Layered cooperative walking aid control method for lower limb exoskeleton robot

Abstract

The invention discloses a lower limb exoskeleton robot layered cooperative walking aid control method. The method comprises the steps of data acquisition and preprocessing; human-machine coupling dynamics modeling is carried out, and required moments tau < hip >, tau < knee > and tau < ankle > of the hip joint, the knee joint and the ankle joint are obtained; low-layer joint control, including hip joint moment, knee / ankle virtual impedance moment and friction compensation calculation; middle-layer multi-axis cooperation and gait phase judgment; high-rise energy is optimally distributed, assistance is distributed to three joints in a phase and a safety boundary, meanwhile, the requirements for labor saving and stability are considered, a minimization problem is written in a unified mode, J serves as the total amount, and torque of the hip joint, torque of the knee joint and torque of the ankle joint are solved in a combined mode; according to the step-by-step self-adaption of myoelectricity driving, in order to cope with human body fatigue changes, step-by-step updating is carried out on the virtual stiffness within the movement distance. The method has the advantages of improving man-machine cooperation level, man-machine adaptability and walking aid efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot technology, in particular to a layered cooperative walking assistance control method of lower limb exoskeleton robot. BACKGROUND

[0002] Lower limb exoskeleton robot is a wearable robot, which aims to assist or enhance the lower limb movement ability of the wearer through mechanical structure and control system. This robot can be used for rehabilitation training, to enhance the work efficiency of physical laborers, or to provide walking assistance for people with difficulty in moving. It relies on sensing, driving and control to provide gait assistance or ability enhancement for the wearer. In the prior art, centralized control ignores the multi-joint cooperation and dynamic interaction of the human body, resulting in unnatural human-machine cooperation and poor adaptability. Early layered control has level division, but the cooperation between layers is insufficient. For example, the high layer maps the surface electromyography (sEMG) signal statically and does not realize adaptive gain adjustment. The middle layer lacks a synchronization mechanism for multi-joint coordination, resulting in misalignment of gait timing. The bottom layer has insufficient feedback accuracy of human-machine interaction force, affecting the effect of compliant assistance. There are obvious shortcomings in the layered utilization of multi-sensor fusion and the cooperative optimization of control targets of each layer. In terms of multi-joint cooperation and individualization: first, the communication and sampling are not synchronized or only locally aligned, which easily causes inconsistent joint phases, resulting in misalignment of parameters and actual gait in the support / swing phase; second, the control layer mostly adopts fixed trajectory or joint-specific constant, which is difficult to balance the "stability of weight bearing in the support phase" and "easy release in the swing phase", and the energy efficiency and comfort are difficult to balance; third, most parameters are global constants, which respond slowly to individual differences and fatigue changes and rely on manual adjustment; fourth, safety constraints are scattered and lack unified boundary and slope management, which easily causes mutations during switching. In summary, the current lower limb exoskeleton robot has poor human-machine cooperation, walking assistance efficiency and user experience. SUMMARY

[0003] The present application overcomes the above-mentioned shortcomings and provides a layered cooperative walking assistance control method of lower limb exoskeleton robot, which can improve the level of human-machine cooperation, human-machine adaptability and walking assistance efficiency.

[0004] The layered cooperative walking assistance control method of lower limb exoskeleton robot of the present application, wherein the specific steps include:

[0005] Step 1: Data acquisition and preprocessing: collecting sensing data, including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot events, surface electromyography (sEMG) signal and communication synchronization state;

[0006] Step 2: Human-machine coupled dynamics modeling to obtain the required torques of hip, knee and ankle joints hip ,τ knee ,τ ankle ;

[0007] Step 3: Bottom joint control, including:

[0008] Step 3.1: Hip joint torque: according to the required torque τ hip of the hip joint, the set target torque, the hip joint feedback torque, the hip joint target angle, and the hip joint encoder feedback angle, the hip joint torque is calculated;

[0009] Step 3.2: Knee and ankle virtual impedance torque: due to the high sensitivity of the knee and ankle joints to contact compliance and foot comfort, a virtual spring-damper model is used for calculation to obtain the knee and ankle virtual impedance torque;

[0010] Step 3.3: Friction compensation: static friction and viscous resistance at low speed can cause small signal hysteresis and micro-vibration, and a virtual torque is used to compensate for the friction of each joint;

[0011] Step 4: Middle layer multi-axis coordination and gait phase determination: the gait phase is determined by the zero-crossing point of the hip joint angle from extension to flexion and the rate threshold, and the foot pressure and inertial measurement unit (IMU) events are used as auxiliary determination basis; when in the support phase, the knee joint virtual stiffness k knee-virt , ankle joint virtual stiffness k ankle-virt , and damping coefficient B virt are increased to enhance support and disturbance rejection; in the swing phase, the impedance is reduced to release the swing and reduce the intervention on natural gait;

[0012] Step 5: High-level energy optimal distribution, including the following steps:

[0013] Step 5.1: Objective function definition: under the unified synchronous signal (SYNC) beat, the assist energy is distributed within the movement distance of the hip, knee, and ankle joints, and the goal is to minimize the total score J under the premise of meeting the phase and safety limits;

[0014] Step 5.2: Quadratic programming: according to the distribution of assist force to the three joints within the phase and safety boundary, while considering the requirements of energy saving and smoothness, a minimization problem is written, with J as the total quantity, and the joint torques of the hip, knee, and ankle are jointly solved;

[0015] Step 5.3: Constraint condition: according to steps 3 and 4, the upper and lower bounds of the joint torques are obtained as constraint conditions, and the hip, knee, and ankle joint torque combinations that minimize J within the feasible region are obtained as output, completing the high-level energy optimal distribution and achieving the effect of energy saving and smoothness;

[0016] Step 6: Step-by-step adaptation of muscle-driven

[0017] To cope with the changes in human fatigue, the virtual stiffness is updated step by step within the movement distance, and the specific steps are as follows:

[0018] Step 6.1 Feature extraction: process the surface electromyogram sEMG, including rectification, filtering, envelope, and combine gait phase and joint state to form a state feature vector;

[0019] Step 6.2 Gain update: according to the state feature vector, the virtual stiffness of the knee and ankle joints is updated step by step within the movement distance in the form of baseline and gain;

[0020] Step 6.3 Step-by-step adaptation: according to the gain update, update the joint torque of k+1 steps within the movement distance to realize step-by-step adaptation, and set the amplitude and slope upper and lower limit values of the gait phase to ensure parameter slow change and avoid comfort problems caused by sudden changes.

[0021] The above-mentioned hierarchical collaborative walking control method of lower extremity exoskeleton robot, wherein in step 1, the data acquisition and preprocessing: the specific steps are as follows:

[0022] Step 1.1 Collect sensor data: including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot event, surface electromyogram sEMG, communication synchronization state;

[0023] Step 1.2 Signal preprocessing: including denoising, filtering, zero point calibration, gain calibration, and cross-device time alignment;

[0024] Step 1.3 Data buffering and timestamp: establish a ring buffer queue, write according to the synchronization beat SYNC, and ensure that the multi-joint data is aligned within the same control cycle.

[0025] The above-mentioned hierarchical collaborative walking control method of lower extremity exoskeleton robot, wherein: in step 2, the human-machine coupling dynamics modeling is carried out to obtain the required torque τ hip ,τ knee ,τ ankle of the hip, knee and ankle joints; the specific steps are as follows:

[0026] Step 2.1 Establish the generalized coordinates of the series or parallel system of the exoskeleton and the human body =[q hip ,q knee ,q ankle ]

[0027] Step 2.2 Establish the dynamics equation:

[0028]

[0029] In the formula: M is the inertia matrix; C is the Coriolis and centrifugal term; G is the gravity term; τexo is the driving torque of the exoskeleton; τ human is the active torque of the human muscle; θ is the joint angle, is the joint angular velocity;

[0030] Step 2.3 Desired torque solution τ req :

[0031]

[0032] where τ hip , τ knee , τ ankle are the required torques of the hip, knee, and ankle joints, respectively, M hip , M knee、 M ankle are the effective inertia terms of the hip, knee, and ankle joints, respectively, hip , knee , ankle are the angular accelerations of the hip, knee, and ankle joints, respectively, hip , knee , ankle are the angular velocities of the hip, knee, and ankle joints, respectively.

[0033] The hierarchical collaborative walking control method of the lower extremity exoskeleton robot, wherein: in step 3.1, the hip joint torque: according to the required torque τ hip of the hip joint, the target torque set according to the target torque set, the hip joint feedback torque, the hip joint target angle, the hip joint encoder feedback angle calculation hip joint torque, expressed as:

[0034]

[0035] where τ hip-cmd is the hip joint torque, τ hip-dcs is the target torque set according to τ hip of the hip joint, τ sensor is the hip joint feedback torque, K p is the proportional gain of torque error, K d is the gain of angle error derivative; and represent the hip joint target angular velocity and the hip joint encoder feedback angular velocity, respectively.

[0036] The hierarchical collaborative walking control method of the lower extremity exoskeleton robot, wherein: in step 3.2, the knee joint virtual impedance torque is calculated as:

[0037]

[0038] wherein τ knee-virtual is the virtual torque of the knee joint, k knee-virt is the virtual stiffness of the knee joint, B virt is the damping coefficient, θ knee-dcs and θ knee-enc are the target angle and encoder feedback angle of the knee joint, respectively, is the encoder feedback angular velocity of the knee joint;

[0039] The virtual impedance torque of the ankle joint is calculated as:

[0040]

[0041] wherein τ ankle-virtual is the virtual torque of the ankle joint, k ankle-virt is the virtual stiffness of the ankle joint, B virt is the damping coefficient, θ ankle-dcs and θ ankle-enc are the target angle and encoder feedback angle of the ankle joint, respectively, is the encoder feedback angular velocity of the ankle joint.

[0042] The hierarchical collaborative walking control method of the lower extremity exoskeleton robot, wherein: in step 3.3, the friction compensation: the static friction and viscous resistance at low speed will cause small signal hysteresis and micro vibration, and the virtual torque is used to compensate the friction of each joint; comprising

[0043] The knee joint friction compensation is:

[0044]

[0045] wherein F knee-cmd is the command force of the knee joint, τ knee-virtual is the virtual torque of the knee joint, F c and F v are the static friction and viscous friction coefficients, respectively, is the speed of the electric cylinder of the knee joint;

[0046] The ankle joint friction compensation is:

[0047]

[0048] wherein F ankle-cmd is the command force of the ankle joint, τ ankle-virtual is the virtual torque of the ankle joint, F c and F v are the static friction and viscous friction coefficients, respectively, is the speed of the electric cylinder of the ankle joint.

[0049] The lower extremity exoskeleton robot hierarchical cooperative walking control method, wherein: in step 4, the middle layer multi-axis cooperation and gait phase determination: the gait phase is determined by the zero-crossing point of the hip joint angle from extension to flexion and the rate threshold, and the foot pressure and inertial measurement unit (IMU) event are used as auxiliary determination basis; when in the support phase, the k knee-virt 、k ankle-virt 、B virt is increased to enhance support and anti-interference; in the swing phase, the impedance is reduced to release swing and reduce the intervention on natural gait, and the specific steps are as follows:

[0050] Step 4.1 bus synchronization: the bus CANopen is used, the main station periodically issues a synchronization signal SYNC, and the key measurement data and control command are received and sent through the synchronous process data object (PDO);

[0051] Step 4.2 gait phase determination: the support phase and swing phase are determined based on the zero-crossing point of the hip joint angle from extension to flexion and the rate threshold, and the foot pressure and IMU event can be fused;

[0052] Step 4.3 phase scheduling: the k knee-virt 、k ankle-virt 、B virt is increased in the support phase to enhance support and anti-interference, and is reduced in the swing phase to release swing.

[0053] The lower extremity exoskeleton robot hierarchical cooperative walking control method, wherein: in step 5.1, the target function is defined: the assistance energy is distributed within the motion distance of the hip, knee and ankle joints under the unified SYNC beat, and the target is to minimize the total score under the premise of meeting the phase and safety limit, and the target function is represented as:

[0054]

[0055] Wherein, J is the total score value, the smaller J represents more energy-saving assistance and more stable motion; w1, w2 are weights; τ i is the torque variable of the i-th joint; is the angular velocity of the i-th joint.

[0056] The lower extremity exoskeleton robot hierarchical cooperative walking control method, wherein: in step 6.2, the gain is updated: according to the state characteristic vector, the virtual stiffness of the knee and ankle joints is updated step by step within the motion distance in the form of baseline and gain, which is represented as:

[0057] K knee-virt,k+1 =K knee-base +α knee RMS knee-EMG,k

[0058] Wherein, Kknee-virt,k+1 Let K be the virtual stiffness of the knee joint at the (k+1)th step within the movement distance. knee-base For the baseline stiffness of the knee joint, α knee RMS is the adaptive gain coefficient for the knee joint. knee-EMG,k is the state feature vector of the surface electromyography signal of the knee joint within the k-th step of the movement distance;

[0059] K ankle-virt,k+1 =K ankle-base +α ankle RMS ankle-EMG,k

[0060] Among them, K ankle-virt,k+1 Let K be the virtual stiffness of the ankle joint at the (k+1)th step within the movement distance. ankle-base For ankle baseline stiffness, α ankle RMS is the adaptive gain coefficient for the ankle joint. ankle-EMG,k It is the state feature vector of the surface electromyography signal of the ankle joint during the k-th step of the movement distance.

[0061] Compared with existing technologies, this invention has significant advantages. As can be seen from the above scheme, the hierarchical collaborative walking assistance control method proposed in this invention forms a closed-loop link of "bottom-level execution and compensation—middle-level phase / boundary scheduling—high-level assistance allocation—gradual electromyographic adaptation—state monitoring and safety degradation": Under the CANopen SYNC cycle, phase determination and parameter distribution are completed in steps as the time unit; the bottom layer uses a direct torque channel for the hip joint and a virtual impedance channel for the knee / ankle, combined with necessary compensations such as low-speed friction; the high layer completes the overall allocation of assistance to the three joints within a unified time base and boundary constraints; gradual adaptation is statistically analyzed at step k, updated at the end of step k, and takes effect overall at step k+1, and is constrained by upper limits of amplitude and slope; the entire link is continuously monitored and smoothly degraded according to the strategy. Compared with existing technologies, it has at least the following significant advantages:

[0062] 1. Coordinated allocation across the entire system. The three joints assist in making overall decisions within a unified time base and unified boundary, avoiding each joint acting independently and achieving both "less effort" and "greater stability".

[0063] 2. Timing Consistency and Phase Scheduling. With SYNC as the backbone, the target angle, stiffness, damping, and their upper and lower limits and slope constraints during the support / swing phase are continuously scheduled with the phase, significantly reducing cross-joint phase mismatch.

[0064] 3. Controllable execution at the underlying level. Direct torque at the hip joint provides fast response and high controllability; the knee / ankle virtual spring-damping provides sufficient load-bearing compliance during the support phase and reduces resistance during the swing phase. Combined with friction compensation, it makes the take-off and landing smoother and prevents small disturbances from being amplified.

[0065] 4. Stepwise EMG adaptation. Statistic EMG intensity with step as window, update at step end and take effect in next step; update limited by phase-dependent amplitude and slope upper bound, smooth change with active effort and fatigue level, reduce artificial parameter tuning.

[0066] 5. Unified safety-comfort boundary and degradation. Torque, speed, temperature rise, communication, etc. unified monitoring; once out of boundary, immediately enter safety mode such as zero torque or high damping, and gradually recover in appropriate phase according to set slope, still keep controllable and stable in abnormal situation.

[0067] 6. Strong engineering adaptability. Method and actuator form / mechanism layout decoupled, rely on parameter mapping to support multiple modes and scenes, easy to deploy, maintain and extend without changing hardware.

[0068] 7. Consistency and comfort improvement. Combination of unified time base, phase scheduling, overall allocation and stepwise adaptation, reduce inter-joint phase drift and parameter mutation, improve gait consistency and wearing comfort.

[0069] The beneficial effects of the present application are further illustrated by the specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 is a flowchart of the present application;

[0071] Figure 2 is a detailed flowchart of motion control (bottom layer) in Figure 1

[0072] Figure 3 is a detailed flowchart of coordination (middle layer) in Figure 1

[0073] Figure 4 is a detailed flowchart of supervisory interaction (high layer) in Figure 1

[0074] Figure 5 is a structural schematic diagram of the present application.

[0075] Markings in the figure: 1, joint motor of hip joint; 2, knee joint servo cylinder; 3, ankle joint servo cylinder DETAILED DESCRIPTION

[0076] The specific embodiments, features and effects of a kind of lower limb exoskeleton robot hierarchical collaborative walking control method according to the present application are described in detail as follows in combination with the preferred embodiments and the drawings.

[0077] As Figures 1 to 5 ​​​As shown, the lower limb exoskeleton robot hierarchical cooperative walking control method of the application mainly includes the following driving elements: a hip joint motor (1), a knee joint servo cylinder (2), and an ankle joint servo cylinder (3). Figure 1 As shown, the specific implementation process is that sensors (sEMG, plantar pressure, motion capture) provide signals, the upper layer completes gait phase recognition and sEMG adaptive gain, generates the required stiffness / damping / torque scheduling; the middle layer coordinates the target parameters into adaptive stiffness / damping of each joint and synchronously issues them through CANopen SYNC; and the bottom layer adopts hip joint PD, knee / ankle impedance control (including friction compensation) to drive the three-joint actuator to complete gait assistance. The specific steps of the method include:

[0078] Step 1: Data acquisition and preprocessing: collect sensor data, including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot events, surface electromyogram (sEMG), and communication synchronization state; the specific steps are as follows:

[0079] Step 1.1: Collect sensor data, including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot events, surface electromyogram (sEMG), and communication synchronization state;

[0080] Step 1.2: Signal preprocessing, including denoising, filtering, zero point calibration, gain calibration, and cross-device time alignment;

[0081] Step 1.3: Data buffering and timestamping: establish a ring buffer queue, write according to the synchronization beat (SYNC), and ensure that the multi-joint data is aligned within the same control period.

[0082] Step 2: Human-machine coupling dynamics modeling to obtain the required torques τ hip ,τ knee ,τ ankle of the hip, knee, and ankle joints; the specific steps are as follows:

[0083] Step 2.1: Establish the generalized coordinates of the series or parallel system of the exoskeleton and the human body =[q hip ,q knee ,q ankle ]

[0084] Step 2.2: Establish the dynamics equation:

[0085] (1)

[0086] where: M is the inertia matrix; C is the Coriolis and centrifugal term; G is the gravity term; τ exo is the exoskeleton driving torque; τ humanθ represents the active torque of human muscles; θ is the joint angle. Joint angular velocity;

[0087] Step 2.3 Solving for the desired torque τ req :

[0088] (2)

[0089] Where, τ hip τ knee τ ankle M represents the required torque for the hip, knee, and ankle joints, respectively. hip M knee、 M ankle These are the effective inertial terms for the hip, knee, and ankle joints, respectively. hip , knee , ankle These are the angular accelerations of the hip, knee, and ankle joints, respectively. hip , knee , ankle These are the angular velocities of the hip, knee, and ankle joints, respectively.

[0090] The joint desired torque vector τ obtained from equations (1) and (2) req =[τ hip , τ knee , τ ankle ] T As common inputs for subsequent calculations: ① Input the optimal energy allocation at higher levels (Equation 8) and solve for the optimal torque τ in conjunction with phase / safety constraints; ② Combine the current gait phase (support phase / swing phase) and reference angle / angular velocity to generate the target angle θ from the middle layer and send it down to the lower layer. dcs With impedance parameter K virt B virt (See Figure 3 ).

[0091] Step 3: Bottom-level joint control ensures real-time controllability and compliance of torque and angle, thereby providing an executable foundation for phase decoupling in the middle layer and energy distribution in the upper layer (e.g., Figure 2 );

[0092] Step 3.1 Hip joint torque: Based on the required torque τ of the hip joint hip The set target torque is used to calculate the hip joint torque based on the set target torque, hip joint feedback torque, hip joint target angle, and hip joint encoder feedback angle, expressed as:

[0093] (3)

[0094] Where, τ hip-cmd For hip joint torque, τ hip-dcs According to τ hip The set target torque for the hip joint, τ sensor K is the feedback torque of the hip joint. p K is the proportional gain of the torque error. d The gain is the derivative of the angle error; and These represent the target angular velocity of the hip joint and the feedback angular velocity of the hip joint encoder, respectively. K... p K d It must satisfy: (1) closed-loop stability and phase consistency constraints; (2) driver output amplitude and slope constraints; (3) scheduling relationship with gait phase, walking speed, load or temperature rise. K p K d Updates can be made between strides and take effect in the next stride, where the support phase K... d Not less than the oscillation phase K d To enhance support damping, K p Not lower than the minimum value required to reach the predetermined steady-state tracking error threshold.

[0095] Step 3.2 Virtual resistance torque at the knee and ankle:

[0096] To address the high sensitivity of the knee and ankle joints to contact compliance and foot comfort, a virtual spring-damping model was used to calculate the virtual impedance torque of the knee and ankle.

[0097] The virtual impedance torque of the knee joint is calculated as follows:

[0098] (4)

[0099] Where, τ knee-virtual k represents the virtual torque at the knee joint. knee-virt For the virtual stiffness of the knee joint, B virt θ is the damping coefficient. knee-dcs and θ knee-enc These are the target angle of the knee joint and the encoder feedback angle, respectively. The knee joint encoder provides feedback angular velocity.

[0100] The virtual impedance torque of the ankle joint is calculated as follows:

[0101] (5)

[0102] Where, τ ankle-virtual k represents the virtual torque of the ankle joint. ankle-virt For the virtual stiffness of the ankle joint, B virt θ is the damping coefficient.ankle-dcs and θ ankle-enc are ankle target angle and encoder feedback angle, respectively, is ankle encoder feedback angular velocity.

[0103] Virtual impedance balances the system between landing cushioning and swing release, and provides natural "tunable mechanical properties" for mid-level phase scheduling.

[0104] Step 3.3 Friction compensation and execution side consistency (electric cylinder / transmission):

[0105] The static friction and viscous resistance of the transmission link (such as electric cylinder) at low speed will cause small signal hysteresis and micro-vibration. Formulas (6, 7) are used for compensation:

[0106] Knee joint friction compensation is:

[0107] (6)

[0108] where, F knee-cmd is the knee joint command force, τ knee-virtual is the knee joint virtual torque, F c and F v are the static friction and viscous friction coefficients, respectively, is the knee joint electric cylinder speed.

[0109] Ankle joint friction compensation is:

[0110] (7)

[0111] where, F ankle-cmd is the ankle joint command force, τ ankle-virtual is the ankle joint virtual torque, F c and F v are the static friction and viscous friction coefficients, respectively, is the ankle joint electric cylinder speed.

[0112] By geometric mapping, the consistency of the execution end (force) and the joint side (torque) is maintained, and the low-speed following and comfort are improved.

[0113] The three designs at the bottom level jointly ensure fast, stable, and perceptible assist output, and provide controllable "execution surface" for the coordination and optimization at the upper level.

[0114] Step 4 Mid-level multi-axis coordination and gait phase determination

[0115] The middle layer focuses on timing consistency and phase decoupling. CANopen or other fieldbus is used, and the master station periodically issues SYNC. Key commands and measurements are transmitted and received on the synchronization PDO to ensure that the hip, knee, and ankle work on the same beat, reducing cross-joint phase drift. Figure 3 ).

[0116] The gait phase is determined by the zero-crossing point of the hip joint angle from extension to flexion and the rate threshold, and the foot pressure and inertial measurement unit (IMU) events are used as auxiliary determination basis. When in the support phase, k knee-virt , k ankle-virt , and B virt are adjusted upward to enhance support and anti-interference. In the swing phase, the impedance is reduced to release the swing and reduce the intervention on natural gait. This layer converts discrete gait events into continuous adjustable control parameter flow through "phase-parameter mapping". The specific steps are as follows:

[0117] Step 4.1 Bus synchronization: CANopen bus is used, and the master station periodically issues synchronization signal SYNC. Key measurement data and control commands are transmitted and received through synchronous process data object (PDO).

[0118] Step 4.2 Gait phase determination: the zero-crossing point of the hip joint angle from extension to flexion and its rate threshold are used to determine the support phase and swing phase, which can be combined with foot pressure and IMU events.

[0119] Step 4.3 Phase scheduling: k knee-virt , k ankle-virt , and B virt are adjusted upward in the support phase to enhance support and anti-interference, and are reduced in the swing phase to release the swing. If necessary, phase-related limiters and slope constraints are introduced for hip torque closed-loop control.

[0120] 4. In the SYNC beat, the phase constraint set (joint torque / velocity / slope, temperature rise, etc. boundaries) is formed according to the current gait phase (support phase / swing phase), and is transmitted to the low-level joint target angle, virtual stiffness, and damping coefficient used in formulas (3), (4), and (5). The phase / safety constraints are transmitted to the high-level (as the feasible region of formula (8)) to match the optimal solution with the current gait stage.

[0121] Step 5: High-level energy optimal allocation (e.g. Figure 4 ), including:

[0122] Step 5.1 Objective function definition: in the unified SYNC beat, the assist energy is allocated within the movement distance of the hip, knee, and ankle joints. The goal is to minimize the overall score while meeting the phase and safety constraints. The objective function is represented as:

[0123] (8)

[0124] where J is the total score, the smaller the J, the more energy-saving and stable the motion; w1, w2 are weights; τ i is the torque variable of the i-th joint; is the angular velocity of the i-th joint;

[0125] Step 5.2 Quadratic Programming QP: According to the three joints within the phase, safety boundary, while taking into account the requirements of energy saving and stability, write a minimization problem, use J as the total amount, jointly solve the torque of the hip, knee and ankle joints;

[0126] Step 5.3 Constraint condition: According to step 4, the torque upper and lower bounds of τ i and the necessary slope, power / temperature rise are taken as constraint conditions, and when solving, the torque combination τ=[τ hip , τ knee , τ ankle ] that makes J minimum is automatically selected in the feasible region as output, completing the high-level energy optimal allocation, and optimally allocating the hip, knee and ankle joints to achieve energy-saving and stable effect;

[0127] Step 6: Step-by-step adaptation of electromyographic driving

[0128] In order to cope with the differences between subjects and changes in human fatigue, the virtual stiffness is updated step by step within the movement distance, and the specific steps are as follows:

[0129] Step 6.1 Feature extraction: Process the surface electromyographic signal sEMG, including rectification, filtering, envelope, and form a state feature vector combining gait phase and joint state;

[0130] Step 6.2 Gain update: According to the state feature vector, use the baseline and gain form to update the virtual stiffness of the knee and ankle joints within the movement distance step by step, which is represented as:

[0131] K knee-virt,k+1 =K knee-base +α knee RMS knee-EMG,k (9)

[0132] Where K knee-virt,k+1 is the knee joint virtual stiffness at the k+1 step within the movement distance, K knee-base is the knee joint baseline stiffness, α knee is the knee joint adaptive gain coefficient, and RMS knee-EMG,k is the state feature vector of the knee joint surface electromyographic signal at the k step within the movement distance;

[0133] K ankle-virt,k+1 =K ankle-base+ a ankle RMS ankle-EMG,k (10)

[0134] wherein, K ankle-virt,k+1 is the virtual ankle stiffness of the k+1 step within the motion distance, K ankle-base is the baseline ankle stiffness, a ankle is the ankle adaptive gain coefficient, RMS ankle-EMG,k is the state feature vector of the ankle surface electromyogram signal of the k step within the motion distance;

[0135] Step 6.3 Step-by-step adaptation:

[0136] According to the gain update, update the joint torque of the k+1 step within the motion distance, realize step-by-step adaptation, and set the amplitude, slope upper and lower limit value of the gait phase, ensure the parameter to change slowly, and avoid the comfort problem caused by sudden change.

[0137] Step 7: Execute and security degradation

[0138] The method has a built-in safety-comfort parallel mechanism: real-time monitoring of communication synchronization, torque / speed / temperature rise and packet loss, etc. Once the limit is exceeded, trigger smooth degradation (zero torque or high damping) and perform gradual recovery strategy. All transitions follow the maximum slope and phase boundary constraints to ensure stability and controllability in emergency situations. The specific steps are as follows:

[0139] 1. Abnormal detection: communication asynchronization / packet loss, torque / speed / temperature rise exceeding limit;

[0140] 2. Smooth degradation: switch to zero torque or high damping mode;

[0141] 3. Gradual recovery: after the anomaly is recovered, restore to the nominal parameters according to the set slope.

[0142] Effect analysis:

[0143] To verify the effectiveness of the post-support hierarchical collaborative walking control described in the application in engineering implementation and application coverage, representative lower limb exoskeleton systems in recent years are selected for comparison. The comparison dimensions include system weight, degrees of freedom (DOF), gait task scenarios, control strategies, and key elements such as actual assistance effect. All entries are based on the original text of each literature, and "not reported / less than" and other symbols are used to maintain objective consistency. The RSLE prototype of the application is one of the controls, focusing on its collaborative control ability in multiple scenarios (flat ground / slope) and the implementability of hybrid control (prediction + impedance + EMG adjustment). Table 1 gives an overview of the information and control strategy points of each system.

[0144] Table 1 Comparison of representative lower limb exoskeleton systems

[0145] System name System weight Degrees of freedom (DOF) Gait task scenario Control strategy Joint torque variation Gait consistency variation Muscle activation Metabolic cost RSLE exoskeleton 15 kg 9 Flat + slope Predictive + impedance + EMG regulation Flat / slope (hip) 56% / 22.5%↓ (knee) 44% / 43.8%↓ (ankle) 25% / 21.4%↓ GSI: (flat) 8.82%↑ (slope) 14.4%↑ RMS (flat / slope): RF: 20.62% / 36.38%↓ GM: 30.82% / 35.16%↓ VL: 24.7% / 34.13%↓ GC: 8.41% / 35.16%↓ SO: 18% / 28.22%↓ Not reported Hip-knee-ankle whole leg exoskeleton simulation platform 13.5 kg Not reported Flat 1.0 m / s 1.25 m / s 1.5 m / s Hardware-in-the-loop optimization 11%-55%↓ Not reported Not reported 26%↓47%↓50%↓ AI-driven general lower limb exoskeleton system 5.8 kg 6 Flat + slope + stairs AI intent recognition + gait adaptation Not reported Not reported Not reported 6.5%↓ HUMA exoskeleton 10 kg 12 Flat + running Gait phase-based lower limb load assistance strategy Not reported Not reported Not reported Not reported Fully actuated wearable lower limb exoskeleton 35.4 kg 10 Flat + turning + obstacle crossing Obstacle / hardware-in-the-loop control Not reported Not reported Not reported Not reported Lower limb rehabilitation exoskeleton system < 20 kg 10 Walking, standing, sitting, stair ascent / descent Impedance + gait phase estimation Not reported Not reported Not reported Not reported ALEXO active lower limb exoskeleton Not reported 8 Gentle slope + flat Computed torque control (CTC) + trajectory control Not reported Not reported MVC VL: 47.29%↓ VM: 41.94%↓ GM: 59.66%↓ TA: 11.86%↓ Not reported

[0146] As can be seen from Table 1, under the condition of only maintaining the upper limit constraint of the predetermined hardware configuration and parameters, the RSLE has better comprehensive performance in scene coverage and control synergy; compared with the representative method of only using HIL / CTC, the adaptive effective reduction of the EMG across each time reduces the dependence on artificial adjustment and model priori.

[0147] Under the consistent contrast dimension, the RSLE realizes stable adaptation to the flat ground / slope with 9 DOF + hybrid control; by combining synchronous communication and phase scheduling, the controllability of multi-joint cooperative assistance is maintained under the premise of ensuring wearing comfort and safety.

[0148] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Any simple modification, equivalent change and modification of the above embodiment without departing from the technical solution content of the present application, according to the technical essence of the present application, still belongs to the scope of the technical solution of the present application.

Claims

1. A hierarchical collaborative walking control method for a lower limb exoskeleton robot, characterized in that: The specific steps include: Step 1 Data Acquisition and Preprocessing: Collect sensor data, including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot events, surface electromyography (sEMG) signals, and communication synchronization status. Step 2: Perform human-machine coupled dynamics modeling to obtain the required torque τ at the hip, knee, and ankle joints. hip ,τ knee ,τ ankle ; Step 3: Low-level joint control, including: Step 3.1 Hip joint torque: Based on the required torque τ of the hip joint hip The set target torque is used to calculate the hip joint torque based on the set target torque, hip joint feedback torque, hip joint target angle, and hip joint encoder feedback angle. Step 3.2 Virtual impedance torque of knee and ankle: Considering the high sensitivity of the knee and ankle joints to contact compliance and foot comfort, a virtual spring-damping model is used to calculate the virtual impedance torque of the knee and ankle. Step 3.3 Friction Compensation: Static friction and viscous resistance at low speeds can cause small signal hysteresis and micro-vibration. Virtual torque is used to compensate for friction in each joint. Step 4: Mid-level multi-axis coordination and gait phase determination: Gait phase is primarily determined by the zero-crossing point of the hip joint angle from extension to flexion and the velocity threshold, supplemented by plantar pressure and IMU events as auxiliary criteria; when in the stance phase, the virtual stiffness k of the knee joint is increased. knee-virt Ankle joint virtual stiffness k ankle-virt Damping coefficient B virt Enhanced support and disturbance rejection; reduced impedance during the oscillation phase to release oscillation and minimize interference with the natural gait; Step 5: Optimal allocation of energy at higher levels, the specific steps are as follows: Step 5.1 Objective function definition: Under the unified synchronization signal SYNC beat, the assist energy is distributed to the hip, knee and ankle joints within the movement distance. The objective is to minimize the overall score J of the machine while satisfying phase and safety constraints. Step 5.2 Quadratic Programming: Based on the requirements of allocating assistance to the three joints within the phase and safety boundary, while taking into account the requirements of saving effort and stability, we write it as a unified minimization problem, using J as the total amount, and solve the torque of the hip, knee and ankle joints together. Step 5.3 Constraints: Based on Steps 3 and 4, the upper and lower bounds of the torques of each joint are obtained as constraints. During the solution process, the combination of hip, knee, and ankle joint torques that minimizes J is obtained as the output within the feasible region. This completes the optimal allocation of high-level energy and the optimal distribution of assistance to the hip, knee, and ankle joints to achieve a labor-saving and stable effect. Step 6: Stepwise Adaptation of Electromyographic Drive To address changes in human fatigue, the virtual stiffness is updated incrementally over the movement distance. The specific steps are as follows: Step 6.1 Feature extraction: Process the surface electromyography (sEMG) signal, including rectification, filtering, and envelope, and combine it with gait phase and joint state to form a state feature vector; Step 6.2 Gain Update: Based on the state feature vector, the virtual stiffness of the knee and ankle joints is updated stepwise over the motion distance using the form of baseline and gain. Step 6.3 Gradual Adaptation: Based on the gain update, update the joint torques of each step within the k+1 steps of the movement distance to achieve gradual adaptation, and set the upper and lower limits of the amplitude and slope of the gait phase to ensure that the parameters change slowly and avoid comfort problems caused by sudden changes.

2. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 1, the data acquisition and preprocessing involves the following steps: Step 1.1 Acquire sensor data: including hip joint torque or equivalent feedback, knee angle and angular velocity, ankle angle and angular velocity, gait and foot events, surface electromyography (sEMG) signals, and communication synchronization status; Step 1.2 Signal preprocessing: including denoising, filtering, zero-point calibration, gain calibration, and cross-device time alignment; Step 1.3 Data Buffering and Timestamps: Establish a circular buffer queue and write data according to the SYNC clock to ensure that data from multiple joints are aligned within the same control cycle.

3. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 2, the human-machine coupled dynamics modeling is performed to determine the required torque τ of the hip, knee, and ankle joints. hip ,τ knee ,τ ankle The specific steps are as follows: Step 2.1 Establish the generalized coordinate system of the exoskeleton and the human body in series or parallel. =[q hip ,q knee ,q ankle ] Step 2.2 Establish the dynamic equations: ； In the formula: M is the inertia matrix; C is the Coriolis and centrifugal term; G is the gravity term; τ exo The driving torque for the exoskeleton; τ human θ represents the active torque of human muscles; θ is the joint angle. Joint angular velocity; Step 2.3 Solving for the desired torque τ req : ； Where, τ hip τ knee τ ankle M represents the required torque for the hip, knee, and ankle joints, respectively. hip M knee、 M ankle These are the effective inertial terms for the hip, knee, and ankle joints, respectively. hip , knee , ankle These are the angular accelerations of the hip, knee, and ankle joints, respectively. hip , knee , ankle These are the angular velocities of the hip, knee, and ankle joints, respectively.

4. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 3.1, the hip joint torque is determined based on the required hip joint torque τ. hip The set target torque is used to calculate the hip joint torque based on the set target torque, hip joint feedback torque, hip joint target angle, and hip joint encoder feedback angle, expressed as: ； Where, τ hip-cmd For the hip joint torque, τ hip-dcs According to τ hip The set target torque for the hip joint, τ sensor K is the feedback torque of the hip joint. p K is the proportional gain of the torque error. d The gain is the derivative of the angle error; and These represent the target angular velocity of the hip joint and the feedback angular velocity of the hip joint encoder, respectively.

5. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 3.2, the virtual resistance torque of the knee joint is calculated as follows: ； Where, τ knee-virtual k represents the virtual torque at the knee joint. knee-virt For the virtual stiffness of the knee joint, B virt θ is the damping coefficient. knee-dcs and θ knee-enc These are the target angle of the knee joint and the encoder feedback angle, respectively. The knee joint encoder provides feedback on angular velocity. The virtual impedance torque of the ankle joint is calculated as follows: ； Where, τ ankle-virtual k represents the virtual torque of the ankle joint. ankle-virt For the virtual stiffness of the ankle joint, B virt θ is the damping coefficient. ankle-dcs and θ ankle-enc These are the target angle of the ankle joint and the encoder feedback angle, respectively. The ankle joint encoder provides feedback angular velocity.

6. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 3.3, the friction compensation is as follows: static friction and viscous resistance at low speeds can cause small signal hysteresis and micro-vibration. Virtual torque is used to compensate for friction in each joint. include Knee joint friction compensation is: ； Among them, F knee-cmd For the knee joint command force, τ knee-virtual For the virtual torque of the knee joint, F c and F v These are the coefficients of static friction and viscous friction, respectively. The speed of the electric cylinder in the knee joint; Ankle friction compensation is: ； Among them, F ankle-cmd For the ankle joint command force, τ ankle-virtual For the virtual torque of the ankle joint, F c and F v These are the coefficients of static friction and viscous friction, respectively. The speed of the ankle joint electric cylinder.

7. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 4, the mid-layer multi-axis coordination and gait phase determination: gait phase is primarily determined by the zero-crossing point of the hip joint angle from extension to flexion and the rate threshold, with plantar pressure and IMU events used as auxiliary criteria; when in the stance phase, k is increased. knee-virt k ankle-virt B virt This enhances support and disturbance rejection; during the oscillation phase, impedance is reduced to release the oscillation and minimize interference with the natural gait. The specific steps are as follows: Step 4.1 Bus Synchronization: The CANopen bus is used. The master station periodically sends a synchronization signal SYNC. Key measurement data and control commands are transmitted and received via the synchronization process data object PDO. Step 4.2 Gait phase determination: Based on the zero-crossing point of the hip joint angle from extension to flexion and its rate threshold, the support phase and swing phase are determined, which can be combined with plantar pressure and IMU events; Step 4.3 Phase Scheduling: Supporting Upward Adjustment of k knee-virt k ankle-virt B virt To enhance support and resistance to disturbances, the oscillation phase is reduced to release the oscillation.

8. The hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 5.1, the objective function is defined as follows: Under a uniform SYNC beat, the assist energy is distributed to the hip, knee, and ankle joints within the movement distance. The goal is to minimize the overall machine score while satisfying phase and safety constraints. The objective function is expressed as: ； Where J is the total score, a smaller J indicates less effort required and a smoother movement; w1 and w2 are the weights; τ i Let be the torque variable of the i-th joint; Let be the angular velocity of the i-th joint.

9. A hierarchical collaborative walking control method for a lower limb exoskeleton robot as described in claim 1, characterized in that: In step 6.2, the gain update involves progressively updating the virtual stiffness of the knee and ankle joints over the motion distance based on the state feature vector, using baseline and gain parameters. This is expressed as follows: K knee-virt,k+1 =K knee-base +α knee RMS knee-EMG,k ； Among them, K knee-virt,k+1 Let K be the virtual stiffness of the knee joint at the (k+1)th step within the movement distance. knee-base For the baseline stiffness of the knee joint, α knee RMS is the adaptive gain coefficient for the knee joint. knee-EMG,k is the state feature vector of the surface electromyography signal of the knee joint within the k-th step of the movement distance; K ankle-virt,k+1 =K ankle-base +α ankle RMS ankle-EMG,k ； Among them, K ankle-virt,k+1 Let K be the virtual stiffness of the ankle joint at the (k+1)th step within the movement distance. ankle-base For ankle baseline stiffness, α ankle RMS is the adaptive gain coefficient for the ankle joint. ankle-EMG,k It is the state feature vector of the surface electromyography signal of the ankle joint during the k-th step of the movement distance.

Images