AI-based walking aid and intelligent navigation system thereof

Abstract

The invention discloses an AI-based walking aid and an intelligent navigation system thereof, and belongs to the technical field of intelligent rehabilitation auxiliary instruments. The walking aid based on the AI comprises a body, a multi-mode sensing module, a positioning module, an AI decision center, an execution module and a man-machine interaction unit. The intelligent navigation system of the walking aid based on the AI comprises the walking aid and a remote monitoring platform. According to the invention, comprehensive acquisition of environment, user and equipment states is realized through the multi-mode sensing module, indoor and outdoor seamless positioning is realized by combining with a fusion positioning technology, active path planning and dynamic obstacle avoidance are realized through a deep reinforcement learning model of an AI decision center, and safety early warning and remote intervention are realized by relying on a remote monitoring platform. Compared with the prior art, the problems of passive following, single environment perception, insufficient user state adaptation and poor scene compatibility of a traditional walking aid are solved, and the walking aid safety and autonomy are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent rehabilitation assistive devices, and in particular to an AI-based walking aid and its intelligent navigation system. Background Technology

[0002] Walking aids are devices that help people with mobility impairments (such as the elderly, disabled, and those recovering from surgery) support their bodies, maintain balance, and assist in walking. Their core function is to provide stable support points, reduce physical exertion and the risk of falls while walking, and help users move more safely and independently.

[0003] As a device to assist people with mobility impairments, the core function of a walking aid is to help users reduce physical exertion and the risk of falls by providing support and auxiliary power. Existing technology, such as the patent with publication number CN118948545A, discloses a walking aid and its assistive system. This system collects data through a distance sensor and an inertial measurement unit, and uses an AI model to calculate a suggested speed to achieve assisted movement. However, this technology has the following drawbacks: Passive assistance lacks active navigation capabilities: it only adjusts speed based on the distance to the user and the slope of the terrain, without autonomous path planning or active obstacle avoidance functions, and cannot guide the user to the designated destination. In essence, it is "passive following" rather than "active guidance".

[0004] The environmental perception dimension is limited: it relies on distance sensors and inertial measurement units, and does not integrate multimodal data such as lidar and vision. The recognition accuracy of dynamic obstacles (such as pedestrians and moving vehicles) and complex environments (such as dark light and slippery road surfaces) is insufficient, and the obstacle avoidance safety is limited.

[0005] Lack of user status awareness: The system does not monitor the user's real-time physiological status (such as heart rate and fatigue level), and cannot dynamically adjust the navigation strategy according to changes in the user's physical strength, which can easily lead to the user overexerting their physical strength.

[0006] Poor positioning and scene adaptability: Indoor and outdoor positioning technologies are fragmented, and positioning drift is prone to occur in transitional scenes (such as entrances and exits of apartment buildings and shopping mall passages). Furthermore, there are no customized navigation strategies for different scenes (such as hospitals and communities).

[0007] There is a lack of safety emergency mechanisms: there is a lack of automatic early warning and rescue linkage functions when users experience sudden discomfort (such as falls or palpitations) or equipment failures, and safety relies on manual monitoring.

[0008] Therefore, there is a need for an AI-powered walking aid system that integrates multimodal perception, active navigation, user state adaptation, and remote collaboration functions. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides an AI-based walking aid and its intelligent navigation system, solving the problems mentioned in the background section.

[0010] Technical Solution: To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, an AI-based walking aid, comprising: Body: Provides a supporting structure and installation carrier, adopts a carbon fiber frame (load-bearing capacity of 180kg), and is equipped with drive wheels (with built-in magnetorheological shock absorbers), casters (with self-cleaning brushes) and emergency support feet (automatically pop out in case of power failure); the drive wheels and casters work together to achieve movement and steering, and the emergency support feet are used for stable support in case of emergencies.

[0011] Multimodal perception module: used to collect environmental data, user physiological data, and ontological state data, including: LiDAR (detection range 0.3-15m): Acquires 3D point cloud data of surrounding obstacles, identifies static obstacles (such as walls and furniture) and dynamic obstacles (such as pedestrians and vehicles); The LiDAR has a power level down to the milliwatt level and adopts a 16-line distributed design, clearly meeting the Class 1 laser safety standard. In medical clinical scenarios (rehabilitation training, hospital barrier-free navigation), it can not only provide accurate environmental perception and data support, but also reduce harm to patients from the design source, fully meeting the clinical needs of "long-term safe use" for the elderly, postoperative patients and other groups; Binocular vision sensor (1280×720 resolution, supports infrared mode): recognizes traffic signs, ground textures (such as tactile paving and anti-slip markings), dynamic target movement trajectories, and low-light environment features; Inertial measurement unit: Collects body attitude (three-axis angles), motion data (velocity, acceleration) and road surface conditions (slope, bumpiness); Biosensors: including heart rate sensor (sampling rate 100Hz), pressure sensor (detects grip strength) and palm sweat sensor (determines fatigue status), monitor the user's real-time physiological status; Environmental sensors: collect data on temperature and humidity, light intensity, air quality and odors (such as smoke and gas), and sounds (such as horns and cries for help). Equipment status sensors: monitor battery power (accuracy 1%), drive wheel temperature, and braking system sensitivity.

[0012] Positioning module: Integrates GPS, BeiDou, UWB indoor positioning, visual SLAM, and landmark-assisted positioning technologies to achieve seamless indoor and outdoor positioning. Outdoors, GPS / BeiDou dual-mode positioning (accuracy 0.5m) is used, while indoors, UWB positioning (accuracy 5cm) and visual SLAM (loop closure detection accuracy 0.05m) are used. The system adds landmark-assisted positioning (identifying artificial landmarks such as wall QR codes and elevator signs) and corrects positioning drift in transitional scenarios (such as indoor and outdoor entrances) to ensure continuous and error-free positioning.

[0013] AI Decision Center: Includes storage and processing units to handle data processing and decision output. Storage unit: Stores pre-trained deep reinforcement learning models, indoor and outdoor map databases (including accessible pathways and elevator locations), user behavior feature database (historical gait and physical strength thresholds), scene strategy database (hospital priority accessible routes and rainy day priority shading routes), and user health model (physiological parameter thresholds associated with medical history). Processing Unit: Employs an NVIDIA Jetson AGX Orin processor, supporting parallel computing: Local real-time processing of obstacle avoidance and speed adjustment; Collaborates with cloud AI for long-distance path planning, dynamically optimizing paths based on real-time traffic and weather data; Based on multimodal perception data, outputs obstacle avoidance commands, path planning results, and speed adjustment parameters through a deep reinforcement learning model.

[0014] Execution module: Includes drive motor, steering mechanism and braking device, executes actions according to instructions from AI decision center: The drive motor adopts vector control technology (speed response time of 0.3 seconds) and supports stepless speed regulation from 0 to 5 km / h; The steering mechanism supports a micro-steering mode (minimum steering radius 0.8m), making it suitable for narrow spaces; The braking system adopts a dual redundancy design of electromagnetic braking + mechanical parking brake (locking the wheels within 0.5 seconds in the event of power failure) to ensure emergency braking safety.

[0015] Human-computer interaction unit: Enables information exchange between the user and the walking aid. Voice module: Supports offline Chinese recognition (wake word "Xiaozhu", accuracy rate 95%), dialect recognition and semantic understanding (e.g., "go to get medicine" is associated with community health centers), and has emotional voice (gentle prompts when fatigued, and urgent warnings when in an emergency); Haptic feedback device: The handle has a built-in eccentric motor that provides 3 vibration frequencies (200Hz for left turn, 400Hz for right turn, and 800Hz for obstacle avoidance) and 2 vibration intensities (weak vibration for mild reminder and strong vibration for emergency). Display: 4.3-inch anti-glare touchscreen, displays navigation map and status parameters (speed, heart rate), supports large font mode and one-key help physical button (mis-touch rate <0.1%).

[0016] AI-based intelligent navigation system for walking aids The aforementioned walking aid and remote monitoring platform achieve data interaction through a wireless communication module (5G + Bluetooth dual-mode): Walking aid: Sends real-time data (location, environmental status, user physiological data, device status) to the remote monitoring platform and receives destination updates and remote commands from the platform.

[0017] Remote monitoring platform: Based on Alibaba Cloud servers, it supports access via web and mobile app, including: Map server: Connects to third-party accessible map APIs to update temporary obstacles (such as road construction) in real time and supports guardians to remotely draw customized routes; User management module: Stores user medical history, exercise preferences, and frequently used destinations; generates rehabilitation progress reports (combined with gait data); and links medication reminders (voice prompts when medication time is reached). Anomaly Warning Module: Establishes a three-level warning mechanism: Level 1 warning (minor anomaly, such as slightly elevated heart rate) is provided locally; Level 2 warning (moderate anomaly, such as deviation from the path) is pushed to the guardian's APP; Level 3 warning (emergency anomaly, such as fall, equipment failure) automatically dials the guardian's and community emergency center's phone numbers and sends a help request message with location information. Third-party integration interface: Connect to the hospital HIS system (obtain the location of the clinic and the queue progress) and the community service platform (book home visits by elderly care workers).

[0018] The beneficial effects of the AI-based walking aid and its intelligent navigation system of the present invention are as follows: (1) The present invention realizes the upgrade from "passive following" to "active navigation": through path planning and active obstacle avoidance, it supports users to specify the destination through voice commands, and the system will autonomously guide to the target location, solving the limitation of "no target following" of traditional walking aids and improving mobility autonomy.

[0019] User status adaptive enhancement improves user comfort: By monitoring physiological states such as heart rate and grip strength through biosensors, and combining with the user's health model, the speed is dynamically adjusted (such as automatically slowing down when the heart rate is abnormal) to avoid excessive physical exertion and adapt to the physical strength threshold of different users.

[0020] Full-scene positioning and customized navigation: Fusion positioning technology solves the problem of drift in transitional scenes, and the scene strategy library provides customized routes for scenarios such as hospitals, communities, and rainy days, adapting to the navigation needs of complex environments.

[0021] Remote collaboration and safety backup mechanism: A three-level early warning mechanism and third-party linkage enable automatic rescue in case of user emergencies or equipment failures, solving the shortcomings of traditional technology safety that relies on manual monitoring. Attached Figure Description

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0023] Figure 1This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0024] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0025] 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.

[0026] Reference Figure 1 An AI-based walking aid is implemented as follows: I. Hardware Configuration Main body: carbon fiber frame (weight 3.5kg, load capacity 180kg), drive wheels (20cm diameter, magnetorheological shock absorber), omnidirectional wheels (with self-cleaning brush), emergency support feet (spring-triggered); equipped with a 24V lithium battery (10 hours of battery life).

[0027] Multimodal sensing module: LiDAR: 16-line lidar, detection range 0.3-15m, angular resolution 0.5°; Binocular vision sensor: 1280×720 resolution, 30fps frame rate, supports infrared mode (90% accuracy in low light). Biosensors: The handle has a built-in heart rate sensor (sampling rate 100Hz, error ±2 times / min), a pressure sensor (grip force detection range 0-50N), and a sweat sensor (skin resistance detection accuracy ±5%). Environmental sensors: temperature and humidity sensor (accuracy ±0.5℃ / ±2%RH), light sensor (0-100000lux), odor sensor (detects smoke and gas); Equipment status sensors: battery power monitoring (accuracy 1%), drive wheel temperature sensor (-40℃ to 85℃ range).

[0028] Positioning module: GPS / BeiDou dual-mode positioning (outdoor accuracy 0.5m), UWB indoor positioning (accuracy 5cm), visual SLAM (loop closure detection accuracy 0.05m), landmark recognition (QR code recognition distance 0.5-3m).

[0029] AI decision-making center: NVIDIA Jetson AGX Orin processor (32GB memory), storage unit contains deep reinforcement learning model (state space: obstacle distance, user heart rate, speed, path deviation; action space: speed, steering angle, braking) and map database (including indoor and outdoor barrier-free passages).

[0030] Human-computer interaction unit: voice module (95% accuracy in offline Chinese recognition), haptic feedback (3 frequencies + 2 intensities), 4.3-inch touchscreen (anti-glare), one-click help button.

[0031] An AI-based intelligent navigation system for walking aids includes the following operational flow: Indoor navigation scenarios (e.g., from bedroom to living room): The user's voice command "Go to the living room" is parsed by the voice module and transmitted to the AI ​​decision-making center. The positioning module determines the current location (bedroom), and the AI ​​decision center calls the indoor map database to plan the optimal path to avoid the furniture; The execution module controls the movement of the drive wheel and prompts the direction of turn through the vibration of the handle (200Hz for left turn, 400Hz for right turn). Upon arrival, a voice announcement will say "Arrived in the living room," and the display screen will simultaneously show the arrival information.

[0032] Outdoor obstacle avoidance scenarios (such as encountering pedestrians on community roads): The lidar detected a pedestrian (dynamic obstacle) at a distance of 3m, and the point cloud data was transmitted to the AI ​​decision-making center. A dynamic environment model predicts pedestrian trajectories, while a deep reinforcement learning model calculates detour paths (offset 1.5m to the right). The execution module slows down to 1km / h and adjusts the direction, with a voice prompt saying "There is a pedestrian ahead, you are about to turn right," and the handlebars provide auxiliary prompts with 400Hz vibration. After completing the detour, resume the original speed and path.

[0033] User fatigue adaptive scenario: The biosensor detected a user's heart rate of 110 beats per minute (exceeding the health model threshold), and the data was transmitted to the AI ​​decision center. The processing unit combines the user behavior feature database (historical heart rate-speed correlation data) to output a speed reduction command (from 3km / h to 1.5km / h). The voice prompt says "Speed ​​has been reduced, please rest for a while," and the display shows the heart rate and recommended rest duration. Once the heart rate returns to below 80 beats per minute, it will automatically increase to a suitable speed.

[0034] Emergency warning scenarios (such as a user falling): The inertial measurement unit detected a severe tilt of the body (angle > 45°), and the biosensor simultaneously detected that the grip force had returned to zero, triggering a level three warning. The walker immediately stops the vehicle and turns on its hazard lights, asking "Do you need help?" via voice prompt. The remote monitoring platform automatically dials the guardian's and the community emergency center's phone numbers and sends a request for help containing location information; Guardians can view real-time environmental videos through the platform and remotely send comforting voice messages to the walking aid.

[0035] 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 AI-based walking aid, comprising a body, a multimodal perception module, a positioning module, an AI decision-making center, an execution module, and a human-computer interaction unit, characterized in that: The main body provides a supporting structure and an installation carrier; The multimodal perception module includes a lidar, a binocular vision sensor, an inertial measurement unit, a biosensor, and an environmental sensor, used to collect environmental data, user physiological data, and body state data. The positioning module integrates GPS, BeiDou, UWB, visual SLAM and landmark-assisted positioning technologies to achieve seamless indoor and outdoor positioning; The AI ​​decision-making center includes a storage unit and a processing unit. The storage unit stores a deep reinforcement learning model, a map database, a user behavior feature library, and a scene strategy library. The processing unit outputs obstacle avoidance commands, path planning results, and speed adjustment parameters based on the inputs from the multimodal perception module and the positioning module. The execution module moves, turns, and brakes according to the instructions of the AI ​​decision center; the human-computer interaction unit realizes information interaction between the user and the walking aid.

2. The AI-based walking aid according to claim 1, characterized in that, The biosensor includes a heart rate sensor, a pressure sensor, and a palm sweat sensor; the heart rate sensor is used to monitor the user's heart rate, the pressure sensor is used to detect the user's grip strength, and the palm sweat sensor is used to determine the user's fatigue state.

3. The AI-based walking aid according to claim 1, characterized in that, The landmark-assisted positioning technology of the positioning module achieves positioning correction in transitional scenarios by recognizing QR codes on walls and elevator signs, with a positioning error of ≤0.1m after correction.

4. The AI-based walking aid according to claim 1, characterized in that, The processing unit of the AI ​​decision-making center uses an NVIDIA Jetson AGX Orin processor, which supports parallel operation of environment modeling and reinforcement learning algorithms. The state space of the reinforcement learning model includes obstacle distance, user heart rate, current speed and path deviation, while the action space includes speed, steering angle and braking signal.

5. The AI-based walking aid according to claim 1, characterized in that, The execution module includes a drive motor, a steering mechanism, and a braking device; the drive motor adopts vector control technology with a speed adjustment response time of ≤0.3 seconds; the braking device adopts a dual redundancy design of electromagnetic braking and mechanical parking braking, with a lock-up time of ≤0.5 seconds when power is off.

6. The AI-based walking aid according to claim 1, characterized in that, The human-computer interaction unit includes a voice module, a haptic feedback device, and a display screen; the voice module supports dialect recognition and semantic understanding with a recognition accuracy of ≥95%; the haptic feedback device provides three vibration frequencies and two vibration intensities, corresponding to steering, obstacle avoidance, and alert levels, respectively; the display screen is an anti-glare touchscreen that supports large font mode and one-click help function.

7. A smart navigation system for an AI-based walking aid, comprising the walking aid as described in claim 6, characterized in that, It also includes a remote monitoring platform; The walking aid connects to the remote monitoring platform via a 5G+Bluetooth dual-mode wireless communication module, sending real-time location, environmental data, user physiological data, and device status data. The remote monitoring platform includes a map server, a user management module, and an anomaly warning module. The map server supports dynamic map updates and customized route drawing. The user management module stores user medical history and rehabilitation data. The anomaly warning module implements three-level warnings and links with third-party rescue.

8. The intelligent navigation system for an AI-based walking aid according to claim 7, characterized in that, The three-level early warning system of the anomaly early warning module includes: Level 1 warning indicates a minor anomaly, and is issued locally. A Level 2 alert indicates a moderate abnormality and will be sent to the guardian's app. A Level 3 alert indicates an emergency or abnormal situation. It will automatically call the guardian and the community emergency center and send a location-based help request.

9. The intelligent navigation system for an AI-based walking aid according to claim 7, characterized in that, It also includes third-party integration interfaces, which connect to the hospital's HIS system and community service platform to obtain medical information and schedule community services.

Images