Method for controlling knee joint of artificial limb

Abstract

The invention relates to a control method of an artificial limb knee joint. According to the technical scheme, the control method comprises the steps that a heel pressure sensor signal Ph, a sole pressure sensor signal Pf and an inclination angle theta and an angular velocity omega output by a shank inertial measurement unit are obtained in real time; when it is detected that the Pf decline rate exceeds a preset multiple threshold value of the heel pressure decline rate, dynamically calibrating an inclination angle reference theta base and calculating an offset theta cal; dividing six stages of a gait period based on the theta < cal > and the pressure signal, and calculating period duration T; a step speed-damping mapping knowledge base is constructed, and the damping opening degree in the early stage of swing is optimized through iterative learning; finally, the adjusted U value is applied in the early stage of swinging, and fixed damping is kept in other stages. The method is mainly used for improving the walking naturalness and terrain adaptability of the artificial limb at different speeds.

Description

Technical Field

[0001] The present invention relates to the technical field of lower limb prosthetic motion control, and more particularly to a control method for a prosthetic knee joint. Background Art

[0002] Currently, some prosthetic knee joint control devices utilize a fixed damping parameter strategy. This strategy has limitations when the wearer's gait speed changes: when walking speed increases significantly, the prosthetic knee joint's flexion angle during the swing phase may be insufficient, resulting in a shortened stride and a stiff gait. When walking speed slows significantly, excessive flexion may occur during the swing phase, affecting gait stability. These issues arise from the inability of fixed damping parameters to adapt to the changing biomechanical demands of knee joint motion at different gait speeds.

[0003] The main reason for this limitation is that during physiological gait, the damping force required by the knee joint during the swing phase exhibits a nonlinear relationship with gait speed. Conventional methods struggle to accurately detect changes in gait speed in real time and dynamically adjust the damping parameters accordingly. Furthermore, due to individual gait differences and the variability of walking environments, establishing a universally applicable gait speed-damping mapping is challenging. Existing methods have yet to effectively address these issues in achieving adaptive matching of damping parameters to gait speed. Summary of the Invention

[0004] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0005] Another object of the present invention is to provide a control method for a prosthetic knee joint, which solves the problem of unnatural gait and insufficient safety caused by the inability of traditional prostheses to adapt to changes in gait speed.

[0006] In order to achieve these purposes and other advantages according to the present invention, a control method for a prosthetic knee joint is provided, comprising: S1. Real-time acquisition of heel pressure sensor signals P h , Foot pressure sensor signal P f and the tilt angle output by the calf inertial measurement unit i and angular velocity ω; S2, when detected P f The rate of descent v f > k ⋅ v h When the real-time tilt angle i now Defined as the new 0° reference, denoted as i base = i now ; v h for P h Descent rate, k =2 is the preset multiple threshold; calculate the tilt angle offset after calibration i cal , i cal = i - i base ; S3, based on i cal The gait cycle is divided into early stance, mid-stance, late stance, early swing, mid-swing and late swing based on the dual pressure signals; Calculate the gait cycle duration based on the interval between adjacent early support starting points T ; S4. Establish and call the speed-damping mapping knowledge base to construct the gait cycle duration T and the early damping opening of the swing U Mapping relationship library; detect the current T value and query the pace-damping mapping knowledge base, if there is a match T , directly call the corresponding U value; if unknown T , then load T The closest known pace corresponds to U As initial value; Perform iterative learning on the unknown T: U k+1 =U k +L(θ target - i kmax ) in U is the damping opening, L is the learning rate, i target The target flexion angle is 65°. i kmax For the k The maximum flexion angle measured during the period; When the buckling angle error converges to i target - i kmax When |≤5°, stop iteration and store the mapping relationship; S5, damping opening after adjustment in the early stage of swing U; Maintain fixed damping in the early support stage, middle support stage, late support stage and middle and late swing stage.

[0007] Preferably, the control method of the prosthetic knee joint is based on i cal The gait cycle is divided into early support, mid support, late support, early swing, mid swing and late swing based on the dual pressure signals. Specifically: when P h ≥ P hmax and i cal When the angle is less than -3° (heel touching the ground → full foot bearing pressure starts), it is judged as the early support stage; when P h >50%× P hmax 、 P f >50%× P fmax and i cal When the rate of change is <1° / s (the whole foot is in a stable pressure-bearing stage), it is determined to be the mid-stance stage; P hmax is the current gait cycle P h Peak, P fmax is the current gait cycle P f Peak value; when P h <30%× P hmax and i cal When it is >0° (heel off → before toe off), it is judged as the late support stage; when P f =0 and i cal When the increase was ≥10° within 100ms (toe off → calf accelerated forward swing), it was judged as the early swing stage; when i cal Continue to increase to the peak value and the angular velocity | oh ∣<1° / s When the lower leg swings forward to the maximum flexion angle, it is judged as the mid-swing period; when i cal The peak value detected from the current swing period i maxThe descent value is ≥3° and the angular velocity oh< Time 0 (calf deceleration and extension until next ground contact) is determined to be the late swing phase.

[0008] Preferably, the control method for the prosthetic knee joint performs the following optimization steps during the iterative learning process in step S4: S41. Using 65° as the initial target flexion angle, calculate in real time the personalized target flexion angle baseline value: baseline target flexion angle = adaptation coefficient × (step length / calf length); where step length is calculated based on the geometric relationship between the spatial distance between adjacent early support starting points and the calibration tilt angle; calf length is a preset value; and the adaptation coefficient ranges from 0.4 to 0.6. S42, when |reference target flexion angle -65°|>10°, replacing the target flexion angle in iterative learning with the reference target flexion angle; S43. When it is detected that the slope angle β is greater than 3° for more than 2 seconds, in the uphill scenario, the corrected target flexion angle = current target flexion angle × (1 + 0.2 × β); in the downhill scenario, the corrected target flexion angle = the maximum value (50°, current target flexion angle × (1 - 0.1 × |β|)).

[0009] Preferably, in the control method of the prosthetic knee joint, step S4 further comprises adjusting the learning rate according to the rate of change of the gait cycle duration: When the standard deviation of the gait cycle length of three consecutive cycles is less than 0.05s, the standard learning rate L is used; When the standard deviation is ≥ 0.05 seconds, the adaptive learning rate L is enabled. adaptive :L adaptive =L×[1+0.5×tanh(5×| T k - T k-1 | / T k-1 )];middle T k is the current cycle duration, T k-1 The duration of the previous cycle.

[0010] Preferably, in the control method for the prosthetic knee joint, step S4 further comprises, when the number of records in the pace-damping mapping knowledge base is less than 5, executing the following steps: For unknown gait cycle duration T Additional safety margin for the initial damping opening value: Initial damping opening = U closest ×min(1.3,1+0.1×| T - T closest | / Tclosest ); U closest is the closest damping opening, T closest is the closest gait cycle duration; Set the maximum number of iteration cycles to 10; When the number of iterations k>5 and |65°- i kmax When |>15°, reset the damping opening U k =0.5×( U min + U max );in U min and U max are the preset safe working range boundary values ​​respectively.

[0011] Preferably, the control method of the prosthetic knee joint, step S4 further comprises real-time monitoring of the pressure signal mutation characteristics: if the heel pressure P h Drop > 80% in 20ms P hmax , and the angular velocity is greater than 50° / s, the current iterative learning is interrupted and switched to the safety damping mode: U emergency =0.7× U min +0.3× U max , U emergency The emergency damping opening degree is restored after the terrain stabilizes. U k Continue iterating, the terrain stability determination must meet the following requirements: a) P h ≥40%× P hmax Up to 500ms; b) i cal The rate of change is <2° / s and the angular velocity oh The sampling value is <10° / s for three consecutive times; c) The starting point of the early support stage is detected.

[0012] Preferably, the control method of the prosthetic knee joint adds a damping gradient control in the transition stage from the late support stage to the early swing stage: When the calibration tilt angle is detected in the later stage of support i cal When the angle is greater than 20°, the linear transition of the damping opening is started: U transition = U support +(U - U support )×( i cal -20°) / ( i swingstart -20°), where U transition is the damping opening during the transition period, U support is the fixed damping value opening during the support period, U is the damping opening during the early swing period output by the current iterative learning, i swingstart The starting angle of the early swing phase of the current cycle; the transition process lasts until the end of the early swing phase, and then the full U value.

[0013] Preferably, the control method for the prosthetic knee joint adds sensor failure redundancy processing in S1: Real-time calculation of the heel pressure signal validity index, if 5 consecutive cycles P hmax <10N or signal variance <0.1N 2 , determine that the heel pressure sensor is faulty, and perform the following steps: The vertical acceleration component a is output by the inertial measurement unit installed at the proximal end of the thigh prosthesis. z Alternative P h , the touchdown threshold is set to |a z |>2g; The heel impact sound sensor installed on the inside of the prosthetic ankle is activated simultaneously. When the 200-500Hz sound pressure level is detected to be greater than 80dB and the sound pressure level is greater than 80dB, the z When the peak time difference is less than 50 ms, a touchdown event is confirmed; the backup signal touchdown event triggers the reference angle reset in step S2 and the stage division in step S3.

[0014] The present invention has at least the following beneficial effects: 1. The control method of the present invention effectively improves the adaptability of prosthetic walking by sensing changes in the wearer's gait speed in real time and dynamically adjusting the knee joint damping parameters. The core lies in establishing an adaptive mapping mechanism between gait cycle duration and damping opening during the swing phase, and introducing a closed-loop iterative learning strategy. When the wearer's gait speed changes, the system can autonomously optimize the damping opening, allowing the knee joint to approach the physiological flexion trajectory during the critical swing phase, significantly reducing the risk of gait disorders caused by sudden changes in gait speed. The control method of the present invention overcomes the inherent limitations of fixed damping parameters in variable speed scenarios, allowing the prosthesis to maintain natural swing characteristics at different walking rhythms. 2. For complex usage environments, this invention integrates multi-dimensional safety strategies to enhance system robustness. By monitoring the standard deviation of the gait cycle, adaptive learning rate adjustment is achieved to avoid control instability during drastic speed fluctuations. For unknown speed scenarios, additional damping safety margins and iteration interruption protection are implemented to prevent safety hazards caused by excessive flexion. Especially in sloping terrain, the target flexion angle is dynamically corrected based on the inclination angle to enhance knee support stability during slope walking. This control method collaboratively ensures a smooth transition from flat roads to complex terrain, reducing the need for active intervention by the wearer. 3. The present invention further improves reliability through sensor redundancy design and fine control of movement stages. When the heel pressure sensor fails, the acceleration of the inertial measurement unit and the heel impact soundprint signal are used to cross-verify the touchdown event to maintain the accuracy of gait phase division; damping gradient control is applied during the transition from the stance phase to the swing phase to eliminate the mechanical impact when the joint movement state is switched; the control method of the present invention reduces the dependence on a single sensor, so that the prosthesis can still maintain continuous and stable damping output characteristics in critical scenarios such as sensor abnormalities or movement phase conversion.

[0015] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. DETAILED DESCRIPTION

[0016] The present invention is further described in detail below with reference to the embodiments so that those skilled in the art can implement the invention with reference to the description.

[0017] It should be understood that terms such as “having”, “including” and “comprising” used herein do not preclude the existence or addition of one or more other elements or combinations thereof.

[0018] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0019] To achieve these objectives and other advantages, the present invention provides a method for controlling a prosthetic knee joint, comprising: S1. Real-time acquisition of heel pressure sensor signals P h , Foot pressure sensor signal P f and the tilt angle output by the calf inertial measurement unit i and angular velocity ω; S2, when detected P f The rate of descent v f >k ⋅ v h When the real-time tilt angle i now Defined as the new 0° reference, denoted as i base = i now ; v h for P h Descent rate, k =2 is the preset multiple threshold; calculate the tilt angle offset after calibration i cal , i cal = i - i base ; S3, based on i cal The gait cycle is divided into early stance, mid-stance, late stance, early swing, mid-swing and late swing based on the dual pressure signals; Calculate the gait cycle duration based on the interval between adjacent early support starting points T ; S4. Establish and call the speed-damping mapping knowledge base to construct the gait cycle duration T and the early damping opening of the swing U Mapping relationship library; detect the current T value and query the pace-damping mapping knowledge base, if there is a match T , directly call the corresponding U value; if unknown T , then load T The closest known pace corresponds to U As initial value; Perform iterative learning on the unknown T: U k+1 =U k +L(θ target - i kmax ) in U is the damping opening, L is the learning rate, i target The target flexion angle is 65°. i kmax For the k The maximum flexion angle measured during the period; When the buckling angle error converges to i target - i kmax When |≤5°, stop iteration and store the mapping relationship; S5, damping opening after adjustment in the early stage of swing U ; Maintain fixed damping in the early support stage, middle support stage, late support stage and middle and late swing stage.

[0020] In the above technical solution, the gait cycle duration refers to the time interval between two consecutive heel strike events, reflecting the walking speed; the damping opening is the flow adjustment parameter of the hydraulic valve or magnetorheological fluid device, and the damping decreases as the opening increases; the tilt angle offset is the angle between the calf axis and the vertical direction, and the installation error is eliminated after dynamic benchmark calibration; iterative learning is an adaptive process of cyclically optimizing the damping parameters based on historical gait data.

[0021] In this technical solution, heel and sole pressure sensors are installed in the prosthetic shank, along with an integrated inertial measurement unit. When the system detects that the rate of pressure drop at the sole significantly exceeds that at the heel (determined by a preset multiple threshold), it immediately sets the current inclination angle as the new reference angle. This is used to calculate a real-time offset. Based on this offset and the changing characteristics of the dual-path pressure signals, the gait is decomposed into six consecutive phases, and the gait cycle duration is calculated from the interval between the starting points of adjacent stance phases.

[0022] In this technical solution, the system incorporates a built-in knowledge base of gait speed-damping mappings, storing verified correlations between gait cycle durations and damping openings during the early swing phase. For known gait speeds, the system directly calls the matching damping value. For unknown gait speeds, the system loads the damping opening closest to the historical value as the initial value, and initiates iterative learning. Based on the deviation between the measured maximum flexion angle during the current swing phase and the target flexion angle, the system gradually adjusts the damping opening at the learning rate until the flexion angle deviation converges to a physiologically acceptable range over multiple consecutive cycles. Ultimately, the optimized damping value is applied only during the early swing phase, while the preset fixed damping is maintained during the rest of the phase.

[0023] Existing technologies use multiple preset damping modes (e.g., slow / medium / fast), which require manual switching or rely on crude acceleration threshold triggering. These limitations include: an inability to distinguish individual gait differences; wearers may exhibit over- or under-flexion in the same gear; parameter switching lags during variable speed walking, resulting in discontinuous gait; and a lack of continuous self-optimization capability, requiring repeated manual calibration.

[0024] This technical solution eliminates individual installation differences through dynamic reference angle calibration, uses gait cycle duration to accurately quantify gait speed, and combines closed-loop iterative learning to achieve autonomous progressive optimization of damping parameters, fundamentally overcoming the mechanical response defects of fixed gear switching.

[0025] This technical solution significantly improves the prosthesis's adaptability to changes in the wearer's walking rhythm through closed-loop control of gait speed sensing and damping adjustment. In scenarios with fluctuating gait speed, the knee joint's swing flexion angle automatically approaches the physiological range of motion, eliminating gait stiffness or instability caused by damping mismatch. An iterative learning mechanism ensures that the system completes parameter optimization within several gait cycles, reducing the need for manual intervention. A phased damping strategy balances swing flexibility and support stability, making the prosthetic movement trajectory closer to a natural gait.

[0026] In the above technical solution, the control method of the prosthetic knee joint includes: Signal acquisition and dynamic calibration: The prosthetic calf component integrates a piezoresistive thin film pressure sensor and a nine-axis inertial measurement unit. P h and foot pressure signals P f The data is collected in real time at a sampling rate of 100Hz and the mechanical vibration noise is eliminated by Kalman filtering. The inertial measurement unit calculates the tilt angle through the quaternion fusion algorithm. i and angular velocity oh , the output frequency is 200Hz. During dynamic calibration: calculate P f Descent rate v f =Δ P f / Δ t and P h Descent rate v h =Δ P h / Δ t (Time window Δ t =100ms); When satisfied v f >2 v h and v h When the speed is >5N / s, the tilt angle at that moment is locked. i now As a new benchmark i base ; Output calibration offset i cal = i - i base , eliminating the inclination deviation of prosthetic installation.

[0027] Gait cycle division and duration calculation: based on i cal And the dual pressure signal builds a six-stage state machine: Support early starting point: when P h ≥0.9 P hmax and i cal Timestamp when <-3° t 1 ; Mid-term support: Dual-path pressure > 50% of peak value and tilt angle change rate | d i cal / d t ∣<1° / s; Early swing phase: P f =0 and i cal Increase ≥10° within 100ms.

[0028] The interval between the early starting points of adjacent supports T = t 2- t 1 is the gait cycle duration, reflecting the real-time walking speed.

[0029] Damping Mapping and Iterative Learning The pace-damping knowledge base uses a hash table to store key-value pairs ( T i , U i ), initially presetting three sets of benchmark mappings (the damping opening U refers to the flow control parameter of the hydraulic valve, U=0 indicates that the damping is completely closed (maximum damping force 150N・m・s / rad), U=1 indicates that the damping is fully open (minimum damping force 20N・m・s / rad), and Usupport=0.2 corresponds to a fixed damping force of 100N・m・s / rad during the support period): T =1.5s→U=0.8 (slow) T =1.0s→U=0.5 (medium speed) T =0.6s→U=0.3 (fast) For unknown pace T : Load closest T closest Corresponding U closest As initial U 0; Perform iterative learning: U k+1= U k +λ(65°- i kmax ) ( l =0.1) in i kmax Indicates the k Calibrate the tilt angle offset during the swing phase of the gait cycle i cal The measured maximum value of When 3 consecutive cycles are satisfied, the new mapping is stored ( T , U k ).

[0030] Staged damping execution The hydraulic proportional valve switches the damping opening according to the gait stage: Stance phase (early / middle / late): maintain a fixed opening U support = 0.2 (high damping); in the early swing phase, the learning output value U∈[0.3,0.8] is applied; in the middle and late swing phase, the opening is fixed at Uswing = 0.7 (low damping). The phase switching command is issued within 5 ms of the boundary event trigger, and the PID controller ensures that the opening tracks the target value within 20 ms (error ± 0.02 ± 0.02).

[0031] In the above technical solution, dynamic reference calibration solves the problem of prosthesis wearing tilt i Measurement error (e.g. 5° deviation in varus and valgus); cycle duration T It is more resistant to stride length changes than acceleration amplitude and accurately quantifies stride speed. Iterative learning enables the damping parameters to converge within 3-5 steps under unknown stride speed. The six-stage damping distribution takes into account both support stability (high damping to prevent collapse) and swing efficiency (dynamic damping adapts to stride speed).

[0032] In another technical solution, the control method of the prosthetic knee joint is based on i cal The gait cycle is divided into early support, mid support, late support, early swing, mid swing and late swing based on the dual pressure signals. Specifically: when P h ≥ P hmax and i cal When the angle is less than -3° (heel touching the ground → full foot bearing pressure starts), it is judged as the early support stage; when P h >50%× P hmax 、 P f>50%× P fmax and i cal When the rate of change is <1° / s (the whole foot is in a stable pressure-bearing stage), it is determined to be the mid-stance stage; P hmax is the current gait cycle P h Peak, P fmax is the current gait cycle P f Peak value; when P h <30%× P hmax and i cal When it is >0° (heel off → before toe off), it is judged as the late support stage; when P f =0 and i cal When the increase was ≥10° within 100ms (toe off → calf accelerated forward swing), it was judged as the early swing stage; when i cal Continue to increase to the peak value and the angular velocity | oh ∣<1° / s When the lower leg swings forward to the maximum flexion angle, it is judged as the mid-swing period; when i cal The peak value detected from the current swing period i max The descent value is ≥3° and the angular velocity oh< At 0 (the moment the calf decelerates and extends until the next contact with the ground), it is determined to be in the late swing phase. In the above technical solution, the dynamic threshold for peak pressure is a percentage criterion based on the maximum pressure value measured by the sensor during the current gait cycle, eliminating the influence of individual pedaling force differences; the angle change rate is the rate of change of the tilt angle offset per unit time, which is used to identify motion steady-state; phase boundary events, such as heel strike and toe-off, mark key movements during gait phase transitions.

[0033] The above technical solution designs a multi-condition joint judgment criterion by synchronously analyzing three types of signals: heel pressure, sole pressure, and calibrated tilt angle offset: Early support judgment: The heel pressure must reach the cycle peak and the calf must be in a backward tilt state to ensure accurate ground contact event recognition; Mid-term support judgment: the dual-path pressure exceeds the half-peak value and the tilt angle change rate approaches zero, filtering out body shaking interference; Late support judgment: The heel pressure drops significantly to the low threshold range and the calf turns forward, accurately capturing the moment when the heel leaves the ground; Early swing judgment: the foot pressure returns to zero and the tilt angle increases rapidly, marking the starting point of the calf's accelerated forward swing; Mid-swing judgment: the tilt angle increases to the peak value and the angular velocity approaches zero, locking the maximum flexion position; Late swing judgment: The tilt angle decreases from the peak value and is accompanied by the extension angular velocity, predicting the next touchdown preparation posture.

[0034] Existing technologies use single-signal threshold segmentation (e.g., pressure > a fixed value is considered the stance phase). This has the following limitations: differences in pressure sensor sensitivity can easily lead to missed touchdown events (e.g., light touchdown scenarios); the angle change rate is not combined, making it impossible to distinguish between static pressure bearing (mid-stance phase) and dynamic transition (late-stance phase); and the swing phase division relies on the absolute value of the angle, ignoring individual differences in flexion range.

[0035] In this technical solution, the coupling of pressure and angle dual signals and the dynamic threshold design significantly improve the robustness of phase boundary recognition, especially for abnormal gait (such as limping) and complex terrain (such as soft ground); In this technical solution, the multi-dimensional signal fusion mechanism greatly improves the accuracy of gait phase division, provides a reliable timing benchmark for damping parameter switching, supports mid-term steady-state judgment to avoid false triggering of damping adjustment by small shaking, and ensures stability during the load-bearing period; rapid identification in the early stage of swing ensures timely reduction of damping and releases the flexion freedom of the knee joint; extension prediction in the late stage of swing optimizes the touchdown preparation posture; and overall achieves a high degree of synchronization between the prosthetic movement rhythm and the human gait, reducing energy loss caused by sudden changes in joint movement.

[0036] In the above technical solution, gait phase division includes: Signal synchronization and preprocessing: The prosthetic control system collects three signals synchronously at a sampling rate of 200Hz: heel pressure P h , sole pressure P f , calibrate the tilt angle offset i cal The pressure signal is filtered with a 10Hz low-pass filter to eliminate shock oscillations, and the tilt angle signal is smoothed with a fourth-order Butterworth filter (cut-off frequency 5Hz). Three types of derived parameters are calculated in real time: pressure peak P hmax =max( P h )and P fmax =max( P f ) (within the current cycle); rate of change of tilt angle ohθ = dth cal / d t (Calculated by central difference method); pressure drop rate v h =- dP h / dt , v f =- dP f / dt (100ms moving window).

[0037] Six-stage division logic implementation Support early judgment: when both of the following are met: P h ≥0.9 P hmax ; i cal<-3°; v h >5N / s; determine the heel touchdown event and mark the timestamp t heel-strike This condition ensures that light touchdown interference is excluded.

[0038] Support mid-term switching conditions: dual stability must be continuously met: P h >0.5 P hmax ; P f >0.5P fmax ; ∣oh θ ∣ <1° / s (lasting ≥300ms); the tilt angle change rate threshold filters out slight body movements.

[0039] Capturing the early swing boundary: meeting the characteristics of rapid departure: P f =0 (lasting ≥50ms); Δθ cal ≥10°(t∈[t0,t0+100ms]) and ω>30° / s; where Δ i cal = i cal ( t 0 +100ms) - i cal ( t 0 ).

[0040] Anti-interference mechanism design Peak dynamic update: WhenP h >1.1 P hmax When reset P hmax = P h ; Prevent misjudgment caused by failure of historical peak value. Better yet, when P h >1.1 P hmax And when the duration is greater than 20ms, reset P hmax = P h , to avoid instantaneous interference triggering updates.

[0041] State maintenance constraint: To enter the mid-term support phase, the condition must be maintained stably for ≥300ms; avoid phase oscillation caused by transient signal jumps.

[0042] Cross-cycle check: The end of the swing needs to be detected P h ≥0.85 P hmax Only then is it allowed to switch to the early support stage; eliminate the influence of abnormal gait such as single-leg jumping.

[0043] This technical solution improves the accuracy of heel strike detection: the pressure peak ratio threshold (90%) combined with the negative inclination angle condition, the missed detection rate is <0.5% (measured data); it improves the anti-interference performance in the mid-support period: the dual pressure threshold + inclination angle steady-state maintenance requirement filters out more than 85% of body shaking; and it ensures the real-time toe-off: the pressure returns to zero combined with a 10° inclination angle change within 100ms, the detection delay is ≤15ms.

[0044] In another technical solution, the control method of the prosthetic knee joint performs the following optimization steps during the iterative learning process in step S4: S41. Take 65° as the initial target flexion angle (65° is based on the average maximum flexion angle of the knee joint during the swing phase of healthy adults walking on flat ground), and calculate the personalized target flexion angle benchmark value in real time: benchmark target flexion angle = adaptation coefficient × (step length / calf length); where the step length is calculated based on the geometric relationship between the spatial distance between adjacent early support starting points and the calibration tilt angle; the calf length is the preset value, and the step length is the default value. S = L ×tan( i cal)× r ,in, L is the calf length, i cal is the calibration tilt angle offset, the coefficient r The value is 1.2-1.5, determined based on the gait statistics of healthy people; the adaptation coefficient ranges from 0.4 to 0.6; S42, when |reference target flexion angle -65°|>10°, replacing the target flexion angle in iterative learning with the reference target flexion angle; S43, the slope angle β is calculated by fusing the pitch angle and roll angle of the thigh segment inertial measurement unit using a complementary filter algorithm (cutoff frequency 10 Hz). The formula is: ,in a x 、 a y 、 a z are the three-axis acceleration components of the thigh inertia unit.

[0045] When it is detected that the slope angle β is greater than 3° for more than 2 seconds, in the uphill scene, the corrected target flexion angle = current target flexion angle × (1 + 0.2 × β); in the downhill scene, the corrected target flexion angle = the maximum value (50°, current target flexion angle × (1-0.1 × |β|)).

[0046] In the above technical solution, the stride length reference value is the horizontal projection distance between adjacent heel contact points, which reflects the individual stride characteristics; the adaptation coefficient is the proportional factor of the associated stride length and calf length, reflecting the personalized biomechanical needs; the slope persistence angle is the inclination angle of the slope on which the prosthesis is located, and the over-threshold time needs to be continuously detected to filter out instantaneous interference.

[0047] In the above technical solution, the personalized target flexion angle is calculated in real time during the gait cycle: Individualized benchmark calculation: Calculate the actual stride length through spatial geometric relationships, combine it with the preset calf length, and generate a benchmark target angle based on the adaptation coefficient ratio (baseline value = coefficient × stride length / leg length). The adaptation coefficient is set in a reasonable range based on biomechanical research. Benchmark correction mechanism: When the benchmark target angle deviates from the initial reference value beyond the allowable range, the target angle in iterative learning is replaced; Dynamic Slope Compensation: When a sustained slope is detected, the target angle is proportionally increased on uphill slopes to enhance push-off assistance. On downhill slopes, the target angle is reduced, but with a safety limit to prevent excessive buckling. The amount of slope compensation varies linearly with the incline angle, ensuring a smooth transition.

[0048] Existing technologies use multiple preset target angle modes (e.g., flat ground / climbing modes). These limitations include: manual mode switching, which prevents real-time response to sudden slope changes; fixed target angles within each mode, which fail to account for dynamic stride length changes (e.g., transitioning from brisk walking to slow walking); and global high damping for downhill safety restrictions, sacrificing gait naturalness. This solution achieves individual adaptation through step-length-driven reference angle calculation, combined with online slope-sensitive target angle correction, to maximize freedom of movement while ensuring safety.

[0049] In this technical solution, the stride-length-linked reference target angle is designed to precisely match individual gait characteristics, addressing the issues of limited stride length or excessive flexion caused by a uniform target value. The slope compensation mechanism dynamically optimizes biomechanical requirements: increasing the target angle on uphill slopes improves propulsion, while limiting the target angle to a lower limit on downhill slopes to prevent instability. This also avoids a rigid gait caused by high damping. The system automatically maintains the target angle's rationality during stride length changes and slope transitions, reducing the need for manual recalibration.

[0050] In another technical solution, the control method of the prosthetic knee joint further comprises, in step S4, adjusting the learning rate according to the rate of change of the gait cycle duration: When the standard deviation of the gait cycle length of three consecutive cycles is less than 0.05s, the standard learning rate L is used; When the standard deviation is ≥ 0.05 seconds, the adaptive learning rate L is enabled. adaptive :L adaptive =L×[1+0.5×tanh(5×| T k - T k-1 | / T k-1 )];middle T k is the current cycle duration, T k-1 The duration of the previous cycle.

[0051] In the above technical solution, the degree of gait speed fluctuation is the statistical discreteness of the length of continuous gait cycles, reflecting the stability of walking rhythm; the adaptive learning rate is the parameter correction intensity dynamically adjusted with the rate of change of gait speed; the smoothing constraint function is a nonlinear conversion mechanism that limits the amplitude of learning rate change to prevent mutations.

[0052] The above technical solution monitors the fluctuation characteristics of consecutive gait cycle durations in real time: A stability criterion is used: When the degree of gait cycle duration fluctuation is below a set threshold, the standard learning rate is maintained to ensure the accuracy of parameter fine-tuning; a fluctuation response mechanism is used: When significant gait speed fluctuations are detected, an adaptive learning rate is generated based on the relative rate of change of adjacent cycle durations. This learning rate scales smoothly with the magnitude of gait speed changes; greater rates of change increase the learning rate, but the maximum amplification factor is constrained by a nonlinear function; a smooth transition design is used: The adaptive learning rate automatically decays to the standard value when gait speed returns to stability, avoiding step changes in the damping parameter.

[0053] Existing technologies use a constant learning rate iteration strategy, which has limitations: the learning rate is decoupled from the gait state, resulting in low optimization efficiency in fast-slow transition scenarios; a fixed learning rate causes knee flexion angle oscillations when gait speed changes suddenly, requiring multiple falls to recover before convergence; and the artificially preset learning rate cannot adapt to individual gait differences. This technical solution achieves a balance between optimization efficiency and stability through adaptive adjustment of the learning rate driven by gait speed fluctuations, breaking through the rigid constraints of the fixed learning rate strategy.

[0054] In another technical solution, the control method of the prosthetic knee joint further comprises, in step S4, executing the following steps when the number of records in the pace-damping mapping knowledge base is less than 5: For unknown gait cycle duration T Additional safety margin for the initial damping opening value: Initial damping opening = U closest ×min(1.3,1+0.1×| T - T closest | / T closest ); U closest is the closest damping opening, T closest is the closest gait cycle duration; Set the maximum number of iteration cycles to 10; When the number of iterations k>5 and |65°- i kmax When |>15°, reset the damping opening U k =0.5×( U min + U max );in U min and U max These are the preset safe operating range boundary values; Umin=0.3 (corresponding to a damping force of 80N・m・s / rad) and Umax=0.8 (corresponding to a damping force of 30N・m・s / rad), which are determined based on the mechanical strength test of the knee joint to ensure that the damping force is within the safe range of 20-150N・m・s / rad.

[0055] In the above technical solution, the damping safety margin is a protective amplification of the initial damping value to prevent the risk of underdamping; the iterative fuse mechanism is a safety strategy that interrupts and resets parameters when the learning process is abnormal; and the safe working range is the damping opening boundary value that ensures the stability of knee joint movement.

[0056] In the above technical solution, when the system detects that the pace-damping mapping record is insufficient, it activates triple protection: Initial Damping Enhancement: Based on the damping value that is closest to the historical pace, the initial value is amplified in proportion to the actual pace difference, but the maximum increase threshold is limited to avoid over-conservatism; Learning progress monitoring: Continuously track the buckling angle error during the iteration process. If it exceeds the safety tolerance and the number of iterations exceeds half, the learning process is immediately interrupted and the damping is reset to the midpoint of the preset safety range; Resource hard constraints: Set a maximum limit on the number of iterations, forcibly terminate non-convergent processes, and release computing resources.

[0057] Existing technologies use a global conservative damping strategy (e.g., setting high damping for all unknown gait speeds). This has limitations: sacrificing gait naturalness for safety, leading to rigid swing phases; lacking an iterative progress monitoring mechanism, which can lead to mechanical damage caused by abnormal learning; and inflexible resource allocation, with inefficient iterations hindering system response. This technical solution maximizes freedom of movement while ensuring safety through the coordinated design of neighbor value scaling, circuit breaker resets, and resource constraints.

[0058] In this technical solution, a safety margin design provides robust initial damping for unknown gait speeds, preventing the risk of knee joint loss of control. A fuse mechanism promptly interrupts parameter divergence, resetting the system to a safe range and eliminating the risk of mechanical overload. Iteration constraints optimize computing resource allocation, ensuring the system's real-time responsiveness. These three elements work together to ensure a balance between learning safety and efficiency in scenarios with sparse knowledge bases.

[0059] In another technical solution, the control method of the prosthetic knee joint, step S4 further includes real-time monitoring of the pressure signal mutation characteristics: if the heel pressure P h Drop > 80% in 20ms P hmax , and the angular velocity is greater than 50° / s, the current iterative learning is interrupted and switched to the safety damping mode: U emergency =0.7× U min +0.3× U max , U emergency The emergency damping opening degree is restored after the terrain stabilizes. U k Continue iterating, the terrain stability determination must meet the following requirements: a) P h ≥40%× P hmax Up to 500ms; b) i cal The rate of change is <2° / s and the angular velocity oh The sampling value is <10° / s for three consecutive times; c) The starting point of the early support stage is detected.

[0060] In the above technical solution, the pressure mutation feature represents the signal form in which the heel pressure decays significantly in a very short time, indicating a sudden change in the support surface; the emergency damping opening refers to a compromise damping value that takes into account both braking requirements and motion continuity; and the terrain steady-state criterion refers to a multi-dimensional recovery condition that integrates pressure recovery, angle change, and phase events.

[0061] In this technical solution, the coupled mutation characteristics of heel pressure and angular velocity are monitored in real time: Real-time instability identification: When a sharp decrease in heel pressure accompanied by rapid extension of the knee joint is detected, it is immediately identified as an instability risk; Safety mode switch: interrupts the learning process and switches to the preset emergency damping opening, which is between high damping braking and freedom of movement; Intelligent recovery mechanism: Three conditions must be met simultaneously to resume learning: heel pressure continues to return to a safe range, joint movement is stable, and the starting point of a new gait cycle is detected to ensure that the terrain is truly stable.

[0062] Existing technologies use safety freeze strategies (e.g., locking the joint after triggering). These limitations include: relying solely on angle thresholds to determine instability, missing the risk of slips caused by sudden pressure changes; completely disabling adaptive learning in emergency situations, requiring a system restart for recovery; and employing a single recovery condition (e.g., simply meeting the pressure requirement), leading to erroneous recovery in shaky terrain. This solution enhances instability detection sensitivity through coupled pressure-angular velocity recognition, employs compromised emergency damping to ensure basic mobility, and establishes multi-dimensional recovery criteria to enable seamless learning.

[0063] In this technical solution, the pressure-motion coupling mechanism accurately detects signs of sudden instability, giving the wearer critical reaction time. A compromised emergency damping design strikes a balance between fall prevention and mobility, avoiding the secondary risk of joint lock. Multi-dimensional recovery criteria ensure automatic continuation of the learning process once the terrain is fully stabilized, maintaining the continuity of adaptive control and overall enhancing the prosthesis's proactive protection against sudden terrain changes.

[0064] In another technical solution, the control method of the prosthetic knee joint adds damping gradual change control in the transition stage from the late support stage to the early swing stage: When the calibration tilt angle is detected in the later stage of support i cal When the angle is greater than 20°, the linear transition of the damping opening is started: U transition = U support +( U - U support )×( i cal -20°) / ( i swingstart -20°), where U transition is the damping opening during the transition period, U support is the fixed damping value opening during the support period, U is the damping opening during the early swing period output by the current iterative learning, i swingstart The starting angle of the early swing phase of the current cycle; the transition process lasts until the end of the early swing phase, and then the full U value.

[0065] In the above technical solution, the transition starting angle refers to the critical point of the inclination angle that triggers the linear change of damping; the fixed damping in the support period refers to the high damping value used in the load-bearing stage of the prosthesis; and the target damping in the swing period refers to the ideal damping value in the early stage of swing optimized by iterative learning.

[0066] In the above technical solution, when it is detected that the support late tilt angle offset exceeds the transition starting angle: Linear transition trigger: Calculates the transition value between the fixed damping during the support phase and the target damping during the swing phase based on the current tilt angle's position ratio within the transition interval. Dynamic Gradual Control: As the tilt angle increases, the damping value continuously transitions from high damping during the stance phase to low damping during the swing phase, with the transition rate matching the joint movement speed. Seamless switching is guaranteed: the transition continues until the end of the early swing phase, after which the target damping value is fully applied.

[0067] Existing technologies use phase-boundary step switching, which has limitations: sudden damping changes can cause knee joint clacking, accelerating mechanical wear; the sudden release of push-off energy causes the calf to swing forward unnaturally, affecting gait fluidity; and it fails to account for the varying transition duration requirements due to varying gait speeds. This solution achieves dynamic synchronization of damping force and joint motion through a linear transition mechanism driven by angle proportionality.

[0068] In this technical solution, proportional control of the tilt angle allows the damping force to vary continuously with joint flexion, completely eliminating mechanical shock during transitions between movement phases. Push-off kinetic energy is smoothly released through the damping gradient, improving calf forward swing efficiency and gait continuity. The transition duration adapts to the joint's movement speed, ensuring a natural transition at varying speeds. This overall reduces prosthetic movement noise and the wearer's metabolic burden.

[0069] In another technical solution, the control method of the prosthetic knee joint adds sensor failure redundancy processing in S1: Real-time calculation of the heel pressure signal validity index, if 5 consecutive cycles P hmax <10N or signal variance <0.1N 2, determine that the heel pressure sensor is faulty, and perform the following steps: The vertical acceleration component a is output by the inertial measurement unit installed at the proximal end of the thigh prosthesis. z Alternative P h , the touchdown threshold is set to |a z |>2g; The heel impact sound sensor installed on the inside of the prosthetic ankle is activated simultaneously. When the 200-500Hz sound pressure level is detected to be greater than 80dB and the sound pressure level is greater than 80dB, the z When the peak time difference is less than 50 ms, a touchdown event is confirmed; the backup signal touchdown event triggers the reference angle reset in step S2 and the stage division in step S3.

[0070] In the above technical solution, the signal validity indicator refers to a combination of features (such as peak level and fluctuation intensity) that quantifies the credibility of the pressure signal; The vertical impact acceleration is the transient acceleration component borne by the prosthesis in the axial direction, reflecting the impact strength of the ground contact; Voiceprint event synchronization refers to the spatiotemporal consistency verification of the same event detected by different physical quantity sensors.

[0071] In the above technical solution, the credibility of the heel pressure signal is evaluated in real time. When abnormal signal characteristics are detected for multiple consecutive cycles: multi-source signal switching - the vertical acceleration signal of the thigh segment inertial unit is enabled to replace the heel pressure, and the touchdown event is determined by the impact acceleration threshold; voiceprint collaborative verification - synchronously analyze the heel impact characteristics captured by the acoustic sensor at the ankle, and confirm the touchdown only when the sound pressure characteristics are consistent with the acceleration peak in time and space; functional lossless migration - the backup sensor touchdown event triggers the base angle reset and gait stage division of the original process to ensure the continuity of system functions.

[0072] Existing technologies use a single backup sensing channel (e.g., only an accelerometer). This has the following limitations: the acceleration threshold is easily affected by impacts outside of walking (e.g., bumps), leading to false touchdowns; there is no multi-source signal cross-validation mechanism, resulting in false positives in fault scenarios exceeding safety limits; and the backup signal only maintains basic motion, losing the ability to finely segment gait phases. This solution utilizes a spatiotemporal coupling verification mechanism for both acoustic and acceleration dual physical quantities, enabling highly reliable touchdown detection and maintaining full functionality in fault scenarios.

[0073] In this technical solution, signal validity monitoring provides early warning of sensor failure, preventing erroneous data from contaminating the control system. The acoustic-acceleration spatiotemporal coupling verification mechanism eliminates interference outside of walking, ensuring accurate ground contact event determination. The lossless function migration design ensures that gait classification and damping adjustment are fully supported even in the event of sensor failure, significantly improving the system's fault tolerance and continuous operation time.

[0074] The number of devices and processing scales described herein are intended to simplify the description of the present invention. Applications, modifications, and variations of the present invention will be readily apparent to those skilled in the art.

[0075] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to specific details.

Claims

1. A method for controlling a prosthetic knee joint, characterized in that: include: S1. Real-time acquisition of heel pressure sensor signals P h , Foot pressure sensor signal P f and the tilt angle output by the calf inertial measurement unit θ and angular velocity ω; S2, when detected P f The rate of descent v f > k ⋅ v h When the real-time tilt angle θ now Defined as the new 0° reference, denoted as θ base = θ now ; v h for P h Descent rate, k =2 is the preset multiple threshold; Calculate the tilt angle offset after calibration θ cal , θ cal = θ - θ base ; S3, based on θ cal The gait cycle is divided into early stance, mid-stance, late stance, early swing, mid-swing and late swing based on the dual pressure signals; Calculate the gait cycle duration based on the interval between adjacent early support starting points T ; S4. Establish and call the speed-damping mapping knowledge base to construct the gait cycle duration T and the early damping opening of the swing U Mapping relationship library; detect the current T value and query the pace-damping mapping knowledge base, if there is a match T , directly call the corresponding U value; if unknown T , then load T The closest known pace corresponds to U As initial value; Perform iterative learning on the unknown T: U k+1 =U k +L(θ target - θ kmax ) in U is the damping opening, L is the learning rate, θ target The target flexion angle is 65°. θ kmax For the k The maximum flexion angle measured during the period; When the buckling angle error converges to θ target - θ kmax When |≤5°, stop iteration and store the mapping relationship; S5, damping opening after adjustment in the early stage of swing U ; Maintain fixed damping in the early support period, middle support period, late support period and middle and late swing period.

2. The control method of the prosthetic knee joint according to claim 1, characterized in that: based on θ cal The gait cycle is divided into early support, mid support, late support, early swing, mid swing and late swing based on the dual pressure signals. Specifically: when P h ≥ P hmax and θ cal When the angle is less than -3°, it is determined to be in the early stage of support; when P h >50%× P hmax 、 P f >50%× P fmax and θ cal When the rate of change is <1° / s, it is determined to be in the middle stage of support; P hmax is the current gait cycle P h Peak, P fmax is the current gait cycle P f Peak value; when P h <30%× P hmax and θ cal When it is >0°, it is determined to be in the late support stage; when P f =0 and θ cal When the increase was ≥10° within 100ms, it was determined to be in the early stage of swing; when θ cal Continue to increase to the peak value and the angular velocity | ω ∣<1° / s When , it is judged to be in the middle stage of swing; when θ cal The peak value detected from the current swing period θ max The descent value is ≥3° and the angular velocity ω< When 0, it is judged to be in the late swing stage.

3. The control method of the prosthetic knee joint according to claim 1, characterized in that: The following optimization steps are performed during the iterative learning process in step S4: S41. Using 65° as the initial target flexion angle, calculate in real time the personalized target flexion angle baseline value: baseline target flexion angle = adaptation coefficient × (step length / calf length); where step length is calculated based on the geometric relationship between the spatial distance between adjacent early support starting points and the calibration tilt angle; calf length is a preset value; and the adaptation coefficient ranges from 0.4 to 0.

6. S42, when |reference target flexion angle -65°|>10°, replacing the target flexion angle in iterative learning with the reference target flexion angle; S43. When it is detected that the slope angle β is greater than 3° for more than 2 seconds, in the uphill scenario, the corrected target flexion angle = current target flexion angle × (1 + 0.2 × β); in the downhill scenario, the corrected target flexion angle = the maximum value (50°, current target flexion angle × (1 - 0.1 × |β|)).

4. The control method of a prosthetic knee joint according to claim 1, wherein: Step S4 also includes adjusting the learning rate according to the rate of change of the gait cycle duration: When the standard deviation of the gait cycle length of three consecutive cycles is less than 0.05s, the standard learning rate L is used; When the standard deviation is ≥ 0.05 seconds, the adaptive learning rate L is enabled. adaptive :L adaptive =L×[1+0.5×tanh(5×| T k - T k-1 | / T k-1 )]; middle T k is the current cycle duration, T k-1 The duration of the previous cycle.

5. The control method of the prosthetic knee joint according to claim 4, characterized in that: Step S4 further includes executing the following steps when the number of records in the pace-damping mapping knowledge base is less than 5: For unknown gait cycle duration T Additional safety margin for the initial damping opening value: Initial damping opening = U closest ×min(1.3,1+0.1×| T - T closest | / T closest ); U closest is the closest damping opening, T closest is the closest gait cycle duration; Set the maximum number of iteration cycles to 10; When the number of iterations k>5 and |65°- θ kmax When |>15°, reset the damping opening U k =0.5×( U min + U max );in U min and U max are the preset safe working range boundary values ​​respectively.

6. The control method of the prosthetic knee joint according to claim 5, characterized in that: Step S4 also includes real-time monitoring of the pressure signal mutation characteristics: If heel pressure P h Drop > 80% in 20ms P hmax , and the angular velocity is greater than 50° / s, the current iterative learning is interrupted and switched to the safety damping mode: U emergency =0.7× U min +0.3× U max , U emergency is the emergency damping opening; After the terrain stabilizes, it returns to the pre-interruption state. U k Continue iterating, the terrain stability determination must meet the following requirements: a) P h ≥40%× P hmax Up to 500ms; b) θ cal The rate of change is <2° / s and the angular velocity ω The sampling value is <10° / s for three consecutive times; c) the starting point of the early support stage is detected.

7. The control method of a prosthetic knee joint according to claim 2, wherein: In the transition stage from the late support stage to the early swing stage, add damping gradient control: When the calibration tilt angle is detected in the later stage of support θ cal When the angle is greater than 20°, the linear transition of the damping opening is started: U transition = U support +( U - U support )×( θ cal -20°) / ( θ swingstart -20°), where U transition is the damping opening in the transition period, U support is the fixed damping value opening during the support period, U is the damping opening during the early swing period output by the current iterative learning, θ swingstart The starting angle of the early swing phase of the current cycle; the transition process lasts until the end of the early swing phase, and then the full U value.

8. The control method of a prosthetic knee joint according to claim 1, wherein: Add sensor failure redundancy processing in S1: Real-time calculation of the heel pressure signal validity index, if 5 consecutive cycles P hmax <10N or signal variance <0.1N 2 , determine that the heel pressure sensor is faulty, and perform the following steps: The vertical acceleration component a is output by the inertial measurement unit installed at the proximal end of the thigh prosthesis. z Alternative P h , the touchdown threshold is set to |a z |>2g; The heel impact sound sensor installed on the inside of the prosthetic ankle is activated simultaneously. When the 200-500Hz sound pressure level is detected to be greater than 80dB and the sound pressure level is greater than 80dB, the z When the peak time difference is less than 50 ms, a touchdown event is confirmed; the backup signal touchdown event triggers the reference angle reset in step S2 and the stage division in step S3.