Self-adaptive position constraint exoskeleton rehabilitation robot on-demand auxiliary control system and method

Abstract

The invention discloses a self-adaptive position constraint exoskeleton rehabilitation robot on-demand auxiliary control system and method, and relates to the technical field of rehabilitation robot control. Comprising the following steps: collecting historical motion time sequence data of a patient, and recording a corresponding motion intention of the patient; establishing a deep learning network model, taking the historical motion time sequence data as an input sample, taking the corresponding motion intention as a label, and training the model; collecting real-time motion time sequence data of the patient, inputting the real-time motion time sequence data into the trained model to obtain real-time motion intention data of the patient, and calculating and generating target state parameters based on the real-time motion intention data; establishing a state parameter adaptive adjustment model, collecting a current state parameter, and analyzing and generating control output of the motion joint based on the generated target state parameter; and collecting real-time heart rate data of the patient, correcting the data in the control output, and sending a control instruction to the exoskeleton rehabilitation robot based on the corrected data.

Description

Technical Field

[0001] The present invention relates to the technical field of rehabilitation robot control, and in particular to an on-demand auxiliary control system and method for an exoskeleton rehabilitation robot with adaptive position constraints. Background Art

[0002] In recent years, with the aging society and the increasing number of patients with sports injuries, the demand in the field of rehabilitation therapy has increased significantly. As an advanced auxiliary treatment tool, exoskeleton rehabilitation robots have attracted widespread attention due to their potential in enhancing patients' strength, improving motor function and accelerating the rehabilitation process. These robots can provide adjustable motion assistance, help patients with rehabilitation training, and reduce the workload of medical staff to a certain extent. However, existing exoskeleton rehabilitation robots still face many technical challenges and limitations in practical applications.

[0003] First, existing exoskeleton rehabilitation robots often use fixed-parameter control strategies in the design of control systems. This method is difficult to adapt to the individual differences of different patients and their changing physiological states during the rehabilitation process. Factors such as the patient's muscle strength, range of joint motion, and motor control ability may vary due to individual differences, which means that the same control parameters may not be able to meet the needs of all patients, thereby affecting the rehabilitation effect. Moreover, traditional exoskeleton robots lack a real-time feedback mechanism during training, and patients cannot obtain timely physiological status monitoring and sports performance feedback, making it difficult to make effective adjustments based on their own conditions. This lack of personalized and adaptive control design not only makes patients feel frustrated during the rehabilitation process, but may also lead to inefficient training and prolong the rehabilitation cycle.

[0004] Secondly, existing technologies also have deficiencies in multi-sensor data fusion and dynamic control. Although some exoskeleton robots are equipped with multiple sensors, they often lack effective sensor data processing and fusion algorithms, resulting in insufficient real-time and accuracy of data. In this case, the patient's motion state assessment is not accurate enough, and the control system is difficult to make timely adjustments when responding to the patient's movements, which in turn affects the overall rehabilitation effect.

[0005] In the prior art, the publication number CN116386811A discloses an on-demand auxiliary control method and system for a rehabilitation robot with adaptive position constraints, the method comprising: step S1, collecting the current position, angular velocity and human-machine interaction torque of the human-machine interaction system; step S2, performing position constraint transformation according to the expected trajectory, position and angular velocity to obtain the position error conversion amount of the human-machine interaction system; step S3, linearly combining the position error conversion amount and the human-machine interaction torque to obtain the human motion performance function; step S4, using the human motion performance function as an input variable, designing a robot assistance level function with dead zone characteristics, saturation characteristics and continuous derivability; step S5, using the robot assistance level function as a weight factor of the human-machine interaction system controller, and designing a controller with position constraints. However, step S3 combines the position error conversion amount with the human-machine interaction torque in a linear combination manner to obtain the human motion performance function. However, in the actual human motion process, the relationship between the torque and the position error is often nonlinear and is affected by a variety of complex factors. Therefore, a simple linear combination may not accurately reflect the true performance of human motion, resulting in error accumulation or unsatisfactory control effect. At the same time, the patient's physiological parameters (such as heart rate, muscle fatigue, etc.) are not taken into account. Therefore, the effectiveness and practicality of the system are reduced by only general monitoring of the current position, angular velocity and human-computer interaction torque of the human-computer interaction system.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The object of the present invention is to provide an on-demand auxiliary control system and method for an exoskeleton rehabilitation robot with adaptive position constraints, so as to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An on-demand auxiliary control system for an exoskeleton rehabilitation robot with adaptive position constraints, specifically comprising:

[0010] A historical data acquisition module is used to set a comprehensive monitoring unit at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed.

[0011] The model training module is used to establish a deep learning network model, take the historical motion time series data collected by the comprehensive monitoring unit as input samples, and take the patient's corresponding motion intention as a label to train the deep learning network model and obtain a motion intention judgment model;

[0012] A target state analysis module is used to collect the patient's real-time motion timing data, input it into the motion intention judgment model for completing the training, obtain the patient's real-time motion intention data, and calculate the target state parameters based on the real-time motion intention data. The target state parameters include target joint angle, target motor speed, target joint torque and target applied force;

[0013] A control output generation module is used to establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate control outputs of the motion joints of the exoskeleton rehabilitation robot based on the generated target state parameters. The control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces.

[0014] The control output correction module is used to collect the patient's real-time heart rate data, correct the parameters in the control output according to the control output of the motion joint, obtain the control adaptive output data, and issue control instructions to the exoskeleton rehabilitation robot based on the control adaptive output data.

[0015] Furthermore, the patient's historical motion timing data is collected, wherein the motion timing data includes: three-axis acceleration data, electromyographic signals, vertical and horizontal forces applied by the exoskeleton to the ground, and image data; the patient's corresponding motion intention is recorded, wherein the motion intention includes action type, motion direction, and motion speed, wherein the action type includes walking, standing, and squatting, and the motion direction includes forward, backward, and sideways movement, and the action type and motion direction are encoded, wherein standing, walking, and squatting are encoded as 0, 1, and 2, respectively, and forward, backward, and sideways movement are calibrated as 3, 4, and 5, respectively.

[0016] Furthermore, a deep learning network model is established based on the combination of convolutional neural networks and recurrent neural networks. The deep learning network model has a 6-layer network structure, including an input layer, a convolutional layer, a pooling layer, a recurrent layer, a fully connected layer and an output layer, wherein the convolutional layer is a convolutional neural network layer, and the recurrent layer is a recurrent neural network layer. Finally, a motion intention judgment model is obtained, whose input is motion timing data and whose output is motion intention.

[0017] Furthermore, target state parameters are calculated based on the real-time motion intention data, and the target state parameters include joint angle, motor speed, joint torque and applied force, wherein the logic for calculating and generating the target state parameters is:

[0018] For the target joint angle θ(t), the formula for calculating the target joint angle θ(t) is:

[0019]

[0020] Where θ(t) is the target joint angle at time t, θ0 is the initial joint angle, and k θ is the scaling factor of the joint angle, d type (t) is the code of the action type at time t, d dir (t) is the code of the motion direction at time t, v des (t) represents the speed at time t;

[0021] For the target motor speed V(t), the formula for calculating the target motor speed V(t) is:

[0022]

[0023] Where V(t) represents the target motor speed at time t, k V is the adjustment factor of the motor speed, Δt is the time step, θ pre represents the joint angle of the previous time step;

[0024] The force F(t) applied to the target is calculated based on the formula:

[0025] F(t)=k f *(d type (t)+v des (t))

[0026] Where F(t) represents the applied force at time t, k V is the scaling factor of the applied force;

[0027] For the target joint torque τ(t), the formula for calculating the target joint torque τ(t) is:

[0028] τ(t)=J(θ(t)) T *(F(t)-c*V(t))

[0029] Where τ(t) represents the target joint torque at time t, J(θ(t)) represents the Jacobian matrix, c represents the damping coefficient, and T represents the transpose of the matrix;

[0030] Based on the calculated target state parameters, a target state vector is generated. The target state vector X ref (t)={θ(t),V(t),F(t),τ(t)}.

[0031] Furthermore, a state parameter adaptive adjustment model is established, the current state parameters are collected, and based on the generated target state parameters, the logic based on which the control output of the exoskeleton rehabilitation robot motion joint is analyzed is: the current state parameters are collected, and the current state vector is generated, wherein the current state vector X(t)={θ s (t),V s (t),F s (t),τ s (t)}, based on the current state vector X(t) and the target state vector X ref The calculation error vector of (t) is based on the formula:

[0032] e(t+p)=X ref (t+p)-X(t+p)

[0033] In the formula, e(t+p) represents the deviation between the current state and the target state of the system at time t+p, and p represents the number of future steps starting from the current time t.

[0034] Furthermore, a state parameter adaptive adjustment model is established, and the control output is obtained through the model predictive control algorithm according to the error vector, wherein the cost function Q of the model predictive control algorithm is constructed, and the expression of the cost function Q is:

[0035]

[0036] Where N is the set prediction time domain, which means the number of steps to be predicted in the future within the current time t, H is the state penalty matrix, R is the control penalty matrix, and u(t+p) is the control output vector at time t+p, which includes four-dimensional data of joint angle, motor speed, joint torque and applied force.

[0037] In each control cycle, based on the cost function Q, the formula for solving the optimized control output vector is:

[0038]

[0039] In the formula, st represents the constraint that after solving the optimized control output vector, the first control input vector u(t) = {θ u (t),V u (t),F u (t),τ u (t)} as the optimized control output vector, where θ u (t),V u (t),F u (t) and τ u(t) represent the output joint angle, output motor speed, output joint torque and output applied force respectively.

[0040] Furthermore, the patient's real-time heart rate data is collected, and the motor speed and applied force in the control output are corrected according to the control output vector of the motion joint. The formulas for the correction are:

[0041]

[0042] In the formula, F u (t) and V u (t) are the applied force and motor speed in the optimized control output vector, respectively, F adj (t) and V adj (t) are the corrected applied force and motor speed, HR represents the current heart rate, HR THR The heart rate threshold is set as HR max The maximum heart rate is obtained by combining the corrected applied force and motor speed with the control output vector to obtain the control adaptive output data, wherein the control adaptive output data Y(t)={θ Y (t),V adj (t),F adj (t),τ Y (t)}, where θ Y (t) = θ u (t), τ Y (t) = τ u (t); Based on the control adaptive output data, control instructions are issued to the exoskeleton rehabilitation robot.

[0043] The present invention also provides an on-demand auxiliary control method for an exoskeleton rehabilitation robot with adaptive position constraints, the on-demand auxiliary control method for an exoskeleton rehabilitation robot with adaptive position constraints is used to control the on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints, and the specific steps include:

[0044] A comprehensive monitoring unit is set up at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed.

[0045] A deep learning network model is established, the historical motion time series data collected by the integrated monitoring unit is used as input samples, and the corresponding motion intention of the patient is used as a label to train the deep learning network model to obtain a motion intention judgment model;

[0046] Collecting the patient's real-time motion timing data, inputting it into the completed training motion intention judgment model to obtain the patient's real-time motion intention data, and calculating the target state parameters based on the real-time motion intention data, wherein the target state parameters include the target joint angle, the target motor speed, the target joint torque and the target applied force;

[0047] Establishing a state parameter adaptive adjustment model, collecting current state parameters, and analyzing and generating control outputs of motion joints of the exoskeleton rehabilitation robot based on generated target state parameters, wherein the control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces;

[0048] The patient's real-time heart rate data is collected, and according to the control output of the motion joint, the parameters in the control output are modified to obtain the control adaptive output data. Based on the control adaptive output data, control instructions are issued to the exoskeleton rehabilitation robot.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] A deep learning network model is established, and the historical motion time series data collected by the comprehensive monitoring unit is used as input samples, and the corresponding motion intention of the patient is used as a label to train the deep learning network model; the real-time motion time series data of the patient is collected, and it is input into the motion intention judgment model that has been trained to obtain the real-time motion intention data of the patient, and the target state parameters are calculated and generated based on the real-time motion intention data; a state parameter adaptive adjustment model is established, the current state parameters are collected, and the control output of the motion joint of the exoskeleton rehabilitation robot is analyzed and generated based on the generated target state parameters; the real-time heart rate data of the patient is collected, and the motor speed and applied force in the control output are corrected, and control instructions are issued to the exoskeleton rehabilitation robot based on the corrected data. This scheme predicts the patient's motion intention through a deep learning network. The deep learning network is usually composed of a multi-layer network structure. This hierarchical structure enables the network to learn and express highly complex nonlinear relationships. At the same time, by real-time monitoring of the patient's heart rate data, the system can make appropriate corrections in the control output stage. When the patient's physiological load exceeds the safe range, the system can automatically reduce the motor speed or reduce the applied force, thereby preventing excessive fatigue or potential sports injuries. This physiological state monitoring and feedback mechanism improves the safety of training. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the overall structure of the device of the present invention;

[0052] Figure 2 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0054] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0055] Example:

[0056] See also Figure 1 , the present invention provides a technical solution:

[0057] An on-demand auxiliary control system for an exoskeleton rehabilitation robot with adaptive position constraints, specifically comprising:

[0058] A historical data acquisition module is used to set a comprehensive monitoring unit at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed.

[0059] The patient's historical motion timing data is collected, where the motion timing data includes: three-axis acceleration data, electromyographic signals, the vertical and horizontal forces applied by the exoskeleton to the ground, and image data; where the center of gravity of the patient when standing is taken as the origin, the patient's current right hand direction is the positive direction of the x-axis, the direction vertically pointing to the sky is the positive direction of the z-axis, and the y-axis direction is determined by the right-hand rule to establish a spatial coordinate system. The three-axis acceleration data refers to different acceleration data in the x-axis, y-axis and z-axis directions. The electromyographic signals (usually voltage signals) reflect the contraction and relaxation states of the muscles. The vertical and horizontal forces applied by the exoskeleton to the ground are used to analyze the patient's force application ability; the image data is used to capture the patient's dynamic posture and motion trajectory, and monitor the patient's movement patterns, including walking, standing, squatting, etc.

[0060] The patient's corresponding movement intention was recorded, and the movement intention included movement type, movement direction and movement speed, wherein the movement type included walking, standing and squatting, and the movement direction included forward, backward and sideways movement. The movement type and movement direction were encoded, and standing, walking and squatting were encoded as 0, 1 and 2 respectively, and forward, backward and sideways movement were marked as 3, 4 and 5 respectively, for the convenience of subsequent calculation.

[0061] The model training module is used to establish a deep learning network model, take the historical motion time series data collected by the comprehensive monitoring unit as input samples, and take the patient's corresponding motion intention as a label to train the deep learning network model and obtain a motion intention judgment model;

[0062] A deep learning network model is established based on the combination of convolutional neural network and recurrent neural network. The deep learning network model has a 6-layer network structure, including input layer, convolution layer, pooling layer, recurrent layer, fully connected layer and output layer, wherein the convolution layer is the convolutional neural network layer, which is mainly used to extract local features of time series data, and the recurrent layer is the recurrent neural network layer. The features extracted by the convolution layer will be input into the recurrent neural network to process the time dependency in the sequence. The model uses the ReLU activation function to introduce nonlinearity. Finally, a motion intention judgment model with motion time series data as input and motion intention as output is obtained.

[0063] A target state analysis module is used to collect the patient's real-time motion timing data, input it into the motion intention judgment model for completing the training, obtain the patient's real-time motion intention data, and calculate the target state parameters based on the real-time motion intention data. The target state parameters include target joint angle, target motor speed, target joint torque and target applied force;

[0064] The target state parameters are calculated and generated based on the real-time motion intention data. The target state parameters include joint angle, motor speed, joint torque and applied force. The logic for calculating and generating the target state parameters is as follows:

[0065] For the target joint angle θ(t), the formula for calculating the target joint angle θ(t) is:

[0066] θ(t)=θ0+k θ *(d type (t)+d dir (t)+v des (t))

[0067] Where θ(t) is the target joint angle at time t, θ0 is the initial joint angle, and k θ is the scaling factor of the joint angle, d type (t) is the code of the action type at time t, d dir(t) is the code of the motion direction at time t, v des (t) represents the movement speed at time t; the scaling factor k of the joint angle θ Used to adjust the response of the desired joint angle in the control system, usually used to improve control accuracy and smoothness, usually between 0.1 and 1.0.

[0068] For the target motor speed V(t), the formula for calculating the target motor speed V(t) is:

[0069]

[0070] Where V(t) represents the target motor speed at time t, k V is the adjustment factor of the motor speed, Δt is the time step, θ pre Indicates the joint angle of the previous time step; the adjustment factor k of the motor speed V Determines the degree of response of the motor speed to changes in joint angles, affecting the dynamic response of the robot. The general range is roughly between 0.5 and 2.0, depending on the response characteristics of the motor and the load.

[0071] The force F(t) applied to the target is calculated based on the formula:

[0072] F(t)=k f *(d type (t)+v des (t))

[0073] Where F(t) represents the applied force at time t, k f is the scaling factor of the applied force; the scaling factor k of the applied force f Used to adjust the response degree of the force applied to the joint, commonly used in force control systems. Usually between 0.1 and 1.0.

[0074] For the target joint torque τ(t), the formula for calculating the target joint torque τ(t) is:

[0075] τ(t)=J(θ(t)) T *(F(t)-c*V(t))

[0076] Where τ(t) represents the target joint torque at time t, J(θ(t)) represents the Jacobian matrix, c represents the damping coefficient, and T represents the transpose of the matrix;

[0077] Based on the calculated target state parameters, a target state vector is generated. The target state vector X ref (t)={θ(t),V(t),F(t),τ(t)}.

[0078] A control output generation module is used to establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate control outputs of the motion joints of the exoskeleton rehabilitation robot based on the generated target state parameters. The control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces.

[0079] Establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate the logic based on the control output of the exoskeleton rehabilitation robot motion joint based on the generated target state parameters: collect current state parameters and generate a current state vector, wherein the current state vector X(t)={θ s (t),V s (t),F s (t),τ s (t)}, based on the current state vector X(t) and the target state vector X ref The calculation error vector of (t) is based on the formula:

[0080] e(t+p)=X ref (t+p)-X(t+p)

[0081] In the formula, e(t+p) represents the deviation between the current state and the target state of the system at time t+p, and p represents the number of future steps starting from the current time t.

[0082] A state parameter adaptive adjustment model is established, and the control output is obtained through the model predictive control algorithm according to the error vector, wherein the cost function Q of the model predictive control algorithm is constructed, and the expression of the cost function Q is:

[0083]

[0084] Where N is the set prediction time domain, which means the number of steps to be predicted in the future within the current time t, H is the state penalty matrix, R is the control penalty matrix, and u(t+p) is the control output vector at time t+p, which includes four-dimensional data of joint angle, motor speed, joint torque and applied force.

[0085] The state penalty matrix H is usually a symmetric positive definite matrix, which represents the weight of each state variable. The weight of each state is determined by analyzing the impact of each state on the final performance in the system. The control penalty matrix R is also a symmetric positive definite matrix, which represents the weight of the control input. The two penalty matrices can be obtained by adjusting the values ​​of H and R through experiments and simulations, observing the system performance, or using empirical rules to set H and R. This is a common practice.

[0086] In each control cycle, based on the cost function Q, the formula for solving the optimized control output vector is:

[0087]

[0088] In the formula, st represents the constraint that after solving the optimized control output vector, the first control input vector u(t) = {θ u (t),V u (t),F u (t),τ u (t)} as the optimized control output vector, where θ u (t),V u (t),F u (t) and τ u (t) represent the output joint angle, output motor speed, output joint torque and output applied force respectively.

[0089] After solving the optimization problem, a series of control output vectors u(t), u(t+1),…, u(t+N-1) are obtained. Since MPC is a rolling horizon control strategy, only the first control output u(t) needs to be executed and then re-optimized in the next control cycle.

[0090] In the formula for solving the optimized control output vector, different restrictions can be added, such as the joint angle and the range of applied force. Restrictions can be added and the constraint range can be changed according to the conditions of different patients.

[0091] The control output correction module is used to collect the patient's real-time heart rate data, correct the parameters in the control output according to the control output of the motion joint, obtain the control adaptive output data, and issue control instructions to the exoskeleton rehabilitation robot based on the control adaptive output data.

[0092] The patient's real-time heart rate data is collected, and the motor speed and applied force in the control output are corrected according to the control output vector of the motion joint. The formulas for the correction are:

[0093]

[0094] In the formula, F u (t) and V u (t) are the applied force and motor speed in the optimized control output vector, respectively, F adj (t) and V adj (t) are the corrected applied force and motor speed, HR represents the current heart rate, HR THR The heart rate threshold is set as HR maxFor the maximum heart rate, when the current heart rate is greater than the set heart rate threshold, the applied force and motor speed are reduced to avoid excessive fatigue or potential sports injuries. When the heart rate is less than the set heart rate threshold, the applied force and motor speed can be increased to improve the efficiency of rehabilitation exercise.

[0095] The corrected applied force and motor speed are combined with the control output vector to obtain the control adaptive output data, wherein the control adaptive output data Y(t)={θ u (t),V adj (t),F adj (t),τ u (t)}, based on the control adaptive output data, control instructions are issued to the exoskeleton rehabilitation robot.

[0096] See also Figure 2 The present invention also provides an on-demand auxiliary control method for an exoskeleton rehabilitation robot with adaptive position constraints. The on-demand auxiliary control method for an exoskeleton rehabilitation robot with adaptive position constraints is used to control the on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints. The specific steps include:

[0097] Step 1: A comprehensive monitoring unit is set up at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed.

[0098] Step 2: Establish a deep learning network model, use the historical motion time series data collected by the comprehensive monitoring unit as input samples, and use the patient's corresponding motion intention as a label to train the deep learning network model and obtain a motion intention judgment model;

[0099] Step 3: Collect the patient's real-time motion timing data, input it into the completed training motion intention judgment model, obtain the patient's real-time motion intention data, and calculate the target state parameters based on the real-time motion intention data, wherein the target state parameters include target joint angle, target motor speed, target joint torque and target applied force;

[0100] Step 4: Establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate control outputs of the exoskeleton rehabilitation robot's motion joints based on the generated target state parameters. The control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces.

[0101] Step 5: Collect the patient's real-time heart rate data, modify the parameters in the control output according to the control output of the motion joint, obtain the control adaptive output data, and issue control instructions to the exoskeleton rehabilitation robot based on the control adaptive output data.

[0102] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0103] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An on-demand auxiliary control system for an exoskeleton rehabilitation robot with adaptive position constraints, characterized in that: Specifically include: A historical data acquisition module is used to set a comprehensive monitoring unit at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed. The model training module is used to establish a deep learning network model, take the historical motion time series data collected by the comprehensive monitoring unit as input samples, and take the patient's corresponding motion intention as a label to train the deep learning network model and obtain a motion intention judgment model; A target state analysis module is used to collect the patient's real-time motion timing data, input it into the motion intention judgment model for completing the training, obtain the patient's real-time motion intention data, and calculate the target state parameters based on the real-time motion intention data. The target state parameters include target joint angle, target motor speed, target joint torque and target applied force; A control output generation module is used to establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate control outputs of the motion joints of the exoskeleton rehabilitation robot based on the generated target state parameters. The control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces. The control output correction module is used to collect the patient's real-time heart rate data, correct the parameters in the control output according to the control output of the motion joint, obtain the control adaptive output data, and issue control instructions to the exoskeleton rehabilitation robot based on the control adaptive output data.

2. The on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints according to claim 1 is characterized in that: The patient's historical motion timing data was collected, where the motion timing data included: three-axis acceleration data, electromyographic signals, vertical and horizontal forces applied by the exoskeleton to the ground, and image data; the patient's corresponding motion intention was recorded, where the motion intention included action type, motion direction, and motion speed, where the action type included walking, standing, and squatting, and the motion direction included forward, backward, and sideways movement. The action type and motion direction were encoded, with standing, walking, and squatting encoded as 0, 1, and 2, respectively, and forward, backward, and sideways movement calibrated as 3, 4, and 5, respectively.

3. The on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints according to claim 2 is characterized in that: A deep learning network model is established based on the combination of convolutional neural network and recurrent neural network. The deep learning network model has a 6-layer network structure, including input layer, convolution layer, pooling layer, recurrent layer, fully connected layer and output layer, wherein the convolution layer is the convolutional neural network layer, and the recurrent layer is the recurrent neural network layer. Finally, a motion intention judgment model is obtained, whose input is motion timing data and whose output is motion intention.

4. The on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints according to claim 3 is characterized in that: The target state parameters are calculated and generated based on the real-time motion intention data. The target state parameters include joint angle, motor speed, joint torque and applied force. The logic for calculating and generating the target state parameters is as follows: For the target joint angle θ(t), the formula for calculating the target joint angle θ(t) is: θ(t)=θ0+k θ *(d type (t)+d air (t)+v des (t)) Where θ(t) is the target joint angle at time t, θ0 is the initial joint angle, and K θ is the scaling factor of the joint angle, d type (t) is the code of the action type at time t, d dir (t) is the code of the motion direction at time t, v des (t) represents the speed at time t; For the target motor speed V(t), the formula for calculating the target motor speed V(t) is: Where V(t) represents the target motor speed at time t, k V is the adjustment factor of the motor speed, Δt is the time step, θ pee represents the joint angle of the previous time step; The force F(t) applied to the target is calculated based on the formula: F(t)=k f *(d type (t)+v des (t)) Where F(t) represents the applied force at time t, k V is the scaling factor of the applied force; For the target joint torque τ(t), the formula for calculating the target joint torque τ(t) is: τ(t)=J(θ(t)) T *(F(t)-c*V(t)) Where τ(t) represents the target joint torque at time t, J(θ(t)) represents the Jacobian matrix, c represents the damping coefficient, and T represents the transpose of the matrix; Based on the calculated target state parameters, a target state vector is generated. The target state vector X ref (t)={θ(t),V(t),F(t),τ(t)}.

5. The on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints according to claim 4 is characterized in that: Establish a state parameter adaptive adjustment model, collect current state parameters, and analyze and generate the logic based on the control output of the exoskeleton rehabilitation robot motion joint based on the generated target state parameters: collect current state parameters and generate a current state vector, wherein the current state vector X(t)={θ s (t),V s (t),F s (t),τ s (t)}, based on the current state vector X(t) and the target state vector X ref The calculation error vector of (t) is based on the formula: e(t+p)=X ref (t+p)-X(t+p) In the formula, e(t+p) represents the deviation between the current state and the target state of the system at time t+p, and p represents the number of future steps starting from the current time t.

6. The on-demand auxiliary control system for the exoskeleton rehabilitation robot with adaptive position constraints according to claim 5 is characterized in that: A state parameter adaptive adjustment model is established, and the control output is obtained through the model predictive control algorithm according to the error vector, wherein the cost function Q of the model predictive control algorithm is constructed, and the expression of the cost function Q is: Where N is the set prediction time domain, which means the number of steps to be predicted in the future within the current time t, H is the state penalty matrix, R is the control penalty matrix, and u(t+p) is the control output vector at time t+p, which includes four-dimensional data of joint angle, motor speed, joint torque and applied force. In each control cycle, based on the cost function Q, the formula for solving the optimized control output vector is: In the formula, st represents the constraint that after solving the optimized control output vector, the first control input vector u(t) = {θ u (t),V u (t),F u (t),τ u (t)} as the optimized control output vector, where θ u (t),V u (t),F u (t) and τ u (t) represent the output joint angle, output motor speed, output joint torque and output applied force respectively.

7. The on-demand auxiliary control system of the exoskeleton rehabilitation robot with adaptive position constraints according to claim 6 is characterized in that: The patient's real-time heart rate data is collected, and the motor speed and applied force in the control output are corrected according to the control output vector of the motion joint. The formulas for the correction are: In the formula, F u (t) and V u (t) are the applied force and motor speed in the optimized control output vector, respectively, F adj (t) and V adj (t) are the corrected applied force and motor speed, HR represents the current heart rate, HR THR The heart rate threshold is set as HR max The maximum heart rate is obtained by combining the corrected applied force and motor speed with the control output vector to obtain the control adaptive output data, wherein the control adaptive output data Y(t)={θ Y (t),V adj (t),F adj (t),τ Y (t)}, where θ Y (t) = θ u (t), τ Y (t) = τ u (t); Based on the control adaptive output data, control instructions are issued to the exoskeleton rehabilitation robot.

8. An on-demand auxiliary control method for an exoskeleton rehabilitation robot with adaptive position constraints, characterized in that: The method for on-demand assistance control of an exoskeleton rehabilitation robot with adaptive position constraints is used to control the on-demand assistance control system of an exoskeleton rehabilitation robot with adaptive position constraints according to any one of claims 1 to 7, comprising: A comprehensive monitoring unit is set up at the motion joints of the exoskeleton rehabilitation robot to collect the patient's historical motion time series data and record the patient's corresponding motion intention. The comprehensive monitoring unit is composed of an acceleration sensor, an electromyography sensor, a force sensor, and a visual sensor. The motion intention includes the motion type, motion direction, and motion speed. A deep learning network model is established, the historical motion time series data collected by the integrated monitoring unit is used as input samples, and the corresponding motion intention of the patient is used as a label to train the deep learning network model to obtain a motion intention judgment model; Collecting the patient's real-time motion timing data, inputting it into the completed training motion intention judgment model to obtain the patient's real-time motion intention data, and calculating the target state parameters based on the real-time motion intention data, wherein the target state parameters include the target joint angle, the target motor speed, the target joint torque and the target applied force; Establishing a state parameter adaptive adjustment model, collecting current state parameters, and analyzing and generating control outputs of motion joints of the exoskeleton rehabilitation robot based on generated target state parameters, wherein the control outputs include output joint angles, output motor speeds, output joint torques, and output applied forces; The patient's real-time heart rate data is collected, and according to the control output of the motion joint, the parameters in the control output are modified to obtain the control adaptive output data. Based on the control adaptive output data, control instructions are issued to the exoskeleton rehabilitation robot.

Images