NMS model-based front fork postoperative robot and control method

Abstract

The application provides a front fork postoperative robot based on an NMS model and a control method. The actual joint torque of a diseased leg under the same action and the joint torque distribution of a healthy leg under the current action are subjected to torque difference, so as to obtain the torque deficiency of the diseased leg. The surface electromyogram of the diseased leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, so as to obtain a diseased leg correction model. The torque deficiency of the diseased leg is input into the diseased leg correction model, so as to obtain the predicted theoretical torque of the diseased leg under the current action. Then, the compensation torque instruction of the front fork postoperative robot is generated through the torque difference between the predicted theoretical torque and the actual joint torque of the diseased leg. The compensation torque instruction is used to drive the joint module in the front fork postoperative robot to output auxiliary force, so as to realize the cooperative auxiliary training of the front fork postoperative robot. Based on the above scheme, the torque deficiency compensation of the front fork postoperative robot can be realized.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and more specifically, to a postoperative robot and control method for anterior sclerosis surgery based on the NMS model. Background Technology

[0002] The neuromuscular skeletal (NMS) model is a mathematical model that simulates how the human central nervous system recruits muscles through motor neurons to generate activation, and then converts the muscle contraction force into joint torque through the skeletal lever system.

[0003] Current ACL surgery rehabilitation training treats the affected side as an independent control object for trajectory tracking or torque assistance, ignoring the coordination mechanism of both limbs in normal human movement. Traditional rehabilitation robots only collect the current state of the affected leg as feedback input in their control strategy, driving the movement of the affected side with a preset standard trajectory or constant torque as the target, without referring to the actual force exertion pattern of the healthy leg in the same movement. The isolated unilateral training method causes the muscles of the affected side to only learn to follow external instructions rather than to complete the movement in coordination with the healthy side. The neural control system cannot establish bilateral coupled motor representation. When patients attempt to perform bilateral coordinated movements such as walking without the robot after training, the timing, amplitude, and rate of force exertion of the affected leg are severely mismatched with those of the healthy leg, manifesting as gait asymmetry, abnormal center of gravity transfer, and compensatory pelvic tilt. Even if the strength of the muscles on the affected side recovers, the temporal coordination ability and torque matching ability between the affected and healthy sides are still lacking. Patients cannot smoothly complete alternating walking movements, ultimately forming a pathological bilateral uncoordinated gait. Therefore, how to compensate for the loss of torque on the affected side after ACL surgery, thereby improving the coordination of the torque distribution of both lower limbs, has become a challenge for the industry. Summary of the Invention

[0004] This application provides a robot and control method for postoperative anterior sclerosis (AFS) surgery based on the NMS model, which can compensate for the loss of torque on the affected side of the robot after AFS surgery, thereby improving the coordination of the torque distribution of both lower limbs.

[0005] Firstly, this application provides a control method for anterior cruciate ligament (ACL) postoperative robot based on an NMS model, including: Collect surface electromyography signals and joint motion trajectories of the healthy leg during the movement, input them into the neuromuscular skeletal model in the robot after ACL surgery, and inversely solve the joint torque distribution of the healthy leg under the current movement. The torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement is obtained to determine the torque loss of the affected leg. Then, the surface electromyography signal of the affected leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model. The torque loss of the affected leg is input into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action. Then, the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg is used to generate the compensation torque command of the robot after anterior cruciate fascia surgery. The compensation torque command is used to drive the joint modules in the post-operative anterior cruciate ligament robot to output auxiliary force, thereby realizing the collaborative assisted training of the post-operative anterior cruciate ligament robot.

[0006] In some embodiments, the torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement is calculated to obtain the torque loss of the affected leg, specifically including: Align the joint torque distribution of the healthy leg with the actual joint torque of the affected leg along the time axis, and then calculate the difference between the torque value of the healthy side and the actual torque value of the affected side at each time frame; The torque loss of the affected leg was selected from all the differences.

[0007] In some embodiments, the electromyography (EMG) signal from the affected leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, thereby obtaining the affected-side corrected model, specifically including: Extract the amplitude envelope of surface electromyographic signals of each muscle in the affected leg during movement; The amplitude envelope is compared with the theoretical electromyography corresponding to the current muscle recruitment parameters in the NMS model to obtain multiple deviation ratios; The recruitment threshold and gain coefficient in the NMS model are adjusted in reverse according to the various deviation ratios. The adjusted parameter set is then written into the NMS model to generate the affected side correction model.

[0008] In some embodiments, inputting the torque loss of the affected leg into the affected leg correction model to obtain the predicted theoretical torque of the affected leg under the current action specifically includes: The torque loss of the affected leg is used as an excitation input signal and loaded into the input layer of the motor neuron of the affected side correction model; The affected-side modified model simulates the activation response of each muscle after receiving the stimulus, based on the modified muscle recruitment parameters; The activation response of each muscle is mapped to the theoretical output value in the joint space, thereby obtaining the predicted theoretical torque of the affected leg under the current movement.

[0009] In some embodiments, generating a compensating torque command for the robot after ACL surgery based on the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg specifically includes: At the same time frame, calculate the difference between the predicted theoretical torque and the actual joint torque of the affected leg; Determine the auxiliary gain coefficient of the robot after the anterior fork surgery, and perform low-pass filtering on all real-time torque differences; Based on the filtered torque difference and the auxiliary gain coefficient, the auxiliary torque of the robot after the forklift surgery is limited and compensated to obtain the compensation torque command of the robot after the forklift surgery.

[0010] In some embodiments, using the compensation torque command to drive the joint module in the post-operative anterior cruciate sclerosis robot to output an auxiliary force specifically includes: The compensation torque command is sent to the joint servo driver of the robot after the fork surgery. The driver converts the torque command into a corresponding current command and drives the motor to output the corresponding auxiliary torque. The actual output torque is monitored by the torque sensor built into the joint, and the actual output torque and the compensation torque command are combined to form a closed-loop feedback adjustment to obtain the auxiliary torque command for each joint. After ACL surgery, the robot's joints coordinate to perform actions according to the corresponding auxiliary torque commands, applying auxiliary force to the corresponding segments of the affected leg.

[0011] In some embodiments, the post-anterior cruciate ligament (ACL) robot is a knee joint rehabilitation training robot based on a lower limb exoskeleton.

[0012] Secondly, this application provides a postoperative robot for anterior sclerosis surgery based on the NMS model, including a control unit, the control unit comprising: The acquisition module is used to acquire surface electromyography signals and joint motion trajectories of the healthy leg during the movement, input them into the neuromuscular model in the robot after ACL surgery, and inversely solve the joint torque distribution of the healthy leg under the current movement. The processing module is used to perform torque difference analysis between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement, to obtain the torque loss of the affected leg, and to use the surface electromyography signal of the affected leg to perform electromechanical correction on the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model. The processing module is also used to input the torque loss of the affected leg into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action, and then generate the compensation torque command of the robot after anterior cruciate fasciculations by the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg. The execution module is used to drive the joint modules in the post-operative anterior cruciate fascia robot to output auxiliary force using the compensation torque command, thereby realizing the collaborative assisted training of the post-operative anterior cruciate fascia robot.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device executes the above-described control method for anterior sclerosis postoperative robot based on the NMS model.

[0014] Fourthly, this application provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the above-described control method for anterior cruciate fasciculations (ACL) robot based on the NMS model.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: This application provides a post-anterior ACL surgery robot and its control method based on the NMS model. The method involves collecting surface electromyography (EMG) signals and joint motion trajectories of the healthy leg during movement, inputting them into the neuromuscular skeletal model of the post-anterior ACL surgery robot, and resolving the joint torque distribution of the healthy leg under the current movement. The method then performs torque difference analysis between the actual joint torque of the affected leg under the same movement and the joint torque distribution of the healthy leg to obtain the torque loss of the affected leg. The EMG signals of the affected leg are then used to electromechanically correct the muscle recruitment parameters in the NMS model, resulting in a corrected model for the affected leg. The torque loss of the affected leg is input into the corrected model to obtain the predicted theoretical torque of the affected leg under the current movement. The torque difference between the predicted theoretical torque and the actual joint torque of the affected leg is used to generate a compensation torque command for the post-anterior ACL surgery robot. This compensation torque command is then used to drive the joint modules in the post-anterior ACL surgery robot to output auxiliary forces, achieving collaborative assisted training of the post-anterior ACL surgery robot.

[0016] Therefore, in this application, the compensation torque command is used to drive the joint modules in the ACL surgery robot to output auxiliary force, thereby achieving collaborative auxiliary training of the ACL surgery robot. First, by determining the affected side correction model, a personalized neural drive mapping reflecting the true recruitment ability of the affected side muscles can be obtained. After ACL surgery, the affected side muscles of patients are often in a state of neural inhibition, with an increased recruitment threshold and a decreased gain coefficient. The general NMS model cannot characterize this pathological excitation-contraction coupling change. The basis for robot-assisted training is changed from external preset standards to the real-time state estimation of the patient's own nervous system, avoiding the overestimation or underestimation of the affected side's ability by the general model, and establishing a model-level foundation for achieving accurate physiological matching torque compensation. Then, by determining the compensation torque command, a double-layer closed auxiliary system integrating physiological prediction and real-time sensing can be obtained. The robot's output assist force fills the torque gap between the active force output of the affected side and the physiological baseline of the healthy side. The torque difference between the predicted theoretical torque and the actual joint torque separates the active loss due to neural inhibition on the affected side and the passive deviation due to other factors. After removing high-frequency noise through low-pass filtering, it is multiplied by the assist gain coefficient and then subjected to amplitude limiting to generate a compensation torque command. When the robot executes the compensation torque command, it is equivalent to bridging the gap between the existing recruitment capacity of the affected side and the physiological baseline of the healthy side at the neural drive level. This allows the actual torque on the affected side to continuously approach the torque distribution of the healthy side under closed-loop regulation, thereby reconstructing the synergy of the bilateral lower limb torque spectrum in both the time domain and amplitude dimensions. In summary, based on the above scheme, the robot can achieve torque loss compensation on the affected side after ACL surgery, thereby improving the synergy of the bilateral lower limb motion torque distribution. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an exemplary flowchart of a postoperative robot control method for anterior fork surgery based on an NMS model, according to some embodiments of this application; Figure 2 This is a flowchart illustrating the process of determining the compensation torque command according to some embodiments of this application; Figure 3 This is a schematic diagram of the structure of a control unit according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a computer device for implementing a control method for an anterior sclerosis postoperative robot based on an NMS model, according to some embodiments of this application. Detailed Implementation

[0019] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] refer to Figure 1 The figure is an exemplary flowchart of a control method for an anterior scaffold robot based on an NMS model according to some embodiments of this application. The control method for an anterior scaffold robot based on an NMS model mainly includes the following steps: In step 101, surface electromyography signals and joint motion trajectories of the healthy leg during the movement are collected and input into the neuromuscular skeletal model in the robot after ACL surgery to inversely solve the joint torque distribution of the healthy leg under the current movement.

[0021] It should be noted that, in this application, surface electromyography signal is a physiological signal characterizing the intensity of electrical activity of muscles during movement; joint motion trajectory is a spatial position sequence describing the change of angle of joint over time during movement; neuromuscular model is a mathematical model simulating the activation and regulation of muscles by the human nervous system and the process of muscles generating joint torque; joint torque distribution is a vector set representing the torque values ​​generated by each joint of the healthy leg under different movement phases.

[0022] In practice, firstly, when the patient performs standard rehabilitation movements on their unaffected leg, surface electromyography (SEMG) patches are used to collect SEMG signals from the quadriceps and hamstring muscles. Simultaneously, angle sensors installed at the robot joints collect the knee joint's motion trajectory. After aligning the collected SEMG signals and joint motion trajectories in time, they are input into a neuromuscular model stored in the robot controller after ACL surgery. This neuromuscular model includes a muscle activation dynamics module and a musculoskeletal geometry module. The muscle activation dynamics module calculates muscle activation based on the SEMG signals, while the musculoskeletal geometry module calculates muscle lever arms and joint torques based on the joint motion trajectory. Then… Subsequently, the neuromuscular model performs forward computation on the input signal, converting the surface electromyography signal into muscle activation. The muscle activation range is between 0 and 1, where 0 represents no activation and 1 represents maximum activation. Combining the joint motion trajectory, the musculoskeletal geometry module calculates the lever arm value of each muscle at the current joint angle. The muscle activation is multiplied by the maximum isometric contraction force of the muscle and the lever arm, and the contribution values ​​of all muscles crossing the same joint are accumulated. This allows the real-time torque values ​​of each joint of the healthy leg under the current movement to be deduced. All torque values ​​are arranged in time frame order to form the joint torque distribution, which serves as the benchmark for the normal force exertion pattern of the healthy leg in standard rehabilitation movements.

[0023] It should be noted that in this application, the thresholds involved in the neuromuscular skeletal model include the minimum detection threshold for muscle activation, which is preset to 0.05. In other embodiments, the threshold is modified as follows: The background noise amplitude of the surface electromyography signal is collected while the patient is in a relaxed state, and its root mean square value is calculated. Three times this root mean square value is used as the minimum detection threshold. When the calculated muscle activation value is lower than 0.05, the model considers it as muscle inactivation, and the output torque is 0 to eliminate noise interference. This threshold is automatically calibrated once before each training session using a 5-second resting data acquisition during the patient's relaxed state.

[0024] In step 102, the torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same action is obtained to obtain the torque loss of the affected leg. Then, the surface electromyography signal of the affected leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model.

[0025] In some embodiments, the torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement can be used to obtain the torque loss of the affected leg. This can be achieved through the following steps: Align the joint torque distribution of the healthy leg with the actual joint torque of the affected leg along the time axis, and then calculate the difference between the torque value of the healthy side and the actual torque value of the affected side at each time frame; The torque loss of the affected leg was selected from all the differences.

[0026] It should be noted that, in this application, the joint torque distribution of the healthy leg is a vector set representing the torque values ​​generated by each joint of the healthy leg under different action phases; the actual joint torque of the affected leg is a sequence of torque values ​​measured by sensors at each joint of the affected leg when performing the same action; the healthy side torque value is a single torque value extracted from the joint torque distribution of the healthy leg in a single time frame; the affected side actual torque value is a single torque value extracted from the actual joint torque of the affected leg in the same time frame; the difference is a value that measures the magnitude difference between the healthy side torque value and the affected side actual torque value in the same time frame; the torque missing amount is a subset of differences describing the insufficient force exerted by the affected leg relative to the healthy leg.

[0027] In specific implementation, firstly, the joint moment distribution of the healthy leg and the actual joint moment of the affected leg are aligned on the time axis. Then, at each time frame, the difference between the moment value of the healthy side and the actual moment value of the affected side can be calculated in the following way: After ACL surgery, the robot records the trigger time of the start of the action when collecting data of the healthy leg, and similarly records the trigger time of the start of the action when collecting data of the affected leg. Taking the start time of the action as the zero point, the joint moment distribution of the healthy leg and the actual joint moment of the affected leg are interpolated and resampled at the same time sampling frequency of 100 Hz. Taking the end time of the shorter action duration as the alignment endpoint, the two moment sequences are aligned frame by frame from the zero point to the endpoint to complete the time axis alignment. The difference between the moment value of the healthy side and the actual moment value of the affected side is calculated at each time frame. Starting from the first aligned time frame, the difference is calculated sequentially. The healthy side torque value and the actual torque value of the affected side in the frame are obtained. The actual torque value of the affected side is subtracted from the healthy side torque value, and the result is used as the difference for the frame. The above subtraction operation is repeated frame by frame in chronological order to obtain a difference sequence equal to the number of time frames. Then, the torque deficiency of the affected leg can be screened from all the differences in the following way: Iterate through each difference in the difference sequence, judge the magnitude of the difference, keep the differences greater than 0 and mark them as torque deficiency, indicating that the affected leg outputs less torque than the healthy leg in the time frame; discard the differences less than or equal to 0, indicating that the force exerted by the affected leg in the frame has reached or exceeded the level of the healthy leg and does not require additional compensation. Arrange all the retained differences greater than 0 in the original chronological order as the torque deficiency sequence, which is the part of the affected leg that needs robot assistance to compensate for the insufficient torque in the current action.

[0028] In some embodiments, the electromyography (EMG) signal of the affected leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, thereby obtaining the affected-side corrected model. This can be achieved through the following steps: Extract the amplitude envelope of surface electromyographic signals of each muscle in the affected leg during movement; The amplitude envelope is compared with the theoretical electromyography corresponding to the current muscle recruitment parameters in the NMS model to obtain multiple deviation ratios; The recruitment threshold and gain coefficient in the NMS model are adjusted in reverse according to the various deviation ratios. The adjusted parameter set is then written into the NMS model to generate the affected side correction model.

[0029] It should be noted that, in this application, the surface electromyography (EMG) signals of each muscle in the affected leg during movement are physiological signal sequences representing the electrical activity intensity of each relevant muscle in the affected leg; the amplitude envelope is a numerical sequence representing the outer contour curve of the surface EMG signal intensity changing over time; the NMS model is a mathematical model simulating the activation regulation of muscles by the human nervous system and the process of muscles generating joint torque; the muscle recruitment parameters are a set of adjustable variables controlling the motor unit recruitment threshold and gain coefficient in the NMS model; the theoretical EMG is a numerical value representing the muscle electrical activity intensity simulated and calculated by the NMS model under the current muscle recruitment parameters; the deviation ratio is a percentage measuring the relative difference between the measured amplitude envelope and the theoretical EMG; the recruitment threshold is a parameter that determines the minimum intensity of the input signal required for the muscle motor unit to begin to be activated; the gain coefficient is a parameter controlling the slope of the mapping relationship between the muscle input signal and the output activation intensity; and the affected side correction model is a parameterized model characterizing the personalized NMS model after calibration with measured EMG signals from the affected side.

[0030] In specific implementation, firstly, the amplitude envelope of the surface electromyography (EMG) signals of each muscle in the affected leg during movement can be extracted in the following way: During the standard rehabilitation movement performed on the affected leg, surface EMG signals of the quadriceps and hamstring muscles are collected using surface EMG patches. The collected raw EMG signals are then subjected to full-wave rectification, i.e., all negative values ​​are converted to positive values ​​by taking the absolute value. The rectified signals are then processed using a low-pass filter with a cutoff frequency of 5 Hz to remove high-frequency fluctuation components while retaining the overall intensity variation trend of the signal. The smooth curve obtained after filtering is used as the amplitude envelope of the muscle. The same extraction operation is performed on the quadriceps and hamstring muscles to obtain the amplitude envelope of each muscle. Then, the amplitude envelope and the current muscle recruitment parameters in the NMS model are compared. The deviation ratios can be obtained by comparing the corresponding theoretical electromyography (EMG) values ​​using the following method: Read the currently stored muscle recruitment parameters, including the recruitment threshold and gain coefficient, from the NMS model. Input the amplitude envelope as the input signal to the NMS model. The NMS model calculates the corresponding theoretical EMG output based on the current recruitment threshold and gain coefficient. At each time frame, calculate the difference between the amplitude envelope value and the theoretical EMG value. Divide this difference by the theoretical EMG value and multiply by 100 to obtain the deviation ratio at that time frame. Repeat the above calculation for all time frames and take the arithmetic mean of the deviation ratios at each time frame for the same muscle as the representative deviation ratio for that muscle. The representative deviation ratios for the quadriceps and hamstring muscles are calculated using the above method.

[0031] Finally, in practical implementation, the recruitment threshold and gain coefficient in the NMS model are adjusted in reverse according to each deviation ratio. The adjusted parameter set is then written into the NMS model to generate the affected-side corrected model. This can be achieved in the following way: when the deviation ratio is positive, it indicates that the measured amplitude envelope is greater than the theoretical electromyography (EMG), meaning that the current model parameters underestimate the patient's muscle activation ability. The recruitment threshold is then lowered and the gain coefficient is increased. When the deviation ratio is negative, it indicates that the measured amplitude envelope is less than the theoretical EMG, meaning that the current model parameters overestimate the patient's muscle activation ability. The recruitment threshold is then increased and the gain coefficient is decreased. The adjustment magnitude is proportional to the absolute value of the deviation ratio, and the adjustment coefficient is 0.5. That is, the adjustment amount of the recruitment threshold is the original recruitment threshold multiplied by the absolute value of the deviation ratio and then multiplied by 0.5. The adjustment amount of the gain coefficient is similar. The adjusted parameter set is then written into the NMS model to generate the affected-side corrected model. The adjusted recruitment threshold and gain coefficient are combined to form a new parameter set, which replaces the original muscle recruitment parameters in the NMS model. The NMS model after parameter replacement is then saved as the affected-side corrected model.

[0032] It should be noted that in this application, the proportional adjustment coefficient in the reverse adjustment process of muscle recruitment parameters is preset to 0.5 by default. The specific preset method is as follows: using preliminary experimental data from 10 healthy subjects and 10 patients after ACL surgery, parameter calibration was performed using three adjustment coefficients of 0.2, 0.5, and 0.8, respectively. The minimum root mean square error between the predicted torque and the measured torque after adjustment was used as the evaluation index. The experimental results showed that the average error was the smallest at a coefficient of 0.5, which was 12.3 N·m, and the convergence speed was moderate. Therefore, it was preset to 0.5. This preset value is a fixed parameter when the robot leaves the factory and does not change with individual patient differences.

[0033] In step 103, the torque loss of the affected leg is input into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action. Then, the compensation torque command of the robot after anterior cruciate ligament surgery is generated by the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg.

[0034] In some embodiments, the missing torque of the affected leg is input into the affected leg correction model to obtain the predicted theoretical torque of the affected leg under the current action, which can be achieved by the following steps: The torque loss of the affected leg is used as an excitation input signal and loaded into the input layer of the motor neuron of the affected side correction model; The affected-side modified model simulates the activation response of each muscle after receiving the stimulus, based on the modified muscle recruitment parameters; The activation response of each muscle is mapped to the theoretical output value in the joint space, thereby obtaining the predicted theoretical torque of the affected leg under the current movement.

[0035] It should be noted that, in this application, the motor neuron input layer is the first computational unit used to receive external excitation signals and convert them into a neural impulse firing rate model; the activation response is a value representing the change of normalized activation intensity of the muscle over time after receiving neural excitation; the joint space is a coordinate system describing the relationship between the angles and torques of the joints of the affected leg; the theoretical output value represents the magnitude of the force that a single muscle can generate under a given activation response, as predicted by the model; and the predicted theoretical torque is a vector set representing the predicted torque values ​​that each joint of the affected leg should generate under the current action.

[0036] In specific implementation, firstly, the torque loss of the affected leg is used as an excitation input signal and loaded into the motor neuron input layer of the affected side correction model. This can be achieved in the following way: From the torque loss sequence, each torque loss value is sequentially extracted according to the time frame order. This torque loss value is used as the excitation input signal for the current time frame and input into the motor neuron input layer of the affected side correction model. After receiving the excitation signal, the motor neuron input layer converts it into a change in the membrane potential of the motor neuron. The conversion method is a linear mapping, that is, the change in membrane potential is equal to the excitation input signal multiplied by a unit conversion coefficient of 1. For torque loss... Repeating the above loading operation for each time frame in the stimulus sequence yields a temporally continuous stimulus input stream. Then, the affected-side correction model simulates the activation response of each muscle after receiving this stimulus, based on the corrected muscle recruitment parameters. This can be achieved as follows: The affected-side correction model reads the internally stored corrected muscle recruitment parameters, including the adjusted recruitment threshold and gain coefficient. For each muscle in the quadriceps and hamstrings, the change in motor neuron membrane potential is compared with the recruitment threshold for that muscle. When the change in membrane potential is less than the recruitment threshold, the activation response of that muscle is 0; when the change in membrane potential is greater than or equal to the recruitment threshold, the activation response of that muscle is 0. When the activation response equals the recruitment threshold, it is calculated by subtracting the recruitment threshold from the membrane potential change and multiplying by a gain coefficient. The result is limited to between 0 and 1, with 1 representing the maximum activation response. The activation response values ​​for the quadriceps and hamstring muscles are calculated separately to obtain the activation response value of each muscle in the current time frame. By repeating the simulation calculation for each time frame in the torque loss sequence using the above method, the activation response curves of each muscle over time can be obtained. Finally, the activation response of each muscle is mapped to the theoretical output value in joint space, thus obtaining the predicted theoretical torque of the affected leg under the current movement. This can be achieved using the following method. In other words, the affected side correction model stores the musculoskeletal geometric parameters of the affected leg, including a lookup table of the lever arm of each muscle relative to the knee and hip joints as the joint angle changes. At each time frame, the joint angle value and the activation response value of each muscle are obtained simultaneously. For the quadriceps femoris, its activation response value is multiplied by the maximum isometric contraction force of the quadriceps femoris, 400N, and then multiplied by the lever arm value of the quadriceps femoris at the current joint angle, 0.03m, to obtain the theoretical joint output value of the quadriceps femoris. For the hamstrings, its activation response value is multiplied by the maximum isometric contraction force of the hamstrings, 300N, and then multiplied by the lever arm value of the hamstrings at the current joint angle, 0.03m.At 035m, the theoretical joint output value of the hamstring muscle is obtained. The theoretical output values ​​of all muscles crossing the same joint are then accumulated according to the direction of mechanical action, with positive values ​​for knee extension and negative values ​​for knee flexion. The accumulated values ​​yield the net theoretical output value of the joint in that time frame, i.e., the predicted theoretical torque of the joint. This mapping and accumulation operation is performed on all time frames. The predicted theoretical torques of each time frame are arranged in chronological order to form the predicted theoretical torque sequence of the affected leg under the current movement.

[0037] In some embodiments, the compensation torque command for the robot after anterior cruciate ligament surgery is generated based on the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg, with reference to... Figure 2 The figure is a flowchart illustrating the determination of the compensation torque command in some embodiments of this application. In this embodiment, the determination of the compensation torque command can be achieved through the following steps: In step 1031, at the same time frame, the actual torque difference between the predicted theoretical torque and the actual joint torque of the affected leg is calculated; In step 1032, the auxiliary gain coefficient of the robot after the anterior cruciate ligament surgery is determined, and all real-time torque differences are low-pass filtered. In step 1033, the auxiliary torque of the robot after the anterior fork surgery is limited and compensated according to the filtered torque difference and the auxiliary gain coefficient to obtain the compensation torque command of the robot after the anterior fork surgery.

[0038] It should be noted that, in this application, the actual torque difference represents the numerical value of the difference between the predicted theoretical torque and the actual joint torque of the affected leg on the same time frame; the auxiliary gain coefficient is a proportional factor used to control the magnitude of the robot's output auxiliary force after ACL surgery; and the compensation torque command is a control command used to drive the robot's joint module to output auxiliary force after ACL surgery.

[0039] In specific implementation, firstly, calculating the actual torque difference between the predicted theoretical torque and the actual joint torque of the affected leg in the same time frame can be achieved in the following way: Calculate the actual torque difference between the predicted theoretical torque and the actual joint torque of the affected leg in the same time frame. Obtain the predicted theoretical torque value for the current time frame from the predicted theoretical torque sequence, and simultaneously extract the actual joint torque value of the affected leg in the same time frame from the actual joint torque sequence of the affected leg. Subtract the actual joint torque value of the affected leg from the predicted theoretical torque value as the actual torque difference in that time frame. When the actual torque difference is positive, it indicates that the actual force output of the affected leg is insufficient, and the robot needs to provide positive assistance. When the actual torque difference is negative, it indicates that the actual force exerted by the affected leg has exceeded the predicted value, and robot assistance is not needed or the robot needs to provide resistance. Repeating the above subtraction calculation for each time frame during the movement process yields the time-varying sequence of actual torque differences. Then, the robot's auxiliary gain coefficient after ACL surgery is determined. Low-pass filtering of all real-time torque differences can be achieved as follows: the auxiliary gain coefficient is determined using a segmented preset method. When the absolute value of the actual torque difference is less than 5 N·m, the gain coefficient is set to 0.5; when the absolute value of the actual torque difference is between 5 and 15 N·m, the gain coefficient is set to 0.7; and when the absolute value of the actual torque difference is greater than 15 N·m, the gain coefficient is set to 0. 0.9; Simultaneously, a first-order low-pass filter is used, with a filtering time constant set to 0.05 seconds. The filtering calculation formula is: the current filtered torque difference equals 0.8 multiplied by the torque difference filtered at the previous moment plus 0.2 multiplied by the current real-time torque difference. The filtered torque difference in the first time frame is directly equal to the real-time torque difference of that frame. By performing the above filtering calculation sequentially on each real-time torque difference value, the filtered torque difference sequence can be obtained. Finally, based on the filtered torque difference and the auxiliary gain coefficient, the auxiliary torque of the robot after anterior fork surgery is limited and compensated. The compensation torque command of the robot after anterior fork surgery can be implemented in the following way: multiply the filtered torque difference by the coefficient of friction. Based on the auxiliary gain coefficient determined on the time frame, the initial auxiliary torque value is obtained. Then, the initial auxiliary torque is subjected to amplitude limiting processing, with an upper limit of 30 N·m and a lower limit of 0 N·m. When the initial auxiliary torque is greater than 30 N·m, the compensation torque command is set to 30 N·m; when the initial auxiliary torque is less than 0 N·m, the compensation torque command is set to 0 N·m; when the initial auxiliary torque is in the range of 0 to 30 N·m, the compensation torque command is set to the initial auxiliary torque itself. The value obtained after amplitude limiting processing is output as the compensation torque command for that time frame to the robot joint actuator. By performing the above calculation and amplitude limiting operation sequentially on all time frames, a complete compensation torque command sequence can be obtained.

[0040] In step 104, the compensation torque command is used to drive the joint modules in the post-anterior cruciate scaffold robot to output auxiliary force, thereby realizing the collaborative assisted training of the post-anterior cruciate scaffold robot.

[0041] In some embodiments, driving the joint module in the postoperative anterior cruciate sclerosis robot to output auxiliary force using the compensation torque command can be achieved through the following steps: The compensation torque command is sent to the joint servo driver of the robot after the fork surgery. The driver converts the torque command into a corresponding current command and drives the motor to output the corresponding auxiliary torque. The actual output torque is monitored by the torque sensor built into the joint, and the actual output torque and the compensation torque command are combined to form a closed-loop feedback adjustment to obtain the auxiliary torque command for each joint. After ACL surgery, the robot's joints coordinate to perform actions according to the corresponding auxiliary torque commands, applying auxiliary force to the corresponding segments of the affected leg.

[0042] It should be noted that in this application, the joint servo drive of the robot after anterior fork surgery adopts a three-level cascade control architecture of current-torque-position. After the compensation torque command is input as the target value into the joint servo driver, the driver internally converts the torque command into a given current value of the current loop based on the motor torque constant (N·m / A), driving the auxiliary torque at the output shaft of the brushless DC motor. This auxiliary torque is amplified by the reducer and then transmitted to the joint output end. At the same time, the torque sensor installed between the joint output end and the connecting rod detects the actual output force in real time at a sampling frequency of 1000 Hz. The torque is calculated and the difference between the torque and the original compensation torque command generates a torque error signal. This error signal is then adjusted by a proportional-integral controller to correct the current command, thus completing closed-loop torque control. Multiple joints are synchronized in a master-slave manner. The knee joint, as the master joint, performs flexion and extension assistance according to the compensation torque command. The hip and ankle joints, as slave joints, use the knee joint's movement phase as the trigger reference and independently complete the closed-loop torque adjustment of their respective joints. The straps on each joint linkage apply auxiliary force to the corresponding segments of the affected thigh, calf, and foot, respectively, to achieve multi-joint coordinated auxiliary training.

[0043] Furthermore, in another aspect of this application, in some embodiments, this application provides a post-anesthesia surgery robot based on the NMS model, which includes a control unit, referencing... Figure 3 The figure is a schematic diagram of the structure of a control unit according to some embodiments of this application. The control unit includes: a data acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to acquire the surface electromyography signal and joint motion trajectory of the healthy leg during the action, input the neuromuscular skeletal model in the robot after ACL surgery, and inversely solve the joint torque distribution of the healthy leg under the current action. Processing module 202, in this application, is used to perform torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same action, to obtain the torque loss of the affected leg, and to use the surface electromyography signal of the affected leg to perform electromechanical correction on the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model. It should be noted that the processing module 202 is also used to input the torque loss of the affected leg into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action, and then generate the compensation torque command of the robot after anterior cruciate ligament surgery by means of the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg. The execution module 203 in this application is mainly used to drive the joint modules in the postoperative anterior cruciate fascia robot to output auxiliary force using the compensation torque command, so as to realize the collaborative auxiliary training of the postoperative anterior cruciate fascia robot.

[0044] The foregoing detailed examples of an anterior cruciate fasciculus (ACL) postoperative robot and control method based on the NMS model provided in this application. It is understood that the corresponding device, in order to achieve the above functions, includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0045] In some embodiments, this application also provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device executes the above-described control method for anterior sclerosis postoperative robot based on the NMS model.

[0046] In some embodiments, reference Figure 4 The dashed lines in the figure indicate that the unit or module is optional. This figure is a schematic diagram of the structure of a computer device for implementing a control method for an anterior scaffold postoperative robot based on the NMS model, according to an embodiment of this application. The control method for an anterior scaffold postoperative robot based on the NMS model described in the above embodiments can be achieved through… Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a memory 302 and at least one communication unit 305. The computer device may be a terminal device, a server or a chip.

[0047] Processor 301 can be a general-purpose processor or a special-purpose processor. For example, processor 301 can be a central processing unit (CPU), which can be used to control computer devices, execute software programs, and process data from software programs. The computer device may also include a communication unit 305 for inputting (receiving) and outputting (transmitting) signals.

[0048] For example, the computer device may be a chip, and the communication unit 305 may be the input and / or output circuit of the chip, or the communication unit 305 may be the communication interface of the chip, which may be a component of a terminal device, network device or other device.

[0049] For example, the computer device may be a terminal device or a server, and the communication unit 305 may be a transceiver of the terminal device or the server, or the communication unit 305 may be a transceiver circuit of the terminal device or the server.

[0050] The computer device may include one or more memories 302 storing a program 304. The program 304 can be executed by a processor 301 to generate instructions 303, causing the processor 301 to execute the method described in the above method embodiments according to the instructions 303. Optionally, the memory 302 may also store data (such as a target audit model). Optionally, the processor 301 may also read data stored in the memory 302, which may be stored at the same storage address as the program 304, or it may be stored at a different storage address than the program 304.

[0051] The processor 301 and memory 302 can be configured separately or integrated together, for example, integrated on the system on chip (SOC) of the terminal device.

[0052] It should be understood that each step of the above method embodiment can be completed by hardware logic circuits or software instructions in the processor 301. The processor 301 can be a CPU, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, such as discrete gate, transistor logic devices, or discrete hardware components.

[0053] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] For example, in some embodiments, this application also provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the above-described control method for anterior cruciate fasciculation robots based on the NMS model.

[0055] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0056] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A control method for anterior cruciate fasciculations (ACL) robots based on an NMS model, characterized in that, Includes the following steps: Collect surface electromyography signals and joint motion trajectories of the healthy leg during the movement, input them into the neuromuscular skeletal model in the robot after ACL surgery, and inversely solve the joint torque distribution of the healthy leg under the current movement. The torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement is obtained to determine the torque loss of the affected leg. Then, the surface electromyography signal of the affected leg is used to electromechanically correct the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model. The torque loss of the affected leg is input into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action. Then, the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg is used to generate the compensation torque command of the robot after anterior cruciate fascia surgery. The compensation torque command is used to drive the joint modules in the post-operative anterior cruciate ligament robot to output auxiliary force, thereby realizing the collaborative assisted training of the post-operative anterior cruciate ligament robot.

2. The method as described in claim 1, characterized in that, The torque difference between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement is calculated to obtain the torque loss of the affected leg, which specifically includes: Align the joint torque distribution of the healthy leg with the actual joint torque of the affected leg along the time axis, and then calculate the difference between the torque value of the healthy side and the actual torque value of the affected side at each time frame; The torque loss of the affected leg was selected from all the differences.

3. The method as described in claim 1, characterized in that, Electromyography (EMG) signals from the affected leg were used to electromechanically correct the muscle recruitment parameters in the NMS model, resulting in a modified model for the affected side. Specifically, this included: Extract the amplitude envelope of surface electromyographic signals of each muscle in the affected leg during movement; The amplitude envelope is compared with the theoretical electromyography corresponding to the current muscle recruitment parameters in the NMS model to obtain multiple deviation ratios; The recruitment threshold and gain coefficient in the NMS model are adjusted in reverse according to the various deviation ratios. The adjusted parameter set is then written into the NMS model to generate the affected side correction model.

4. The method as described in claim 1, characterized in that, The missing torque of the affected leg is input into the affected leg correction model to obtain the predicted theoretical torque of the affected leg under the current action, specifically including: The torque loss of the affected leg is used as an excitation input signal and loaded into the input layer of the motor neuron of the affected side correction model; The affected-side modified model simulates the activation response of each muscle after receiving the stimulus, based on the modified muscle recruitment parameters; The activation response of each muscle is mapped to the theoretical output value in the joint space, thereby obtaining the predicted theoretical torque of the affected leg under the current movement.

5. The method as described in claim 1, characterized in that, The generation of compensating torque commands for the robot after anterior cruciate ligament surgery, based on the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg, specifically includes: At the same time frame, calculate the actual torque difference between the predicted theoretical torque and the actual joint torque of the affected leg; Determine the auxiliary gain coefficient of the robot after the anterior fork surgery, and perform low-pass filtering on all real-time torque differences; Based on the filtered torque difference and the auxiliary gain coefficient, the auxiliary torque of the robot after the forklift surgery is limited and compensated to obtain the compensation torque command of the robot after the forklift surgery.

6. The method as described in claim 1, characterized in that, The specific methods for using the aforementioned compensation torque command to drive the joint modules in the post-operative anterior cruciate scaffold robot to output auxiliary forces include: The compensation torque command is sent to the joint servo driver of the robot after the fork surgery. The driver converts the torque command into a corresponding current command and drives the motor to output the corresponding auxiliary torque. The actual output torque is monitored by the torque sensor built into the joint, and the actual output torque and the compensation torque command are combined to form a closed-loop feedback adjustment to obtain the auxiliary torque command for each joint. After ACL surgery, the robot's joints coordinate to perform actions according to the corresponding auxiliary torque commands, applying auxiliary force to the corresponding segments of the affected leg.

7. The method as described in claim 1, characterized in that, The aforementioned post-anterior cruciate ligament (ACL) surgery robot is a knee joint rehabilitation training robot based on a lower limb exoskeleton.

8. A postoperative robot for anterior scaffold surgery based on the NMS model, the postoperative robot for anterior scaffold surgery based on the NMS model includes a control unit, characterized in that, The control unit includes: The acquisition module is used to acquire surface electromyography signals and joint motion trajectories of the healthy leg during the movement, input them into the neuromuscular model in the robot after ACL surgery, and inversely solve the joint torque distribution of the healthy leg under the current movement. The processing module is used to perform torque difference analysis between the actual joint torque of the affected leg and the joint torque distribution of the healthy leg under the same movement, to obtain the torque loss of the affected leg, and to use the surface electromyography signal of the affected leg to perform electromechanical correction on the muscle recruitment parameters in the NMS model, thereby obtaining the affected side correction model. The processing module is also used to input the torque loss of the affected leg into the affected side correction model to obtain the predicted theoretical torque of the affected leg under the current action, and then generate the compensation torque command of the robot after anterior cruciate fasciculations by the torque difference between the predicted theoretical torque and the actual joint torque of the affected leg. The execution module is used to drive the joint modules in the post-operative anterior cruciate fascia robot to output auxiliary force using the compensation torque command, thereby realizing the collaborative assisted training of the post-operative anterior cruciate fascia robot.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to call and run the computer programs from the memory, so that the computer device performs the control method of the anterior cruciate fasciculus robot based on the NMS model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions or code that, when executed on a computer, cause the computer to implement the control method for anterior cruciate fasciculus robot based on the NMS model as described in any one of claims 1 to 7.

Images