Gait training control methods, systems and devices

Abstract

This invention provides a control method, system, and device for gait training, which can be applied to the field of artificial intelligence technology. The control method for gait training includes inputting the patient's current gait data while walking on a treadmill into a prediction model, outputting gait prediction information and center of gravity prediction information; adjusting the weight-reduction control parameters of the weight-reduction device and the lower limb joint angles of the lower limb robot based on the gait prediction information and center of gravity prediction information; determining the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight-reduction control command time of the weight-reduction device based on the gait prediction information, the center of gravity prediction information, and the compensation time of the treadmill, the lower limb robot, and the weight-reduction device; and controlling the treadmill, the lower limb robot, and the weight-reduction device using the speed control command time, the joint control command time, and the weight-reduction control command time respectively, so that the treadmill, the lower limb robot, and the weight-reduction device synchronously execute control commands that match the gait prediction information.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more specifically to a control method, system, and apparatus for gait training. Background Technology

[0002] With the rapid development of lower limb robotics technology, it has shown great potential in the rehabilitation of patients with neurological diseases such as stroke and spinal cord injury. Lower limb robots are often used in conjunction with other rehabilitation equipment to assist patients in gait training (such as weight-reducing suspension systems that can reduce the stress on the lower limbs during gait training), thereby improving the safety and comfort of gait training.

[0003] In lower limb rehabilitation training, multiple devices are needed to assist patients in gait training. However, these devices often have poor coordination and control. Summary of the Invention

[0004] In view of the above problems, the present invention provides a control method, system and apparatus for improving gait training.

[0005] According to a first aspect of the present invention, a control method for gait training is provided, comprising: inputting current gait data of a patient walking on a treadmill into a prediction model, and outputting gait prediction information and center of gravity prediction information, wherein the patient wears a lower limb robot on the lower limbs and a weight-reducing device on the pelvic area, the weight-reducing device being used to reduce the force on the patient's lower limbs during walking, and the lower limb robot being used to correct the lower limb joint angles of the patient's abnormal walking posture, so as to cooperate with the treadmill for gait training; and adjusting the weight-reducing control parameters of the weight-reducing device and the lower limb joints of the lower limb robot according to the gait prediction information and the center of gravity prediction information, respectively. Angle; Based on gait prediction information, center of gravity prediction information, and the compensation time of the treadmill, lower limb robot, and weight reduction device, determine the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device. The compensation time is determined based on the historical delay time of the treadmill, lower limb robot, and weight reduction device. Use the speed control command time, joint control command time, and weight reduction control command time to control the treadmill, lower limb robot, and weight reduction device respectively, so that the treadmill, lower limb robot, and weight reduction device synchronously execute control commands that match the gait prediction information.

[0006] According to an embodiment of the present invention, the current gait data of a patient walking on a treadmill is input into a prediction model, and gait prediction information and center of gravity prediction information are output. This includes: extracting features from the current gait data using a two-layer joint coding layer of the prediction model to obtain local gait features and global gait features. The current gait data includes joint motion parameters, gait cycle, plantar pressure, and center of gravity fluctuation trajectory. The local gait features characterize the correlation between joint motion parameters, plantar pressure, and center of gravity fluctuation trajectory, respectively. The global gait features are the temporal linkage features of the gait cycle and the center of gravity fluctuation trajectory. The plantar pressure is collected by a plantar pressure sensor installed on a lower limb robot. The local gait features and global gait features are fused using a network sharing layer of the prediction model to obtain fused features. The fused features are decoded using a two-branch decoding layer of the prediction model to obtain gait prediction information and center of gravity prediction information.

[0007] According to an embodiment of the present invention, the fused features are decoded using a dual-branch decoding layer of the prediction model to obtain gait prediction information and center of gravity prediction information, including: using the gait decoding branch in the dual-branch decoding layer to perform a first decoding of the fused features to obtain gait prediction information; and using the center of gravity decoding branch in the dual-branch decoding layer to perform a second decoding of the fused features to obtain center of gravity prediction information.

[0008] According to an embodiment of the present invention, the weight reduction device includes: a suspension bracket mounted above a treadmill; a vertically floating axle fitted inside the suspension bracket and extending and retracting vertically along the suspension bracket; a pelvic support strap suspended from the lower end of the vertically floating axle; a motor mounted on the suspension bracket for driving the vertically floating axle to extend and retract vertically to adjust the position of the pelvic support strap in the vertical direction; and a variable damping adjustment unit connected to the vertically floating axle for matching the patient's trunk floating motion inertia during walking.

[0009] According to an embodiment of the present invention, the weight reduction control parameters include the damping coefficient of the variable damping adjustment unit and the response speed of the vertical floating actuation axis; the weight reduction control parameters of the weight reduction device and the lower limb joint angle of the lower limb robot are adjusted forward according to gait prediction information and center of gravity prediction information, respectively, including: adjusting the damping coefficient of the variable damping adjustment unit based on gait phase information and center of gravity prediction information in the gait prediction information; adjusting the response speed of the vertical floating actuation axis based on the predicted center of gravity floating rate in the center of gravity prediction information based on the first correlation between the response speed and the center of gravity floating rate; and adjusting the lower limb joint angle based on the predicted center of gravity change in the center of gravity prediction information based on the second correlation between the center of gravity change and the lower limb joint angle.

[0010] According to an embodiment of the present invention, the damping coefficient of the feedforward variable damping adjustment unit is adjusted based on the predicted gait phase information and center of gravity prediction information in the gait prediction information, including: when the predicted gait phase information is a support phase and the predicted center of gravity fluctuation trajectory corresponding to the support shows a downward trend in the center of gravity, the damping coefficient is increased by a first preset ratio, wherein the center of gravity prediction information includes the predicted center of gravity fluctuation trajectory; when the gait phase information is a swing phase and the predicted inertial acceleration at multiple moments corresponding to the swing shows a decreasing trend, the damping coefficient is decreased by a second preset ratio, wherein the center of gravity prediction information includes the predicted inertial acceleration.

[0011] According to an embodiment of the present invention, after adjusting the weight reduction control parameters of the weight reduction device and the lower limb joint angle of the lower limb robot according to the gait prediction information and the center of gravity prediction information respectively, the method further includes: when the real-time suspension tension obtained by the tension sensor is not within the preset tension range, adjusting the extension length of the vertical floating actuator via motor feedback, wherein the tension sensor is disposed between the pelvic support sling and the suspension bracket.

[0012] According to an embodiment of the present invention, determining the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on gait prediction information, center of gravity prediction information, and the respective compensation times of the treadmill, the lower limb robot, and the weight reduction device includes: determining the running speed of the treadmill based on the predicted gait cycle and predicted cadence in the gait prediction information, and determining the speed control command time based on the compensation time and running speed of the treadmill; determining the joint movements of the lower limb robot based on the predicted gait cycle and the predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and determining the joint control command time based on the compensation time and joint movements of the lower limb robot; and determining the extension and retraction movements of the vertical floating actuation axis based on the predicted center of gravity floating rate and predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and determining the weight reduction control command time based on the compensation time and extension and retraction movements of the vertical floating actuation axis of the weight reduction device.

[0013] A second aspect of the present invention provides a gait training control system, comprising: a treadmill for simulating walking movements for a patient; a lower limb robot worn on the patient's lower limbs for correcting abnormal walking postures of the patient's lower limb joint angles; a weight-reducing device mounted above the treadmill and worn on the patient's pelvis for reducing the force on the patient's lower limbs during walking; and a controller communicatively connected to the treadmill, the lower limb robot, and the weight-reducing device for executing the method of any one of the above.

[0014] A third aspect of the present invention provides a control device for gait training, comprising: an input module for inputting current gait data of a patient walking on a treadmill into a prediction model and outputting gait prediction information and center of gravity prediction information, wherein the patient wears a lower limb robot on their lower limbs and a weight-reducing device on their pelvis, the weight-reducing device being used to reduce the force on the patient's lower limbs during walking, and the lower limb robot being used to correct the lower limb joint angles of the patient's abnormal walking posture, in order to cooperate with the treadmill for gait training; and a feedforward adjustment module for adjusting the weight-reducing control parameters of the weight-reducing device and the lower limb joint angles of the lower limb robot according to the gait prediction information and the center of gravity prediction information, respectively. The system comprises: a determination module, used to determine the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on gait prediction information, center of gravity prediction information, and the compensation time of each of the treadmill, lower limb robot, and weight reduction device; and a control module, used to control the treadmill, lower limb robot, and weight reduction device respectively using the speed control command time, joint control command time, and weight reduction control command time, so that the treadmill, lower limb robot, and weight reduction device synchronously execute control commands that match the gait prediction information.

[0015] A fourth aspect of the present invention provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0016] A fifth aspect of the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0017] A sixth aspect of the present invention also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0018] According to an embodiment of the present invention, by inputting the patient's current gait data into the prediction model, gait prediction information and center of gravity prediction information are output, and then using the gait prediction information and center of gravity prediction information to feedforward and adjust the weight reduction device and the lower limb robot respectively, so as to realize the prediction and early control of the patient's gait.

[0019] To enable the treadmill, lower limb robot, and weight-reduction device to achieve identical and synchronized gait movements, the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight-reduction control command time of the weight-reduction device are determined based on gait prediction information, center of gravity prediction information, and the compensation time of each device. This allows the treadmill, lower limb robot, and weight-reduction device to synchronously execute control commands that match the gait prediction information, avoiding motion interference between multiple devices and improving the coordination between the treadmill, lower limb robot, and weight-reduction device in gait training. Attached Figure Description

[0020] The above-described features, other objects, and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings, in which:

[0021] Figure 1 This diagram illustrates an application scenario of the gait training control method according to an embodiment of the present invention.

[0022] Figure 2 A flowchart illustrating a gait training control method according to an embodiment of the present invention is shown schematically.

[0023] Figure 3 A schematic diagram of a gait training control system according to an embodiment of the present invention is shown.

[0024] Figure 4A A flowchart illustrating a gait training control method according to another embodiment of the present invention is shown schematically;

[0025] Figure 4B A flowchart illustrating the three-axis spatiotemporal integrated synchronous cooperative control according to an embodiment of the present invention is shown schematically.

[0026] Figure 4C A flowchart illustrating phase-separated adaptive constant force control according to an embodiment of the present invention is shown schematically.

[0027] Figure 5 A flowchart illustrating asymmetric gait differentiation compensation control according to an embodiment of the present invention is shown schematically.

[0028] Figure 6 A flowchart illustrating a predictive model training method according to an embodiment of the present invention is shown schematically.

[0029] Figure 7 A schematic diagram illustrating the structure of a control device for gait training according to an embodiment of the present invention is shown.

[0030] Figure 8 A block diagram of an electronic device suitable for implementing a control method for gait training according to an embodiment of the present invention is shown schematically. Detailed Implementation

[0031] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0032] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0033] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0034] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0035] In recent years, with the rapid development of lower limb robotics technology, it has shown great potential in the rehabilitation of patients with neurological diseases such as stroke and spinal cord injury. However, rehabilitation equipment still faces multiple technical bottlenecks, with prominent problems such as poor clinical training effects and slow functional recovery.

[0036] In the relevant embodiments, the constant force control for weight reduction exhibits significant lag. The weight reduction system uses a single tension sensor for real-time feedback adjustment, which can only passively respond to the up-and-down movement of the torso caused by walking. It is difficult to predict fluctuations in the center of gravity in advance, resulting in large fluctuations in the weight reduction force, which can easily cause a pulling or sinking sensation, affecting the safety and comfort of rehabilitation training. The root cause of this problem is that the relevant control strategy is a passive follow-up mode, lacking the forward-looking prediction of the patient's movement trends.

[0037] In related embodiments, the treadmill, lower limb robot, and weight-loss system are controlled independently or in simple linkage, achieving only basic gait phase synchronization. The vertical movement of the human torso is not incorporated as an independent control variable into the collaborative system, leading to motion interference and gait distortion issues. Specifically, the lower limb robot, treadmill, and weight-loss system are often independent structures, requiring manual positioning based on personal experience. When positioning is inaccurate, multiple devices do not align with the human body's motion axis, resulting in poor rehabilitation training effects or even secondary injury to the patient. Furthermore, since the devices are provided by different manufacturers with inconsistent communication protocols, coordination is difficult, making true spatiotemporal synchronization impossible.

[0038] In the relevant embodiments, the constant force control mode is singular, using a fixed weight reduction ratio / fixed tension control throughout, without distinguishing the changes in the center of gravity and force differences between the gait support phase and swing phase, and thus failing to adapt to the biomechanical laws of natural human walking. Studies have pointed out that hemiplegic patients have special rehabilitation needs that are "two sides of the same coin," namely, the affected side needs precise guidance, while the healthy side needs smooth coordination, and the related control strategies are difficult to meet this need with a "one-size-fits-all" approach.

[0039] In summary, the gait prediction technology currently used is only applied to lower limb robot trajectory planning and is not deeply integrated with weight-loss constant force regulation, trunk floating compensation, and treadmill speed control. Therefore, the control precision and synchronization are insufficient to meet clinical rehabilitation needs. The current level of intelligence has not yet reached the requirement of "providing patients with the necessary support at specific times and under specific circumstances," making it difficult to achieve personalized rehabilitation services for different patients.

[0040] In view of this, embodiments of the present invention provide a gait training control method comprising: inputting current gait data of a patient walking on a treadmill into a prediction model, and outputting gait prediction information and center of gravity prediction information, wherein the patient wears a lower limb robot on the lower limbs and a weight-reducing device on the pelvic area. The weight-reducing device is used to reduce the force on the patient's lower limbs during walking, and the lower limb robot is used to correct the lower limb joint angles of the patient's abnormal walking posture, so as to cooperate with the treadmill for gait training; and adjusting the weight-reducing control parameters of the weight-reducing device and the lower limb joint angles of the lower limb robot according to the gait prediction information and the center of gravity prediction information, respectively. Based on gait prediction information, center of gravity prediction information, and the compensation times of the treadmill, lower limb robot, and weight reduction device, the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device are determined. The compensation time is determined based on the historical delay time of each of the treadmill, lower limb robot, and weight reduction device. The speed control command time, joint control command time, and weight reduction control command time are used to control the treadmill, lower limb robot, and weight reduction device respectively, so that the treadmill, lower limb robot, and weight reduction device synchronously execute control commands that match the gait prediction information.

[0041] Figure 1 The diagram illustrates an application scenario of the gait training control method according to an embodiment of the present invention.

[0042] like Figure 1 As shown, the application scenario diagram of the gait training control method according to this embodiment may include a treadmill 110, a lower limb robot 120, and a weight reduction device 130.

[0043] In lower limb weight-loss rehabilitation training, the patient can walk on a treadmill 110, with a lower limb robot 120 worn on the lower limbs and a weight-loss device 130 worn on the pelvic area. The weight-loss device 130 includes a suspension bracket 131, which is mounted above the treadmill 110.

[0044] The weight reduction device 130 is used to reduce the force on the lower limbs of the patient during walking, and the lower limb robot 120 is used to correct the lower limb joint angles of the patient's abnormal walking posture, so as to cooperate with the treadmill 110 for gait training.

[0045] For example, treadmill 110 allows patients to walk, and the walking speed can be adjusted according to the patient's recovery progress.

[0046] For example, the weight reduction device 130 can adjust the vertical position of the pelvic support sling worn on the patient's pelvis by controlling the vertical floating actuator shaft mounted inside the suspension bracket 131 to make vertical extension and retraction movements along the suspension bracket, thereby adjusting the force on the patient's lower limbs during walking.

[0047] For example, the lower limb robot 120 can adopt a lower limb exoskeleton structure, with servo motors and angle sensors at its hip, knee and ankle joints, and foot pressure sensors at the soles of its feet. It can correct the movement trajectory of the hip, knee and ankle joints in real time based on the data from each sensor, adapt to the lower limb posture deviation caused by trunk floating, and support independent control of the trajectory of the left and right joints to adapt to the needs of asymmetrical gait differentiation compensation.

[0048] Figure 2 A flowchart illustrating a gait training control method according to an embodiment of the present invention is shown.

[0049] like Figure 2 As shown, the gait training control method of this embodiment includes operations S210 to S240.

[0050] In operation S210, the patient's current gait data while walking on the treadmill is input into the prediction model, and gait prediction information and center of gravity prediction information are output. The patient wears a lower limb robot on the lower limbs and a weight reduction device on the pelvis. The weight reduction device is used to reduce the force on the patient's lower limbs during walking, and the lower limb robot is used to correct the lower limb joint angles of the patient's abnormal walking posture in order to cooperate with the treadmill for gait training.

[0051] According to embodiments of the present invention, the patient can be a user who needs gait training to achieve voluntary movement. For example, patients with neurological diseases such as stroke and spinal cord injury can achieve normal walking through gait training.

[0052] According to an embodiment of the present invention, the treadmill can be an intelligent treadmill, which can adjust the speed of the treadmill according to gait prediction information and center of gravity prediction information.

[0053] According to embodiments of the present invention, the current gait data can be gait-related data collected from a lower limb robot, a treadmill, and a weight-loss device. For example, the current gait data may include lower limb joint angles, treadmill speed, weight loss, etc., and may also include the patient's height, weight, weight loss percentage, training speed, and intensity.

[0054] According to embodiments of the present invention, the prediction model can be an intelligent algorithm such as a neural network model, a support vector machine model, or a random forest.

[0055] According to an embodiment of the present invention, the gait prediction information includes the gait pose at each moment within multiple prediction cycles. For example, the gait prediction information may include the prediction cycle, the predicted angle of the lower limb joints, and the phase information of each cycle.

[0056] According to embodiments of the present invention, the center of gravity prediction information may be information related to the patient's center of gravity. For example, the vertical height of the center of gravity, trunk inertial acceleration, etc.

[0057] According to an embodiment of the present invention, the lower limb robot can adopt a lower limb exoskeleton structure, with servo motors and angle sensors installed at the patient's hip, knee, and ankle joints, and foot pressure sensors installed at the soles of the feet, which can obtain real-time hip, knee, and ankle joint movement trajectories.

[0058] According to an embodiment of the present invention, the weight-reduction device may include a device that suspends the patient and is worn on the patient's pelvic area to reduce the stress on the patient's lower limbs during walking.

[0059] During operation of S220, the weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot are adjusted forward based on gait prediction information and center of gravity prediction information, respectively.

[0060] According to embodiments of the present invention, feedforward adjustment is an adjustment method that anticipates disturbances and compensates for them in advance. For example, when disturbance signals such as patient movement posture, tension waveform, and walking inertia are detected from gait prediction information and center of gravity prediction information, the weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot can be adjusted in advance to improve the patient's walking training effect while making the training process more comfortable.

[0061] According to embodiments of the present invention, the weight reduction control parameters can be control parameters relating to the weight reduction device. For example, the weight reduction control parameters can be weight reduction tension, suspension stroke, etc.

[0062] According to embodiments of the present invention, the lower limb joint angles may include hip joint angles, knee joint angles, and ankle joint angles. By adjusting the lower limb joint angles of the lower limb robot using feedforward adjustment, abnormal gait postures of patients can be corrected.

[0063] In operation S230, based on gait prediction information, center of gravity prediction information, and the compensation time of the treadmill, lower limb robot, and weight reduction device, the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device are determined. The compensation time is determined based on the historical delay time of the treadmill, lower limb robot, and weight reduction device.

[0064] According to an embodiment of the present invention, the compensation time can be the average delay time of multiple historical delay times of the treadmill / lower limb robot or weight-loss device. For example, if the multiple historical delay times of the treadmill are 220 milliseconds, 200 milliseconds, and 230 milliseconds, the average delay is calculated to be 220 milliseconds. The average delay time is used as the compensation time.

[0065] According to embodiments of the present invention, speed control commands for the treadmill, joint control commands for the lower limb robot, and weight reduction control commands for the weight reduction device can be determined based on gait prediction information and center of gravity prediction information. For example, the speed control command may include the treadmill speed in each prediction cycle, and the joint control command may include the hip joint angle, knee joint angle, and ankle joint angle in each prediction cycle. The weight reduction control command may include weight reduction control parameters in each prediction cycle.

[0066] The speed control command timing can be determined based on the speed control commands and the treadmill's compensation time. The joint control command timing can be determined based on the lower limb robot's compensation time and joint control commands. The weight reduction control command timing can be determined based on the weight reduction control commands and the weight reduction device's compensation time.

[0067] In operation S240, the speed control command time, joint control command time, and weight reduction control command time are used to control the treadmill, lower limb robot, and weight reduction device respectively, so that the treadmill, lower limb robot, and weight reduction device synchronously execute control commands that match the gait prediction information.

[0068] According to an embodiment of the present invention, the treadmill, lower limb robot, and weight-reduction device synchronously execute control commands that match the gait prediction information, which can avoid motion interference between multiple devices and improve the coordination between the treadmill, lower limb robot, and weight-reduction device in gait training.

[0069] According to an embodiment of the present invention, by inputting the patient's current gait data into the prediction model, gait prediction information and center of gravity prediction information are output, and then using the gait prediction information and center of gravity prediction information to feedforward and adjust the weight reduction device and the lower limb robot respectively, so as to realize the prediction and early control of the patient's gait.

[0070] To enable the treadmill, lower limb robot, and weight-reduction device to achieve identical and synchronized gait movements, the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight-reduction control command time of the weight-reduction device are determined based on gait prediction information, center of gravity prediction information, and the compensation time of each device. This allows the treadmill, lower limb robot, and weight-reduction device to synchronously execute control commands that match the gait prediction information, avoiding motion interference between multiple devices and improving the coordination between the treadmill, lower limb robot, and weight-reduction device in gait training.

[0071] According to an embodiment of the present invention, the current gait data of a patient walking on a treadmill is input into a prediction model, and gait prediction information and center of gravity prediction information are output. This includes: extracting features from the current gait data using a two-layer joint coding layer of the prediction model to obtain local gait features and global gait features. The current gait data includes joint motion parameters, gait cycle, plantar pressure, and center of gravity fluctuation trajectory. The local gait features characterize the correlation between joint motion parameters, plantar pressure, and center of gravity fluctuation trajectory, respectively. The global gait features are the temporal linkage features of the gait cycle and the center of gravity fluctuation trajectory. The plantar pressure is collected by a plantar pressure sensor installed on a lower limb robot. The local gait features and global gait features are fused using a network sharing layer of the prediction model to obtain fused features. The fused features are decoded using a two-branch decoding layer of the prediction model to obtain gait prediction information and center of gravity prediction information.

[0072] The prediction model can be a custom-built LSTM / GRU deep learning model. LSTM (Long Short-Term Memory) is a long short-term memory network. GRU (Gated Recurrent Unit) is a simplified version of LSTM with fewer parameters and faster computation.

[0073] The prediction model can achieve multi-step advance prediction in two dimensions: gait and center of gravity. Unlike the existing general single-dimensional prediction model, it adopts a dual-input, dual-output architecture to achieve deep fusion and synchronous prediction of gait and center of gravity features, providing advance prediction basis for subsequent control.

[0074] Current gait data can include both gait dimension data and center of gravity dimension data. For example, gait dimension data may include time-series signals of plantar pressure at a sampling frequency of 100 Hz, time-series signals of hip / knee / ankle joint angles in the lower limbs, and historical data from the last 5-10 gait cycles. Center of gravity dimension data may include time-series data such as trunk inertial acceleration and angular velocity, vertical height of the center of gravity, and suspension tension. It should be noted that current gait data can also include auxiliary data such as the patient's height, weight, weight loss percentage, training speed, and intensity.

[0075] The system includes a torso IMU (Inertial Measurement Unit) attitude sensor, joint angle sensors, plantar pressure sensors, and tension sensors. Each sensor is connected to the controller to collect real-time data on human gait phase, vertical center of gravity height, joint angles, plantar force, suspension tension, and other multi-dimensional data. For example, a torso IMU attitude sensor can be installed at each joint of a lower limb robot, and a plantar pressure sensor can be installed at the sole of the foot. Tension sensors can be positioned between the pelvic support sling and the suspension frame.

[0076] The preprocessing of the current gait data may include: using a sliding window of one gait cycle dynamically matched to the patient's real-time gait frequency to extract continuous time-series data; filtering high-frequency noise through adaptive Kalman filtering; using max-min normalization to map all data to the [0,1] interval to eliminate dimensional differences; concatenating gait and center-of-gravity related features into a 12-dimensional time-series feature vector (such as 3-dimensional plantar pressure + 3-dimensional joint angle + 2-dimensional IMU acceleration + 2-dimensional center-of-gravity height + 1-dimensional tension + 1-dimensional gait frequency), finally forming a time-series feature sequence of length T (T is the number of sampling points in one gait cycle), which is then input into the prediction model.

[0077] A dual-layer LSTM / GRU network (either can be selected, but the processing logic is consistent) is used to construct a "joint encoding-dual-branch decoding" architecture.

[0078] For example, the first layer of the dual-layer joint coding layer, LSTM / GRU, is used to extract short-term local temporal dependencies and learn the correlation between joint motion parameters and plantar pressure within gait segments and the trajectory of center of gravity fluctuations, so as to output 64-dimensional gait local features.

[0079] The second layer of the dual-layer joint coding layer, LSTM / GRU, is used to extract the temporal linkage features between long-period gait rhythm and center of gravity fluctuation trajectory, outputting 128-dimensional global gait features.

[0080] The network sharing layer of a two-layer LSTM / GRU network can share some weights to achieve deep fusion of two-dimensional features.

[0081] A dropout layer (probability 0.2) can be set between the two encoding layers to prevent overfitting, and batch normalization can be performed through a BN layer to accelerate model convergence and improve the stability of feature extraction.

[0082] The dual-branch decoding layer can include two parallel decoding branches: one decoding branch is used to decode gait prediction information, and the other decoding branch is used to decode center of gravity prediction information.

[0083] The time-series regression layer of a two-layer LSTM / GRU network predicts the results for the next 1-3 gait cycles at once, and then restores them to the actual physical quantities through inverse normalization to meet subsequent control requirements.

[0084] The online self-learning process of a two-layer LSTM / GRU network may include: using the mean squared error as the loss function, employing the Adam optimizer, and fine-tuning the model weights online after every 5 gait cycles of prediction; introducing a personalized self-learning mechanism to memorize the patient's gait and center of gravity characteristics, and dynamically correcting parameters to improve prediction accuracy.

[0085] The dual-layer LSTM / GRU network outputs gait prediction information and center of gravity prediction information in two dimensions simultaneously. The gait prediction information can include the support phase / swing phase switching time (accuracy 0.01 seconds), gait frequency, gait cycle duration, and phase duty cycle for the next 1-3 gait cycles.

[0086] The center of gravity prediction information can include the center of gravity fluctuation trajectory for the next 1-3 gait cycles (0.01 seconds / data point), the predicted center of gravity floating rate, the predicted inertial acceleration, and the predicted center of gravity extreme value, etc.

[0087] The prediction model constructs a dual-dimensional fusion feature of gait and center of gravity at its input, rather than a single feature; it adopts a joint encoding-dual-branch decoding architecture to ensure simultaneous prediction of both dimensions; it has online self-learning capabilities to adapt to different patients and rehabilitation stages; it is specifically designed for weight loss and rehabilitation scenarios, and the prediction results are directly connected to the control link, solving the problems of single prediction, poor linkage, and insufficient adaptability.

[0088] According to an embodiment of the present invention, the fused features are decoded using a dual-branch decoding layer of the prediction model to obtain gait prediction information and center of gravity prediction information, including: using the gait decoding branch in the dual-branch decoding layer to perform a first decoding of the fused features to obtain gait prediction information; and using the center of gravity decoding branch in the dual-branch decoding layer to perform a second decoding of the fused features to obtain center of gravity prediction information.

[0089] The dual-branch decoding layer can include a parallel gait decoding branch and a centroid decoding branch. The gait decoding branch (2 fully connected layers + 1 temporal regression layer) outputs gait prediction information; the centroid decoding branch (structured symmetrical to the gait decoding branch) outputs centroid prediction information. The two branches share global features to ensure prediction synchronization.

[0090] According to an embodiment of the present invention, the weight reduction device includes: a suspension bracket mounted above a treadmill; a vertically floating axle fitted inside the suspension bracket and extending and retracting vertically along the suspension bracket; a pelvic support strap suspended from the lower end of the vertically floating axle; a motor (such as a servo motor) mounted on the suspension bracket for driving the vertically floating axle to extend and retract vertically to adjust the position of the pelvic support strap in the vertical direction; and a variable damping adjustment unit connected to the vertically floating axle for matching the patient's trunk floating motion inertia during walking.

[0091] Patients have different paces, and their trunk floating inertia and undulation amplitudes are completely different. Fixed damping cannot be adapted to everyone and requires adaptive parameter adjustment.

[0092] If the damping is too large, problems such as mechanical lag, slow following, dragging of the body, and gait distortion will occur. If the damping is too small, problems such as system resonance, up-and-down swaying, and fluctuating weight reduction force will occur. Therefore, adaptive damping needs to be precisely within the critical optimal range to achieve a state of keeping up without hesitation or oscillation.

[0093] The variable damping adjustment unit can adaptively adjust the damping coefficient according to the floating inertia of the torso, and match the dynamic inertial characteristics of the vertical undulation of the torso during human walking in real time, effectively suppressing the resonant vibration of the suspension mechanism and eliminating mechanical following jamming and lag phenomena.

[0094] Simultaneously, the vertical floating actuator axis independently achieves trunk floating and following, without interfering with the body's horizontal gait, providing hardware support for decoupled control of trunk vertical floating. Complete decoupling of the body's horizontal gait and vertical floating motion avoids interference from suspension tension on normal walking and pelvic posture, significantly reducing gravity fluctuations and improving the smoothness, comfort, and safety of rehabilitation training, providing reliable hardware support for decoupled control of trunk vertical floating.

[0095] The weight-reduction device may include a tension sensor positioned between the pelvic support sling and the suspension bracket to collect suspension tension signals in real time and feed them back to the controller. The controller can be used to implement gait training control methods.

[0096] According to an embodiment of the present invention, the weight reduction control parameters include the damping coefficient of the variable damping adjustment unit and the response speed of the vertical floating actuation axis; the weight reduction control parameters of the weight reduction device and the lower limb joint angle of the lower limb robot are adjusted forward according to gait prediction information and center of gravity prediction information, respectively, including: adjusting the damping coefficient of the variable damping adjustment unit based on gait phase information and center of gravity prediction information in the gait prediction information; adjusting the response speed of the vertical floating actuation axis based on the predicted center of gravity floating rate in the center of gravity prediction information based on the first correlation between the response speed and the center of gravity floating rate; and adjusting the lower limb joint angle based on the predicted center of gravity change in the center of gravity prediction information based on the second correlation between the center of gravity change and the lower limb joint angle.

[0097] Gait phase information can include the support phase and the swing phase. For example, in the support phase, the leg sinks with a large impact, requiring increased damping for cushioning and shock absorption; in the swing phase, the leg lifts with less inertia, requiring reduced damping for easier following. Fixed damping cannot achieve phase-specific adaptation and can only provide rigid following. Therefore, by adjusting the damping coefficient of the variable damping unit based on the gait phase information and center of gravity prediction information from the gait prediction information, the dynamic inertial characteristics of the vertical undulation of the torso during human walking can be matched in real time.

[0098] The formula for the first correlation between response speed and center of gravity float rate is as follows:

[0099] v axis =k×v z + c (1)

[0100] v z It is the velocity of the center of gravity (unit: m / s), v z It is calculated from IMU sensor and laser ranging data. The laser ranging data is the target distance between the suspension bracket and the top of the patient's head, which is measured by the laser sensor on the suspension bracket. The center of gravity height can be obtained by subtracting the target distance and the patient's torso length from the height of the laser sensor.

[0101] v z =Δh / Δt, Δh=h1-h2, Δt=t1-t2, then h1 and h2 are the heights of the center of gravity at times t1 and t2 respectively, and Δh is the change in the center of gravity.

[0102] v axisThis can be the response speed (in m / s) of the vertical floating execution axis. k is the adaptation coefficient. k is adjusted according to the patient's weight and rehabilitation stage. For example, if the patient is overweight and in the early stage of rehabilitation (with large fluctuations), k can be 1.1~1.2; if the patient is underweight and in the later stage of rehabilitation (with smaller fluctuations), k can be 0.8~0.9.

[0103] k×v z As a dynamic adjustment term, it changes in real time with the floating rate of the center of gravity, so that the faster the center of gravity floats, the faster the vertical floating execution axis responds, thus solving the problem of lag in the response of the vertical floating execution axis.

[0104] c is the base response speed (default 0.05m / s), which ensures the minimum following accuracy of the vertical floating axis and avoids stuttering.

[0105] Formula (1) can be used to feedforward adjust the response speed of the vertical floating actuator based on the predicted center of gravity floating rate in the center of gravity prediction information.

[0106] By using center of gravity floating rate feedback and model prediction, dynamic matching between the vertical floating actuator response speed and the center of gravity floating rate is achieved, breaking the limitations of the existing fixed response speed. By using a quantization formula to link the center of gravity floating rate and the actuator response speed, a closed-loop control of prediction-follow-feedback is realized.

[0107] For example, the center of gravity floats at a small rate (v z ≤0.03m / s), such as in the later stages of recovery, the fluctuation is gradual, v axis =0.9×0.03+0.05=0.077m / s. Adjust the response speed of the vertical floating actuator to 0.077m / s, and slowly float with the center of gravity to avoid pulling on the human body, while maintaining constant force stability without jamming.

[0108] For example, the center of gravity floats at a large rate (vz>0.03m / s), such as in the early stages of rehabilitation, where the floating is intense.

[0109] v axis =1.1×0.04+0.05=0.094m / s. Increasing the response speed of the vertical floating actuator to 0.094m / s accelerates the following rhythm, counteracts the tension fluctuations caused by the center of gravity floating, and, in conjunction with damping adjustment, prevents the actuator from jamming.

[0110] For example, in extreme cases (such as when the patient is overweight and their center of gravity fluctuates wildly), if (v z If the velocity is greater than 0.05 m / s, then k is taken as 1.2, and v axis =1.2×0.05+0.05=0.11m / s, the vertical floating actuator responds quickly, avoiding tension fluctuations caused by the floating center of gravity, and at the same time, the k value is adjusted through online self-learning to ensure adaptability.

[0111] The formula based on the second correlation between the change in center of gravity and the lower limb joint angles is as follows:

[0112] θ = arcsin(Δh / L) (2;

[0113] Δh is the change in center of gravity, L is the thigh length, and θ is the lower limb joint angle.

[0114] Formula (2) can be used to feedforward and adjust the lower limb joint angles based on the predicted change in the center of gravity in the center of gravity prediction information, so as to ensure that the trunk floating and the lower limb movement do not conflict and avoid interference with the gait phase due to vertical floating.

[0115] By decoupling the horizontal gait of human walking from the vertical up-and-down floating motion, an independent vertical floating execution axis is established to specifically match the natural undulation rhythm of the torso. Based on the predicted center of gravity floating rate, the response speed of the vertical floating execution axis is dynamically adjusted to achieve trend prediction and early following, eliminate mechanical following lag, and prevent the problems of stuttering and uneven force caused by torso floating.

[0116] According to an embodiment of the present invention, the damping coefficient of the feedforward variable damping adjustment unit is adjusted based on the predicted gait phase information and center of gravity prediction information in the gait prediction information, including: when the predicted gait phase information is a support phase and the predicted center of gravity fluctuation trajectory corresponding to the support shows a downward trend in the center of gravity, the damping coefficient is increased by a first preset ratio, wherein the center of gravity prediction information includes the predicted center of gravity fluctuation trajectory; when the gait phase information is a swing phase and the predicted inertial acceleration at multiple moments corresponding to the swing shows a decreasing trend, the damping coefficient is decreased by a second preset ratio, wherein the center of gravity prediction information includes the predicted inertial acceleration.

[0117] The first preset ratio can be 15%-20%. For example, in the support phase, and when the predicted center of gravity fluctuation trajectory corresponding to the support phase shows a downward trend in the center of gravity (torso sinking, large impact), the damping coefficient can be increased by 15%-20% 100ms in advance to buffer the sinking impact.

[0118] The second preset ratio can be 10%-15%. For example, in the swing phase, if the predicted inertial acceleration at multiple moments corresponding to the swing phase shows a decreasing trend (the torso floats and the impact is smaller), the damping coefficient can be reduced by 10%-15% in advance to avoid excessive pulling.

[0119] According to an embodiment of the present invention, after adjusting the weight reduction control parameters of the weight reduction device and the lower limb joint angle of the lower limb robot according to the gait prediction information and the center of gravity prediction information respectively, the method further includes: when the real-time suspension tension obtained by the tension sensor is not within the preset tension range, adjusting the extension length of the vertical floating actuator via motor feedback, wherein the tension sensor is disposed between the pelvic support sling and the suspension bracket.

[0120] The preset tension range can be ±5 of the base tension. For example, if the base tension is 300 Newtons, the preset tension range could be 295 to 305 Newtons. The real-time suspension tension acquired by the tension sensor is 308 Newtons, exceeding the upper limit of the preset tension range of 305 Newtons. By adjusting the extension and retraction length of the vertical floating actuator shaft (fine-tuning the suspension stroke) via motor feedback, the tension is brought back to 303 Newtons (still within the preset tension range). Combined with the next state period predicted by the LSTM, the damping is adjusted to the appropriate value in advance to avoid exceeding the range again.

[0121] If the predicted gait phase information is the support phase, and the predicted center of gravity fluctuation trajectory corresponding to the support phase shows a downward trend, the damping coefficient can be increased first according to the first preset ratio, and the damping coefficient of the feedforward adjustable variable damping unit can be used. If the real-time suspension tension obtained by the tension sensor is not within the preset tension range, the suspension tension is immediately increased to about 300 Newtons via the servo motor. For example, if the patient weighs 60 kg and has lost 50% of their weight, the torso sinks during the support phase, and the tension sensor detects a tension of 292 Newtons (below the lower limit of the preset tension range), the tension is increased to 298 Newtons to maintain constant force stability.

[0122] When the predicted gait phase information is the swing phase, and the predicted inertial acceleration at multiple moments corresponding to the swing phase shows a decreasing trend (torso rises, impact is smaller), shorten the suspension travel by 1-2 cm and reduce the suspension tension to 290-295 Newtons (lower than the base tension of 300 Newtons) to avoid pulling on the body during the rise. For example, during the swing phase, if the torso rises by 2 cm, fine-tune the suspension tension from 300 Newtons to 292 Newtons to ensure smooth following without any jerking.

[0123] According to an embodiment of the present invention, determining the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on gait prediction information, center of gravity prediction information, and the respective compensation times of the treadmill, the lower limb robot, and the weight reduction device includes: determining the running speed of the treadmill based on the predicted gait cycle and predicted cadence in the gait prediction information, and determining the speed control command time based on the compensation time and running speed of the treadmill; determining the joint movements of the lower limb robot based on the predicted gait cycle and the predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and determining the joint control command time based on the compensation time and joint movements of the lower limb robot; and determining the extension and retraction movements of the vertical floating actuation axis based on the predicted center of gravity floating rate and predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and determining the weight reduction control command time based on the compensation time and extension and retraction movements of the vertical floating actuation axis of the weight reduction device.

[0124] Centered on a unified temporal benchmark for predicted gait cycles, this approach binds treadmill speed to gait cycles and center of gravity fluctuation depth, rather than operating them independently. The predicted gait cycles and cadences output by the prediction model are translated into speed control commands for the treadmill. For example, if the predicted gait cycle T = 4 seconds (1.5 seconds for the stance phase and 2.5 seconds for the swing phase), the treadmill speed = stride length × cadence, and is synchronized with the gait cycle. Specifically, during the stance phase (center of gravity lowering), the treadmill speed is slightly reduced by 0.1 m / s to avoid strain; during the swing phase (center of gravity rising), the treadmill speed is synchronously increased by 0.05 m / s to ensure a perfect match between the patient's stride rhythm and the treadmill speed.

[0125] Using a unified time-series reference for predicted gait cycles as the core, the joint movements of the lower limb robot are bound to the predicted gait cycles and predicted center of gravity fluctuation trajectories, rather than operating independently. For example, during the support phase (t∈[0,1.5 seconds]), the robot's joint movements are locked, coordinating with the sinking of the center of gravity to provide stable support; during the swing phase (t∈[1.5 seconds,4 seconds]), the joint movements of the lower limb robot flex and extend synchronously with the rising center of gravity, linked with the treadmill speed and center of gravity height, avoiding movement lag.

[0126] Centered on the predicted center of gravity fluctuation trajectory based on a unified timing reference, the extension and retraction of the vertical floating actuator are linked to the predicted center of gravity fluctuation trajectory and the predicted center of gravity floating rate, rather than operating independently. For example, when the predicted center of gravity fluctuation trajectory is in the center of gravity sinking phase, the extension and retraction of the vertical floating actuator are adjusted synchronously to adjust the suspension stroke, precisely corresponding to the time nodes of the unified timing reference. When t∈[0,1.5 seconds] (support phase), the suspension maintains a fixed stroke, cooperating with the center of gravity sinking to achieve stable tension; when t∈[1.5 seconds,4 seconds] (oscillation phase), the suspension stroke is adjusted synchronously with the center of gravity rising, ensuring complete alignment with gait and center of gravity timing.

[0127] The speed control command time, joint control command time, and weight reduction control command time are the times when the controller issues the speed control command, joint control command (used to control the joint movements of the lower limb robot), and extension and retraction control command, respectively.

[0128] Taking the weight reduction control command time as an example, the duration of a single gait cycle within the predicted gait cycle is used as the base cycle and the core time unit for the clock reference. Key timing nodes within the gait cycle are marked, including the start and end times of the support phase and the swing phase, serving as critical scales for the clock reference. Timing data of the predicted center of gravity fluctuation trajectory is bound, including the time nodes corresponding to the predicted center of gravity change and the predicted center of gravity floating rate, achieving a linked clock scale for "gait phase + center of gravity floating". Integrating the above cycles, nodes, and related data forms a quantifiable and synchronizeable clock reference, providing a unified time scale for the timing linkage of the treadmill, lower limb robot, and weight reduction device, ensuring precise alignment of their movements with the gait phase.

[0129] The current timestamp is 8:00:00.000 (i.e., t=0 of the unified timing reference, corresponding to the start of the support phase and the imminent sinking of the torso). The historical delay times of the telescopic motion control command in the past three similar scenarios were 220 milliseconds, 200 milliseconds, and 230 milliseconds, respectively, with an average delay of 220 milliseconds. Using this average delay time as the compensation time, the control command is issued 220 milliseconds earlier, i.e., execution starts at 7:59:59.780, ensuring the device completes its action response by 8:00:00.000, precisely aligned with the unified timing reference. Data acquisition is strictly performed according to the actual execution time after compensation (starting from 7:59:59.780), synchronously recording data such as center of gravity height and suspension tension. This data is used both to verify the compensation effect and as input parameters for the LSTM model's online self-learning, dynamically optimizing subsequent timing synchronization accuracy.

[0130] Meanwhile, the treadmill and lower limb robot also operate in conjunction with this unified timing reference. The treadmill maps the gait cycle and step frequency in the reference into speed adjustment commands, which are synchronized with the gait rhythm. The lower limb robot binds the gait phase switching time in the reference to the joint movement, realizing the synchronous movement of the support phase and the swing phase. The three work together to form a three-axis control system with unified timing, which completely solves the problem of timing disconnection in traditional control.

[0131] Figure 3 A schematic diagram of a gait training control system according to an embodiment of the present invention is shown.

[0132] like Figure 3 As shown, the gait training control system may include a treadmill 110, a lower limb robot 120, a weight reduction device 130, and a controller 140.

[0133] Treadmill 110 is used to allow patients to simulate walking exercises.

[0134] According to an embodiment of the present invention, the treadmill 110 may be equipped with a speed servo adjustment unit and a gait synchronization communication interface, the gait synchronization communication interface being signal-connected to the controller 140; the treadmill abandons the fixed-speed operation mode and can achieve micro-amplitude dynamic speed adjustment based on the predicted gait phase information and predicted center of gravity fluctuation trajectory output by the controller, accurately matching the walking rhythm of the human body's gait support phase and swing phase, and assisting in three-axis spatiotemporal integrated synchronous control.

[0135] The lower limb robot 120 is worn on the patient's lower limbs to correct the lower limb joint angles of the patient's abnormal walking posture.

[0136] According to an embodiment of the present invention, the lower limb robot 120 can adopt a lower limb exoskeleton structure, with servo motors and angle sensors mounted at the hip, knee, and ankle joints, and foot pressure sensors mounted at the soles of the feet; the lower limb robot is connected to the controller and can correct the movement trajectory of the hip, knee, and ankle joints in real time according to the predicted gait information output by the controller, adapting to the lower limb posture deviation caused by trunk floating, and supporting independent control of the left and right joint trajectories to adapt to the asymmetrical gait difference compensation requirements.

[0137] The weight reduction device 130 is mounted above the treadmill and worn on the patient's pelvic area to reduce the stress on the lower limbs during walking.

[0138] According to an embodiment of the present invention, the weight reduction device 130 may include: a suspension bracket mounted above the treadmill; a vertically floating axle fitted inside the suspension bracket and extending and retracting vertically along the suspension bracket; a pelvic support strap suspended from the lower end of the vertically floating axle; a motor (such as a servo motor) mounted on the suspension bracket for driving the vertically floating axle to extend and retract vertically to adjust the position of the pelvic support strap in the vertical direction; and a variable damping adjustment unit connected to the vertically floating axle for matching the patient's trunk floating motion inertia during walking.

[0139] The controller 140 is connected to the treadmill, the lower limb robot, and the weight reduction device to execute the control method for gait training.

[0140] According to an embodiment of the present invention, the controller 140 can serve as the core control unit of the entire system. It has a built-in gait-center of gravity dual-dimensional advanced prediction model (which may be an LSTM or GRU deep learning model), a phase-separated constant force control unit, a three-axis synchronous coordination unit, and an asymmetric gait compensation unit. It is responsible for receiving real-time data from multiple types of sensors, completing dual-dimensional prediction calculations, outputting precise control commands, and realizing the coordinated linkage and precise regulation of each module.

[0141] For example, various types of sensors may include trunk IMU posture sensors, joint angle sensors, plantar pressure sensors, and tension sensors. Tension sensors are positioned between the pelvic support sling and the suspension bracket to collect suspension tension signals in real time and feed them back to the controller. All sensors are signal-connected to the controller to collect multi-dimensional data such as gait phase, trunk center of gravity height, joint angles, plantar force, and suspension tension in real time. The collected data is transmitted to the controller in real time, providing precise data support for gait-center of gravity dual-dimensional advanced prediction, phase-separated constant force control, and three-axis synchronous coordination.

[0142] In lower limb weight-loss rehabilitation training, weight-loss devices, treadmills, and lower limb robots are the core equipment. Weight-loss devices offset part of the patient's weight, reducing the load on the lower limbs. Combined with treadmills and lower limb robots, standardized gait training is achieved, which is the core means of rehabilitation for patients with lower limb dysfunction.

[0143] Figure 4A A flowchart illustrating a gait training control method according to another embodiment of the present invention is shown.

[0144] like Figure 4A As shown, another embodiment of the gait training control method of the present invention includes operations S410~S470.

[0145] According to an embodiment of the present invention, "gait prediction" and "center of gravity prediction" are used as dual-dimensional inputs to construct a "feedforward + feedback" dual closed-loop control architecture.

[0146] Data acquisition is performed while operating the S410.

[0147] For example, gait data can be collected using IMU posture sensors, joint angle sensors, plantar pressure sensors, and tension sensors.

[0148] When operating the S420, gait-center of gravity dual-dimensional multi-step advance prediction is performed.

[0149] For example, by using a deep learning model (LSTM / GRU) to synchronously predict the gait phase switching time and center of gravity fluctuation trajectory for the next 1-3 gait cycles, a time window can be provided for feedforward control.

[0150] For example, a personalized gait self-learning mechanism is introduced to dynamically correct prediction parameters and adapt to the gait characteristics of patients with different conditions and at different stages of rehabilitation. To address the problem of limited training data samples in the early stages of rehabilitation, a "transfer learning + online fine-tuning" strategy is adopted: the model is pre-trained offline using a gait database of healthy individuals, and then fine-tuned online using real-time collected individual patient data. This reduces the data requirements in the cold start phase and ensures the prediction accuracy of the model in the early stages of training.

[0151] When operating S430, phase-adaptive constant force control is used.

[0152] For example, to distinguish between the swing phase (trend of the torso rising) and the support phase (impact of the torso sinking), different weight reduction compensation strategies are implemented to achieve a constant force fluctuation error of ≤±1.5%. Vertical torso floating decoupling control can decouple horizontal gait movement from vertical floating control, establish an independent vertical floating execution axis, specifically match the natural undulation rhythm of the torso, and achieve "trend prediction and early following".

[0153] Phase-adaptive constant force dual-feedback closed-loop control is mainly used for weight-reducing suspension equipment to achieve precise constant force control of individual devices (by using feedforward prediction and feedback correction to ensure that the weight-reducing tension fluctuation error is ≤±1.5%), which is the basic prerequisite for achieving three-axis coordination. Only by ensuring the control accuracy and motion stability of individual devices (especially weight-reducing devices) through this closed-loop control can they be integrated with treadmills and lower limb robots into a unified timing reference to achieve three-axis spatiotemporal integrated control. Its core function is to achieve integrated linkage of the weight-reduction device, lower limb robot, and treadmill in time (time synchronization) and space (motor coordination) by binding the action timing of three devices through a unified timing reference, based on the precise constant force control of a single device. This avoids the problem of precise control of a single device but disjointed timing and conflicting actions of multiple devices, ensuring the coordinated adaptation of the patient's gait, weight-reduction support, and joint assistance, further improving the safety, smoothness, and adaptability of rehabilitation training. At the same time, it enhances the effect of phase adaptive constant force control, forming a closed-loop system of "precise control of a single device - coordinated linkage of multiple devices", which is different from the defects of the existing technology where the control of a single device is disconnected from the coordination of multiple devices.

[0154] When operating the S440, three-axis spatiotemporal integrated synchronous and coordinated control is used.

[0155] For example, a unified gait phase clock reference is established, integrating the treadmill, lower limb robot, and weight-reduction device into the same timing control system. The treadmill dynamically adjusts its speed based on the gait phase, slightly reducing speed during the support phase to stabilize the gait and increasing speed by matching the stride length during the swing phase. The lower limb robot corrects its joint trajectories in real time to compensate for posture deviations caused by torso floating. The weight-reduction device synchronously completes constant force adjustment and vertical floating follow, achieving interference-free and fully synchronized movement of multiple devices. To achieve high-precision synchronization, a real-time communication protocol is adopted, and a timestamp compensation mechanism is designed. When sending commands, the controller compensates for the timestamp offset of the commands based on the historical delay time of each device, keeping the synchronization error of multiple devices within ≤5 milliseconds to ensure interference-free movement and fully synchronized timing of multiple devices.

[0156] Real-time feedback and model parameter correction during S450 operation.

[0157] In operation S460, determine whether the model training has ended. If yes, execute operation S470; otherwise, execute operation S410.

[0158] When operating S470, the system terminates.

[0159] Figure 4B A flowchart illustrating the three-axis spatiotemporal integrated synchronous collaborative control according to an embodiment of the present invention is shown.

[0160] like Figure 4B As shown, the three-axis spatiotemporal integrated synchronous collaborative control includes operations S441~S445.

[0161] In operation of S441, gait phase clock reference.

[0162] S442 is used to correct the trajectory of the lower limb robot.

[0163] When operating S443, the treadmill's dynamic speed is adjusted.

[0164] When operating S444, the weight reduction device is synchronously controlled.

[0165] When operating S445, real-time status feedback is used for closed-loop adjustment.

[0166] Figure 4C A flowchart illustrating phase-separated adaptive constant force control according to an embodiment of the present invention is shown.

[0167] like Figure 4C As shown, the phase-separated adaptive constant force control includes operations S431~S435.

[0168] In operation S431, gait phase prediction and center of gravity prediction.

[0169] When operating S432, support phase control (anticipate sinking, stabilize and buffer).

[0170] When operating S433, the oscillation phase control (predicting buoyancy and pre-compensating for tension) is used.

[0171] The S434 is operated using a dual closed-loop constant force control system with feedforward and feedback.

[0172] When operating the S435, the tension sensor provides real-time feedback.

[0173] Figure 5 A flowchart illustrating asymmetric gait differentiation compensation control according to an embodiment of the present invention is shown.

[0174] like Figure 5 As shown, the asymmetric gait differentiation compensation control includes operations S510~S560.

[0175] In operating the S510, gait phase prediction and center of gravity prediction are performed.

[0176] Asymmetric gait recognition and determination during S520 operation.

[0177] When operating S530, the left (affected side) independent compensation control trajectory correction and tension adjustment are performed.

[0178] When operating the S540, the right side (healthy side) has independent compensation control correction and tension adjustment.

[0179] When operating the S550, the lower limb robot and the weight reduction device work together for control.

[0180] When operating the S560, real-time feedback and dynamic correction are provided.

[0181] According to embodiments of the present invention, for patients with hemiplegia or unilateral lower limb injury, differentiated control of gait, center of gravity fluctuation, and weight-reduction pulling force on both sides is achieved: the robot's assist torque is increased on the affected side, and the assist torque is decreased on the healthy side; the weight-reduction pulling force is increased by 5% to 10% on the affected side to compensate for insufficient muscle strength on the affected side. Through the above differentiated compensation, the robot trajectory and local weight-reduction force of the unilateral lower limb are individually corrected, pelvic tilt and gait asymmetry problems are corrected, and the pertinence and effectiveness of rehabilitation training are improved.

[0182] Figure 6 A flowchart illustrating a prediction model training method according to an embodiment of the present invention is shown.

[0183] like Figure 6 As shown, the prediction model training method includes operations S610~S650.

[0184] When operating the S610, historical gait data is retrieved from the historical gait database.

[0185] In operation of S620, gait feature extraction and pre-classification recognition are performed.

[0186] When operating the S630, the prediction model undergoes self-learning iteration.

[0187] The S640 is operated using phase-separated adaptive constant force control and three-axis spatiotemporal integrated synchronous and coordinated control.

[0188] During the operation of S650, the control effect is evaluated and deviation feedback is provided.

[0189] Figure 7 A schematic block diagram of a gait training control device according to an embodiment of the present invention is shown.

[0190] like Figure 7 As shown, the gait training control device 700 of this embodiment includes an input module 710, a feedforward adjustment module 720, a determination module 730, and a control module 740.

[0191] The input module 710 is used to input the patient's current gait data while walking on the treadmill into the prediction model and output gait prediction information and center of gravity prediction information. The patient wears a lower limb robot on their lower limbs and a weight-reducing device on their pelvis. The weight-reducing device reduces the force on the patient's lower limbs during walking, and the lower limb robot corrects abnormal gait postures of the patient's lower limb joints to facilitate gait training in conjunction with the treadmill. In one embodiment, the input module 710 can be used to perform the operation S210 described above, which will not be repeated here.

[0192] The feedforward adjustment module 720 is used to adjust the weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot based on gait prediction information and center of gravity prediction information, respectively. In one embodiment, the feedforward adjustment module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0193] The determining module 730 is used to determine the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on gait prediction information, center of gravity prediction information, and the compensation time of each of the treadmill, lower limb robot, and weight reduction device. The compensation time is determined based on the historical delay time of each of the treadmill, lower limb robot, and weight reduction device. In one embodiment, the determining module 730 can be used to perform the operation S230 described above, which will not be repeated here.

[0194] The control module 740 is used to control the treadmill, lower limb robot, and weight reduction device respectively using speed control command timing, joint control command timing, and weight reduction control command timing, so that the treadmill, lower limb robot, and weight reduction device synchronously execute control commands that match the gait prediction information. In one embodiment, the control module 740 can be used to execute the operation S240 described above, which will not be repeated here.

[0195] According to an embodiment of the present invention, the input module 710 includes a feature extraction submodule, a feature fusion submodule, and a decoding submodule. The feature extraction submodule is used to extract features from the current gait data using a two-layer joint coding layer of the prediction model to obtain gait local features and gait global features. The current gait data includes joint motion parameters, gait cycle, plantar pressure, and center of gravity fluctuation trajectory. Gait local features characterize the correlation between joint motion parameters, plantar pressure, and center of gravity fluctuation trajectory, respectively. Gait global features are the temporal linkage features of gait cycle and center of gravity fluctuation trajectory. Plantar pressure is collected by a plantar pressure sensor installed on the lower limb robot. The feature fusion submodule is used to fuse the gait local features and gait global features using a network sharing layer of the prediction model to obtain fused features. The decoding submodule is used to decode the fused features using a two-branch decoding layer of the prediction model to obtain gait prediction information and center of gravity prediction information.

[0196] According to an embodiment of the present invention, the decoding submodule includes a first decoding unit and a second decoding unit. The first decoding unit is used to perform a first decoding on the fused features using the gait decoding branch in the dual-branch decoding layer to obtain gait prediction information; the second decoding unit is used to perform a second decoding on the fused features using the centroid decoding branch in the dual-branch decoding layer to obtain centroid prediction information.

[0197] According to an embodiment of the present invention, the weight reduction device includes: a suspension bracket mounted above a treadmill; a vertically floating axle fitted inside the suspension bracket and extending and retracting vertically along the suspension bracket; a pelvic support strap suspended from the lower end of the vertically floating axle; a motor mounted on the suspension bracket for driving the vertically floating axle to extend and retract vertically to adjust the position of the pelvic support strap in the vertical direction; and a variable damping adjustment unit connected to the vertically floating axle for matching the patient's trunk floating motion inertia during walking.

[0198] According to an embodiment of the present invention, the weight reduction control parameters include the damping coefficient of the variable damping adjustment unit and the response speed of the vertical floating actuation axis; the feedforward adjustment module 720 includes a first feedforward adjustment submodule, a second feedforward adjustment submodule, and a third feedforward adjustment submodule. The first feedforward adjustment submodule is used to adjust the damping coefficient of the variable damping adjustment unit based on the gait phase information and center of gravity prediction information in the gait prediction information; the second feedforward adjustment submodule is used to adjust the response speed of the vertical floating actuation axis based on the predicted center of gravity floating rate in the center of gravity prediction information, based on a first correlation between the response speed and the center of gravity floating rate; the third feedforward adjustment submodule is used to adjust the lower limb joint angle based on the predicted center of gravity change in the center of gravity prediction information, based on a second correlation between the center of gravity change and the lower limb joint angle.

[0199] According to an embodiment of the present invention, the first feedforward adjustment submodule includes an amplification unit and a depletion unit. The amplification unit is used to increase the damping coefficient by a first preset ratio when the predicted gait phase information is a support phase and the predicted center of gravity fluctuation trajectory corresponding to the support shows a downward trend in the center of gravity, wherein the center of gravity prediction information includes the predicted center of gravity fluctuation trajectory; the depletion unit is used to decrease the damping coefficient by a second preset ratio when the gait phase information is a swing phase and the predicted inertial acceleration at multiple moments corresponding to the swing shows a decreasing trend, wherein the center of gravity prediction information includes the predicted inertial acceleration.

[0200] According to an embodiment of the present invention, after the feedforward adjustment module 720, the device further includes a feedback adjustment module. The feedback adjustment module is used to adjust the extension and retraction length of the vertical floating actuator via motor feedback when the real-time suspension tension acquired by the tension sensor is not within a preset tension range, wherein the tension sensor is disposed between the pelvic support sling and the suspension bracket.

[0201] According to an embodiment of the present invention, the determining module 730 includes a first determining submodule, a second determining submodule, and a third determining submodule. The first determining submodule is used to determine the running speed of the treadmill based on the predicted gait cycle and predicted step frequency in the gait prediction information, and to determine the speed control command time based on the compensation time and running speed of the treadmill. The second determining submodule is used to determine the joint movements of the lower limb robot based on the predicted gait cycle and the predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and to determine the joint control command time based on the compensation time and joint movements of the lower limb robot. The third determining submodule is used to determine the extension and retraction movements of the vertical floating actuation axis based on the predicted center of gravity floating rate and predicted center of gravity fluctuation trajectory in the center of gravity prediction information, and to determine the weight reduction control command time based on the compensation time and extension and retraction movements of the weight reduction device and the vertical floating actuation axis.

[0202] According to embodiments of the present invention, any plurality of modules among the input module 710, feedforward adjustment module 720, determination module 730, and control module 740 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules can be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the input module 710, feedforward adjustment module 720, determination module 730, and control module 740 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the input module 710, feedforward adjustment module 720, determination module 730, and control module 740 can be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0203] Figure 8 A block diagram of an electronic device suitable for implementing a control method for gait training according to an embodiment of the present invention is shown schematically.

[0204] like Figure 8 As shown, an electronic device 800 according to an embodiment of the present invention includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0205] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 executes various operations of the method flow according to embodiments of the present invention by executing programs in ROM 802 and / or RAM 803. It should be noted that programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in one or more memories.

[0206] According to an embodiment of the present invention, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0207] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0208] According to embodiments of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803 described above.

[0209] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to enable the computer system to implement the gait training control method provided in the embodiments of the present invention.

[0210] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this invention. According to embodiments of the invention, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0211] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0212] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of this embodiment of the invention. According to embodiments of the invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0213] According to embodiments of the present invention, program code for executing the computer programs provided in the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0214] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0215] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0216] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A control method for gait training, characterized in that, The method includes: The patient's current gait data while walking on the treadmill is input into the prediction model, which outputs gait prediction information and center of gravity prediction information. The patient wears a lower limb robot on his / her lower limbs and a weight reduction device on his / her pelvis. The weight reduction device is used to reduce the force on the patient's lower limbs during walking. The lower limb robot is used to correct the lower limb joint angles of the patient's abnormal walking posture in order to cooperate with the treadmill for gait training. The weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot are adjusted forward according to the gait prediction information and the center of gravity prediction information, respectively. Based on the gait prediction information, the center of gravity prediction information, and the compensation time of the treadmill, the lower limb robot, and the weight reduction device, the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device are determined. The compensation time is determined based on the historical delay time of the treadmill, the lower limb robot, and the weight reduction device. The treadmill, the lower limb robot, and the weight reduction device are controlled by the speed control command time, the joint control command time, and the weight reduction control command time, respectively, so that the treadmill, the lower limb robot, and the weight reduction device synchronously execute control commands that match the gait prediction information.

2. The method according to claim 1, characterized in that, The process of inputting the patient's current gait data while walking on the treadmill into the prediction model and outputting gait prediction information and center of gravity prediction information includes: The current gait data is feature extracted using a two-layer joint coding layer of the prediction model to obtain gait local features and gait global features. The current gait data includes joint motion parameters, gait cycle, plantar pressure, and center of gravity fluctuation trajectory. The gait local features characterize the correlation between the joint motion parameters, the plantar pressure, and the center of gravity fluctuation trajectory, respectively. The gait global features are the temporal linkage features between the gait cycle and the center of gravity fluctuation trajectory. The plantar pressure is collected by a plantar pressure sensor installed on the lower limb robot. By using the network sharing layer of the prediction model, feature fusion is performed on local and global gait features to obtain fused features; The fused features are decoded using a two-branch decoding layer of the prediction model to obtain the gait prediction information and the center of gravity prediction information.

3. The method according to claim 2, characterized in that, The step of decoding the fused features using a dual-branch decoding layer of the prediction model to obtain the gait prediction information and the center of gravity prediction information includes: The fusion features are first decoded using the gait decoding branch in the dual-branch decoding layer to obtain the gait prediction information. The centroid decoding branch in the dual-branch decoding layer is used to perform a second decoding on the fused features to obtain the centroid prediction information.

4. The method according to claim 1, characterized in that, The weight reduction device includes: A suspension bracket is installed above the treadmill; A vertically floating actuator is mounted on the inner side of the suspension bracket and moves vertically along the suspension bracket; A pelvic support sling is suspended from the lower end of the vertical floating actuator shaft; The motor, mounted on the suspension bracket, is used to drive the vertical floating actuator to perform telescopic movements in the vertical direction, so as to adjust the position of the pelvic support sling in the vertical direction; A variable damping adjustment unit, connected to the vertical floating actuation axis, is used to match the motion inertia of the patient's torso floating during walking.

5. The method according to claim 4, characterized in that, The weight reduction control parameters include the damping coefficient of the variable damping adjustment unit and the response speed of the vertical floating actuator. The step of adjusting the weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot based on the gait prediction information and the center of gravity prediction information includes: Based on the gait phase information and the center of gravity prediction information in the gait prediction information, the damping coefficient of the variable damping adjustment unit is adjusted forward. Based on the first correlation between response speed and center of gravity floating rate, the response speed of the vertical floating axle is adjusted forward according to the predicted center of gravity floating rate in the center of gravity prediction information. Based on the second correlation between the change in center of gravity and the lower limb joint angle, the lower limb joint angle is adjusted feedforward according to the predicted change in center of gravity in the predicted center of gravity information.

6. The method according to claim 5, characterized in that, The step of adjusting the damping coefficient of the variable damping adjustment unit based on the predicted gait phase information and the predicted center of gravity information in the gait prediction information includes: When the predicted gait phase information is a support phase and the predicted center of gravity fluctuation trajectory corresponding to the support shows a downward trend of the center of gravity, the damping coefficient is increased by a first preset ratio, wherein the center of gravity prediction information includes the predicted center of gravity fluctuation trajectory. When the gait phase information is a swing phase and the predicted inertial acceleration at multiple moments corresponding to the swing shows a decreasing trend, the damping coefficient is reduced according to a second preset ratio, wherein the center of gravity prediction information includes the predicted inertial acceleration.

7. The method according to claim 5, characterized in that, After adjusting the weight reduction control parameters of the weight reduction device and the lower limb joint angles of the lower limb robot based on the gait prediction information and the center of gravity prediction information respectively, the method further includes: If the real-time suspension tension obtained by the tension sensor is not within the preset tension range, the extension and retraction length of the vertical floating actuator is adjusted via feedback from the motor. The tension sensor is located between the pelvic support sling and the suspension bracket.

8. The method according to claim 4, characterized in that, The step of determining the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on the gait prediction information, the center of gravity prediction information, and the compensation time of the treadmill, the lower limb robot, and the weight reduction device includes: Based on the predicted gait cycle and predicted step frequency in the gait prediction information, the running speed of the treadmill is determined, and based on the compensation time of the treadmill and the running speed of the treadmill, the speed control command time is determined. Based on the predicted gait cycle and the predicted center of gravity fluctuation trajectory in the predicted center of gravity information, the joint movements of the lower limb robot are determined, and based on the compensation time of the lower limb robot and the joint movements of the lower limb robot, the joint control command time is determined. Based on the predicted center of gravity floating rate and predicted center of gravity fluctuation trajectory in the predicted center of gravity information, the extension and retraction action of the vertical floating actuation shaft is determined, and based on the compensation time of the weight reduction device and the extension and retraction action of the vertical floating actuation shaft, the weight reduction control command time is determined.

9. A control system for gait training, characterized in that, The system includes: Treadmills are used to simulate walking exercises for patients; A lower limb robot, worn on the patient's lower limbs, is used to correct the lower limb joint angles of the patient's abnormal walking posture; The weight-reduction device is mounted above the treadmill and worn on the patient's pelvic area to reduce the stress on the patient's lower limbs during walking. The controller is communicatively connected to the treadmill, the lower limb robot, and the weight reduction device, and is used to execute the method described in any one of claims 1 to 8.

10. A control device for gait training, characterized in that, The device includes: The input module is used to input the patient's current gait data while walking on the treadmill into the prediction model and output gait prediction information and center of gravity prediction information. The patient wears a lower limb robot on his / her lower limbs and a weight reduction device on his / her pelvis. The weight reduction device is used to reduce the force on the patient's lower limbs during walking. The lower limb robot is used to correct the lower limb joint angles of the patient's abnormal walking posture in order to cooperate with the treadmill for gait training. The feedforward adjustment module is used to adjust the weight reduction control parameters of the weight reduction device and the lower limb joint angle of the lower limb robot based on the gait prediction information and the center of gravity prediction information, respectively. The determination module is used to determine the speed control command time of the treadmill, the joint control command time of the lower limb robot, and the weight reduction control command time of the weight reduction device based on the gait prediction information, the center of gravity prediction information, and the compensation time of the treadmill, the lower limb robot, and the weight reduction device, wherein the compensation time is determined based on the historical delay time of the treadmill, the lower limb robot, and the weight reduction device. The control module is used to control the treadmill, the lower limb robot, and the weight reduction device respectively using the speed control command time, the joint control command time, and the weight reduction control command time, so that the treadmill, the lower limb robot, and the weight reduction device synchronously execute control commands that match the gait prediction information.

Images