Personalized wheelchair head control and obstacle avoidance scheme based on multiple models

Abstract

The invention provides a personalized wheelchair head control and obstacle avoidance scheme based on multiple models, and belongs to the field of rehabilitation auxiliary equipment and intelligent control, and the method comprises the steps: collecting the pitch angle and roll angle information of the head in real time through an inertial sensor module fixed right above the head of a user; transmitting the data to a data processing module for attitude calculation and fusion to obtain a head space attitude; establishing a control model according to individual differences, and selecting according to user reaction capability during advancing; the head posture information is converted into a wheelchair speed instruction through a PID algorithm; and when the obstacle is less than a safety threshold value, forced parking and alarming are carried out. According to the multi-model-based personalized wheelchair head control and obstacle avoidance scheme, the problem that traditional rocker control is unfriendly to people with disabled upper limbs is solved, the defect of a visual control scheme in the aspect of environmental adaptability is overcome, and more convenient, safer and more comfortable wheelchair use experience is provided for users with mobility difficulties.

Description

Technical Field

[0001] The present invention relates to the fields of rehabilitation assistive devices and intelligent control, and in particular to a personalized wheelchair head control and obstacle avoidance solution based on multiple models. Background Art

[0002] Wheelchairs, as assistive devices widely used in hospitals, homes, rehabilitation centers, and public spaces, offer significant convenience and freedom of movement for people with disabilities, the elderly, and those undergoing rehabilitation. However, despite years of advancements in wheelchair technology, the operating systems of most wheelchairs on the market still rely primarily on simple joystick controls. While intuitive and easy to use, this approach presents significant challenges for those with upper limb disabilities or high-level paraplegia. Due to limited or no upper limb function, these users are unable to perform fine manual manipulation, making traditional joystick controls inadequate, limiting their ability to move independently and their quality of life.

[0003] In addition to joystick control, recent approaches have emerged that use cameras or depth-of-field cameras to recognize human posture and thereby control the wheelchair. These visual control solutions utilize advanced image processing and machine learning techniques to attempt to drive the wheelchair's motion by capturing the user's body movements or postures. However, despite their theoretical innovation and potential, these solutions face numerous challenges in practical application. In particular, visual control solutions are highly susceptible to factors such as lighting, obstructions, and environmental changes, resulting in unstable control or even failure, thus limiting their widespread application in real-world scenarios.

[0004] More critically, the vast majority of wheelchairs currently designed fail to fully consider the user's safety needs. While in motion, users often rely solely on their vision and judgment to avoid obstacles, which undoubtedly increases the risk for users with limited mobility. This is especially true in complex or unpredictable environments, such as crowded public places, narrow corridors, or stairwells. Misjudgment or inability to react quickly can lead to accidents such as collisions or falls, causing physical harm and psychological distress.

[0005] At the same time, inertial sensors, as core sensors for human posture estimation and motion capture, have been widely used in fields such as health monitoring, smart wearables, and virtual reality. By measuring physical quantities such as acceleration and angular velocity, inertial sensors can capture and record an object's motion and posture changes in real time. By fusing, processing, and analyzing the raw data collected by inertial sensors, an object's spatial posture and motion trajectory can be accurately estimated, providing reliable data support for a variety of application scenarios. Summary of the Invention

[0006] The purpose of this invention is to provide a personalized wheelchair head control and obstacle avoidance solution based on multiple models, which solves the problem that traditional joystick control is unfriendly to people with upper limb disabilities, overcomes the shortcomings of visual control solutions in environmental adaptability, and provides users with limited mobility with a more convenient, safe and comfortable wheelchair usage experience.

[0007] To achieve the above objectives, the present invention provides a personalized wheelchair head control and obstacle avoidance solution based on multiple models, comprising the following steps:

[0008] S1. Head posture data collection: The inertial sensor module collects the user's head posture data in real time. The inertial sensor module is fixed directly above the user's head to collect the pitch and roll angle information of the head.

[0009] S2. Data transmission and fusion: The collected head posture data is transmitted to the data processing module via wired or wireless means, and posture calculation and data fusion are performed in the data processing module to obtain the spatial posture of the head;

[0010] S3. Establish and select a personalized control model: Establish a control model based on head motion data according to individual differences, and select the corresponding control model based on the user's reaction ability during the wheelchair movement;

[0011] S4, posture data processing and control command generation: The calculated head posture information is converted into a speed command for the wheelchair through a PID algorithm, where the pitch angle controls the wheelchair to move straight, and the roll angle controls the wheelchair to turn;

[0012] S5. Obstacle detection and safe obstacle avoidance: During the movement of the wheelchair, the distance to obstacles around the wheelchair is monitored in real time through the ultrasonic radar ranging module. When the obstacle distance is detected to be less than the safety threshold, the wheelchair movement is forced to stop and an alarm is issued.

[0013] Preferably, S1 also includes a process for calibrating the initial installation error of the inertial sensor module, which includes static measurement, nodding movement and shaking movement, recording the gravity vector and calculating the rotation axis direction vector, and then constructing a rotation matrix to solve the attitude quaternion and compensate for the fixed installation error between the IMU coordinate system and the user coordinate system.

[0014] Preferably, the same posture data in the control model in S3 has different turning and straight-ahead accelerations to adapt to use by patients with different degrees of disability, and after the system is started, adaptive control parameters are generated according to the user's head rotation range.

[0015] Preferably, in S4, multi-model incremental PID control and multi-model standard PID control are respectively adopted for the straight-moving and turning states of the wheelchair, and a gain selection function is defined according to the current usage mode and current speed to dynamically adjust the PID parameters.

[0016] Preferably, the ultrasonic radar ranging module includes at least four ultrasonic ranging sensors, which continuously detect the distance between the wheelchair and surrounding obstacles. The detection frequency is 10Hz, the safety threshold is set to 50cm~70cm, and the emergency braking mechanism is immediately triggered when the obstacle distance is detected to be less than the threshold.

[0017] Preferably, the error calibration process is as follows: the user sits upright and does not move, and records the static gravity vector measured by the accelerometer. The user nods his head up and down once, and the gravity vector is recorded after the action is completed. The user shakes his head left and right once, and the gravity vector is recorded after the action is completed.

[0018] Preferably, the rotation axis direction vector is They correspond to the x, y, and z axes in the user coordinate system respectively;

[0019]

[0020] in, is the direction of the nodding rotation axis in the IMU coordinate system, is the direction of the head shaking rotation axis in the IMU coordinate system, is the IMU coordinate system The third orthogonal axis, for The normalized unit vector, for The normalized unit vector, for Normalized unit vector.

[0021] Preferably, the construction of the rotation matrix to solve the quaternion includes letting the rotation matrix R transform the axis of the IMU coordinate system to the user coordinate system

[0022]

[0023] e x =[1,0,0] T ;

[0024] e y =[0,1,0] T ;

[0025] e x =[0,0,1]T ;

[0026] The attitude quaternion expression is:

[0027]

[0028] in, is the inverse quaternion of the installation error quaternion, q imu (t) is the IMU attitude quaternion obtained by complementary filtering.

[0029] Preferably, the gain selection function includes:

[0030] (K p ,K i ,K d =SelectGains(mode,v(k));

[0031]

[0032] Among them, mode∈{1,2,3} is the current usage mode, v(k) is the current speed, L is low speed, M is medium speed, and H is high speed.

[0033] Therefore, the present invention adopts the above-mentioned personalized wheelchair head control and obstacle avoidance solution based on multiple models, and the technical effects are as follows:

[0034] 1. Personalized control based on head posture: Inertial sensors are applied to wheelchair control. By collecting and analyzing head posture data in real time, personalized control based on head posture is achieved, solving the problem that traditional wheelchair control is unfriendly to people with upper limb disabilities.

[0035] 2. Active obstacle avoidance mechanism: An ultrasonic ranging radar is integrated into the wheelchair to achieve real-time obstacle detection and active obstacle avoidance functions, improving the safety and reliability of the wheelchair in complex environments.

[0036] 3. Multi-model PID control: Different PID control strategies are used in straight-line and turning control, and multiple sets of PID gain combinations are set according to the user's response ability, which improves the flexibility and adaptability of wheelchair control. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is the overall flow chart of the present invention;

[0038] Figure 2 This is a schematic diagram of the electrical control of the wheelchair of the present invention;

[0039] Figure 3 This is a communication diagram of the wheelchair module of the present invention;

[0040] Figure 4This is a flow chart of the control program and a schematic diagram of the UI operation interface of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0042] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0043] Example 1

[0044] like Figure 1-Figure 4 As shown, the present invention provides a personalized wheelchair head control and obstacle avoidance solution based on multiple models. The spatial posture of the head is determined by performing posture calculation and data fusion on inertial sensor data. The pitch angle and roll angle control the wheelchair's straight movement and steering, respectively. The control data is processed by a computing platform on the wheelchair. A control model based on head motion data can be established based on individual differences. This personalized model outputs corresponding control signals to control the wheelchair. During travel, an ultrasonic ranging radar monitors surrounding obstacles in real time. If the distance between the wheelchair and an obstacle is less than a threshold, the safety obstacle avoidance mechanism is triggered, forcing the wheelchair to stop moving and sounding an alarm.

[0045] The following steps are involved:

[0046] S1. Head posture data collection: The inertial sensor module collects the user's head posture data in real time. The inertial sensor module is fixed directly above the user's head to collect the pitch and roll angle information of the head.

[0047] S2. Data transmission and fusion: The collected head posture data is transmitted to the data processing module via wired or wireless means, and posture calculation and data fusion are performed in the data processing module to obtain the spatial posture of the head;

[0048] S3. Establish and select a personalized control model: Establish a control model based on head motion data according to individual differences, and select the corresponding control model based on the user's reaction ability during the wheelchair movement;

[0049] S4, posture data processing and control command generation: The calculated head posture information is converted into a speed command for the wheelchair through a PID algorithm, where the pitch angle controls the wheelchair to move straight, and the roll angle controls the wheelchair to turn;

[0050] S5. Obstacle detection and safe obstacle avoidance: During the movement of the wheelchair, the distance to obstacles around the wheelchair is monitored in real time through the ultrasonic radar ranging module. When the obstacle distance is detected to be less than the safety threshold, the wheelchair movement is forced to stop and an alarm is issued.

[0051] The inertial sensor module is placed directly above the head using a fixed device such as a hat. This device collects head posture in real time and transmits the collected data to the wheelchair's data processing module via Bluetooth or a wired connection. The inertial sensor acquisition module's MCU fuses the raw data to determine the head's spatial posture. The pitch angle is used as the wheelchair's control command for straight movement, and the roll angle as the command for turning. Both posture information is processed and converted into wheelchair speed commands using a PID algorithm. Once the wheelchair is powered on and enabled, the ultrasonic radar continuously detects the distance between the wheelchair and surrounding obstacles in real time. Before issuing the wheelchair motor speed command, it determines whether the wheelchair is within a safe distance from surrounding obstacles.

[0052] The inertial sensor uses the JY60 module from Witt Intelligent, which has an accelerometer, gyroscope, and magnetometer. The inertial sensor module is fixed directly above the user's head using a hat or elastic bandage. The raw data of the inertial sensor module is read by an STM32 (or low-power microcontrollers such as Arduino and ESP32). After the microcontroller reads the data, it sends it to the computing platform via a wired serial port or wireless methods such as Bluetooth.

[0053] The complementary filtering in AHRS is used for data processing, and the raw data is fused in the MCU that reads the raw data to obtain the specific posture.

[0054] When the system is powered on, it is initialized to the current posture at the zero position. After initialization, this position is the starting position of the head posture. If you want to change the zero position, you should re-initialize the program to allow the user to determine the initial position to ensure normal subsequent motion control.

[0055] To address the issue of magnetometer drift caused by environmental interference in the fusion filtering algorithm, we fix the magnetometer's orientation so that the direction the human-machine is facing is always considered north (false north). Using only accelerometer observations, complementary filtering can calculate roll and pitch angles while fixing the yaw angle to zero. By keeping the heading angle constant (defining the IMU's forward direction as "false north"), the complementary filtering result always aligns the IMU's forward direction with the preset reference direction, eliminating heading drift and providing a stable attitude estimate.

[0056] When the user wears the posture sensor, wearing errors will occur, and the human-machine coordinate system and the inertial sensor coordinate system cannot completely coincide. Therefore, it is necessary to perform initial installation error calibration: the user sits upright and does not move, and records the static gravity vector measured by the accelerometer. At this time, it is assumed that the IMU coordinate system is facing upward in the user coordinate system (gravitational acceleration [0,0,-g]).

[0057] The user nods his head up and down once, and the gravity vector is recorded after the action is completed. calculate and The cross product of : The vector obtained by the cross product operation is always perpendicular to the two original vectors, here That is the direction of the nod (pitch) rotation axis in the IMU coordinate system.

[0058] The user shakes his head left and right once, and the gravity vector is recorded after the action is completed. Same calculation Get the direction vector of the pan (roll) rotation axis. Ideally, Should be aligned with the roll axis in the user's head coordinate system.

[0059] Will and Normalized to unit vector Then calculate the third orthogonal axis And normalized to In this way, three mutually orthogonal reference axes are obtained in the IMU coordinate system. They correspond to the forward (x-axis), nod / pitch (y-axis), and shake / yaw (z-axis) directions in the user coordinate system.

[0060] Let the rotation matrix R transform the axis of the IMU coordinate system to the user coordinate system, that is, satisfy where e x =[1,0,0] T , e y =[0,1,0] T , e x =[0,0,1] T . Can be directly taken Then calculate the quaternion q according to the rotation matrix R e By installing the error quaternion q e , in the actual attitude solution, the output quaternion is compensated. If q imu (t) is the IMU attitude quaternion obtained by complementary filtering, and the attitude quaternion of the user's head under the navigation system is:

[0061]

[0062] in, is the inverse quaternion of the installation error quaternion.

[0063] In order to enable patients with different degrees of disability to easily control their wheelchairs using this control scheme, a control model based on head motion data is established according to individual differences, and the corresponding control model is selected according to the user's reaction ability during the movement of the wheelchair. The same posture data in the control model has different turning and straight-line accelerations to adapt to the use of patients with different degrees of disability, and adaptive control parameters are generated according to the user's head rotation range after the system is started.

[0064] The defined gain selection functions include:

[0065] (K p ,K i ,K d =SelectGains(mode,v(k));

[0066]

[0067] Among them, mode∈{1,2,3} is the current usage mode, v(k) is the current speed, L is low speed, M is medium speed, and H is high speed.

[0068] Add hysteresis speed judgment logic to the algorithm for switching, set V1, V2 (low-speed-medium speed boundary and medium-speed-high speed boundary) and hysteresis band ∈> 0. According to the current speed v(k) and the previous state, the gear switching logic is as follows:

[0069]

[0070] In each control cycle, the current speed gear is determined based on v(k) and combined with mode, it is input into SelectGains() to obtain new PID parameters.

[0071] After the gain is selected, the data increment of the speed channel controller is:

[0072] Δu(k)=K p (e(k)-e(k-1))+K i e(k)+k d (e(k)-2e(k-

[0073] 1)+e(k-2));

[0074] The updated output of the controller is u(k)=u(k-1)+Δu(k).

[0075] To ensure the wheelchair's safety during travel, it's equipped with four ultrasonic ranging sensors that continuously monitor the distance to obstacles. The sensors have a maximum detection range of 3 meters and a minimum detection range of 10 cm. The safety threshold between the wheelchair and obstacles is 50 cm to 70 cm, and the frequency of obstacle detection is 10 Hz. Before transmitting speed commands to the motors, the wheelchair first determines whether the distance to the obstacle is below the safety threshold. If so, emergency braking is applied.

[0076] The hardware of this system uses 485 communication based on modbus-RTU and Raspberry Pi 4B as the computing platform, which is responsible for data transmission, control command issuance and logical judgment of each part.

[0077] The operation process is:

[0078] After startup, the system first asks the user to rotate their head once. The system then determines the range of head postures that can be achieved. Based on this range of head movement, it generates appropriate adaptive control parameters. Different control models are selected based on the user's different response abilities.

[0079] After the model is selected, the posture is initialized. The user maintains the initial posture for about 3 seconds. The initial posture is sensed through the calibration algorithm to establish the initial posture.

[0080] After the initial posture is established, the wheelchair's motion is controlled by the enable button.

[0081] Therefore, the present invention adopts the above-mentioned multi-model-based personalized wheelchair head control and obstacle avoidance solution. By integrating hardware components such as inertial sensors, ultrasonic ranging radars, and computing platforms, and adopting advanced algorithms and control strategies, a multi-model-based personalized wheelchair head control and obstacle avoidance solution is realized. This not only improves the control accuracy and comfort of the wheelchair, but also enhances the safety and reliability of the wheelchair in complex environments, providing a more convenient, safe and comfortable wheelchair usage experience for disabled people with limited mobility.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A personalized wheelchair head control and obstacle avoidance solution based on multiple models, characterized by: The following steps are involved: S1. Head posture data collection: The inertial sensor module collects the user's head posture data in real time. The inertial sensor module is fixed directly above the user's head to collect the pitch and roll angle information of the head. S2. Data transmission and fusion: The collected head posture data is transmitted to the data processing module via wired or wireless means, and posture calculation and data fusion are performed in the data processing module to obtain the spatial posture of the head; S3. Establish and select a personalized control model: Establish a control model based on head motion data according to individual differences, and select the corresponding control model based on the user's reaction ability during the wheelchair movement; S4, posture data processing and control command generation: The calculated head posture information is converted into a speed command for the wheelchair through a PID algorithm, where the pitch angle controls the wheelchair to move straight, and the roll angle controls the wheelchair to turn; S5. Obstacle detection and safe obstacle avoidance: During the movement of the wheelchair, the distance to obstacles around the wheelchair is monitored in real time through the ultrasonic radar ranging module. When the obstacle distance is detected to be less than the safety threshold, the wheelchair movement is forced to stop and an alarm is issued.

2. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 1 is characterized in that: The S1 also includes the process of calibrating the initial installation error of the inertial sensor module, which includes static measurement, nodding movement and shaking movement, recording the gravity vector and calculating the rotation axis direction vector, and then constructing the rotation matrix to solve the attitude quaternion and compensate for the fixed installation error between the IMU coordinate system and the user coordinate system.

3. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 1 is characterized in that: The same posture data in the control model in S3 has different turning and straight-ahead accelerations to adapt to use by patients with different degrees of disability, and generates adaptive control parameters according to the user's head rotation range after the system is started.

4. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 1 is characterized in that: In S4, multi-model incremental PID control and multi-model standard PID control are respectively adopted for the straight-moving and turning states of the wheelchair, and a gain selection function is defined according to the current usage mode and current speed to dynamically adjust the PID parameters.

5. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 1 is characterized in that: The ultrasonic radar ranging module includes at least four ultrasonic ranging sensors, which continuously detect the distance between the wheelchair and surrounding obstacles. The detection frequency is 10Hz, and the safety threshold is set at 50cm to 70cm. When the obstacle distance is detected to be less than the threshold, the emergency braking mechanism is immediately triggered.

6. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 2 is characterized in that: The error calibration process is as follows: the user sits upright and does not move, and records the static gravity vector measured by the accelerometer. The user nods his head up and down once, and the gravity vector is recorded after the action is completed. The user shakes his head left and right once, and the gravity vector is recorded after the action is completed.

7. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 2, characterized in that: The rotation axis direction vector is They correspond to the x, y, and z axes in the user coordinate system respectively; in, is the direction of the nodding rotation axis in the IMU coordinate system, is the direction of the head shaking rotation axis in the IMU coordinate system, is the IMU coordinate system The third orthogonal axis, for The normalized unit vector, for The normalized unit vector, for Normalized unit vector.

8. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 2 is characterized in that: The construction of the rotation matrix to solve the quaternion includes letting the rotation matrix R transform the axis of the IMU coordinate system to the user coordinate system e x ＝[1,0,0] T ； e y ＝[0,1,0] T ； e x ＝[0,0,1] T ； The attitude quaternion expression is: in, is the inverse quaternion of the installation error quaternion, q imu (t) is the IMU attitude quaternion obtained by complementary filtering.

9. The multi-model-based personalized wheelchair head control and obstacle avoidance solution according to claim 4, characterized in that: The gain selection function includes: (K p ,K i ,K d =SelectGains(mode,v(k)); Among them, mode∈{1,2,3} is the current usage mode, v(k) is the current speed, L is low speed, M is medium speed, and H is high speed.

Images