Anti-tilt and fall protection system
By utilizing hip joint motors for coordinated drive and a safety locking mechanism within the exoskeleton system, the problem of falls caused by instability in the exoskeleton system's center of gravity is solved, achieving high-precision forward-looking prediction and two-way protection, and improving the system's simplicity and safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Filing Date
- 2026-06-11
- Publication Date
- 2026-07-14
AI Technical Summary
Existing exoskeleton systems pose a risk of instability when providing flexion and extension power assistance, leading to irreversible problems such as forward and backward tilting and falling. Existing technical solutions also suffer from problems such as perception lag, hardware redundancy, and excessive weight.
By employing a posture detection module, a center of gravity calculation unit, and a risk prediction unit, and utilizing existing hip joint motors for coordinated drive, combined with a safety locking mechanism, it achieves high-precision prediction and seamless active intervention for forward and backward tilting falls, avoiding the need for additional physical actuators.
It achieves accurate prediction and proactive prevention of falls without increasing system weight, reduces the number of motors by 50%, improves system simplicity and portability, ensures response time within 50ms, and provides reliable physical safety protection.
Smart Images
Figure CN122378809A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent robots and human-machine collaborative equipment technology, and in particular to a fall protection system to prevent forward and backward tilting. Background Technology
[0002] With the deep integration of robotics technology and intelligent control algorithms, hip exoskeletons, as typical human-machine collaborative intelligent wearable devices (mainly involving high-end manufacturing fields such as exoskeleton robots and intelligent assistive equipment), play an irreplaceable and crucial role in assisting vulnerable groups in daily walking, gait rehabilitation for hemiplegic patients, and heavy-duty work scenarios.
[0003] However, while exoskeleton systems provide flexion and extension power assistance to human joints, users (especially those with impaired lower limb muscle strength or deteriorated balance perception networks) often face the risk of instability. During complex dynamic gait transitions or in unstructured environments (such as climbing stairs or overcoming obstacles), if the user's overall center of gravity shifts forward or backward, and its projected boundary exceeds the range of the stable support polygon formed by the two feet, irreversible forward or backward tilting and falling are highly likely, resulting in serious secondary physical injuries.
[0004] Regarding the fall protection issue during exoskeleton operation, existing technical solutions mainly follow two evolutionary paths, but both have significant technical shortcomings: The first evolutionary approach focuses on passive protection and delayed response. Such systems typically monitor the absolute tilt angle using only a single sensor. Only when the torso tilt angle exceeds a pre-set physical limit threshold does the system respond passively by triggering a mechanical dead zone (hard stop) or directly cutting off system power. Because this approach lacks advanced predictive algorithms based on human multibody dynamics models, the system's control loop suffers from severe perception lag. By the time the exoskeleton responds, the irreversible gravitational overturning moment has often already been fully formed. This "post-incident" intervention has an extremely low success rate and fails to provide substantial fall protection.
[0005] The second evolutionary path shifts to an active protection architecture employing multiple independent actuators. This type of solution, in addition to the conventional left and right hip joint gait assist motors, adds dedicated forward / backward tilt protection motors in series or parallel (e.g., constructing an independent four-motor drive system). While this approach improves the proactiveness of protective intervention, the electromechanical defects it introduces are fatal: First, in terms of the overall physical architecture, the additional independent motor and its matching reduction transmission components will cause the overall mass and outer volume of the exoskeleton system to expand dramatically, which seriously violates the ergonomic development trend of lightweight and compact smart wearable equipment. Secondly, redundant physical actuators significantly increase the rotational inertia of active nodes such as the hip joint, which severely deteriorates the electromechanical responsiveness and torque transparency of the exoskeleton during normal walking, making it very easy to cause additional human-machine interaction fatigue. Finally, the independent operation of multiple high-power actuators not only increases hardware manufacturing costs and power consumption, but also greatly increases the complexity of the underlying multi-motor collaborative control and the overall mechanical failure rate of the system.
[0006] In summary, the core technical problem that urgently needs to be solved in the field of intelligent exoskeletons is how to make full use of the existing human-machine collaborative drive architecture, break through the technical bottlenecks of existing perception lag and hardware redundancy, and achieve high-precision forward prediction of fall risk and seamless two-way (forward / backward) active intervention protection without increasing the burden of additional physical actuators and keeping the system extremely lightweight. Summary of the Invention
[0007] The purpose of this invention is to address the shortcomings of existing exoskeleton systems, such as excessive reliance on additional independent anti-fall protection motors leading to cumbersome and heavy systems, and a lack of forward-looking kinematic prediction resulting in serious lag in protective control. Therefore, this invention proposes an anti-forward and backward tilting fall protection system.
[0008] To achieve the above objectives, the present invention adopts the following technical solution: A fall protection system that prevents forward and backward tilting includes: The posture detection module includes at least one inertial measurement unit for real-time acquisition of posture information of the user's body parts; A center of gravity calculation unit is used to calculate the overall center of gravity position of the user in real time based on the posture information. The risk prediction unit is used to predict the risk of leaning forward or backward based on the overall center of gravity position and movement trend, and to calculate the risk index of leaning forward or backward respectively. The hip joint dual-motor coordinated drive unit is configured with only a left hip joint motor and a right hip joint motor as the drive power source and fall protection actuator. The left hip joint motor and the right hip joint motor are used to drive the flexion or extension movements of the left and right hip joints, respectively. When the forward tilt risk index exceeds a first threshold, the two hip joint motors work together to output a reverse extension torque. When the backward tilt risk index exceeds the first threshold, the two hip joint motors work together to output a reverse flexion torque. The safety locking mechanism is structurally installed on both sides of the lumbar support frame, including an electromagnet and a wedge-shaped locking block, as well as a locking rack distributed in the inner circle of the lumbar support frame. When the forward tilt risk index or the backward tilt risk index exceeds a second threshold, the safety locking mechanism controls the electromagnet to de-energize to release the wedge-shaped locking block, so that the wedge-shaped locking block forms a self-locking engagement with the locking rack under the action of a spring, thereby rigidly connecting the exoskeleton to the user's waist.
[0009] Furthermore, the posture detection module includes multiple inertial measurement units distributed on the user's torso and both lower limbs; The plurality of inertial measurement units includes five inertial measurement units, which are respectively installed on the user's waist center, the outer side of the left thigh, the outer side of the right thigh, the front side of the left calf, and the front side of the right calf; The center of gravity calculation unit adopts a calculation method based on a multibody dynamics model, which simplifies the human body into five rigid body segments: torso, left thigh, right thigh, left calf, and right calf. The formulas for calculating the anterior-posterior position Xcog and the height Zcog of the overall center of gravity in the sagittal plane are as follows: Xcog=[mt×rt×sin(θt)+mlt×rlt×sin(θlt)+mrt×rrt×sin(θrt)+mls×rls×sin(θls)+mrs×rrs×sin(θrs)] / Mtotal; Zcog=[mt×rt×cos(θt)+mlt×rlt×cos(θlt)+mrt×rrt×cos(θrt)+mls×rls×cos(θls)+mrs×rrs×cos(θrs)] / Mtotal; Wherein, mt, mlt, mrt, mls, and mrs are the masses of the torso, left thigh, right thigh, left calf, and right calf, respectively; rt, rlt, rrt, rls, and rrs are the distances from the center of mass of each segment to the corresponding joint; and θt, θlt, θrt, θls, and θrs are the angles between the torso, left thigh, right thigh, left calf, and right calf and the vertical direction, respectively.
[0010] Furthermore, the posture detection module includes three inertial measurement units, which are respectively installed on the user's waist, the outer side of the left thigh, and the outer side of the right thigh; The center of gravity calculation unit simplifies the human body into a three-rigid-body segment model, which includes a torso segment, a left leg merging segment, and a right leg merging segment. The left leg merging segment is the merging of the left thigh, left calf, and left foot, and the right leg merging segment is the merging of the right thigh, right calf, and right foot. The formulas for calculating the anterior-posterior position Xcog and the height Zcog of the overall center of gravity in the sagittal plane are as follows: Xcog=[mt'×rt'×sin(θt)+ml'×rl'×sin(θlt)+mr'×rr'×sin(θrt)] / Mtotal'; Zcog=[mt'×rt'×cos(θt)+ml'×rl'×cos(θlt)+mr'×rr'×cos(θrt)] / Mtotal'; Where mt', ml', and mr' are the masses of the torso segment, the left leg combined segment, and the right leg combined segment, respectively; rt', rl', and rr' are the distances from the center of mass of each combined segment to the center of the hip joint, respectively; and θt, θlt, and θrt are directly measured by the three inertial measurement units. Furthermore, the posture detection module includes three inertial measurement units, which are respectively installed on the user's waist, the outer side of the left thigh, and the outer side of the right thigh, and the system also includes a plantar pressure sensor installed on the user's foot. The center of gravity calculation unit adopts a five-rigid-body segment calculation method based on a multibody dynamics model, and estimates the angles of the left and right lower legs based on the gait phase identified by the plantar pressure sensor through the knee joint locking angle model in the support phase and the knee joint flexion angle model in the swing phase, thereby calculating the overall center of gravity position. The plantar pressure sensor communicates wirelessly via Bluetooth.
[0011] Furthermore, the attitude detection module includes a single inertial measurement unit mounted on the user's waist; The center of gravity calculation unit uses the extended Kalman filter algorithm to estimate the overall center of gravity position trend based on the torso posture information measured by the individual inertial measurement unit. The sampling frequency of the single inertial measurement unit is 200 Hz.
[0012] Furthermore, the center of gravity calculation unit defines the convex hull region formed by the heels and toes of the left and right feet when standing on both feet as a stable support polygon; When the center of gravity projection exceeds the outward extension boundary of the stable support polygon by 0.1m, it is determined to be a dangerous state and the safety locking mechanism is triggered. The risk prediction unit uses an extended Kalman filter algorithm to predict the centroid position within the next 200ms. The calculation formulas for the forward tilt risk index (RFI) and the backward tilt risk index (RBI) are as follows: When Xcog > Xfront, RFI = (Xcog - Xfront) / (Xfront_limit - Xfront); When Xcog < Xback, RBI = (Xback - Xcog) / (Xback - Xback_limit); Among them, Xcog is the position before and after the predicted center of gravity, Xfront and Xback are the front boundary coordinates and the back boundary coordinates of the stable support polygon respectively, and Xfront_limit and Xback_limit are the front boundary coordinates and the back boundary coordinates of the extended boundary respectively.
[0013] Furthermore, both the left hip joint motor and the right hip joint motor are brushless DC motors, and both are equipped with harmonic reducers with a reduction ratio of 20:1; The first threshold is 0.3, and the second threshold is 0.8; In the safety locking mechanism, the electromagnet is a push-pull type DC electromagnet, the wedge angle of the wedge-shaped lock block is 15°, and the effective locking length of the locking rack is 30% of the perimeter of the waist support frame; When the hip joint dual-motor cooperative driving unit performs anti-fall protection, it adopts a progressive torque output strategy, and its torque output formula is: T = Tpeak × k × R; Among them, Tpeak is the peak torque of the motor, k is the gain coefficient, and R is the forward tilt risk index or the backward tilt risk index.
[0014] Furthermore, the hip joint dual-motor cooperative driving unit adopts a three-layer closed-loop control architecture; Among them, the outer loop is a position loop that calculates the target attitude angle based on the attitude information and outputs a torque command, the middle loop is a speed loop that calculates the target angular velocity according to the torque command output by the position loop, and the inner loop is a current loop that controls the motor current; When the forward tilt risk index or the backward tilt risk index exceeds the first threshold, the control law switches to the torque output mode, and the target attitude angle of the position loop automatically switches to the anti-fall target angle.
[0015] Furthermore, the system is built-in with a dual-motor cooperative fault detection mechanism; When it is detected that the current output difference between the left hip joint motor and the right hip joint motor exceeds 50% and the duration exceedsWhen a collaborative drive failure is detected, the wedge-shaped locking block on either side forms a self-locking engagement with the locking rack, which can provide sufficient torque to support the user.
[0016] Another objective of this invention is to provide a method for precisely controlling the execution of fall protection by the aforementioned system, thereby overcoming the shortcomings of slow response in existing passive protection and achieving a highly efficient closed-loop fall prevention intervention. Therefore, based on the above technical solution, this invention also proposes the following technical solution: A method for preventing forward and backward tilting falls, applied to the aforementioned method for preventing forward and backward tilting falls, the method comprising the following steps: The posture detection module collects posture information of the user's body parts in real time. The center of gravity calculation unit calculates the user's overall center of gravity position in real time based on the posture information; The risk prediction unit predicts the forward or backward tilt risk based on the overall center of gravity position and movement trend, and calculates the forward tilt risk index or backward tilt risk index respectively. When a risk of forward tilting is detected and the forward tilting risk index exceeds a first threshold, the left hip joint motor and the right hip joint motor are controlled to output a counter-extension torque in coordination. When a risk of backward tilting is detected and the risk index of backward tilting exceeds the first threshold, the left hip joint motor and the right hip joint motor are controlled to output a counter-flexion torque in coordination. When the forward tilt risk index or the backward tilt risk index exceeds the second threshold, the safety locking mechanism is triggered, the electromagnet power supply is cut off to release the wedge-shaped locking block, so that the wedge-shaped locking block and the locking rack form a self-locking mechanism, thereby rigidly connecting the exoskeleton to the user's waist.
[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: The design does not rely on independent protection motors. Instead, the assist motors of the left and right hip joints are used as fall protection actuators. Through a dual-motor collaborative control strategy, the system automatically adjusts the hip joint torque output when it detects a risk of leaning forward or backward. This eliminates the need for two additional independent protection motors and their associated drive circuits, transmission mechanisms, and mounting brackets. Compared to the traditional four-motor solution, this design achieves bidirectional fall protection while reducing the number of motors by 50%, significantly reducing the overall weight of the system and improving the simplicity and portability of the exoskeleton system. By introducing a multibody dynamics model based on five rigid body segments of the human body and combining it with the extended Kalman filter (EKF) algorithm to perform in-depth calculation of the historical trajectory of the center of gravity, the system can accurately predict the position of the center of gravity within the next 200ms. By calculating the forward tilt risk index or backward tilt risk index in real time, the system can intervene in advance before the center of gravity completely leaves the stable support polygon, thereby ensuring that the total response time from risk prediction calculation, motor drive response to mechanical transmission delay is strictly controlled within 50ms. The dual-motor collaborative control of the hip joint has constructed a three-layer closed-loop control architecture consisting of a position loop, a speed loop, and a current loop. When the risk index exceeds the trigger threshold, the control law can instantly switch from the normal state to the anti-fall torque output mode. During the protective torque release phase, the system strictly follows the progressive torque output strategy with muscle-like force characteristics. The torque output is relatively gentle in the early stage and dynamically increases as the risk index increases, effectively preventing sudden peak torque output from causing strong rigid impact and secondary injury to the user's hip bones and muscles. To cope with extreme unstable operating conditions, this system is specially designed with dual redundancy defenses. On the one hand, it has a built-in dual-motor collaborative fault detection mechanism. By comparing the differences in output between the two motors in real time (when the difference exceeds 50% and lasts for more than 200ms), it can forcibly determine the fault and trigger locking when a collaborative abnormality is detected. On the other hand, it is equipped with a safety locking mechanism consisting of an electromagnet and a wedge-shaped locking block. When the risk index exceeds the emergency threshold or a serious fault is triggered, the control unit cuts off the power to the electromagnet. The locking block quickly wedges into the locking rack under the action of the spring. Utilizing the self-locking effect of the 15° wedge angle, it can provide strong support of more than 50Nm within a response time of less than 10ms, forming a reliable rigid lock between the exoskeleton and the user's waist. It can still provide absolute physical safety in the event of any power or motor failure. Attached Figure Description
[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.
[0019] Figure 1 This is a diagram showing the overall hardware topology of the anti-tipping and fall protection system of the present invention. Figure 2 This is a schematic diagram of the sensor arrangement of the attitude detection module of the present invention; Figure 3 This is a flowchart of the center of gravity calculation and risk prediction algorithm of the present invention; Figure 4 This is a cross-sectional structural diagram of the safety locking mechanism of the present invention; Figure 5 This is a block diagram illustrating the principle of the three-layer closed-loop control and cooperative drive of the present invention. Figure 6This is a flowchart of the anti-tipping and fall protection method of the present invention.
[0020] In the diagram: 11. Waist inertial measurement unit; 12. Left thigh inertial measurement unit; 13. Right thigh inertial measurement unit; 14. Left calf inertial measurement unit; 15. Right calf inertial measurement unit; 50. Waist support frame; 51. Electromagnet; 52. Wedge-shaped locking block; 53. Locking rack; 54. Spring. Detailed Implementation
[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0022] The anti-tipping and fall protection system of the present invention includes a waist support frame 50, a left hip joint, a right hip joint, a left thigh connecting rod, a right thigh connecting rod, a posture detection module, a control unit, a left hip joint motor, a right hip joint motor, and a safety locking mechanism.
[0023] Specifically, the overall hardware topology of the system is as follows: Figure 1 As shown below, in conjunction with Figure 1 The interaction relationships of the various components of this system are summarized below. This invention does not set up independent forward tilt protection motors and backward tilt protection motors, but makes full use of the existing assist motors of the left and right hip joints to achieve the fall prevention function through coordinated control.
[0024] When a risk of forward tilting is detected, the bilateral hip joint motors simultaneously output torque in the extension direction, generating an effect that pulls the torso backward. When a risk of backward leaning is detected, the bilateral hip joint motors simultaneously output torque in the flexion direction, generating an effect that pushes the torso forward.
[0025] The lumbar support frame is made of carbon fiber composite material, with a thickness of 3mm and a total weight of approximately 800g. A flexible, breathable pad, 8mm thick, is attached to the inner surface of the frame. This pad is made of highly elastic sponge material with a hardness of 25~35 Shore A.
[0026] The lumbar frame is secured to the user's waist with a three-point binding strap. The binding strap is 50mm wide and uses Velcro to adjust the tightness. The left and right hip joints have the same structure and both adopt a dual-degree-of-freedom design, supporting flexion or extension movements and abduction or adduction movements.
[0027] The flexion / extension degrees of freedom are driven by a brushless DC motor with a rated torque of 10 Nm and a reduction ratio of 20:1, using a Harmonic Drive reducer to minimize backlash. The hip joint range of motion is limited to flexion 0°~120°, extension 0°~15°, abduction 0°~45°, and adduction 0°~30°.
[0028] The control unit is mounted on the back of the waist support frame, measuring 80mm×60mm×30mm. It uses an ARM Cortex-M4 processor with a main frequency of 168MHz and has a built-in FPU (floating-point unit) for running real-time attitude calculation and control algorithms. The control unit is equipped with a CAN bus interface supporting a communication rate of 1Mbps and a UART debugging interface. It is powered by a 2000mAh lithium battery pack and can work continuously for more than 8 hours. Compared with the solution that requires 4 motors, this system saves 2 motors and their drive circuits, transmission mechanisms and mounting brackets, and is expected to reduce the weight by about 400g.
[0029] like Figure 2 As shown, in one multi-sensor implementation, the attitude detection module includes five inertial measurement units (IMUs), which are respectively installed at the center of the waist, the outer side of the left thigh, the outer side of the right thigh, the front side of the left calf, and the front side of the right calf.
[0030] The installation locations and orientations of each IMU are as follows: The lumbar inertial measurement unit 11 is installed in the center of the back of the lumbar support frame 50, with its X-axis pointing directly in front of the human body, its Y-axis pointing to the left side of the human body, and its Z-axis pointing upwards. The left thigh inertial measurement unit 12 is installed on the outer side of the upper end of the left thigh connecting rod, and the right thigh inertial measurement unit 13 is installed on the outer side of the upper end of the right thigh connecting rod. The X-axis of the left thigh inertial measurement unit 12 and the right thigh inertial measurement unit 13 both point to the front of the human body, the Y-axis both point to the outer side of the thigh, and the Z-axis both point upward. The left lower leg inertial measurement unit 14 is installed at the front end of the left lower leg connecting rod, and the right lower leg inertial measurement unit 15 is installed at the front end of the right lower leg connecting rod. The X-axis of both the left lower leg inertial measurement unit 14 and the right lower leg inertial measurement unit 15 points directly in front of the human body, the Y-axis points to the outside of the lower leg, and the Z-axis points downward.
[0031] Each IMU uses an MPU-6050 six-axis inertial measurement chip, with a gyroscope range of ±2000° / s and a resolution of 16.65° / s; an accelerometer range of ±16g and a resolution of 0.488mg; and an IMU sampling frequency of 100Hz. The raw data is transmitted to the control unit via the I2C bus.
[0032] The plantar pressure sensor is embedded in the heel and arch area of the user's insole and communicates wirelessly with the control unit via Bluetooth at a sampling frequency of 50Hz. It is used to detect the foot's ground contact status in real time and identify the support phase and the swing phase.
[0033] The installation orientation of the three IMUs is consistent with the five-sensor implementation: the X-axis of each IMU points directly in front of the human body, and the Z-axis points upwards. Each IMU has a sampling frequency of 100Hz and communicates with the control unit via a CAN bus.
[0034] The center of gravity calculation unit still adopts the five-rigid-body segment calculation method based on the multibody dynamics model, simplifying the human body into five rigid-body segments: torso, left thigh, right thigh, left calf, and right calf. The difference from the five-sensor implementation method is that the posture of the left and right calf is not directly measured by the IMU, but is estimated based on the gait phase identified by the plantar pressure sensor, through the following kinematic constraints and empirical model phase state estimation.
[0035] Estimation of lower leg angle in the support phase: When the plantar pressure sensor detects that one foot is in the support phase, the knee joint on that side is in a nearly extended locked state. The angle θshank between the lower leg and the vertical direction and the thigh angle θthigh satisfy: θshank=θthigh-φlock, where φlock is the knee joint locking angle in the support phase, with a typical value range of 2°~8°, and a default value of 5°, which can be finely adjusted through calibration experiments.
[0036] Estimation of lower leg angle during the swing phase: When the plantar pressure sensor detects that one foot is in the swing phase, the knee flexion angle on that side changes sinusoidally with the progress of the swing: φknee(t) = φmax × sin(π × t / Tswing), where φmax is the maximum knee flexion angle, typically 65°; Tswing is the swing phase period, estimated in real time by the gait phase recognition unit based on historical gait data; and t is the timing from when the foot leaves the ground. Therefore, the lower leg angle during the swing phase is: θshank = θthigh - φknee(t).
[0037] After obtaining the angles of the left and right lower legs, the anterior-posterior position Xcog and height Zcog of the overall center of gravity in the sagittal plane are calculated using the weighted centroid formula of the five-rigid-body model: Xcog=[mt×rt×sin(θt)+mlt×rlt×sin(θlt)+mrt×rrt×sin(θrt)+mls×rls×sin(θls)+mrs×rrs×sin(θrs)] / Mtotal; Zcog=[mt×rt×cos(θt)+mlt×rlt×cos(θlt)+mrt×rrt×cos(θrt)+mls×rls×cos(θls)+mrs×rrs×cos(θrs)] / Mtotal; The mass and centroid position parameters of each segment are the same as those in the five-sensor implementation method.
[0038] Since the lower leg angle is an estimated value, an extended Kalman filter (EKF) is introduced to correct the center of gravity position online. The EKF observations use the trunk tilt angle and the position of the plantar pressure center measured by the lumbar IMU. The state vector includes the forward and backward position of the center of gravity, velocity, and acceleration, which can control the center of gravity estimation error of the three-sensor scheme within ±2cm.
[0039] In another embodiment, when no plantar pressure sensor is configured, the center of gravity calculation unit simplifies the human body into a three-rigid-body segment model, merging the mass of the left lower leg and left foot into the left thigh to form the left leg merged segment, and merging the mass of the right lower leg and right foot into the right thigh to form the right leg merged segment; the overall center of gravity position is calculated directly based on the trunk angle measured by the waist IMU and the thigh angle measured by the left and right thigh IMUs, without the need to estimate the lower leg angle.
[0040] The distances rl' from the center of mass of the left leg combined segment to the center of the hip joint and rr' from the center of mass of the right leg combined segment to the center of the hip joint were determined through a static calibration experiment: the subject stood upright on one leg, the trunk tilt angle θt was measured by the lumbar IMU, and the thigh tilt angle θlt or θrt was measured by the thigh IMU. The position of the ground reaction force application point was obtained by the force table, and rl' and rr' were obtained by combining the inverse solution of the overall center of gravity balance equation. In the absence of personalized calibration, the typical values can be taken as rl'=rr'=0.35m~0.40m.
[0041] Attitude calculation uses a direction cosine matrix algorithm to fuse data. The attitude angle output frequency is 100Hz, the angle resolution is better than 0.01°, and the dynamic angle error is less than 0.5°. The raw data is preprocessed by a first-order Butterworth low-pass filter with a cutoff frequency of 20Hz.
[0042] Before the system is powered on, zero bias calibration is required. Users are required to maintain an upright posture for 3 seconds. The system automatically records the zero bias values of each sensor and stores them in the EEPROM, which is automatically loaded each time the system is started.
[0043] In a simplified single-sensor implementation, only a single waist IMU is configured, with its sampling frequency increased to 200Hz. An extended Kalman filter (EKF) algorithm is used to predict the torso posture, thereby estimating the overall center of gravity position trend. This is suitable for application scenarios with relatively low protection requirements.
[0044] like Figure 3 As shown, the center of gravity calculation unit adopts a calculation method based on a multibody dynamics model: The human body is simplified into five rigid body segments: torso, left thigh, right thigh, left calf, and right calf. The parameters are configured as follows: The trunk mass is 0.5 times the body weight, and the center of gravity is located at the height of the third lumbar vertebra, approximately 0.15m from the center of the hip joint; The mass of one thigh is 0.1 times the body weight, and the center of mass is about 0.2m away from the center of the hip joint; the mass of one calf is 0.05 times the body weight, and the center of mass is about 0.15m away from the center of the knee joint.
[0045] Let θt be the angle between the torso and the vertical (a positive value indicates forward lean), θlt be the left thigh, θrt be the right thigh, θls be the left calf, and θrs be the right calf. The formulas for calculating the anterior-posterior position and height of the overall center of gravity in the sagittal plane are as follows: Xcog=[mt×rt×sin(θt)+mlt×rlt×sin(θlt)+mrt×rrt×sin(θrt)+mls×rls×sin(θls)+mrs×rrs×sin(θrs)] / Mtotal; Zcog=[mt×rt×cos(θt)+mlt×rlt×cos(θlt)+mrt×rrt×cos(θrt)+mls×rls×cos(θls)+mrs×rrs×cos(θrs)] / Mtotal.
[0046] In this system, the Z-axis is positive upwards and the X-axis is positive forwards. The center of gravity height Zcog and position Xcog are calculated every 10ms. The stable support polygon is defined as the convex hull area formed by the heels and toes of the left and right feet when standing with both feet. The front boundary is approximately 0.1m in front of the toes, the rear boundary is approximately 0.05m behind the heels, and the left and right boundaries are approximately half the width of the foot (approximately 0.05m). When the center of gravity projection is within this polygon, it is considered to be in a stable state. When it exceeds the polygon boundary but does not exceed the extended boundary by 0.1m, it is considered to be in a warning state. When it exceeds the extended boundary, it is considered to be in a dangerous state.
[0047] The risk prediction unit uses the Extended Kalman Filter (EKF) algorithm to predict the centroid position within the next 200ms. The state vector of the EKF is defined as follows: Where Xcog is the forward / backward position of the center of gravity, Vx is the forward / backward velocity of the center of gravity, Ax is the forward / backward acceleration of the center of gravity, and Zcog is the height of the center of gravity; the typical value of the process noise covariance matrix is Q=diag([0.01,0.1,1.0,0.01]), and the typical value of the observation noise covariance matrix is R=diag([0.05,0.05]); the state prediction model adopts a uniform acceleration model, and the observation values are the trunk tilt angle and the forward / backward position of the plantar pressure center measured by the waist IMU.
[0048] The formulas for the forward lean risk index RFI and the backward lean risk index RBI are as follows: When Xcog > Xfront, RFI = 10×(Xcog - Xfront); When Xcog < Xback, RBI = 10×(Xback - Xcog); When RFI or RBI exceeds the first threshold (e.g., 0.3), active protection is triggered; when it exceeds the second threshold (e.g., 0.8), emergency locking is triggered; the prediction time step of the extended Kalman filter algorithm is 50 ms, the total prediction duration is 200 ms, and prediction updates are performed every 10 ms; The left and right hip motors each have a rated torque of 10 Nm, a peak torque of 15 Nm, and a reduction ratio of 20:1. In the forward lean protection mode, the bilateral hip motors cooperate to output an extension torque to pull the thighs backward to prevent the body from leaning forward; In the backward lean protection mode, the bilateral hip motors cooperate to output a flexion torque to pull the thighs forward to prevent the body from leaning backward. To generate sufficient anti - fall torque, the torque output of the hip motors needs to satisfy: Thip = Tmotor×η×N; where Thip is the hip demand torque, Tmotor is the motor output torque, η is the transmission efficiency of about 0.8, and N is the reduction ratio of 20:!
[0049] The torque output adopts a progressive strategy: T = Tpeak×k×R; where Tpeak is the peak torque of 15 Nm, k is the gain coefficient (usually taken as 3), and R is the risk index. The total system response time from risk detection to torque output is less than 50 ms As Figure 4 shown, the safety locking mechanism includes push - pull DC electromagnets 51 (stroke 10 mm, pulling force 50 N, rated voltage 12 V, rated power 15 W) installed on both sides and wedge - shaped lock blocks 52. The wedge - shaped lock blocks 52 are made of quenched steel material, with a wedge angle of 15°, and cooperate with the locking rack 53 inside the waist support frame 50 to form self - locking. The effective locking length of the locking rack 53 is 30% of the circumference of the waist support frame 50.
[0050] When the risk index exceeds the second threshold, the electromagnet 51 is powered off, and the wedge - shaped lock block 52 pops forward under the action of the spring 54 and wedges into the locking rack 53. Since the 15° wedge angle is less than the friction angle, a self - locking effect is generated between the wedge - shaped lock block 52 and the locking rack 53, providing a support torque greater than 50 Nm. The locking response time is less than 10 ms, and the locking can be released by manually pulling the wedge - shaped lock block 52 backward.
[0051] The system state machine includes a normal state, a warning state, an active protection state, and an emergency locking state; The system also has multiple built-in safety mechanisms, including: automatically entering standby mode when the battery level is below 5% and retaining the last protection capability; A CAN communication timeout of 500ms is identified as a fault and the system switches to independent operation mode. When the motor temperature exceeds 70°C or the current exceeds 120% of the rated value for 1 minute, the output power will be automatically reduced to 50%. The output torque is automatically reduced when the motor stalls for more than 500ms.
[0052] The system supports zero-bias calibration and personalized adjustment of the stability boundary. In a scenario where a person leans forward and trips while walking on flat ground, if the risk prediction unit predicts that the center of gravity will exceed the stability support boundary within 200ms, the dual motors will output an extension torque of approximately 8Nm; if the RFI rises above 0.8, the electromagnet will be de-energized for rigid locking. In the scenario of losing balance while leaning backward while descending stairs, the dual motors immediately output the same flexion torque of about 7Nm, helping the user to regain a stable gait within 1 to 2 seconds; In rehabilitation training scenarios, the dual motors output a reverse torque of approximately 3~5Nm to provide auxiliary support.
[0053] like Figure 5 As shown in the further optimized embodiment, the underlying drive logic of the hip joint dual-motor cooperative drive unit adopts a three-layer closed-loop control architecture. Specifically, this architecture includes a position loop, a velocity loop, and a current loop from the outside in.
[0054] The outer ring, acting as the position ring, calculates the target attitude angle in real time based on the attitude information obtained by the attitude detection module. The middle ring acts as a velocity ring, calculating the target angular velocity in real time based on the torque command. The inner loop acts as a current loop, directly controlling the phase current output of the motor.
[0055] When the system is in normal state, the position loop is controlled by the user's normal walking gait phase; when the RFI or RBI is determined to exceed the first threshold and enters the active protection state, the control law will seamlessly switch to torque output mode. At this time, the outer loop's position target angle will be forcibly overwritten and automatically switched to the anti-fall target angle. Through the three-layer cascaded bottom-level switching mechanism, the power output conversion from "normal assistance" to "fall intervention" is completed in microseconds, eliminating torque abrupt changes or gaps during the transition phase.
[0056] In addition, to address the potential risks of unilateral failure or loss of synchronization in the dual-motor multiplexing architecture, this embodiment constructs an independent dual-motor collaborative fault detection mechanism and a physical redundancy locking architecture.
[0057] The control unit's underlying monitor continuously compares the real-time current feedback between the left and right hip joint motors. When the difference in current output (or torque difference) between the two sides exceeds 50% and this state persists for more than 200ms, the system determines that a "cooperative drive failure" has occurred (e.g., single-sided motor overheating shutdown or reducer tooth breakage). At this time, the redundancy protection logic has the highest execution priority, skipping the calculation results of the risk prediction unit and directly sending a power-off trigger command to the safety locking mechanism. Simultaneously, the safety locking mechanism employs a dual-electromagnetic-dual-locking-block redundancy design, with each side of the waist support frame 50 equipped with a set of locking block assemblies monitored by an independent channel. In extreme conditions such as cooperative failure or even single-sided mechanical structure damage, as long as the wedge-shaped locking block 52 on either side successfully forms a self-locking engagement with the locking rack 53, the structural resistance provided by that side is still sufficient to support the user's entire weight and prevent falls, thus achieving aviation-grade safety redundancy in the motor multiplexing scheme.
[0058] See Figure 6 This invention also provides a method for preventing forward and backward tilting falls. This method is applied to the aforementioned method for preventing forward and backward tilting falls, and its core execution flow is as follows: Figure 6 As shown, the specific steps include: Step S1: Collect posture information of the user's body parts in real time through the posture detection module; Step S2: The center of gravity calculation unit calculates the user's overall center of gravity position in real time based on the posture information; Step S3: Based on the overall center of gravity position and movement trend, the risk prediction unit predicts the forward or backward tilt risk and calculates the forward tilt risk index or backward tilt risk index respectively. Step S4: Determine whether the risk index exceeds the threshold. When a forward tilt risk is detected and the forward tilt risk index exceeds the first threshold, control the left hip joint motor and the right hip joint motor to output a counter-extension torque in coordination. When a backward tilt risk is detected and the backward tilt risk index exceeds the first threshold, control the left hip joint motor and the right hip joint motor to output a counter-flexion torque in coordination. Step S5: When the forward tilt risk index or the backward tilt risk index exceeds the second threshold, the safety locking mechanism is triggered, the power supply of the electromagnet 51 is cut off to release the wedge-shaped locking block 52, so that the wedge-shaped locking block 52 forms a self-locking with the locking rack 53 under the action of the spring 54, thereby rigidly connecting the waist support frame 50 to the user's waist.
[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A fall protection system to prevent forward and backward tilting, characterized in that, include: The posture detection module includes at least one inertial measurement unit for real-time acquisition of posture information of the user's body parts; A center of gravity calculation unit is used to calculate the overall center of gravity position of the user in real time based on the posture information. The risk prediction unit is used to predict the risk of leaning forward or backward based on the overall center of gravity position and movement trend, and to calculate the risk index of leaning forward or backward respectively. The hip joint dual-motor coordinated drive unit is configured with only a left hip joint motor and a right hip joint motor as the drive power source and fall protection actuator. The left hip joint motor and the right hip joint motor are used to drive the flexion or extension movements of the left and right hip joints, respectively. When the forward tilt risk index exceeds a first threshold, the two hip joint motors work together to output a reverse extension torque. When the backward tilt risk index exceeds the first threshold, the two hip joint motors work together to output a reverse flexion torque. The safety locking mechanism is structurally installed on both sides of the lumbar support frame and includes an electromagnet, a spring, and a wedge-shaped locking block, as well as a locking rack distributed in the inner circle of the lumbar support frame. When the forward tilt risk index or the backward tilt risk index exceeds a second threshold, the safety locking mechanism controls the electromagnet to de-energize to release the wedge-shaped locking block, so that the wedge-shaped locking block forms a self-locking engagement with the locking rack under the action of the spring, thereby rigidly connecting the lumbar support frame to the user's waist.
2. The anti-fall-over protection system according to claim 1, characterized in that, The posture detection module includes multiple inertial measurement units distributed on the user's torso and both lower limbs; The plurality of inertial measurement units includes five inertial measurement units, which are respectively installed on the user's waist center, the outer side of the left thigh, the outer side of the right thigh, the front side of the left calf, and the front side of the right calf; The center of gravity calculation unit adopts a calculation method based on a multibody dynamics model, which simplifies the human body into five rigid body segments: torso, left thigh, right thigh, left calf, and right calf. The formulas for calculating the anterior-posterior position Xcog and the height Zcog of the overall center of gravity in the sagittal plane are as follows: Xcog=[mt×rt×sin(θt)+mlt×rlt×sin(θlt)+mrt×rrt×sin(θrt)+mls×rls×sin(θls)+mrs×rrs×sin(θrs)] / Mtotal; Zcog=[mt×rt×cos(θt)+mlt×rlt×cos(θlt)+mrt×rrt×cos(θrt)+mls×rls×cos(θls)+mrs×rrs×cos(θrs)] / Mtotal; Wherein, mt, mlt, mrt, mls, and mrs are the masses of the torso, left thigh, right thigh, left calf, and right calf, respectively; rt, rlt, rrt, rls, and rrs are the distances from the center of mass of each segment to the corresponding joint; and θt, θlt, θrt, θls, and θrs are the angles between the torso, left thigh, right thigh, left calf, and right calf and the vertical direction, respectively.
3. The anti-fall-over protection system according to claim 1, characterized in that, The posture detection module includes three inertial measurement units, which are respectively installed at the midpoint of the user's waist, the outer side of the left thigh, and the outer side of the right thigh; The center of gravity calculation unit simplifies the human body into a three-rigid-body segment model. The three-rigid-body segments include a torso segment, a combined left leg segment, and a combined right leg segment. The combined left leg segment is the combination of the left thigh, left calf, and left foot, and the combined right leg segment is the combination of the right thigh, right calf, and right foot; The calculation formulas for the front-back position Xcog and height Zcog of the overall center of gravity in the sagittal plane are respectively: Xcog = [mt'×rt'×sin(θt) + ml'×rl'×sin(θlt) + mr'×rr'×sin(θrt)] / Mtotal'; Zcog = [mt'×rt'×cos(θt) + ml'×rl'×cos(θlt) + mr'×rr'×cos(θrt)] / Mtotal'; Where, mt', ml', and mr' are the masses of the torso segment, the combined left leg segment, and the combined right leg segment respectively, rt', rl', and rr' are the distances from the centers of mass of each combined segment to the hip joint center respectively, and θt, θlt, and θrt are directly measured by the three inertial measurement units.
4. The anti-fall-over protection system according to claim 1, characterized in that, The posture detection module includes three inertial measurement units, which are respectively installed at the midpoint of the user's waist, the outer side of the left thigh, and the outer side of the right thigh, and the system further includes a plantar pressure sensor installed on the user's sole; The center of gravity calculation unit adopts a five-rigid-body segment calculation method based on a multi-body dynamics model, and estimates the left calf angle and the right calf angle through a support-phase knee joint locking angle model and a swing-phase knee joint flexion angle model based on the gait phase identified by the plantar pressure sensor, and then calculates the overall center of gravity position; The plantar pressure sensor performs wireless communication via Bluetooth.
5. The anti-fall-over protection system according to claim 1, characterized in that, The posture detection module includes a single inertial measurement unit, which is installed on the user's waist; The center of gravity calculation unit adopts an extended Kalman filter algorithm to estimate the trend of the overall center of gravity position according to the torso posture information measured by the single inertial measurement unit; The sampling frequency of the single inertial measurement unit is 200Hz.
6. The anti-fall-over protection system according to claim 1, characterized in that, The center of gravity calculation unit defines the convex hull area formed by the left and right heels and toes during double-foot standing as the stable support polygon; When the center of gravity projection exceeds the extended boundary of the stable support polygon boundary by 0.1m and extends outward, it is determined as a dangerous state and the safety locking mechanism is triggered; The risk prediction unit adopts an extended Kalman filter algorithm to predict the predicted center of gravity position within the next 200ms. The calculation formulas for the forward tilt risk index RFI and the backward tilt risk index RBI are respectively: When Xcog > Xfront, RFI = (Xcog - Xfront) / (Xfront_limit - Xfront); When Xcog < Xback, RBI = (Xback - Xcog) / (Xback - Xback_limit); Where Xcog is the predicted front and back position of the centroid, Xfront and Xback are the front and back boundary coordinates of the stable support polygon, respectively, and Xfront_limit and Xback_limit are the front and back boundary coordinates of the extended boundary, respectively.
7. The anti-fall-over protection system according to claim 1, characterized in that, Both the left hip joint motor and the right hip joint motor are brushless DC motors and are equipped with harmonic reducers with a reduction ratio of 20:
1. The first threshold is 0.3, and the second threshold is 0.8; In the safety locking mechanism, the electromagnet is a push-pull type DC electromagnet, the wedge angle of the wedge-shaped locking block is 15°, and the effective locking length of the locking rack is 30% of the circumference of the waist support frame; The hip joint dual-motor collaborative drive unit adopts a progressive torque output strategy when performing fall protection, and its torque output formula is: T=Tpeak×k×R; Where Tpeak is the peak torque of the motor, k is the gain coefficient, and R is the forward tilt risk index or the backward tilt risk index.
8. The anti-fall-over protection system according to claim 1, characterized in that, The hip joint dual-motor collaborative drive unit adopts a three-layer closed-loop control architecture; The outer loop is a position loop that calculates the target attitude angle based on the attitude information and outputs a torque command; the middle loop is a velocity loop that calculates the target angular velocity based on the torque command output by the position loop; and the inner loop is a current loop that controls the motor current. When the forward tilt risk index or the backward tilt risk index exceeds the first threshold, the control law switches to torque output mode, and the target attitude angle of the position loop automatically switches to the anti-fall target angle.
9. The anti-fall-over protection system according to claim 1, characterized in that, The system has a built-in dual-motor collaborative fault detection mechanism. When the current output difference between the left hip joint motor and the right hip joint motor is detected to exceed 50% and the duration exceeds 200ms, it is determined to be a collaborative drive failure, and the safety locking mechanism is directly triggered independently of the forward tilt risk index and the backward tilt risk index. The safety locking mechanism adopts a redundant design of dual electromagnets and dual locking blocks. Each of the left and right sides of the waist support frame is equipped with an independently monitored electromagnet and a wedge-shaped locking block. When a collaborative drive failure is detected, the wedge-shaped locking block on either side forms a self-locking engagement with the locking rack, which can provide sufficient torque to support the user.
10. A method for preventing forward and backward tilting falls, characterized in that, Applied to the anti-fall-over protection system as described in any one of claims 1 to 9, the method comprises the following steps: The posture detection module collects posture information of the user's body parts in real time. The center of gravity calculation unit calculates the user's overall center of gravity position in real time based on the posture information. The risk prediction unit predicts the forward or backward tilt risk based on the overall center of gravity position and movement trend, and calculates the forward tilt risk index or backward tilt risk index respectively. When a risk of forward tilting is detected and the forward tilting risk index exceeds a first threshold, the left hip joint motor and the right hip joint motor are controlled to output a counter-extension torque in coordination. When a risk of backward tilting is detected and the risk index of backward tilting exceeds the first threshold, the left hip joint motor and the right hip joint motor are controlled to output a counter-flexion torque in coordination. When the forward tilt risk index or the backward tilt risk index exceeds the second threshold, the safety locking mechanism is triggered, the electromagnet power supply is cut off to release the wedge-shaped locking block, and the wedge-shaped locking block forms a self-locking mechanism with the locking rack under the action of the spring, thereby rigidly connecting the waist support frame to the user's waist.