Arm exoskeleton robot control method and apparatus

Abstract

The application provides an arm exoskeleton robot control method and device, and relates to the technical field of wearable robot control.The method comprises the following steps: modeling a motion speed profile based on a ProMP model and motion trajectory data of an arm to obtain a predicted motion speed profile of the arm; the arm is worn with an exoskeleton robot; obtaining an expected motion trajectory of the arm based on the predicted motion speed profile, a predicted motion duration and a predicted motion range; inputting the expected motion trajectory, human-machine interaction force and actual motion trajectory into a model predictive controller to optimize impedance parameters; and inputting the optimized impedance parameters into a variable impedance controller to control the arm exoskeleton robot.The application enables the arm exoskeleton robot to closely follow the motion of the user's arm, improves the transparency of the exoskeleton robot and reduces the human-machine interaction torque.

Description

Technical Field

[0001] This invention relates to the field of wearable robot control technology, and in particular to a control method and device for an arm exoskeleton robot. Background Technology

[0002] Wearable exoskeletons have shown exciting potential in rehabilitation and human enhancement. Exoskeleton transparency is an indicator of the degree to which a wearable exoskeleton affects the user; the less impact the exoskeleton has on the user, the higher its transparency. Researchers recognize that improving exoskeleton transparency and enabling it to closely follow human movement are essential problems to be solved to improve the long-term usability of exoskeletons and promote their widespread use. In a no-load state, if the human-computer interaction force is zero, the exoskeleton is considered to have perfect transparency.

[0003] In an unloaded state, users need to adjust joint angles and body posture to meet subsequent force exertion requirements. During posture adjustment, existing arm exoskeleton robots typically sacrifice the speed of movement to achieve the adjustment process in an unloaded state. This affects the user's movement and force exertion habits, generates harmful human-machine interaction forces, leads to a significant increase in user energy consumption, causes an uncomfortable and fatigue-prone experience, and ultimately makes the assistive exoskeleton unusable for long-term use. Summary of the Invention

[0004] This invention provides a control method and device for an arm exoskeleton robot, which solves the problem of low transparency of arm exoskeleton robots in the prior art when there is no load, and effectively improves the transparency of arm exoskeleton robots and reduces human-machine interaction in the absence of load.

[0005] This invention provides a control system for an arm-shaped exoskeleton robot, comprising:

[0006] Based on the ProMP model and the motion trajectory data of the arm, the motion velocity profile is modeled to obtain the predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot.

[0007] Based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the desired motion trajectory of the arm is obtained;

[0008] The desired motion trajectory, human-computer interaction force, and actual motion trajectory are input into the model prediction controller to optimize the impedance parameters;

[0009] The optimized impedance parameters are input into the variable impedance controller to control the arm exoskeleton robot.

[0010] In some embodiments, the step of inputting the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters includes:

[0011] Based on human-computer interaction, a model predicts the cost function of the controller.

[0012] Substituting the discretized model based on state variables into the cost function yields the optimized cost function; the state variables include the error between the desired motion trajectory and the actual motion trajectory, the derivative of the error, and the human-computer interaction force;

[0013] The optimized cost function is solved to obtain the optimized impedance parameters.

[0014] In some embodiments, before obtaining the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the method further includes:

[0015] Based on the maximum acceleration, maximum velocity, and deceleration to half of the maximum velocity, the predicted motion velocity profile is segmented to determine the segmentation points of the predicted motion velocity profile.

[0016] The predicted motion duration is obtained based on the time taken to reach any segment point;

[0017] The predicted range of motion is obtained based on the activity model and the predicted duration of motion.

[0018] In some embodiments, the method further includes:

[0019] Adaptability parameters are determined based on actual movement time and theoretical running time; these adaptive parameters characterize the degree to which the user adapts to the exoskeleton robot.

[0020] The motion vitality model is scaled based on adaptive parameters.

[0021] In some embodiments, determining the adaptive parameters based on actual movement time and theoretical running time includes:

[0022] When the actual running time is less than or equal to the theoretical running time, the adaptive parameter is determined to be 1;

[0023] When the actual running time is greater than the theoretical running time, the adaptive parameter is determined to be: Where t0 represents the actual motion time, This represents the theoretical motion time.

[0024] In some embodiments, the method further includes:

[0025] The motion velocity profile model is updated based on the Bayesian optimization algorithm and the newly acquired motion velocity profile during the motion process.

[0026] The present invention also provides a control device for an arm exoskeleton robot, comprising:

[0027] The modeling module is used to model the motion velocity profile based on the ProMP model and the motion trajectory data of the arm, so as to obtain the predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot.

[0028] The acquisition module is used to obtain the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range.

[0029] The optimization module is used to input the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters;

[0030] The control module is used to input the optimized impedance parameters into the variable impedance controller to control the arm exoskeleton robot.

[0031] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the arm exoskeleton robot control method described above.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the arm exoskeleton robot control method as described above.

[0033] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the arm exoskeleton robot control method described above.

[0034] The control method and device for an arm exoskeleton robot provided by this invention establishes a motion velocity profile model of the arm based on the ProMP model, thereby obtaining a predicted motion velocity profile; based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the desired motion trajectory of the arm is obtained; based on the model predictive controller, impedance parameters are optimized online, and the impedance parameters are applied to the variable impedance controller to control the arm exoskeleton robot, so that the arm exoskeleton robot closely follows the user's arm movement, improves the transparency of the exoskeleton robot, reduces the human-machine interaction torque, reduces the harmful interaction force when the exoskeleton robot assists, and improves the long-term usability of the exoskeleton. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the neurobiological theory of the "minimum time-cost principle";

[0037] Figure 2 This is a schematic diagram of a sports vitality model;

[0038] Figure 3 This is one of the flowcharts illustrating the control method for the arm exoskeleton robot provided by the present invention;

[0039] Figure 4 This is a segmented schematic diagram of the motion velocity profile provided by the present invention;

[0040] Figure 5 This is the second flowchart illustrating the control method for the arm exoskeleton robot provided by the present invention;

[0041] Figure 6 This is a schematic diagram of the control device for the arm exoskeleton robot provided by the present invention;

[0042] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. It should be noted that, unless otherwise specified, the embodiments and features of the embodiments of this invention can be combined with each other. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0044] To facilitate a clearer understanding of the various embodiments of the present invention, some related technical knowledge will be introduced as follows.

[0045] (1) Neurobiological theory 1, the minimum time-cost principle

[0046] There are many goal-oriented movements in daily life, such as arm grasping movements and eye looking at an object. The speed or duration of these movements is called motor vitality.

[0047] Existing research suggests that the selection of motor activity originates in the basal ganglia of the brain, essentially representing the nervous system's judgment and selection of the value of different movement patterns. Studies indicate that motor activity tends to obtain the greatest reward in the shortest possible time and with the least possible movement cost. Since the reward for movement is usually constant, such as grasping an object, activity can be characterized by a trade-off between time cost and movement cost. One such trade-off model is the Minimum Time-Effort (MTE) principle.

[0048] Figure 1 This is a schematic diagram of the "minimum time-cost principle" in neurobiology, as shown below. Figure 1 As shown, the points marked with an asterisk are the points with the lowest total cost, which are the motion parameters that humans tend to choose.

[0049] In fact, the MTE principle provides a way to determine the vitality of a movement, that is, the relationship between the range of motion and the duration of movement from the arm joint to the point. By solving for the minimum total cost, the relationship between the range of motion and the duration of movement can be determined. In application, scholars have found that under natural and comfortable movement conditions, the relationship between the range of motion and the duration of movement of the arm joint-to-point trajectory can be fitted by an affine function. Figure 2 This is a schematic diagram of the fitting of the exercise vitality model, such as... Figure 2 As shown, by fitting the motion duration and range of motion data points, the resulting affine model is the motion vitality model.

[0050] (2) Neurobiological Theory 2, Bell Curve

[0051] There are two concepts: "motion velocity trajectory" and "motion velocity profile". The "motion velocity trajectory" is the complete elbow joint motion velocity data in the actual point-to-point motion, while the "motion velocity profile" is the normalized "motion velocity trajectory".

[0052] A consensus in arm movement research is that, in point-to-point movements under natural and comfortable conditions, the velocity profile of the arm joint is a smooth bell-shaped curve. Furthermore, for point-to-point movements under specific tasks, the velocity profile remains roughly constant for each individual movement, although individual variations exist in the shape of the velocity profile.

[0053] Furthermore, some research results indicate that exoskeletons with good transparency can maintain the bell-shaped velocity profile of the arm joints relatively unchanged. However, due to limitations in the performance of existing real-time motion trajectory prediction algorithms, the motion speed usually decreases after wearing an exoskeleton. This decrease in speed is similar to the motion pattern of the arm under load, therefore, it can be assumed that exoskeletons with insufficient transparency exert a similar load effect on the arm, which reduces the long-term usability of the exoskeleton.

[0054] Figure 3 This is one of the flowcharts illustrating the control method for the arm exoskeleton robot provided by the present invention, such as... Figure 3 As shown, the present invention provides a control method for an arm exoskeleton robot, comprising the following steps:

[0055] Step 310: Based on the ProMP model and the motion trajectory data of the arm, the motion velocity profile is modeled to obtain the predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot.

[0056] Specifically, the Probabilistic Movement Primitives (ProMP) model is used to model the motion velocity profile. On the one hand, the ProMP model can integrate the features of multiple sample motion velocity profiles; on the other hand, the ProMP model is a probabilistic model, which is beneficial for the design of the model update algorithm.

[0057] The ProMP model uses a Gaussian distribution to process the collected motion trajectory data τ={q t} t＝0,...,T Perform probabilistic modeling, q t Let q represent the motion state at time t, and T represent the sampling period. t and Forming a joint state y t , q t Differentiation with respect to time. Assume joint state. It must meet the following form:

[0058]

[0059] In the formula, y t Indicates the joint state, Φ t This represents the time-dependent basis function matrix, with the superscript T indicating transpose. Where, φ t This represents the vector consisting of the values ​​of the basis functions at time t. φ t The derivative with respect to time, where n represents the number of basis functions and θ represents the weights of the basis functions. ε yThis represents the zero-mean Gaussian noise term. Where, σ y This represents the variance of Gaussian noise.

[0060] Joint state y t Satisfying the mean is The variance is σ y Gaussian distribution:

[0061]

[0062] Assuming that the states at each time t are independent, the likelihood of the motion trajectory data can be calculated as follows:

[0063]

[0064] In the formula, p(τ|θ) represents the probability of generating trajectory τ under the current parameter θ, T represents the sampling period, and y t Indicates the joint state, Φ t This represents the time-dependent basis function matrix, with the superscript T indicating transpose. Where, φ t This represents the vector consisting of the values ​​of the basis functions at time t. φ t The derivative with respect to time, where n represents the number of basis functions and θ represents the weights of the basis functions. ε y This represents the zero-mean Gaussian noise term. Where, σ y This represents the variance of Gaussian noise.

[0065] Since there are multiple teaching trajectories, the concept of variance needs to be introduced to encode the differences in different teaching data. Therefore, it is assumed that the parameter θ follows a parameter... probability distribution

[0066] The likelihood of motion trajectory data can be obtained by integrating the marginal probabilities:

[0067]

[0068] In the formula, The hyperparameters representing the trajectory τ are: The probability distribution, This indicates that the parameter θ follows a hyperparameter of θ. The probability distribution is given by p(τ|θ), which represents the probability of generating trajectory τ under the current parameter θ.

[0069] The expressions for the basis functions are as follows:

[0070]

[0071] In the formula, the superscript G stands for "Gaussian", and z t c represents the basis function input (i.e., phase) at time t. i Let h represent the mean of the i-th Gaussian function, and h represent the bandwidth of the Gaussian function.

[0072] Assumption It follows a Gaussian distribution with mean θ0 and variance Σ0, i.e. The likelihood of motion trajectory data can be calculated as follows:

[0073]

[0074] In the formula, The hyperparameters representing the trajectory τ are: The probability distribution of y t Indicates the joint state, Φ t This represents the time-dependent basis function matrix, with the superscript T indicating transpose. Where, φ t This represents the vector consisting of the values ​​of the basis functions at time t. φ t The derivative with respect to time, where n represents the number of basis functions, and θ0 represents... The mean of the Gaussian distribution, Σ0 represents The variance, σ, of the Gaussian distribution follows y This represents the variance of Gaussian noise.

[0075] Through maximum likelihood estimation, we can obtain the motion trajectory data τ={q t} t＝0,...,T The parameters are estimated in the middle. This yields the predicted motion velocity profile.

[0076] Step 320: Based on the predicted motion speed profile, predicted motion duration, and predicted motion range, the desired motion trajectory of the arm is obtained.

[0077] Specifically, unlike the periodic movements of the lower limbs, arm movements are typically non-periodic, making it impossible to directly derive parameters for the next movement from the previous one. Obtaining the complete motion trajectory of the arm joint requires considering three factors: movement duration, range of motion, and velocity profile. If all three factors are known, multiplying the movement duration and range of motion by the normalized velocity profile will yield a complete motion trajectory.

[0078] Based on the summary of Neurobiology Theory 2, for a given user, the velocity profile of the elbow joint in point-to-point motion is a roughly constant bell-shaped curve. Therefore, we temporarily assume that the velocity profile of the elbow joint is constant. Based on this assumption, the complete arm joint motion trajectory can be determined simply by predicting the duration and range of motion of the elbow joint online.

[0079] The key to further simplifying the problem is the motion vitality model. Neurobiological theory 1 points out that the motion vitality model establishes the relationship between range of motion and duration of motion. With the help of the motion vitality model, the motion trajectory of the elbow joint can be predicted in real time simply by predicting the duration of motion online.

[0080] Therefore, based on the predicted motion velocity profile obtained in the previous step, combined with the predicted motion duration and the motion range predicted based on the motion vitality model, the expected motion trajectory of the arm can be obtained.

[0081] Step 330: Input the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters.

[0082] Specifically, after obtaining the desired motion trajectory of the arm, the subsequent task is to smoothly follow the desired motion trajectory and further reduce the human-computer interaction force, which usually requires the use of impedance control.

[0083] The desired motion trajectory, human-computer interaction force, and actual motion trajectory are input into the model predictive controller. The model predictive controller optimizes the impedance parameters online based on these parameters. Impedance parameters include mass, damping, and stiffness coefficient.

[0084] Step 340: Input the optimized impedance parameters into the variable impedance controller to control the arm exoskeleton robot.

[0085] Specifically, the optimized impedance parameters are input into the variable impedance controller, which outputs the optimized human-machine interaction force to the human-machine interaction system based on the optimized impedance parameters, thereby controlling the arm exoskeleton robot.

[0086] The control method for an arm exoskeleton robot provided by this invention establishes a motion velocity profile model of the arm based on the ProMP model, thereby obtaining a predicted motion velocity profile; based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the desired motion trajectory of the arm is obtained; based on the model predictive controller, impedance parameters are optimized online, and the impedance parameters are applied to the variable impedance controller to control the arm exoskeleton robot, so that the arm exoskeleton robot closely follows the user's arm movement, improves the transparency of the exoskeleton robot, reduces human-machine interaction torque, reduces harmful interaction forces when the exoskeleton robot assists, and improves the long-term usability of the exoskeleton.

[0087] In some embodiments, the desired motion trajectory, human-computer interaction force, and actual motion trajectory are input into the model predictive controller to optimize the impedance parameters, including:

[0088] Based on human-computer interaction, a model predicts the cost function of the controller.

[0089] Substituting the discretized model based on state variables into the cost function yields the optimized cost function; the state variables include the error between the expected trajectory and the actual trajectory, the derivative of the error, and the human-computer interaction force.

[0090] The optimized cost function is solved to obtain the optimized impedance parameters.

[0091] Specifically, impedance control compensates the controlled system (exoskeleton robot system) into the following second-order system:

[0092]

[0093] In the formula, e represents the desired trajectory q. d The error between the actual trajectory q and the actual trajectory q Let e ​​represent the derivative of e with respect to time. Let e ​​represent the second derivative of e with respect to time, M represent the desired mass, D represent the desired damping, K represent the desired stiffness coefficient, and f represent the human-computer interaction force.

[0094] Since this invention relates to only a single elbow joint, the above model can be written in the following form:

[0095]

[0096] In the formula, e represents the error between the desired trajectory and the actual trajectory. Let e ​​represent the derivative of e with respect to time. Let e ​​represent the second derivative of e with respect to time, m represent the expected mass in a single-joint application scenario, d represent the expected damping in a single-joint application scenario, k represent the expected stiffness coefficient in a single-joint application scenario, and f represent the human-computer interaction force.

[0097] Optimization methods based on variable impedance control typically approximate the above model as the following linear state-space model. This approximation assumes... ( The cost is the differential of the human-computer interaction force with respect to time. The expression for the linear state-space model is as follows:

[0098]

[0099] In the formula, ξ represents the state variable. Let ξ represent the derivative of ξ with respect to time, A represent the system matrix, b represent the control matrix, u represent the control input, m represent the desired mass in a single-joint application scenario, d represent the desired damping in a single-joint application scenario, k represent the desired stiffness coefficient in a single-joint application scenario, and e represent the error between the desired motion trajectory and the actual motion trajectory. It is the differential of e with respect to time, and f is the human-computer interaction force.

[0100] As can be seen, the input of the linear state-space model includes impedance parameters (m, d, and k) and state variables ξ. This makes it impossible for the impedance parameters to participate in the cost function as a separate variable in the optimization control based on this system.

[0101] To address this issue, a novel discretization model based on state variables is proposed. The expression for the state variable-based discretization model is shown below:

[0102]

[0103] in,

[0104]

[0105]

[0106] In the formula, ξ t+1 Let A represent the state variable at time t+1, T represent the sampling period, I represent the identity matrix, and ξ represent the system matrix. t Let represent the state variable at time t, b represent the control matrix, β represent the compensation vector generated by uniform acceleration discretization, and the superscript T denotes transpose. t m represents the control input at time t. t d represents the expected mass at time t in a single-joint application scenario. t k represents the expected damping at time t in a single-joint application scenario. t This represents the expected stiffness coefficient at time t in a single-joint application scenario.

[0107] Based on the above system, the human-computer interaction force is introduced into the cost function of the model predictive controller, and the expression of the cost function of the single-step model predictive controller is designed as follows:

[0108]

[0109] Limited by

[0110]

[0111] In the formula, J(u) t ) indicates that when the input is u t The cost function ξ at time t+1Let u represent the state variable at time t+1. t q represents the control input at time t. t and R t It is a symmetric positive definite matrix, the superscript T denotes transpose, W(f t f represents the weight matrix based on human-computer interaction force. t Let A represent the human-computer interaction force at time t, A represent the system matrix, T represent the sampling period, I represent the identity matrix, and ξ represent the interaction force at time t. t Let Ω represent the state variable at time t, b represent the control matrix, β represent the compensation vector generated by uniform acceleration discretization, and Ω represent the state variable at time t. u It is the set of input constraints.

[0112] Input constraint set Ω u The set represented is shown below:

[0113]

[0114] In the formula, m t Let m0 represent the expected mass at time t in a single-joint application scenario, and d represent a positive constant. t d represents the expected damping at time t in a single-joint application scenario. t,min d represents the minimum damping at time t in a single-joint application scenario. t,max k represents the maximum damping at time t in a single-joint application scenario. t k represents the expected stiffness coefficient at time t in a single-joint application scenario. t,min k represents the minimum stiffness coefficient at time t in a single-joint application scenario. t,max This represents the maximum stiffness coefficient at time t in a single-joint application scenario.

[0115] W(f t W(f) is used to further reduce human-computer interaction and improve the flexibility of the exoskeleton. t The expression for ) is as follows:

[0116]

[0117] In the formula, k1, k2, and k3 are positive constants. max(·) represents the maximum value algorithm, f min It is a positive constant, f t The force of human-computer interaction at time t represents the force at which the interaction occurs.

[0118] It can be seen that R t W(f t It is also a positive definite matrix. The greater the human-computer interaction force, the stronger the regularization term. For u t The stronger the constraint, the better. According to m... t =m0, the greater the human-computer interaction force, the more important the parameter k is.t and d t The smaller the size, the smoother the human-computer interaction.

[0119] Substituting the discretized model based on state variables into the cost function of the model predictive controller, we can obtain the optimized cost function:

[0120]

[0121] Limited by u t ∈Ω u ,in

[0122]

[0123] In the formula, J(u) t ) indicates that when the input is u t The cost function at time, u t This represents the control input at time t, where the superscript T indicates transpose, and Q... t and R t It is a symmetric positive definite matrix, where T represents the sampling period, I represents the identity matrix, and ξ t Let W(f) represent the state variable at time t, b represent the control matrix, β represent the compensation vector generated by uniform acceleration discretization, and W(f) represent the state variable at time t. t Ω represents the weight matrix based on human-computer interaction force. u It is the set of input constraints.

[0124] Perform u on the optimized cost function t By solving this problem, the optimized impedance parameters can be obtained.

[0125] The control method for an exoskeleton robot provided by this invention introduces human-machine interaction forces into the cost function of a model predictive controller, and then substitutes a discretized model based on state variables into the cost function to obtain an optimized cost function. Solving the optimized cost function yields optimized impedance parameters, which further improves the accuracy of the impedance parameters and is beneficial for improving the transparency of the exoskeleton robot in the future.

[0126] In some embodiments, before obtaining the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the method further includes:

[0127] Based on the maximum acceleration, maximum velocity, and deceleration to half of the maximum velocity, the predicted motion velocity profile is segmented, and the segmentation points of the predicted motion velocity profile are determined.

[0128] The predicted motion duration is obtained based on the time taken to reach any segment point;

[0129] Based on the exercise vitality model and predicted exercise duration, the predicted exercise range is obtained.

[0130] Specifically, an intuitive way to segment the predicted motion velocity profile is to divide it into four segments based on the maximum acceleration, maximum velocity, and maximum deceleration.

[0131] However, the designed point-to-point motion task involves grasping an object. Based on the actual data collected, it was found that multiple extreme points of deceleration may occur during the deceleration process, which is related to the visual feedback adjustment during arm deceleration. Therefore, the segment point corresponding to the maximum deceleration was replaced with the segment point corresponding to deceleration to half of the maximum speed. That is, the predicted motion velocity profile was segmented based on the maximum acceleration, maximum speed, and deceleration to half of the maximum speed, resulting in four segments.

[0132] Figure 4 This is a segmented schematic diagram of the motion velocity profile provided by the present invention, as shown below. Figure 4 As shown, the motion velocity profile is divided into four segments based on the maximum acceleration, maximum speed, and deceleration to half of the maximum speed. Stage 1 is the stage from 0 to the maximum acceleration, stage 2 is the stage from the maximum acceleration to the maximum speed, stage 3 is the stage from the maximum speed to 50% of the maximum speed, and stage 4 is the stage from 50% of the maximum speed to 0.

[0133] Since it is assumed that the shape of the velocity profile remains approximately constant, the duration of motion can be estimated based on the time taken to reach any segment point. Taking the point of maximum acceleration as an example, the expression for predicting the duration of motion is as follows:

[0134]

[0135] In the formula, T total t1 is the predicted motion duration, t1 is the actual motion duration to reach the point of maximum acceleration, and τ1 is the proportion of time to reach the point of maximum acceleration in the predicted motion velocity profile.

[0136] After obtaining the predicted exercise duration, inputting the predicted exercise duration into the exercise activity model will yield the predicted exercise range.

[0137] The control method for an arm exoskeleton robot provided by this invention segments the predicted motion velocity profile based on the maximum acceleration, maximum speed, and deceleration to half of the maximum speed. The predicted motion duration is obtained based on the time to reach any segment point. The predicted motion duration is then substituted into the motion vitality model to obtain the predicted motion range, which is beneficial for obtaining the desired motion trajectory of the arm in the future.

[0138] In some embodiments, the arm exoskeleton robot control method provided by the present invention further includes:

[0139] Based on actual movement time and theoretical running time, adaptive parameters are determined; these parameters characterize the degree to which the user adapts to the exoskeleton robot.

[0140] The motion activity model is scaled based on adaptive parameters.

[0141] Specifically, multiple studies have shown that human-computer interaction with exoskeleton robots involves an adaptation process. For subjects using an arm exoskeleton robot for the first time, the degree of adaptation is low, typically manifested in slow movement speed. Therefore, an algorithm is needed to promote user adaptation to the exoskeleton, with the goal of restoring the user's affected speed as quickly as possible under the constraints of a motion vitality model.

[0142] First, let's assume the affine model of athletic vitality is...

[0143] t vig =δ1d vig +δ2

[0144] In the formula, t vig It is the duration of exercise, d vig δ1 and δ2 are constants, representing the range of motion.

[0145] After a workout session, the actual range of motion and the actual duration of the workout can be obtained. Further, based on the aforementioned exercise vitality model and theoretical range of motion, the theoretical duration of the workout can be derived, as shown in the following expression:

[0146]

[0147] In the formula, d represents the theoretical motion time, d0 represents the theoretical motion range, and δ1 and δ2 are constants.

[0148] An adaptation parameter α is set to characterize the user's degree of adaptation to the exoskeleton robot. The closer the adaptation parameter α is to 1, the higher the user's degree of adaptation. The adaptation parameter is determined by the actual movement time and the theoretical running time.

[0149] The adaptive parameter α can be used as feedback to accelerate the user's adaptation process to the exoskeleton; that is, the adaptive parameter α is used to scale the motion vitality model. The expression for the scaled motion vitality model is shown below:

[0150]

[0151] In the formula, d represents the scaled motion duration. vig The range of motion is represented by δ1 and δ2, which are constants, and α represents the adaptive parameter.

[0152] Using scaled motion activity models for online trajectory prediction can accelerate the user's adaptation process to exoskeletons.

[0153] The control method for the exoskeleton robot provided by this invention scales the motion vitality model using an adaptive parameter α, which helps users adapt to the exoskeleton robot.

[0154] In some embodiments, determining the adaptation parameters based on actual motion time and theoretical running time includes:

[0155] When the actual running time is less than or equal to the theoretical running time, the adaptive parameter is set to 1;

[0156] When the actual running time is greater than the theoretical running time, the adaptation parameter is determined as follows: Where t0 represents the actual motion time, This represents the theoretical motion time.

[0157] Specifically, the expression for the fitness parameter α is as follows:

[0158]

[0159] In the formula, α represents the adaptation parameter, and t0 represents the actual motion time. d represents the theoretical motion time, d0 represents the theoretical motion range, and δ1 and δ2 are constants.

[0160] For first-time users of arm exoskeletons, if the actual movement time exceeds the theoretical running time, it indicates that the user's movement speed is slow. In this case, the adaptation parameter is... That is, the adaptive parameter α is greater than 1; if the actual movement time is less than or equal to the theoretical running time, it indicates that the user's movement speed is not slow, and the adaptive parameter α is equal to 1.

[0161] In some embodiments, the arm exoskeleton robot control method provided by the present invention further includes:

[0162] The motion velocity profile model is updated based on the Bayesian optimization algorithm and the newly acquired motion velocity profile during the motion process.

[0163] Specifically, while the introduction of the adaptive parameter α has accelerated the user's adaptation to the exoskeleton, it is also necessary to provide reverse adaptation, that is, for the exoskeleton to adapt to the user, so as to achieve bidirectional adaptation between the exoskeleton robot and the user.

[0164] In trajectory prediction, it is assumed that the motion velocity profile remains roughly constant. However, in actual point-to-point grasping motion, the motion velocity profile of the elbow joint is not constant. Furthermore, wearing an exoskeleton on the arm will affect the arm's movement habits, which will also lead to changes in the motion velocity profile.

[0165] To enable the exoskeleton to adapt to the subject, a method for online adaptive adjustment of the motion velocity profile based on newly acquired motion trajectory data is proposed. Considering that the modeled motion velocity profile is a probabilistic model, a Bayesian optimization algorithm can be used to update the motion velocity profile model based on newly acquired motion velocity profiles during the motion process.

[0166] The motion velocity profile is modeled using the ProMP model, and the motion velocity profile model can be expressed by the following formula:

[0167]

[0168] In the formula, It is the joint state trajectory of the motion profile of the complete motion cycle, ∈ y This represents a zero-mean Gaussian noise term, which follows a mean of 1 / 2. variance is Gaussian distribution, ∑ y It is a diagonal matrix, and all diagonal elements are... The variance of Gaussian noise is represented; Let be the Gaussian function matrix, where φ t This represents the vector consisting of the values ​​of the basis functions at time t. φ t The derivative with respect to time, where n represents the number of basis functions and T represents the sampling period. The basis function weights follow a Gaussian distribution with mean θ0 and variance ∑0.

[0169] Joint state trajectory of motion profile It follows the following Gaussian distribution:

[0170]

[0171] In the formula, This represents the joint state trajectory that yields the complete periodic motion profile under the current parameter θ. Assuming the motions are independent, the probability of each motion is given by Bayes' theorem:

[0172]

[0173] In the formula, This represents the velocity profile of m newly acquired complete motion cycles. Represents the profile of a known motion velocity. In the case of θ, the posterior probability of the parameter θ This indicates the velocity profile obtained under the current parameter θ. The probability, Let P(θ) represent the kth of the m newly acquired motion velocity profiles, where P(θ) represents the prior probability distribution of parameter θ, and a represents the first normalization parameter.

[0174] Substituting into a specific expression, we have:

[0175]

[0176] In the formula, a ′ Σ represents the second normalization parameter. y θ0 represents the inherent variance of human motion. The mean of the Gaussian distribution, Σ0 represents The variance of the Gaussian distribution it follows.

[0177] After simplification, we have:

[0178]

[0179] In the formula, a ″ Let represent the third normalization parameter. Since the posterior distribution of θ also follows a Gaussian distribution, let its mean be θ₁ and its variance be ∑₁. Then, after simplification, we have:

[0180]

[0181] The variance and mean of the posterior distribution satisfy the following relationship:

[0182]

[0183] In the formula, θ1 represents The posterior mean of a Gaussian distribution, where ∑1 represents... The posterior variance follows a Gaussian distribution, where the superscript T denotes the transpose, and θ0 represents... The prior mean of the Gaussian distribution, ∑0 represents... The prior variance of the Gaussian distribution.

[0184] The mean θ1 and variance ∑1 of the posterior distribution are the parameters of the new motion velocity profile.

[0185] To update the motion velocity profile model at the frequency specified in the formula, it can be set to update the motion velocity profile model once every 5 motion velocity profile acquisitions.

[0186] The control method for the exoskeleton robot provided by this invention updates the motion velocity profile model based on a Bayesian optimization algorithm and newly acquired motion velocity profiles during the motion process, which helps the exoskeleton adapt to the subject.

[0187] Figure 5This is the second flowchart illustrating the control method for the exoskeleton robot provided by the present invention, as shown below. Figure 5 As shown, the control method for the arm exoskeleton robot provided by the present invention includes:

[0188] The system acquires the user's motion velocity trajectory without wearing the exoskeleton and performs offline motion velocity profile modeling based on this trajectory. This offline modeled motion velocity profile serves as the basis for segmented prediction of arm motion trajectories after the user dons the exoskeleton. Furthermore, this velocity profile model incorporates human arm movement habits, and using it as a reference profile for exoskeleton movement can reduce changes in the user's arm movement habits caused by wearing the exoskeleton.

[0189] Motion velocity profile modeling is performed based on the ProMP model to obtain the predicted motion velocity profile. The predicted motion velocity profile is then segmented based on maximum acceleration, maximum velocity, and deceleration to half of the maximum velocity (using a profile model segmentation prediction algorithm) to obtain the predicted motion duration. The predicted motion duration is then input into the motion activity model to obtain the predicted motion range. Based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the desired motion trajectory is obtained.

[0190] Will (Desired trajectory differential with respect to time), q (actual trajectory) (The derivative of the actual trajectory with respect to time) The second derivative of the actual motion trajectory with respect to time and f (human-computer interaction force) are input into the model predictive controller, and the model predictive controller outputs k. t d t m t (Impedance parameters).

[0191] k t d t m t (Impedance parameters) and The input is a variable impedance controller, and the output of the variable impedance controller is a human-machine interaction force, which controls the human-machine interaction system.

[0192] In addition, the fitness parameter α is calculated based on the actual exercise time and the theoretical exercise time, and then the fitness model is scaled based on the fitness parameter α.

[0193] Based on the newly acquired motion velocity profiles from the human-computer interaction system, the motion velocity profile model is updated using a Bayesian optimization algorithm.

[0194] The control device for the exoskeleton robot provided by the present invention is described below. The control device for the exoskeleton robot described below can be referred to in correspondence with the control method for the exoskeleton robot described above.

[0195] Figure 6 This is a schematic diagram of the control device for the arm exoskeleton robot provided by the present invention, as shown below. Figure 6 As shown, the present invention provides a control device for an arm exoskeleton robot, comprising:

[0196] Modeling module 610 is used to model the motion velocity profile based on the ProMP model and the motion trajectory data of the arm to obtain the predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot.

[0197] The acquisition module 620 is used to obtain the desired motion trajectory of the arm based on the predicted motion velocity profile, the predicted motion duration, and the predicted motion range.

[0198] The optimization module 630 is used to input the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters;

[0199] The control module 640 is used to input the optimized impedance parameters into the variable impedance controller to control the arm exoskeleton robot.

[0200] In some embodiments, the optimization module 630 is specifically used for:

[0201] Based on human-computer interaction, a model predicts the cost function of the controller.

[0202] Substituting the discretized model based on state variables into the cost function yields the optimized cost function; the state variables include the error between the desired motion trajectory and the actual motion trajectory, the derivative of the error, and the human-computer interaction force;

[0203] The optimized cost function is solved to obtain the optimized impedance parameters.

[0204] In some embodiments, the apparatus further includes:

[0205] The segmentation module is used to segment the predicted motion velocity profile based on the maximum acceleration, maximum velocity, and deceleration to half of the maximum velocity, and to determine the segmentation points of the predicted motion velocity profile.

[0206] The first prediction module is used to obtain the predicted motion duration based on the time to reach any segment point;

[0207] The second prediction module is used to obtain the predicted range of motion based on the exercise vitality model and the predicted exercise duration.

[0208] In some embodiments, the apparatus further includes:

[0209] The determination module is used to determine the adaptation parameters based on the actual movement time and the theoretical running time; the adaptation parameters are used to characterize the user's degree of adaptation to the exoskeleton robot;

[0210] The scaling module is used to scale the motion vitality model based on adaptive parameters.

[0211] In some embodiments, the determining module is specifically used for:

[0212] When the actual running time is less than or equal to the theoretical running time, the adaptive parameter is determined to be 1;

[0213] When the actual running time is greater than the theoretical running time, the adaptive parameter is determined to be: Where t0 represents the actual motion time, This represents the theoretical motion time.

[0214] In some embodiments, the apparatus further includes:

[0215] The update module is used to update the motion velocity profile model based on the Bayesian optimization algorithm and the newly acquired motion velocity profile during the motion process.

[0216] It should be noted that the above-mentioned arm exoskeleton robot control device provided by the present invention can realize all the method steps implemented in the above method embodiments and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiments and the beneficial effects will not be described in detail.

[0217] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 can call logic instructions in the memory 730 to execute an arm exoskeleton robot control method. The method includes: modeling the motion velocity profile based on a ProMP model and the motion trajectory data of the arm to obtain a predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot; obtaining the desired motion trajectory of the arm based on the predicted motion velocity profile, the predicted motion duration, and the predicted motion range; inputting the desired motion trajectory, the human-machine interaction force, and the actual motion trajectory into a model prediction controller to optimize the impedance parameters; and inputting the optimized impedance parameters into a variable impedance controller to control the arm exoskeleton robot.

[0218] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0219] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the arm exoskeleton robot control method provided by the above methods. The method includes: modeling the motion velocity profile based on a ProMP model and motion trajectory data of the arm to obtain a predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot; obtaining the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range; inputting the desired motion trajectory, human-computer interaction force, and actual motion trajectory into a model prediction controller to optimize impedance parameters; and inputting the optimized impedance parameters into a variable impedance controller to control the arm exoskeleton robot.

[0220] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the arm exoskeleton robot control method provided by the above methods. The method includes: modeling a motion velocity profile based on a ProMP model and motion trajectory data of the arm to obtain a predicted motion velocity profile of the arm; the arm is equipped with an exoskeleton robot; obtaining a desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range; inputting the desired motion trajectory, human-computer interaction force, and actual motion trajectory into a model prediction controller to optimize impedance parameters; and inputting the optimized impedance parameters into a variable impedance controller to control the arm exoskeleton robot.

[0221] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0222] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0223] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0224] It should be further noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0225] In this invention, "at least one" means one or more, and "more than one" means two or more. The terms "first," "second," "third," "fourth," etc. (if present) in this invention are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0226] In embodiments of the present invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0227] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method for an arm-shaped exoskeleton robot, characterized in that, include: Based on the ProMP model and the arm's motion trajectory data, the motion velocity profile is modeled to obtain the predicted motion velocity profile of the arm. The arm is equipped with an exoskeleton robot; Based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the desired motion trajectory of the arm is obtained; The desired motion trajectory, human-computer interaction force, and actual motion trajectory are input into the model prediction controller to optimize the impedance parameters; The optimized impedance parameters are input into the variable impedance controller to control the arm exoskeleton robot; The step of inputting the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters includes: Based on human-computer interaction, a model predicts the cost function of the controller. Substituting the discretized model based on state variables into the cost function yields the optimized cost function; the state variables include the error between the desired motion trajectory and the actual motion trajectory, the derivative of the error, and the human-computer interaction force; The optimized cost function is solved to obtain the optimized impedance parameters; Before obtaining the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range, the method further includes: Based on the maximum acceleration, maximum velocity, and deceleration to half of the maximum velocity, the predicted motion velocity profile is segmented to determine the segmentation points of the predicted motion velocity profile. The predicted motion duration is obtained based on the time taken to reach any segment point; Based on the exercise vitality model and the predicted exercise duration, the predicted exercise range is obtained; The method further includes: Adaptability parameters are determined based on actual movement time and theoretical running time; these adaptive parameters characterize the degree to which the user adapts to the exoskeleton robot. The motion vitality model is scaled based on adaptive parameters.

2. The control method for the arm exoskeleton robot according to claim 1, characterized in that, The determination of adaptive parameters based on actual movement time and theoretical running time includes: When the actual running time is less than or equal to the theoretical running time, the adaptive parameter is determined to be 1; When the actual running time is greater than the theoretical running time, the adaptive parameter is determined to be: ,in, Indicates the actual time spent in motion. This represents the theoretical motion time.

3. The control method for the arm exoskeleton robot according to claim 1, characterized in that, The method further includes: The motion velocity profile model is updated based on the Bayesian optimization algorithm and the newly acquired motion velocity profile during the motion process.

4. A control device for an arm exoskeleton robot based on the control method for an arm exoskeleton robot according to any one of claims 1-3, characterized in that, include: The modeling module is used to model the motion velocity profile based on the ProMP model and the motion trajectory data of the arm, so as to obtain the predicted motion velocity profile of the arm. The arm is equipped with an exoskeleton robot; The acquisition module is used to obtain the desired motion trajectory of the arm based on the predicted motion velocity profile, predicted motion duration, and predicted motion range. The optimization module is used to input the desired motion trajectory, human-computer interaction force, and actual motion trajectory into the model prediction controller to optimize the impedance parameters; The control module is used to input the optimized impedance parameters into the variable impedance controller to control the arm exoskeleton robot.

5. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the arm exoskeleton robot control method as described in any one of claims 1 to 3.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the arm exoskeleton robot control method as described in any one of claims 1 to 3.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the arm exoskeleton robot control method as described in any one of claims 1 to 3.

Images