specific action sequences for different scenarios such as meetings, teaching, and entertainment; for example, in meeting scenarios, store business etiquette actions such as "nodding to listen" and "spreading hands to explain"; in teaching scenarios, configure educational demonstration actions such as "finger writing on the blackboard" and "sideways demonstration" to achieve precise matching between the functions of the intelligent puppet and the needs of the scene;

[0030] Personalized Action Library: Users can define their own actions through a graphical programming interface or motion capture devices; the system binds the user's action data to their account, and automatically calls up their personalized action configuration when the user interacts with the intelligent avatar, enhancing the user's personalized interactive experience;

[0031] In terms of the invocation mechanism, the processor retrieves matching action data based on the current scene label, user commands, and AI analysis results. For example, when the intelligent avatar determines through sensors that it is in a social gathering scene and receives a "greet" command, the processor will prioritize retrieving appropriate action combinations such as waving and bowing from the social sub-library of the "scene-specific action library" for execution. Simultaneously, the system also has an action data optimization function, automatically fine-tuning action parameters based on feedback from the intelligent avatar's actual action execution to ensure natural and smooth movements.

[0032] The human-computer interaction module is used to realize the interaction between the intelligent puppet and the user, including voice recognition, receiving the user's voice commands, and making the intelligent puppet perform corresponding actions or responses; it can also be equipped with a display screen and indicator lights to display the status information or emotional feedback of the intelligent puppet, etc., to enhance the immersive experience of human-computer interaction;

[0033] The speech recognition system employs a three-tiered processing architecture: front-end noise reduction, keyword detection, and semantic understanding. The front-end utilizes beamforming and noise suppression algorithms to reduce the interference of environmental noise on the speech signal. Keyword detection captures key command words, reducing invalid speech processing. Semantic understanding combines natural language processing (NLP) technology to analyze the deeper intent behind the command. For example, when a user says "I'm a little cold," the intelligent assistant can not only recognize the literal meaning but also understand through semantic analysis that the user may need to close the window or raise the room temperature, and respond accordingly.

[0034] When a user issues a voice command, the smart puppet responds with voice, displays relevant information on the screen, flashes indicator lights to provide status prompts, and uses body movements to enhance the expression effect. For example, when a user asks "How is the weather today?", the smart puppet not only broadcasts the weather information, but also displays weather icons on the screen, and its head will turn slightly to point in the direction of the screen, achieving a multi-sensory interactive experience.

[0035] An emotional speech synthesis model is constructed to generate emotionally charged voice responses based on different interaction scenarios and user emotions. When a user asks a question in an anxious tone, the intelligent avatar answers in a soothing tone and at a slightly slower pace; in entertainment and interactive scenarios, a lively and playful voice style is adopted. In addition, the system also supports dialect recognition and response, automatically switching to the corresponding dialect voice library for users from different regions to enhance the sense of intimacy.

[0036] Through machine learning algorithms, the system continuously optimizes speech recognition and response capabilities; records user interactions and execution results, analyzes errors in instruction recognition or inappropriate responses, and automatically updates the speech recognition model and semantic understanding rules; it supports users to manually correct erroneous instructions, such as by providing feedback like "That's not what I meant, I meant to say..." The system adjusts its response strategy based on user feedback to continuously improve the accuracy of the interaction.

[0037] The exoskeleton frame is connected to the intelligent android using a magnetic quick-release clamp; the clamp body is composed of a metal magnetic component encased in medical-grade silicone.

[0038] The magnetic components calculate the magnetic force distribution at the joints of the smart doll and adhere it without exceeding the limits of the resin material. The thickness of the silicone outer layer is precisely controlled within a preset range, and the nano-texture design increases friction while preventing surface scratches. The clamp adopts a split nested structure, and the adapter module can be replaced according to the size of the BJD (such as 3-point or 6-point smart dolls), reducing the assembly time to within 30 seconds and leaving no residue after disassembly.

[0039] The exoskeleton joints utilize a miniature servo motor array, with each motor measuring ≤10mm×10mm×15mm and weighing only 5g, increasing power density to three times that of traditional motors. The motors achieve high torque output through harmonic reducers, with a reduction ratio of up to 50:1, ensuring the completion of complex movements such as 360° head rotation and 180° arm swing within a small space. The drive module employs distributed wiring, embedding the circuit board within the exoskeleton frame's interlayer, keeping the thickness within 2mm to avoid exposure and maintain aesthetics.

[0040] The exoskeleton frame is covered with a flexible biomimetic polyurethane film, which uses 3D texture printing technology to simulate the matte texture and fine pore structure of BJD skin, with a light transmittance of over 95%, achieving a visually "invisible" effect. The film integrates a microchannel heat dissipation system, which uses phase change material (PCM) to absorb the heat from the motors, avoiding the damage to the appearance caused by traditional heat dissipation holes. The film can be customized with the color of the BJD clothing to ensure that the exoskeleton blends seamlessly with the overall design.

[0041] A topology optimization algorithm was used to design the titanium alloy frame, reducing the frame wall thickness from 1.2mm to 0.8mm while maintaining strength, resulting in a 40% weight reduction. The joints utilize a flexible hinge structure made of carbon fiber reinforced resin (CFRP) and shape memory alloy wire. This structure ensures motion accuracy while mimicking the natural curves of human joints, allowing the exoskeleton to move with a smooth, anthropomorphic posture, completely eliminating the mechanical feel of traditional robots.

[0042] The exoskeleton has a built-in matrix of miniature sensors, including a 0.3mm thick flexible pressure sensor, a 2mm diameter miniature camera, and millimeter-wave radar. The sensors perceive the user's touch pressure, facial expressions, and spatial distance in real time, and perform fusion analysis through an edge AI processor. The BJD autonomously determines the interaction scenario. For example, when a user taps the smart doll's shoulder, the system identifies the pressure through the pressure sensor and combines it with the facial expressions captured by the camera to drive the smart doll to make a natural reaction of turning its head and smiling.

[0043] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention retains the original value of the intelligent doll: It adopts a non-invasive design, using a detachable exoskeleton frame and adjustable clamps to fix the intelligent doll without damaging the original ball joint structure and appearance of the BJD; This invention achieves anthropomorphic dynamic capabilities: With the help of the exoskeleton drive structure, the previously static BJD, which could only be manually placed, possesses the ability to move autonomously and perform multi-degree-of-freedom movements (such as head and upper limb movements), resulting in natural and smooth movements and enhancing the anthropomorphism of its dynamic performance; This invention integrates AI voice interaction, motion generation, and emotional expression technologies, combined with the logic processing of the central control system, allowing the intelligent doll to interact with the user in real-time and immersive, breaking through the limitations of traditional BJDs that can only be statically displayed; This invention has a flexible and easily expandable structure: The exoskeleton drive structure adopts a modular design, supporting quick assembly, disassembly, and replacement of different intelligent dolls; it can also connect to a cloud-based motion library, supporting OTA upgrades, facilitating functional expansion and motion updates, and adapting to diverse scenario needs. Attached Figure Description

[0044] Figure 1 The three-view diagram shows the structure of the exoskeleton control and interaction system for intelligent dolls according to the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Example: Figure 1 As shown, the present invention provides a technical solution for an exoskeleton control and interaction system for intelligent dolls, the system including a central control subsystem and an exoskeleton frame;

[0047] The central control subsystem is responsible for AI interaction logic, action planning, and multi-module collaboration.

[0048] The exoskeleton frame is fixed to the intelligent android by adjustable clamps, enabling motion drive of the head, upper limbs and torso.

[0049] The central control subsystem has signal processing, instruction generation, and module coordination functions, and consists of a processor, sensor module, driver module, communication module, power management module, and storage module.

[0050] The processor is selected from digital signal processors (DSPs) or ARM processors and uses an embedded system PC / 104. The processor is responsible for processing sensor data and performing motion planning and decision-making. Based on user input or AI-generated interactive logic, it coordinates and controls the movements of the exoskeleton.

[0051] The sensor module is used to collect the status information and external environment data of the intelligent puppet, providing a basis for control decisions; the sensor module includes a status sensing unit, a tactile sensing unit, an environmental sensing unit, and a data processing unit;

[0052] The state sensing unit consists of an accelerometer, a gyroscope, and a magnetometer;

[0053] The accelerometer, based on MEMS (Micro-Electro-Mechanical Systems) technology, detects acceleration using the principle of capacitance change. It measures the displacement of a mass block under inertial force and converts it into an electrical signal output, thus acquiring the linear acceleration data of the intelligent puppet. The gyroscope, based on the Coriolis effect, detects the vibration displacement of internal vibrating objects caused by the Coriolis force when the intelligent puppet rotates. The angular velocity is calculated by detecting this displacement. After fusing the two data sets, complementary filtering and Kalman filtering algorithms are used to eliminate noise interference and accurately calculate the intelligent puppet's attitude angles (pitch, roll, and yaw) and trajectory, achieving dynamic attitude stability control. The magnetometer determines the absolute orientation of the intelligent puppet by detecting the direction of the Earth's magnetic field. In complex electromagnetic environments, an adaptive Kalman filtering algorithm dynamically adjusts the weights to reduce magnetic field interference and ensure the accuracy of the intelligent puppet's orientation perception.

[0054] The tactile sensing unit includes a joint force sensor and a tactile sensor array; the joint force sensor is deployed at the exoskeleton joint and includes a variable-size sensor or a piezoresistive force sensor.

[0055] The strain gauge sensor is based on the piezoresistive effect. When the joint is subjected to force, the strain gauge attached to the elastic body deforms, causing a change in resistance. The change in resistance is converted into a voltage signal through a Wheatstone bridge circuit, thereby measuring the magnitude and direction of the force on the joint. The piezoresistive sensor utilizes the resistivity change characteristics of semiconductor materials under pressure to directly convert pressure into an electrical signal, realizing real-time monitoring of joint torque and contact force, and supporting the intelligent puppet to perform precise grasping, collision buffering and other motion control. The tactile sensor array is integrated on the surface of the intelligent puppet and uses piezoresistive materials and capacitive sensing technology. The capacitive tactile sensor consists of upper and lower electrodes and an elastic dielectric layer. When subjected to pressure, the change in the electrode spacing causes a change in capacitance. By scanning the capacitance change of each unit in the array, a pressure distribution image is generated, enabling the intelligent puppet to perceive the contact position, force and object contour, and realize human-like tactile feedback, such as simulating the force of a handshake and perceiving the texture of an object surface.

[0056] The environmental sensing unit includes a visual sensor and an environmental parameter sensor;

[0057] The visual sensors include RGB cameras, depth cameras, or LiDAR. The RGB cameras capture environmental images using a CMOS image sensor and utilize computer vision algorithms (such as YOLO and OpenCV libraries) for target recognition and scene classification. The depth cameras (such as ToF cameras) acquire 3D point cloud data of the environment by emitting infrared light and measuring the time-of-flight of reflected light, enabling obstacle detection, distance measurement, and intelligent android navigation and obstacle avoidance. The LiDAR emits a rotating laser beam and receives reflected signals to construct a high-precision environmental map, suitable for real-time perception in complex dynamic environments. The environmental parameter sensors include temperature and humidity sensors, gas sensors, and light sensors. The temperature and humidity sensors use digital temperature and humidity chips (such as DHT22) and sense environmental temperature and humidity through capacitive humidity sensing elements and thermistors. The gas sensors (such as the MQ series) utilize the conductivity changes of metal oxide semiconductors at different gas concentrations to detect harmful gases (such as CO2 and smoke). The light sensors sense ambient light intensity through photoresistors or silicon photodiodes to adjust the brightness of the intelligent android display or trigger specific scene responses (such as activating night mode).

[0058] Data collected by each sensor is transmitted via SPI and I 2 Transmitted to the processor via C or CAN bus.

[0059] The data processing unit employs multi-sensor data fusion technologies, such as distributed Kalman filtering and Bayesian networks, to synchronize heterogeneous data in time and fuse features, eliminating data conflicts and redundancy, and generating a unified environment and state perception model. Edge computing technology is used to complete data preprocessing (such as noise reduction and feature extraction) at sensor nodes, reducing data transmission volume and processor load, and improving the system's real-time response capability.

[0060] The drive module consists of a motor driver and a corresponding drive circuit; the drive module receives instructions from the processor to drive the joint movement of the exoskeleton frame and the movement of the micro-wheel chassis.

[0061] The communication module is responsible for data transmission between the smart doll and external devices. It uses wireless communication modules, such as Wi-Fi and Bluetooth, to connect the smart doll with the terminal device, making it convenient for users to remotely control the smart doll. It can also be used to download and update the motion library and obtain cloud AI services.

[0062] The power management module provides power to the central control subsystem and exoskeleton, and is responsible for power distribution, management and monitoring; lithium batteries or other materials can be used as power sources. The power management module must have overcharge and over-discharge protection functions to extend the power supply life and ensure the safe operation of the system.

[0063] The storage module is used to store pre-trained AI algorithm models, action libraries, user settings, and other data. For example, it stores different anthropomorphic action data so that it can be called according to different scenarios and interaction needs. It can also store the historical records of interactions between the intelligent puppet and the user to optimize the interactive experience.

[0064] The pre-trained AI algorithm model is a set of algorithm programs and parameters that support the three core functions of the exoskeleton control and interaction system for intelligent dolls: interaction logic parsing, action planning and decision-making, and multi-dimensional perception fusion. It needs to work in collaboration with the processor, sensor module, and driver module. Through the analysis and processing of intelligent doll state data, user interaction data, and external environment data, it generates action commands and interactive feedback that meet human-like requirements, and also has the ability to dynamically iterate and optimize based on interaction history.

[0065] The action library data in the storage module is stored according to a structured classification method to improve calling efficiency and adapt to different application scenarios; specifically, it can be divided into the following dimensions:

[0066] Basic motion library: contains basic body motion data such as nodding, waving, shaking head, and turning around. These motions serve as the basic units for the daily operation of the intelligent doll and can be retrieved through basic commands. For example, when a user is detected, the "waving" motion sequence can be invoked.

[0067] Emotional Expression Action Library: Combining the correspondence between facial expressions and body language in psychology, this library constructs action data that includes different emotions such as joy, anger, and sadness. For example, when the intelligent avatar recognizes a positive user review, it calls up the joyful action combination of "smiling + raising both hands"; when it perceives negative feedback, it presents a frustrated posture of "head down + arms hanging down".

[0068] Scene-specific action library: Preset specific action sequences for different scenarios such as meetings, teaching, and entertainment; for example, in meeting scenarios, store business etiquette actions such as "nodding to listen" and "spreading hands to explain"; in teaching scenarios, configure educational demonstration actions such as "finger writing on the blackboard" and "sideways demonstration" to achieve precise matching between the functions of the intelligent puppet and the needs of the scene;

[0069] Personalized Action Library: Users can define their own actions through a graphical programming interface or motion capture devices; the system binds the user's action data to their account, and automatically calls up their personalized action configuration when the user interacts with the intelligent avatar, enhancing the user's personalized interactive experience;

[0070] In terms of the invocation mechanism, the processor retrieves matching action data based on the current scene label, user commands, and AI analysis results. For example, when the intelligent avatar determines through sensors that it is in a social gathering scene and receives a "greet" command, the processor will prioritize retrieving appropriate action combinations such as waving and bowing from the social sub-library of the "scene-specific action library" for execution. Simultaneously, the system also has an action data optimization function, automatically fine-tuning action parameters based on feedback from the intelligent avatar's actual action execution to ensure natural and smooth movements.

[0071] The human-computer interaction module is used to realize the interaction between the intelligent puppet and the user, including voice recognition, receiving the user's voice commands, and making the intelligent puppet perform corresponding actions or responses; it can also be equipped with a display screen and indicator lights to display the status information or emotional feedback of the intelligent puppet, etc., to enhance the immersive experience of human-computer interaction;

[0072] The speech recognition system employs a three-tiered processing architecture: front-end noise reduction, keyword detection, and semantic understanding. The front-end utilizes beamforming and noise suppression algorithms to reduce the interference of environmental noise on the speech signal. Keyword detection captures key command words, reducing invalid speech processing. Semantic understanding combines natural language processing (NLP) technology to analyze the deeper intent behind the command. For example, when a user says "I'm a little cold," the intelligent assistant can not only recognize the literal meaning but also understand through semantic analysis that the user may need to close the window or raise the room temperature, and respond accordingly.

[0073] When a user issues a voice command, the smart puppet responds with voice, displays relevant information on the screen, flashes indicator lights to provide status prompts, and uses body movements to enhance the expression effect. For example, when a user asks "How is the weather today?", the smart puppet not only broadcasts the weather information, but also displays weather icons on the screen, and its head will turn slightly to point in the direction of the screen, achieving a multi-sensory interactive experience.

[0074] An emotional speech synthesis model is constructed to generate emotionally charged voice responses based on different interaction scenarios and user emotions. When a user asks a question in an anxious tone, the intelligent avatar answers in a soothing tone and at a slightly slower pace; in entertainment and interactive scenarios, a lively and playful voice style is adopted. In addition, the system also supports dialect recognition and response, automatically switching to the corresponding dialect voice library for users from different regions to enhance the sense of intimacy.

[0075] Through machine learning algorithms, the system continuously optimizes speech recognition and response capabilities; records user interactions and execution results, analyzes errors in instruction recognition or inappropriate responses, and automatically updates the speech recognition model and semantic understanding rules; it supports users to manually correct erroneous instructions, such as by providing feedback like "That's not what I meant, I meant to say..." The system adjusts its response strategy based on user feedback to continuously improve the accuracy of the interaction.

[0076] The exoskeleton frame is connected to the intelligent android using a magnetic quick-release clamp; the clamp body is composed of a metal magnetic component encased in medical-grade silicone.

[0077] The magnetic components calculate the magnetic force distribution at the joints of the smart doll and adhere it without exceeding the limits of the resin material. The thickness of the silicone outer layer is precisely controlled within a preset range, and the nano-texture design increases friction while preventing surface scratches. The clamp adopts a split nested structure, and the adapter module can be replaced according to the size of the BJD (such as 3-point or 6-point smart dolls), reducing the assembly time to within 30 seconds and leaving no residue after disassembly.

[0078] The exoskeleton joints utilize a miniature servo motor array, with each motor measuring ≤10mm×10mm×15mm and weighing only 5g, increasing power density to three times that of traditional motors. The motors achieve high torque output through harmonic reducers, with a reduction ratio of up to 50:1, ensuring the completion of complex movements such as 360° head rotation and 180° arm swing within a small space. The drive module employs distributed wiring, embedding the circuit board within the exoskeleton frame's interlayer, keeping the thickness within 2mm to avoid exposure and maintain aesthetics.

[0079] The exoskeleton frame is covered with a flexible biomimetic polyurethane film, which uses 3D texture printing technology to simulate the matte texture and fine pore structure of BJD skin, with a light transmittance of over 95%, achieving a visually "invisible" effect. The film integrates a microchannel heat dissipation system, which uses phase change material (PCM) to absorb the heat from the motors, avoiding the damage to the appearance caused by traditional heat dissipation holes. The film can be customized with the color of the BJD clothing to ensure that the exoskeleton blends seamlessly with the overall design.

[0080] A topology optimization algorithm was used to design the titanium alloy frame, reducing the frame wall thickness from 1.2mm to 0.8mm while maintaining strength, resulting in a 40% weight reduction. The joints utilize a flexible hinge structure made of carbon fiber reinforced resin (CFRP) and shape memory alloy wire. This structure ensures motion accuracy while mimicking the natural curves of human joints, allowing the exoskeleton to move with a smooth, anthropomorphic posture, completely eliminating the mechanical feel of traditional robots.

[0081] The exoskeleton has a built-in matrix of miniature sensors, including a 0.3mm thick flexible pressure sensor, a 2mm diameter miniature camera, and millimeter-wave radar. The sensors perceive the user's touch pressure, facial expressions, and spatial distance in real time, and perform fusion analysis through an edge AI processor. The BJD autonomously determines the interaction scenario. For example, when a user taps the smart doll's shoulder, the system identifies the pressure through the pressure sensor and combines it with the facial expressions captured by the camera to drive the smart doll to make a natural reaction of turning its head and smiling.

[0082] An exoskeleton control and interaction method for intelligent androids is proposed, which is applied to an exoskeleton control and interaction system for intelligent androids.

[0083] In this embodiment, a 6-point resin BJD doll (approximately 25cm tall, with custom surface coating) is designed for home desktop companionship and small-scale social interaction, achieving "human-like movements + natural interaction" while meeting the following requirements: total exoskeleton weight ≤ 500g, does not damage the doll's coating, is visually invisible, and strictly uses specified materials to ensure lightweight, high strength, and adaptability.

[0084] Central Control Subsystem: Its core is a DSP processor (embedded PC / 104 architecture), which coordinates three main tasks:

[0085] Process sensor data (attitude, touch, environment), calculate the puppet's attitude angle using accelerometer + gyroscope + magnetometer, and use complementary filtering to reduce noise;

[0086] Based on user instructions (such as "greet") or AI logic, call the action library to generate instructions;

[0087] It coordinates the driving, communication, and storage modules, such as obtaining weather data from the cloud via Wi-Fi, and then driving actions and voice responses;

[0088] Key configurations of the sensor module: joint force sensor (strain gauge at shoulder / elbow) to measure force, and capacitive tactile array for sensing touch.

[0089] Head-mounted RGB camera (YOLO algorithm for face recognition), chest-mounted ToF camera for obstacle avoidance, and temperature and humidity sensor for adjusting screen brightness;

[0090] Exoskeleton frame (including selection of specified materials):

[0091] Drive joint connector: The connecting shaft (2mm in diameter) is made of titanium alloy (TC4), which has high strength and low density (4.5g / cm³). 3 This ensures the accuracy of the head's 360° rotation and the arm's 180° swing, preventing deformation over a long period.

[0092] Transmission gears / connecting rods: Made of polyoxymethylene (POM), which is self-lubricating, has a low coefficient of friction (0.1-0.3), produces noiseless transmission, and is lightweight (1.4g / cm²). 3 It does not increase the driving burden;

[0093] The main frame is made of titanium alloy with topology optimization, reducing the wall thickness from 1.2mm to 0.8mm and reducing weight by 40%. The joints use carbon fiber reinforced resin (CFRP) flexible hinges to simulate the curves of the human body.

[0094] Clamping and cushioning (damage-proof doll):

[0095] Clamping contact surface: Shore 30-50A silicone textured pad (2mm thick), soft and elastic to fit the curved surface of the doll, does not react with resin / coating, and can be disassembled and reassembled 1000 times without scratching.

[0096] Adjustment mechanism: Elastic PU foam (density 0.1-0.3 g / cm³) 3 The thickness (1-3mm) is suitable for 6 / 4 point figure figures, and it can buffer pressure (≤2N).

[0097] Chassis: ABS engineering plastic (1.05g / cm³) 3 ), with a load-bearing capacity of ≥1kg, drop-proof and waterproof (IP54), supporting the doll + exoskeleton;

[0098] Surface: 3D printed flexible polyurethane film (0.3mm thick), imitating BJD skin, 95% light transmittance and "invisible", with built-in phase change material (PCM) for heat dissipation, and no holes to maintain aesthetics;

[0099] Built-in sensors: 0.3mm flexible pressure-sensitive touch on the shoulder, 2mm miniature camera on the head to capture facial expressions, and millimeter-wave radar on the chest to measure distance;

[0100] Quick release and drive:

[0101] Magnetic quick-release clamp: Silicone-coated titanium alloy magnetic components (magnetic force ≤5N, below the resin limit), 30-second assembly and disassembly, and compatible with 3 / 6 point figure by changing the adapter module;

[0102] Drive: 6-channel micro servo motors (10mm×10mm×15mm, 5g), equipped with harmonic reducers (50:1), sufficient torque, circuit board embedded in frame sandwich (thickness ≤2mm) without being exposed;

[0103] Human-computer interaction:

[0104] Voice interaction: 2-channel microphone array + chip, three-level processing (noise reduction → wake-up → semantic understanding), supports dialects (Cantonese / Sichuan dialect), emotion synthesis (slow and soothing when anxious, lively when entertaining), and can learn from user corrections and optimizations;

[0105] Visual and motor coordination: The 1.2-inch OLED screen on the chest displays the weather / mood, and the three RGB lights on the head indicate the status (blue for standby, green for interaction, and red for low battery); for example, when a user asks "weather", the voice reports the data, the screen displays a sun icon, the head turns the screen in the correct direction, and the user points to the screen.

[0106] Workflow example:

[0107] When a user approaches, the camera recognizes their face, and AI triggers a "greeting" message.

[0108] The processor calls the social motion library to drive titanium alloy joints and POM gears to achieve arm swing (±30°) and head point (±15°);

[0109] When a user says "dance," the voice is analyzed, and an entertainment motion library is invoked. The motor drives the arm / torso to move, and PU foam cushions the mascot to prevent injury from being crushed.

[0110] Slow motion is detected by sensors, and AI fine-tunes motor parameters to ensure smooth operation.

[0111] When the user leaves, the radar detects a distance greater than 1m, the avatar returns to standby mode, and the indicator light turns blue.

[0112] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An exoskeleton control and interaction system for a smart doll, characterized by: The system includes a central control subsystem and an exoskeleton frame; The central control subsystem is responsible for AI interaction logic, action planning, and multi-module collaboration. The exoskeleton frame is fixed to the intelligent android by adjustable clamps, enabling motion drive of the head, upper limbs and torso. The exoskeleton frame is connected to the intelligent android using a magnetic quick-release clamp; the clamp body is composed of a metal magnetic component encased in medical-grade silicone. The magnetic component calculates the magnetic force distribution at the joints of the smart doll and performs adsorption without exceeding the limits of the resin material of the smart doll; the thickness of the silicone outer layer is precisely controlled within a preset range; the clamp adopts a split nested structure, and the adapter module can be replaced according to the body shape of the BJD. The exoskeleton frame is covered with a flexible biomimetic polyurethane film, which uses 3D texture printing technology to simulate the matte texture and fine pore structure of BJD skin; the film integrates a microchannel heat dissipation system, which uses phase change material PCM to absorb the heat from the motor; the exoskeleton has a built-in micro sensor matrix. The central control subsystem has signal processing, instruction generation, and module coordination functions, and consists of a processor, sensor module, driver module, communication module, power management module, and storage module. The storage module is used to store pre-trained AI algorithm models and action library data; The action library data in the storage module is stored in a structured classification manner, divided into the following dimensions: basic action library, emotion expression action library, scene-specific action library, and personalized action library.

2. The smart doll-oriented exoskeleton control and interaction system according to claim 1, characterized in that: The processor is a digital signal processor (DSP) or an ARM processor, using an embedded system PC / 104. The processor is responsible for processing sensor data and performing motion planning and decision-making, coordinating and controlling the exoskeleton's movements based on user input or AI-generated interactive logic.

3. The smart doll-oriented exoskeleton control and interaction system of claim 2, wherein: The sensor module is used to collect the status information and external environment data of the intelligent puppet, providing a basis for control decisions; the sensor module includes a status sensing unit, a tactile sensing unit, an environmental sensing unit, and a data processing unit; The state sensing unit consists of an accelerometer, a gyroscope, and a magnetometer; The accelerometer is based on MEMS (Micro-Electro-Mechanical Systems) technology and uses the principle of capacitance change to detect acceleration. It measures the displacement of the mass block under the action of inertial force and converts it into an electrical signal output to obtain the linear acceleration data of the intelligent puppet. The gyroscope is based on the Coriolis effect. When the intelligent puppet rotates, the internal vibrating object is affected by the Coriolis force and produces vibration displacement. The angular velocity is calculated by detecting the displacement. The magnetometer determines the absolute location of the intelligent doll by detecting the direction of the geomagnetic field. The tactile sensing unit includes a joint force sensor and a tactile sensor array; The joint force sensor is deployed at the exoskeleton joint and includes a variable plate sensor or a piezoresistive force sensor. The variable-strain sensor is based on the piezoresistive effect. When the joint is subjected to force, the strain gauge attached to the elastic body deforms, causing a change in resistance. The change in resistance is converted into a voltage signal through a Wheatstone bridge circuit, thereby measuring the magnitude and direction of the force on the joint. The piezoresistive force sensor utilizes the resistivity change characteristics of semiconductor materials under pressure to directly convert pressure into an electrical signal, realizing real-time monitoring of joint torque and contact force. The tactile sensor array is integrated on the surface of the smart doll and uses piezoresistive materials and capacitive sensing technology. A capacitive tactile sensor consists of upper and lower electrodes and an elastic dielectric layer. When pressure is applied, the change in electrode spacing causes a change in capacitance. By scanning the capacitance changes of each unit in the array, a pressure distribution image is generated, enabling the intelligent dummy to perceive the contact position, force, and object outline. The environmental sensing unit includes a visual sensor and an environmental parameter sensor; The visual sensors include an RGB camera, a depth camera, or a LiDAR. The RGB camera captures environmental images using a CMOS image sensor and performs target recognition and scene classification using computer vision algorithms. The depth camera acquires 3D point cloud data of the environment by emitting infrared light and measuring the time-of-flight of reflected light, enabling obstacle detection, distance measurement, and intelligent robot navigation and obstacle avoidance. The LiDAR emits a rotating laser beam and receives reflected signals to construct a high-precision environmental map. The environmental parameter sensors include a temperature and humidity sensor, a gas sensor, and a light sensor. The temperature and humidity sensor uses a digital temperature and humidity chip and senses environmental temperature and humidity through a capacitive humidity sensing element and a thermistor. The gas sensor detects harmful gases by utilizing the conductivity changes of metal oxide semiconductors at different gas concentrations. The light sensor senses ambient light intensity through a photoresistor or silicon photovoltaic cell. The data collected by each sensor is transmitted to the processor; The data processing unit employs multi-sensor data fusion technology to synchronize and fuse heterogeneous data in time and features, generating a unified environment and state perception model. Edge computing technology is used to perform data preprocessing at sensor nodes.

4. The smart doll-oriented exoskeleton control and interaction system of claim 3, wherein: The drive module consists of a motor driver and a corresponding drive circuit; the drive module receives instructions from the processor to drive the joint movement of the exoskeleton frame and the movement of the micro-wheeled chassis.

5. The smart doll oriented exoskeleton control and interaction system of claim 4, wherein: The communication module is responsible for data transmission between the intelligent doll and external devices, and uses a wireless communication module to achieve connection with the terminal device.

6. The smart doll-oriented exoskeleton control and interaction system of claim 5, wherein: The power management module provides power to the central control subsystem and the exoskeleton, and is responsible for the distribution, management and monitoring of power.

7. The exoskeleton control and interaction system for intelligent dolls according to claim 6, characterized in that: The storage module is used to store pre-trained AI algorithm models and action library data; The action library data in the storage module is stored according to a structured classification method, divided into the following dimensions: Basic Movement Library: Contains basic limb movement data; Emotion Expression Action Database: Combining the psychological correspondence between facial expressions and body language, this database constructs action data containing different emotions; Scene-specific action library: Presets specific action sequences for different scenes; Personalized motion library: Users can define their own motions by entering them through a graphical programming interface or motion capture device; the system binds the user's motion data to their account, and automatically calls up their personalized motion configuration when the user interacts with the intelligent avatar. In terms of the invocation mechanism, the processor retrieves matching action data based on the current scene label, user instructions, and AI analysis results.

8. The smart doll-oriented exoskeleton control and interaction system of claim 7, wherein: The human-computer interaction module is used to enable interaction between the intelligent avatar and the user, including voice recognition, display screen and indicator lights; The speech recognition system employs a three-tiered processing architecture: front-end noise reduction, keyword detection, and semantic understanding. The front-end utilizes beamforming and noise suppression algorithms to reduce the interference of environmental noise on the speech signal; keyword detection captures key command words. Semantic understanding, combined with natural language processing technology, is used to interpret the intent of instructions. When the user issues a voice command, the intelligent puppet responds with voice, displays relevant information on the screen, flashes indicator lights to provide status prompts, and also coordinates with body movements. Construct an emotional speech synthesis model to generate emotionally charged speech responses based on different interaction scenarios and user emotions; Through machine learning algorithms, the system continuously optimizes speech recognition and response capabilities; records user interactions and execution results, analyzes errors in instruction recognition or inappropriate responses, and automatically updates the speech recognition model and semantic understanding rules; it supports users in manually correcting erroneous instructions, and the system adjusts its response strategy based on user feedback.

9. The smart doll oriented exoskeleton control and interaction system of claim 8, wherein: The exoskeleton joints utilize a miniature servo motor array; the motors achieve high torque output through harmonic reducers. The exoskeleton has a built-in matrix of miniature sensors; the sensors perceive the user's touch pressure, facial expressions, and spatial distance in real time, and perform fusion analysis through an edge AI processor, allowing the BJD to autonomously determine the interaction scenario.

10. A method for exoskeleton control and interaction with an intelligent mannequin, characterized in that: This method is applied to the exoskeleton control and interaction system for intelligent dolls as described in any one of claims 1-9.

Images