A machine pet suitable for the elderly and a system thereof

Abstract

The application discloses an old-age adaptation robot pet and a system thereof, and belongs to the technical field of robots. The system comprises a pet robot body, a multi-modal sensing module, a central control unit, an intelligent interaction module, an autonomous navigation module, a health management and emergency response module, and a cloud collaborative platform. The robot body is covered with a waterproof short fluff layer and integrates visual, auditory, tactile and vital sign monitoring units. The central control unit is embedded with a large model and dynamically adjusts the behavior mode according to the user portrait to realize personalized companionship. The intelligent interaction module supports dialect adaptation and emotional response. The autonomous navigation module realizes automatic following, intelligent charging scheduling and lost handling. The health management module realizes fall detection and graded alarm. The cloud platform supports data encryption, remote upgrading and multi-user collaborative management. The application builds an efficient and safe old-age adaptation service system through adaptive interaction and full-scene health monitoring.

Description

Technical Field

[0001] This invention relates to the field of robotics, and more particularly to an age-friendly robotic pet and its system. Background Technology

[0002] With the accelerating aging of the population, the issues of daily care and emotional support for empty-nest elderly and very elderly people are becoming increasingly prominent. Traditional monitoring methods rely on human caregivers or fixed monitoring equipment, which have limitations such as high cost, poor interactivity, and inability to move and follow. Existing pet-type companion robots mostly use simple sensor combinations and preset behavior patterns to achieve basic human-computer interaction, but they still have significant shortcomings in personalized adaptation, accuracy of health monitoring, intelligent emergency response, and multi-scenario collaboration.

[0003] Existing technologies cannot dynamically adjust behavior patterns based on the elderly’s lifestyle, health status, and personality traits, making it difficult to achieve personalized companionship tailored to each individual. In terms of health monitoring, there is a lack of multimodal fusion-based anomaly identification and tiered alarm mechanisms. In terms of emergency response, a closed-loop management system involving multiple parties has not been established. In terms of interactive experience, there is a lack of effective support for dialects commonly used by the elderly. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide an age-appropriate robotic pet and its system to solve the problems of semantic analysis, automated prediction, and closed-loop management throughout the entire process.

[0005] Technical solution: An age-friendly robotic pet and its system, comprising: It includes a pet-type robot body, a multimodal perception module, a central control unit, an intelligent interaction module, an autonomous navigation module, a health management and emergency response module, and a cloud-based collaborative platform.

[0006] The pet-shaped robot body is covered with a waterproof short plush layer, which has waterproof, impact-resistant and lightweight characteristics. The weight of the whole machine is controlled between 5kg and 20kg depending on the pet model.

[0007] The multimodal perception module integrates a visual perception unit (binocular camera, infrared imaging sensor), an auditory perception unit (microphone array, dialect recognition model (the dialect recognition model can be fine-tuned and trained based on the PaddleSpeech framework for dialects such as Cantonese and Sichuanese to achieve adaptive learning)), a tactile perception unit (pressure sensor, touch sensing layer), and a vital signs monitoring unit (contact blood pressure and blood oxygen sensor, non-contact respiratory and body temperature sensor), enabling multi-dimensional real-time perception of people, environment, and interaction.

[0008] The central control unit embeds a large model (Llama 3.2 lightweight language model) decision-making system, receives multi-modal perception data, conducts fusion analysis, constructs user profiles, and dynamically adjusts the robot's behavior patterns through behavioral decision-making algorithms. The behavior patterns include an intimate mode, a following mode, a silent mode, and a coquettish mode, achieving "one-to-one" adaptive interaction based on user personality and status.

[0009] The intelligent interaction module includes a dialect-adaptive voice interaction sub-module and an emotional feedback sub-module, supports multi-local dialect recognition and learning, can recognize user emotions based on voice intonation and facial expressions, and generate corresponding emotional responses. In addition, this module also supports the recognition and execution of commands for picking up items.

[0010] The autonomous navigation module integrates SLAM mapping, path planning, obstacle avoidance algorithms, and human pose recognition technologies to achieve automatic following and autonomous return in complex environments. The automatic charging unit intelligently schedules charging strategies according to the user's self-care ability and daily routine, supports multi-charging point management, and continuously runs a health monitoring function during the charging process.

[0011] The health management and emergency response module is based on a hierarchical alarm mechanism to achieve fall detection, abnormal vital sign recognition, and multi-party linkage alarms. According to the elderly's consciousness state and vital sign data, execute first-level, second-level, or third-level alarm strategies, and simultaneously activate video transmission and audible and visual alarms.

[0012] The cloud collaborative platform adopts an "end-edge-cloud" collaborative architecture, responsible for data encrypted storage, large model training and remote upgrade, multi-user collaborative management, and supports hierarchical access by the elderly, their children, community service providers, and platform operators according to their permissions.

[0013] Beneficial effects: 1. Through multi-modal perception and adaptive behavior regulation, a personalized companionship experience is achieved, enabling the robotic pet to dynamically adjust the interaction method according to the user's habits and preferences; 2. Construct a full-scenario health monitoring and hierarchical alarm closed-loop to improve the accuracy of abnormal recognition and the timeliness of emergency response; 3. The intelligent charging scheduling strategy takes into account both device battery life and user experience, avoiding interference with the elderly's life during the charging process; 4. Dialect-adaptive voice interaction improves the convenience and acceptance of use for elderly users; 5. The cloud collaborative platform realizes secure data storage and continuous model optimization, supports multi-party collaborative management, and constructs a complete aging-friendly service system. Brief Description of the Drawings

[0014] Figure 1 is a schematic diagram of the working process of the present invention. Detailed Embodiments

[0015] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] like Figure 1 As shown, an age-friendly robotic pet and its system include: a pet-type robot body, a multimodal perception module, a central control unit, an intelligent interaction module, an autonomous navigation module, a health management and emergency response module, and a cloud-based collaborative platform.

[0017] Example 1: System Hardware Architecture This embodiment provides a hardware implementation scheme for an age-friendly robotic pet. The robot body adopts a biomimetic design, with its shape modeled after a pet dog. The exterior is covered with a waterproof short plush layer made of hydrophobic material, possessing waterproof and easy-to-clean properties, allowing the elderly to easily wipe and clean it. The total weight is controlled at 8kg (medium-sized pet model), and the outer shell is made of lightweight, high-strength material, capable of withstanding a certain amount of impact without damaging the internal components.

[0018] The internal hardware architecture includes: Main control board: Equipped with a high-performance embedded AI chip, supporting local inference operations for large models, with a main frequency of over 2.0GHz, 4GB of memory and 64GB of storage space.

[0019] Sensor system: 1) Vision Unit: A binocular camera is installed in the robot's eye area, with a resolution of 1080P and supporting infrared night vision; an infrared imaging sensor is installed in the forehead area for body temperature monitoring and body posture recognition. 2) Hearing unit: A circular microphone array is mounted on the head to support sound source localization and noise suppression; 3) Tactile unit: Pressure sensors are distributed on the back, abdomen and limbs, which can sense interactive actions such as stroking and patting; 4) Vital signs monitoring unit: The contact sensor is integrated into the back stroking area, so that blood pressure and blood oxygen can be measured without the elderly touching the pet; the non-contact respiratory monitoring module is based on millimeter-wave radar technology and can monitor respiratory rate within a certain distance.

[0020] The actuators include a quadrupedal walking mechanism, a head turning mechanism, a tail swinging mechanism, and a mouth grasping mechanism. Each mechanism is driven by a servo motor and has flexible movement and operation capabilities.

[0021] Communication module: Supports 4G / 5G cellular networks, Wi-Fi and Bluetooth communication to ensure network connectivity in various environments.

[0022] Power system: It adopts a high-energy-density lithium battery pack with a capacity of 200Wh, which can support continuous operation for more than 8 hours; it is equipped with a wireless charging receiver module and supports automatic recharging.

[0023] Example 2: Adaptive Interactive Behavior Control This embodiment details how the central control unit achieves personalized adaptive interactive behavior control.

[0024] User profile building: When the robot first interacts with a user, a user identity profile is established through facial recognition. During the initial interaction phase, the robot operates in default mode while recording user interaction data, including the frequency of touching, the number of voice interactions, reactions when followed, and tolerance for the robot's approach. Analysis of the interaction data of the female user, Zhang San, revealed that she touches the robot an average of 8 times a day, enjoys hugging the robot, and responds positively to the robot's affectionate behavior. Analysis of the interaction data of the male user, Li Si, revealed that he exhibits avoidance behavior towards the robot's approach, interacts with the robot less frequently, and prefers the robot to follow him from a distance.

[0025] Behavioral pattern matching: Based on the above user profile, the behavioral decision-making submodule assigns a "closeness mode" weight of 0.8 to the female host Zhang San, and the robot will increase the frequency of triggering coquettish behavior, actively approach the user to seek interaction, and actively accompany the user when the user is sitting quietly; assigns a "silent mode" weight of 0.7 to the male host Li Si, and the robot will maintain a following distance of more than 1.5 meters, reduce active interaction, and only approach the user when abnormal situations are detected.

[0026] Dynamic Optimization: The self-learning optimization submodule employs a reinforcement learning algorithm, using user feedback as a reward signal to continuously optimize behavioral strategies. For example, when the robot approaches a user in a certain way, if the user actively touches the robot, it is considered positive feedback, increasing the weight of that behavioral pattern; if the user turns away or issues a rejection command, it is considered negative feedback, reducing the trigger probability of that behavioral pattern. Through continuous learning, the robot's behavioral patterns will increasingly align with the user's personalized preferences.

[0027] Adaptation to user sleep patterns: The system learns from long-term observation of users' sleep patterns. For example, if it detects that a user usually goes to bed around 10:00 PM, the robot will reduce its activity volume and proactive interactions after 10:30 PM, and schedule automatic charging during this period to avoid making noise or disturbing the user during rest time.

[0028] Example 3: Tiered Health Monitoring and Alarm Process This embodiment details the workflow of the health management and emergency response module.

[0029] Routine monitoring mode: The robot continuously performs non-contact health monitoring while following or accompanying the user. An infrared imaging sensor collects the user's body temperature data every 5 minutes, a respiratory monitoring module tracks changes in respiratory rate in real time, and a posture recognition algorithm continuously analyzes the user's posture. All data is processed locally, with normal data summarized and uploaded to the cloud daily, and abnormal data triggering real-time review.

[0030] Fall detection and graded alarm: Scenario 1: An elderly person suddenly falls while moving around in the living room. The posture recognition algorithm detects a change in posture from standing to lying down accompanied by an acceleration impact signal, triggering a fall event. The robot immediately approaches the elderly person and initiates a voice inquiry: "Did you fall? Do you need help contacting your family?" The elderly person answers "Yes." The robot initiates the first-level alarm process: automatically dialing the emergency contact (daughter) and initiating a two-way video call, allowing the daughter to understand the situation and speak directly with the elderly person.

[0031] 2) Scenario Two: The elderly person falls and becomes unresponsive. The robot repeatedly asks three times but detects no voice response and observes no movement of the elderly person's limbs. The system initiates a vital signs check: infrared imaging detects a body temperature of 36.5℃ (normal), and respiratory monitoring shows 16 breaths per minute (normal). The system is determined to be "unconscious but with stable vital signs" and enters the level-two alarm process: it automatically dials the emergency contact number; if no one answers after three calls, it automatically transfers the call to the platform's customer service center, and simultaneously transmits encrypted video of the scene to the platform for intervention.

[0032] 3) Scenario 3: After an elderly person falls and becomes unresponsive, vital signs monitoring shows a rapid drop in body temperature (from 36.5℃ to 35.2℃) and a respiratory rate decreasing to 8 breaths per minute and gradually weakening. This is determined to be "loss of consciousness and unstable vital signs," triggering a Level 3 alarm process: The robot simultaneously dials the emergency contact number and the platform's emergency number; it activates a high-decibel audible and visual alarm (flashing lights and emitting a distress signal) to attract the attention of neighbors or passersby; a video call is automatically initiated, allowing the emergency contact and the platform to view the scene in real time; the robot continuously monitors the elderly person's condition until rescue arrives.

[0033] Monthly Health Report: The system automatically summarizes the elderly person's health monitoring data every month, including the range of body temperature fluctuations, the trend of respiratory rate changes, activity statistics, blood pressure and blood oxygen measurement values, etc., and generates a visual health report, which is pushed to the emergency contact's mobile phone through the cloud platform, so that the children can understand the elderly person's health status.

[0034] Example 4: Intelligent Charging Scheduling This embodiment details the working mechanism of the automatic recharge unit.

[0035] Charging point setup: During initial deployment, the system autonomously explores and creates a map of the home environment. Users place wireless charging pads on the bedside table in the bedroom and next to the coffee table in the living room; the robot automatically marks the locations of these two charging points during the mapping process.

[0036] User self-care ability identification: During the initial system setup, children or community service personnel select the elderly person's self-care ability level (self-sufficient / semi-self-sufficient / unable to care for themselves) on the APP. For self-sufficient elderly people, the system adopts a "feeding mode" charging strategy; for unable to care for themselves elderly people, an "automatic mode" charging strategy is adopted.

[0037] Feeding Mode Example: Grandma Wang is an independent elderly person. When the robot's battery level drops to 30% (the first threshold), it activates its "cuddling" mode. The robot walks to Grandma Wang, making whimpering sounds similar to a real pet, wagging its tail, and displaying the text and animation "Grandma, I'm hungry, I want to eat!" on its screen. Grandma Wang is amused and picks up the robot, placing it on the charging dock. The robot then emits a satisfied sound to express its gratitude while charging. This design solves the charging problem and adds emotional value to the human-computer interaction. When the battery level drops to 10% (the second threshold), the robot no longer waits for human intervention and immediately starts automatically returning to the charging dock, ensuring it doesn't shut down due to depleted battery.

[0038] Example of automatic mode: Grandpa Zhang is a disabled elderly person. When the robot's battery level drops to 30%, it automatically initiates recharging. The system first detects Grandpa Zhang's current state: through visual recognition, it finds that Grandpa Zhang is watching TV on the sofa (stationary state), and the TV is close to the charging point in the living room. The robot determines that this is an ideal time to recharge and slowly moves to the charging dock in the living room to begin charging. During the charging process, the robot keeps its head facing Grandpa Zhang, and infrared imaging and respiratory monitoring continue to operate to ensure uninterrupted health monitoring during charging.

[0039] Multi-charging point selection strategy: The system makes a comprehensive decision based on the user's current location, user status, and distance to the charging point. For example, if the user is resting in the bedroom and the bedroom charging point is available, the bedroom charging point will be selected first; if the user is active and both charging points are available, the charging point corresponding to the area where the user is expected to stay for a longer period of time (such as the living room when watching TV) will be selected, or the system will initiate a return to charging after the user stops.

[0040] Example 5: Dialect Adaptive Interaction This embodiment details the implementation of the dialect-adaptive voice interaction submodule.

[0041] Dialect Recognition Model: The system has pre-built acoustic and linguistic models for various dialects, including major dialect types such as Mandarin, Cantonese, Minnan, Hakka, Sichuanese, and Shanghainese. In the initial stage, users can select the elderly person's commonly used dialect through the APP, or the system can automatically recognize it.

[0042] Adaptive Learning Process: Taking Grandpa Chen, who only speaks Sichuan dialect, as an example. Initially set to Sichuan dialect mode, the robot uses Sichuan dialect for voice interaction. During use, the system continuously collects Grandpa Chen's voice data, analyzing his accent features, common vocabulary, and expression habits. For example, Grandpa Chen habitually says "shazi" instead of "senme," and "yaode" instead of "haode." The dialect adaptive learning submodule dynamically adjusts the parameters of the speech recognition model based on these features, improving recognition accuracy.

[0043] Personalized vocabulary database: The system also builds a personalized vocabulary database for each user, recording the user's unique expressions. For example, Grandpa Chen calls slippers "tuohaier". After learning this, the system establishes a mapping relationship between "bring the tuohaier" and "bring the slippers", ensuring that it can correctly understand the user's instructions.

[0044] Multi-user adaptive: When there are multiple elderly people in a family, the system can distinguish different users based on voiceprint features and load the corresponding dialect model for each user. For example, when Grandpa Chen speaks Sichuan dialect, the Sichuan dialect model is used, and when his grandson speaks Mandarin, the system automatically switches to the Mandarin model, achieving seamless switching between multiple dialects.

[0045] Example 6: Implementation of the item-grabbing function This embodiment details the workflow of the item-picking control submodule.

[0046] Item Recognition Training: In the initial stage of system deployment, children or community service personnel will assist with item recognition training. Users can use the app to mark the location and appearance features of common household items (such as slippers, water cups, remote controls, glasses, etc.), and the system will build an item database. For items that cannot be pre-trained, the robot can perform zero-shot recognition using the visual recognition capabilities of the large model.

[0047] Command Parsing and Execution: Grandpa Chen, sitting on the sofa, says, "Little one, bring me my slippers." The auditory perception unit collects the speech, and the dialect recognition model parses the command as "bring me my slippers," with the target item being "slippers." The system calls the visual recognition module to search for the location of the slippers on the environmental map. After recognizing the slippers as being located beside the bed in the bedroom, the autonomous navigation module plans a path, the robot moves to the bedroom, its mouth-gripping mechanism adjusts its posture, picks up the slippers, and then returns to Grandpa Chen's location, handing the slippers to the old man's feet.

[0048] Complex command processing: The system supports multi-condition command parsing. For example, Grandpa Chen says, "Bring me the water glass on the coffee table in the living room." The robot needs to simultaneously understand the target item (water glass), its location (coffee table), and the action (bring it), and execute them comprehensively. If the water glass is not in the specified location, the robot will reply, "Grandpa, there is no water glass on the coffee table. I see a water glass on the dining table. Do you want me to get that?" After confirming through dialogue, the robot will execute the command.

[0049] Safety features: During the item retrieval process, the robot's obstacle avoidance system operates continuously to prevent collisions with furniture or elderly individuals. The gripping mechanism is equipped with force sensors that adjust the gripping force based on the item's material and weight to prevent damage. For fragile items, the robot employs a more cautious gripping strategy.

[0050] Example 7: Cloud Collaboration and Privacy Protection This embodiment details the data management and privacy protection mechanisms of the cloud-based collaborative platform.

[0051] Encrypted data storage: All video and health data are encrypted locally on the robot using AES-256 encryption before being uploaded to the cloud server. Video data is stored in segments, with each segment encrypted independently. The encryption key is linked to the user account, preventing unauthorized decryption and viewing.

[0052] Hierarchical access control: The system has four levels of access control. 1) The elderly person: Has the highest level of access, can view all data, and set privacy protection rules; 2) Emergency contact (children): With the elderly person's authorization, they can view health reports, real-time video (with the elderly person's consent), and receive alarm information; 3) Community service providers: can only obtain the elderly person's location and necessary health information when the alarm is triggered; 4) Platform operators: They can only access anonymized statistical data for model optimization and have no right to view personal identity information and original videos.

[0053] User consent mechanism: Enabling the video monitoring function requires the elderly person's explicit consent. Upon initial system startup, the system will solicit the elderly person's opinion via voice and on-screen prompts: "Grandma, I need to turn on the camera to better protect your safety. Do you agree?" The system will only activate the camera after the elderly person answers "yes." If the elderly person refuses, the system will only use non-visual sensors and will request activation again in emergency situations.

[0054] Video playback process: When an abnormal event such as a fall occurs, the system automatically marks the video data 5 minutes before and after the event as "event video," which is then stored separately with encryption. Children can request to view the video via the app, and the system sends a confirmation request to the elderly person's mobile phone. Viewing is only permitted after the elderly person's consent. In emergency situations (such as a level 3 alarm), the system automatically authorizes emergency contacts and the platform to view the live video, but this access is automatically revoked after the event ends.

[0055] Example 8: Outdoor Adaptive Navigation and Step Crossing Capability This embodiment details the extended functionality of the autonomous navigation module in outdoor environments.

[0056] To meet the outdoor activity needs of the elderly, the robot, in addition to standard indoor navigation, also integrates a step recognition and climbing control submodule and a complex terrain traversal submodule in its obstacle avoidance unit. Based on binocular cameras, depth sensors, and inertial measurement units, it can perceive steps, slopes, ditches, and uneven surfaces in the outdoor environment in real time.

[0057] When the robot detects a step ahead, it first uses visual and depth data to identify the height and width of the step and determine whether it is within a passable range (e.g., ≤15cm). If passable, the robot adjusts its gait to climbing mode, raising its forelimbs and shifting its center of gravity forward to climb gradually; if the step is too high or there is an obstacle, it automatically plans an alternative route or prompts the elderly person to choose another route.

[0058] In outdoor following mode, the robot can adaptively adjust its cadence and stride based on the elderly person's walking speed to ensure stable following. If the elderly person is going up or down slopes or traversing complex terrain such as grass or gravel paths, the robot adjusts its foot pressure and movement strategy in real time to prevent slipping or tipping over.

[0059] This feature is particularly suitable for elderly people when they are active in outdoor settings such as residential areas and parks. The robot can provide continuous companionship and safety, further expanding the application scenarios of age-friendly robotic pets.

[0060] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An age-friendly robotic pet and its system, characterized in that, include: The pet-type robot body is covered with a waterproof short plush layer on the outside and integrates a multimodal perception module, a central control unit, an autonomous navigation module, an intelligent interaction module, and an actuator on the inside. The multimodal perception module includes a visual perception unit, an auditory perception unit, a tactile perception unit, and a vital signs monitoring unit, which are used to collect elderly people's status data, environmental data, and interactive behavior data in real time. The central control unit, embedded with a large model decision-making system, receives data collected by the multimodal perception module, performs fusion analysis and behavioral decision-making, and outputs control commands to the actuator. The intelligent interaction module, connected to the central control unit, includes a dialect-adaptive voice interaction submodule and an emotional feedback submodule, used to realize natural language dialogue and emotional response with the elderly; The autonomous navigation module, including a SLAM mapping unit, a path planning unit, an obstacle avoidance unit, and an automatic recharging unit, is used to achieve automatic following, autonomous return, and intelligent charging scheduling in complex environments, and supports adaptation to going up and down stairs, over obstacles, and uneven road surfaces in outdoor scenarios. The health management and emergency response module is connected to the vital signs monitoring unit and, based on a tiered alarm mechanism, enables fall detection, identification of abnormal vital signs, and multi-party linkage alarms. The cloud-based collaborative platform communicates with the central control unit and is used for data storage, model training, remote upgrades, and multi-user collaborative management.

2. The age-friendly robotic pet and its system according to claim 1, characterized in that, The visual perception unit includes a binocular camera and an infrared imaging sensor. The binocular camera is embedded in the robot's eye position and is used for video calls, video monitoring, and facial recognition. The infrared imaging sensor is used for non-contact body temperature monitoring and body posture recognition.

3. The age-friendly robotic pet and its system according to claim 1, characterized in that, The auditory perception unit includes a microphone array and a dialect recognition model. The dialect recognition model supports adaptive learning of multiple dialects and can dynamically optimize the recognition accuracy based on the user's voice characteristics.

4. The age-friendly robotic pet and its system according to claim 1, characterized in that, The vital signs monitoring unit includes contact sensors and non-contact sensors. The contact sensors are placed on the robot's back or abdomen and are used to measure blood pressure, blood oxygen, and blood glucose. The non-contact sensors are used to monitor respiratory rate and body temperature changes.

5. The age-friendly robotic pet and its system according to claim 1, characterized in that, The central control unit includes: The user profile building submodule creates personalized interaction profiles for different users based on facial recognition and interaction history data. The behavior decision-making submodule dynamically adjusts the robot's behavior mode based on the user profile, current status, and environmental information. The behavior modes include the closeness mode, the following mode, the silent mode, and the affectionate mode. The self-learning optimization submodule continuously optimizes decision-making strategies based on user feedback and behavioral data through reinforcement learning algorithms.

6. An age-friendly robotic pet and its system, characterized in that, The automatic recharging unit includes: The battery power monitoring submodule monitors the battery power status in real time. The charging strategy decision submodule dynamically sets the charging threshold and recharging timing based on the user's self-care ability and daily routine. For elderly people who can take care of themselves, the module will activate the "pleading mode" to request manual charging when the battery level is below the first threshold and activate automatic recharging when the battery level is below the second threshold. For elderly people who cannot take care of themselves, the module will directly activate automatic recharging when the battery level is below the first threshold. The multi-charging point management submodule supports setting at least two charging stations and selecting the optimal recharge timing and charging point based on the user's location and stationary state. The charging monitoring submodule maintains continuous operation of health monitoring and fall prevention alarm functions during the charging process.

7. The age-friendly robotic pet and its system according to claim 1, characterized in that, The health management and emergency response module includes: The fall detection submodule determines whether a fall has occurred based on body posture recognition and acceleration sensor data. The tiered alarm submodule executes a tiered alarm strategy based on the elderly person's level of consciousness and vital signs data: Level 1 alarm: When the elderly person is conscious and confirms that they need help, contact the emergency contact person; Level 2 alarm: When the elderly person loses consciousness but their vital signs are stable, or they do not get up within three minutes of falling, contact the emergency contact person and, if unable to contact them, alert the platform; Level 3 alarm: When the elderly person loses consciousness and their vital signs are unstable, contact both the emergency contact person and the platform, and activate the audible and visual alarm mode. The video retrospective submodule marks and encrypts video data before and after abnormal events, allowing authorized users to review and view it.

8. The age-friendly robotic pet and its system according to claim 1, characterized in that, The intelligent interaction module also includes: The dialect adaptive learning submodule collects user voice data, identifies dialect types and accent features, and dynamically adjusts the parameters of the speech recognition model. The emotion computing submodule identifies the user's emotional state based on voice tone, facial expressions, and interactive behavior, and generates corresponding emotional responses. The item-grabbing control submodule, based on the instruction recognition capability trained by a large model, parses the item-grabbing and delivery instructions issued by the user and controls the actuator to complete the item recognition, grabbing and delivery actions.

9. The age-friendly robotic pet and its system according to claim 1, characterized in that, The cloud-based collaboration platform includes: The data encryption and storage submodule performs local encryption and cloud backup of video data and health data; The model training and update submodule uses multi-user data to train and optimize large models, and pushes model updates to the robot body via OTA. The multi-user collaborative management submodule supports seniors, emergency contacts, community service providers, and platform operators to access data and receive alarm information according to their respective permissions.

10. The age-friendly robotic pet and its system according to claim 1, characterized in that, The autonomous navigation module also includes a tracking handling submodule. When the distance to the target elderly person exceeds a preset threshold and relocation is not possible, the automatic return mode is activated to return to the last known location or charging point. At the same time, location information and alarm prompts are sent to emergency contacts. It also integrates a step recognition and climbing control submodule and a complex terrain traversal submodule. Based on multimodal perception data, it identifies the height of steps, slope and obstacle type, and dynamically adjusts the robot's gait and movement strategy to achieve stable walking, climbing up and down steps and crossing obstacles in outdoor environments.

Images