High-compliance impedance training method and system based on self-adaptive visual servo control and application of high-compliance impedance training method and system

Abstract

The invention discloses a high-compliance impedance training method and system based on self-adaptive visual servo control and application of the high-compliance impedance training method and system. The training method comprises the steps that historical pose data and historical contact stress data are collected; performing feature processing to obtain stress pose information and generate a corresponding pose mapping relation; constructing a dynamic mapping model associated with the muscle force, the contact stress, the training track and the impedance parameters in combination with the stress pose information and the pose mapping relation; acquiring rehabilitation training information of a current patient, and generating and optimizing a preliminary expected trajectory; calculating deviation data of the real-time trajectory and the expected training trajectory; and an impedance adjusting model is constructed, the input of the impedance adjusting model comprises the deviation data and the real-time contact stress data, and the output of the impedance adjusting model is the impedance adjusting quantity. According to the invention, through visual real-time feedback, fuzzy PID adaptive adjustment and multi-feature fusion modeling, impedance adjustment training with better dynamic adaptability, higher safety, higher flexibility and high personalized precision can be realized.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a high compliant impedance training method, system, and application based on adaptive visual servo control. Background Technology

[0002] With the aging population, the demand for rehabilitation services for chronic diseases and neurological injuries is constantly increasing, and the number of patients with motor dysfunction due to stroke, spinal cord injury, and other conditions is also gradually rising, leading to a growing global demand for rehabilitation. Traditional rehabilitation treatments mainly rely on manual intervention by medical staff, which presents numerous problems, such as: insufficient medical staff to meet the rehabilitation needs of a large number of patients; unstable treatment effects, influenced by factors such as the skill level of medical staff and the patient's own willingness to undergo rehabilitation; and low patient compliance, resulting in prolonged treatment cycles and unsatisfactory rehabilitation outcomes. These patients' urgent need for rehabilitation treatment provides a vast market space for the development of rehabilitation robots.

[0003] To better facilitate patient rehabilitation training, rehabilitation robots need to possess good control compliance and stability. This enhances patient comfort during rehabilitation training and achieves better training results. For example, the invention patent application number 202210470538.3 proposes a compliant control method for a lower limb rehabilitation robot. This method uses pressure and angle sensors to provide real-time feedback of motion data during rehabilitation training. This data serves as input to the impedance control loop and position control loop, allowing for real-time correction of the motion trajectory and adjustment of the posture to ensure the compliance of the rehabilitation movements and achieve stability throughout the entire cyclical rehabilitation training process.

[0004] The above scheme mainly relies on fixed initial trajectory recognition parameters, lacks real-time visual feedback, and is difficult to adapt to the dynamic posture changes of patients during training. In addition, the above scheme only relies on interactive force feedback to correct the trajectory, the feedback is relatively simple, lacks stress feature processing and multi-dimensional safety protection, and may have insufficient training safety issues. Summary of the Invention

[0005] The purpose of this invention is to provide a high-compliance impedance training method, system, and application based on adaptive visual servo control, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A high-compliance impedance training method based on adaptive visual servo control includes:

[0008] Collect historical posture data and historical contact stress data of the end effector of the training institution and the patient's training site;

[0009] The historical pose data and historical contact stress data are subjected to feature processing to obtain stress pose information, and then a corresponding pose mapping relationship is generated by extraction;

[0010] According to the divided muscle strength grade data, the stress pose information and the pose mapping relationship are combined to construct a dynamic mapping model related to muscle strength, contact stress, training trajectory and impedance parameter;

[0011] Rehabilitation training information of a current patient is obtained, a preliminary expected trajectory and corresponding preliminary contact stress information are generated by the dynamic mapping model, and the preliminary expected trajectory is optimized to generate an expected training trajectory;

[0012] Real-time pose data and real-time contact stress data of the current patient are collected in real time based on visual servoing control, real-time trajectory is generated based on the real-time pose data and the pose mapping relationship, and deviation data of the real-time trajectory and the expected training trajectory is calculated.

[0013] An impedance adjustment model is constructed, the input of the impedance adjustment model includes the deviation data and the real-time contact stress data, the output of the impedance adjustment model is an impedance adjustment amount, and impedance adjustment training of an end effector of a training mechanism is performed according to the impedance adjustment amount.

[0014] The application also proposes a system based on the above method, which includes a data acquisition module, a pose mapping module, a model construction module, a trajectory generation module, a visual servoing and deviation calculation module and an impedance adjustment module, the data acquisition module is used to collect historical and real-time pose data and historical and real-time contact stress data; the pose mapping module is used to process data and generate stress pose information and pose mapping relationship; the model construction module is used to construct a dynamic mapping model related to muscle strength, contact stress, training trajectory and impedance parameter; the trajectory generation module is used to generate and optimize an expected training trajectory; the visual servoing and deviation calculation module is used to collect data in real time, calculate deviation data of the trajectory and monitor changes in contact stress; the impedance adjustment module is used to construct an impedance adjustment model and output adjustment instructions related to the impedance adjustment amount to drive control instructions.

[0015] The application also proposes an application of the above system in a rehabilitation training device.

[0016] Compared with the prior art, the beneficial effects of the present application are: the present application is based on visual real-time acquisition of real-time pose data of the patient training part, and generates a real-time trajectory through the pose mapping relationship, then, based on the comparison between the real-time trajectory of visual feedback and the expected trajectory, the corresponding position and attitude deviation is calculated, and the impedance parameter and the trajectory are dynamically corrected, which can adapt to the dynamic attitude change of the patient in the training process, avoiding the limitation of static initialization trajectory; the present application first eliminates the instantaneous interference of historical and real-time contact stress data through feature processing, then extracts the stress features, ensures the reliability of the feedback data, and avoids false correction; by extracting the stress peak value in the preliminary contact stress information first, when the stress peak value exceeds the preset value, the trajectory path is adjusted by using the Bezier curve and the trajectory speed is reduced, which can effectively avoid the risk of excessive stress; by monitoring the contact stress change rate in real time, when a sudden change occurs, the stiffness is immediately reduced, the damping is increased and the trajectory advancement is paused, thereby effectively improving the training safety. BRIEF DESCRIPTION OF DRAWINGS

[0017] Fig. 1 Fig. 1 is a flowchart of a high-compliance impedance training method based on adaptive visual servo control;

[0018] Fig. 2 Fig. 2 is a schematic diagram of a high-compliance impedance training system based on adaptive visual servo control;

[0019] Fig. 3 Fig. 3 is a schematic diagram of the running process of a dynamic mapping model in the high-compliance impedance training system based on adaptive visual servo control. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0021] Please refer to Figs. 1-3 The present application provides a technical solution: a high-compliance impedance training method based on adaptive visual servo control, which comprises the following specific steps:

[0022] S1, collecting historical pose data and historical contact stress data of the end effector of the training mechanism and the training part of the patient;

[0023] S2, feature processing is performed on the historical pose data and the historical contact stress data to obtain stress pose information, and based on the stress pose information, corresponding pose mapping relationship is extracted and generated, wherein the stress pose information includes contact stress feature information and pose feature information;

[0024] S3, constructing a dynamic mapping model of the associated muscle strength, contact stress, training trajectory and impedance parameter according to the divided muscle strength grade data, in combination with the stress posture information and the posture mapping relationship;

[0025] S4, obtaining rehabilitation training information of a current patient, generating a preliminary expected trajectory and corresponding preliminary contact stress information from the dynamic mapping model, and optimizing the preliminary expected trajectory to generate an expected training trajectory, wherein the rehabilitation training information includes muscle strength grade and training target;

[0026] S5, collecting real-time posture data and real-time contact stress data of the current patient in real time based on visual servoing control, generating a real-time trajectory based on the real-time posture data and using the posture mapping relationship, and calculating deviation data of the real-time trajectory and the expected training trajectory;

[0027] S6, constructing an impedance adjustment model, the input of the impedance adjustment model including the deviation data and the real-time contact stress data, and the output of the impedance adjustment model being an impedance adjustment amount, and performing impedance adjustment training of the end effector of the training mechanism according to the impedance adjustment amount.

[0028] In some embodiments, the method for collecting the historical posture data includes collecting the historical posture data covering the full range of motion of the training part of the patient using a binocular vision camera, wherein the training part of the patient can be specifically the upper limbs shoulder, elbow, wrist, lower limbs hip, knee, ankle, etc.

[0029] The method for collecting the historical contact stress data includes attaching an array type flexible pressure sensor to the curved surface area of the end effector of the training mechanism in contact with the patient, and collecting the historical contact stress data through the array type flexible pressure sensor.

[0030] The present application collects real-time posture data of the training part of the patient in real time based on vision, generates a real-time trajectory through a posture mapping relationship, then compares the real-time trajectory with the expected trajectory based on visual feedback, calculates the corresponding position and attitude deviation, and dynamically corrects the impedance parameter and the trajectory, which can adapt to the dynamic attitude change of the patient in real time during the training process, and avoids the limitations of static initialization trajectory.

[0031] In some embodiments, the method for processing the features of the historical posture data includes filtering the motion jitter noise of the historical posture data using a Kalman filtering algorithm, and extracting posture feature information through principal component analysis, wherein the posture feature information can specifically include joint angle, end position coordinates, etc.

[0032] The method for feature processing of historical contact stress data comprises: adopting a sliding window average method to eliminate transient interference of the historical contact stress data, and extracting contact stress feature information through a threshold segmentation algorithm, wherein the contact stress feature information can specifically comprise average stress, stress distribution uniformity, etc. The corresponding relationship between the contact stress feature information and the pose feature information can be fitted through a least square method to obtain a pose mapping relationship.

[0033] The present application firstly eliminates the transient interference of historical and real-time contact stress by feature processing of the contact stress data, and then extracts stress features to ensure the reliability of the feedback data and avoid false correction.

[0034] In some embodiments, the method for constructing a dynamic mapping model comprises: constructing a model using an improved BP neural network, wherein the model input is muscle strength grade data, contact stress feature information and / or pose feature information, and the model output is a training trajectory path point, a trajectory speed, a stiffness parameter and / or a damping parameter. The muscle strength data can be specifically divided into 0-5 levels according to a muscle strength evaluation scale, and each level corresponds to 30-50 groups of historical stress and pose data.

[0035] The network weight is optimized using a gradient descent method with historical data as training samples until the model prediction error is less than a set threshold. The sample size of the training sample can be greater than or equal to 300 groups. In addition, a model update interface can be specifically set to automatically iterate and optimize the model parameters every 100 new data.

[0036] In some embodiments, the method for optimizing the preliminary expected trajectory comprises:

[0037] The stress peak value in the preliminary contact stress information is extracted. If the stress peak value is greater than a preset peak value, the trajectory path is adjusted using a Bezier curve, so that the deviation between the adjusted trajectory path and the original trajectory path is less than a preset deviation, and the trajectory speed is reduced to a set range.

[0038] If the acceleration change rate of adjacent path points of the preliminary expected trajectory is greater than a preset threshold, the trajectory is smoothed through an interpolation algorithm such as a cubic spline interpolation.

[0039] Meanwhile, the trajectory is secondarily adapted in combination with the training target of the patient. Specifically, the training target can include: expanding the trajectory activity range by 10-15% for joint range of motion improvement; increasing the trajectory turning frequency for movement coordination training; and generating an optimized expected training trajectory based on the above method.

[0040] The present application extracts the stress peak value in the preliminary contact stress information first, and then adjusts the trajectory path and reduces the trajectory speed using a Bezier curve when the stress peak value exceeds a preset value, which can effectively avoid the risk of excessive stress.

[0041] In some embodiments, the real-time trajectory includes a real-time position trajectory and a real-time attitude trajectory, and the method for calculating the deviation data includes: calculating the position deviation between the real-time position trajectory and the expected position trajectory by using the Euclidean distance; and calculating the attitude deviation between the real-time attitude trajectory and the expected attitude trajectory by using the angle difference. As a specific scheme, when the position deviation is less than or equal to 0.5 mm and the attitude deviation is less than or equal to 1°, the trajectory is determined to be adapted.

[0042] In some embodiments, the method for constructing the impedance adjustment model includes: designing the impedance adjustment model by using a fuzzy PID algorithm, taking the position deviation, the attitude deviation and the contact stress as input variables, and dividing each variable into a plurality of fuzzy subsets; setting a fuzzy rule base; and outputting the impedance adjustment amount by solving the fuzzy algorithm under the constraint conditions of minimum deviation, trajectory smoothness and non-sudden change of the contact stress. In a specific application process, each input variable can be divided into 5 fuzzy subsets, respectively corresponding to a negative large subset, a negative small subset, a zero subset, a positive small subset and a positive large subset; the fuzzy rule base can specifically include 40-50 rules, for example, when the position deviation is positive large and the contact stress is positive, the output stiffness adjustment amount is positive large; the stiffness adjustment range of the impedance adjustment amount can be set to 50-300 N / m, the damping adjustment range can be set to 5-35 N·s / m, and the adjustment response time can be set to be less than or equal to 6 ms.

[0043] The present application uses a fuzzy PID algorithm for adaptive impedance adjustment, and corrects the impedance adjustment amount according to the deviation data and the stress stability, so that the greater the deviation, the more accurate the adjustment, and the more stable the stress, the higher the compliance, and the dynamic balance is achieved, avoiding the limitations of fixed parameters.

[0044] In some embodiments, the training method for impedance adjustment includes: converting the impedance adjustment amount into a driving control instruction of the training mechanism, and driving the end effector to act through a servo driver; and monitoring the contact stress change rate in real time during the adjustment process, and if the contact stress suddenly changes, triggering an emergency adjustment mechanism, instantaneously reducing the stiffness parameter and increasing the damping parameter, and simultaneously pausing the trajectory advancement.

[0045] After completing each training cycle, the deviation data and the contact stress stability of the real-time trajectory and the expected training trajectory are compared, and if the deviation is greater than a preset threshold or the contact stress suddenly changes, the impedance adjustment amount is corrected based on the deviation data and the stress change data to ensure high compliance in the training process. In a specific application process, in order to ensure high compliance, the contact stress fluctuation can be set to be less than or equal to 5 N, and the contact stress change rate can be set to be less than or equal to 4 kPa / s.

[0046] The present application monitors the contact stress change rate in real time, and when a sudden change occurs, the stiffness is immediately reduced, the damping is increased, and the trajectory advancement is paused, thereby effectively improving the training safety.

[0047] The application also provides a system based on the method, comprising a data acquisition module, a pose mapping module, a model construction module, a trajectory generation module, a visual servoing and deviation calculation module, and an impedance adjustment module, the data acquisition module is used for acquiring historical and real-time pose data and historical and real-time contact stress data; the pose mapping module is used for processing data and generating stress-pose information and a pose mapping relationship; the model construction module is used for constructing a dynamic mapping model related to muscle strength, contact stress, training trajectory, and impedance parameters; the trajectory generation module is used for generating and optimizing a desired training trajectory; the visual servoing and deviation calculation module is used for acquiring data in real time, calculating deviation data of the trajectory, and monitoring contact stress changes; and the impedance adjustment module is used for constructing an impedance adjustment model and outputting adjustment instructions related to an impedance adjustment amount to drive a control instruction.

[0048] The application also provides an application of the system in a rehabilitation training device, which can be specifically applied in an upper limb rehabilitation robot and a lower limb rehabilitation robot, and the application process of the upper limb rehabilitation robot can specifically include: acquiring historical and real-time joint poses of a patient's upper limb shoulder, elbow, wrist, etc. by the data acquisition module, and also acquiring historical and real-time contact stresses of the arm and a contact stress change rate, etc., and executing steps S1-S6 of the high-compliance impedance training method, wherein the constraint conditions of the impedance adjustment model can specifically be that the deviation is less than or equal to 0.5 mm, the velocity fluctuation is less than or equal to 0.1 m / s, the contact stress change rate is less than or equal to 4 kPa / s, and the desired training trajectory is set as a flexion and extension, pronation, and / or supination motion trajectory, so as to ensure that there is no stress mutation in the training process.

[0049] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the involved claims.

Claims

1. A high compliance impedance training method based on adaptive visual servoing control, characterized in that, The method comprises the following steps: Collecting historical position data and historical contact stress data of a training mechanism end effector and a patient training site; Processing the historical position data and the historical contact stress data to obtain stress position information, and then extracting a corresponding position mapping relationship; According to the divided muscle strength grade data, combining the stress position information and the position mapping relationship, a dynamic mapping model related to muscle strength, contact stress, training trajectory and impedance parameters is constructed; Obtaining rehabilitation training information of a current patient, generating a preliminary expected trajectory and corresponding preliminary contact stress information from the dynamic mapping model, and optimizing the preliminary expected trajectory to generate an expected training trajectory; Based on visual servoing control, real-time position data and real-time contact stress data of the current patient are collected, and based on the real-time position data, a real-time trajectory is generated by using the position mapping relationship, and deviation data of the real-time trajectory and the expected training trajectory is calculated; An impedance adjustment model is constructed, the input of the impedance adjustment model includes the deviation data and the real-time contact stress data, and the output of the impedance adjustment model is an impedance adjustment amount, and the impedance adjustment training of the training mechanism end effector is performed according to the impedance adjustment amount.

2. The high-compliance impedance training method based on adaptive visual servoing control according to claim 1, wherein, The collection method of the historical position data comprises: collecting the historical position data in the full range of motion of the patient training site by using a binocular vision camera; The collection method of the historical contact stress data comprises: attaching an array type flexible pressure sensor to a curved surface region of the training mechanism end effector in contact with the patient, and collecting the historical contact stress data by using the array type flexible pressure sensor.

3. The high-compliance impedance training method based on adaptive visual servoing control according to claim 2, wherein, The method for processing the historical position data comprises: filtering the motion jitter noise of the historical position data by using a Kalman filtering algorithm, and extracting position feature information by using a principal component analysis method; The method for processing the historical contact stress data comprises: eliminating the instantaneous interference of the historical contact stress data by using a sliding window average method, and extracting contact stress feature information by using a threshold segmentation algorithm.

4. The high-compliance impedance training method based on adaptive visual servoing control according to claim 3, wherein, The construction method of the dynamic mapping model comprises: constructing a model by using an improved BP neural network, the input of the model is muscle strength grade data, contact stress feature information and / or position feature information, and the output of the model is a training trajectory path point, a trajectory speed, a stiffness parameter and / or a damping parameter; Using the historical data as a training sample, the network weight is optimized by using a gradient descent method, and the model is trained until the prediction error is less than a set threshold value.

5. The high-compliance impedance training method based on adaptive visual servoing control according to claim 4, wherein, The optimization method of the preliminary expected trajectory comprises: Extracting a stress peak value in the preliminary contact stress information, if the stress peak value is greater than a preset peak value, adjusting the trajectory path by using a Bezier curve, so that the deviation between the adjusted trajectory path and the original trajectory path is less than a preset deviation, and the trajectory speed is reduced to a set range; If the acceleration change rate of adjacent path points of the preliminary expected trajectory is greater than a preset threshold value, the trajectory is smoothed by using an interpolation algorithm; At the same time, the trajectory is secondarily adapted to generate an optimized expected training trajectory in combination with the training target of the patient.

6. The high-compliance impedance training method based on adaptive visual servoing control according to claim 1, wherein, The real-time trajectory includes a real-time position trajectory and a real-time attitude trajectory, and the method for calculating the deviation data includes: calculating the position deviation between the real-time position trajectory and the expected position trajectory by using the Euclidean distance; and calculating the attitude deviation between the real-time attitude trajectory and the expected attitude trajectory by using the angle difference.

7. The high-compliance impedance training method based on adaptive visual servoing control according to claim 6, wherein, The method for constructing the impedance adjustment model includes: designing the impedance adjustment model by using a fuzzy PID algorithm, taking the position deviation, the attitude deviation and the contact stress as input variables, and dividing each variable into a plurality of fuzzy subsets; setting a fuzzy rule base; and outputting the impedance adjustment amount by solving the fuzzy algorithm under the constraint conditions of minimum deviation, trajectory smoothness and non-sudden change of the contact stress.

8. The high-compliance impedance training method based on adaptive visual servoing control according to claim 4, wherein, The training method of the impedance adjustment includes: converting the impedance adjustment amount into a driving control instruction of the training mechanism, driving the end effector to move through a servo driver, monitoring the contact stress change rate in real time during the adjustment, triggering an emergency adjustment mechanism if the contact stress suddenly changes, instantaneously reducing the stiffness parameter and increasing the damping parameter, and pausing the trajectory advancement at the same time. After each training cycle is completed, the deviation data and the contact stress stability of the real-time trajectory and the expected training trajectory are compared, and if the deviation is greater than a preset threshold or the contact stress suddenly changes, the impedance adjustment amount is corrected based on the deviation data and the stress change data.

9. A system based on the method according to any one of claims 1 to 8, characterized in that, The system comprises: a data acquisition module configured to acquire historical and real-time position data and historical and real-time contact stress data; a position mapping module configured to process the data and generate stress position information and a position mapping relationship; a model construction module configured to construct a dynamic mapping model associated with muscle strength, contact stress, training trajectory and impedance parameters; a trajectory generation module configured to generate and optimize an expected training trajectory; a visual servoing and deviation calculation module configured to acquire data in real time, calculate the deviation data of the trajectory and monitor the contact stress change; and an impedance adjustment module configured to construct an impedance adjustment model and output adjustment instructions related to the impedance adjustment amount.

10. Application of the system of claim 9 in a rehabilitation training device.

Images