Adaptive position constrained rehabilitation robot on-demand assistance control method and system

Abstract

The application discloses a kind of rehabilitation robot on-demand auxiliary control method and system of self-adapting position constraint, belong to lower limb rehabilitation robot control field, method includes: step S1, the current position of man-machine interaction system, angular velocity and man-machine interaction torque are collected;Step S2, position constraint transformation is carried out according to desired trajectory, position and angular velocity, and the position error conversion quantity of man-machine interaction system is obtained;Step S3, position error conversion quantity and man-machine interaction torque are linearly combined, and the human motion performance function is obtained;Step S4, human motion performance function is used as input variable, and robot auxiliary level function with dead zone characteristics, saturation characteristics and continuous derivable is designed;Step S5, robot auxiliary level function is used as the weight factor of man-machine interaction system controller, and the controller with position constraint is designed.The application can realize seamless switching of man-machine interaction system in person leading mode, robot leading mode and man-machine cooperation mode.

Description

Technical Field

[0001] This invention belongs to the field of lower limb rehabilitation robot control technology, and more specifically, relates to an adaptive position-constrained on-demand auxiliary control method and system for rehabilitation robots. Background Technology

[0002] Patients paralyzed after a stroke require rehabilitation training to improve their muscle strength and motor coordination. In recent decades, robot-assisted therapy has offered unique advantages in improving rehabilitation efficiency and reducing the workload of healthcare workers by providing patients with repetitive training and appropriate assistance. Generally, training methods where rehabilitation robots guide the patient's limbs along a predetermined trajectory, completely ignoring the patient's initiative, are called passive training. This position-controlled assistive strategy helps to isolate the patient's joints and maintain their basic motor abilities during flaccid paralysis, but it does not stimulate the nervous system and has limited effect on restoring the patient's motor abilities. Therefore, rehabilitation medicine encourages patients to actively control their limbs to complete rehabilitation tasks after regaining minimal mobility.

[0003] Improving the compliance of rehabilitation robots is a first attempt at active training scenarios. However, the motion accuracy of impedance controllers is very limited. To address this issue, researchers proposed an "on-demand assistance" (AAN) strategy, in which the rehabilitation robot provides necessary assistance only when the patient is unable to complete the rehabilitation task, thereby encouraging the patient to increase muscle activity and stimulating neural remodeling.

[0004] Existing on-demand assistance technologies mainly fall into two categories. One involves designing a virtual tunnel: when a person can control limb movement within the tunnel, the robot provides no assistance; when the person's motor skills are insufficient to move limbs within the tunnel, the robot provides assistance to enable limb movement. However, this method suffers from different force outputs from the robot inside and outside the tunnel, causing discontinuous hard switching in the controller during transitions, potentially leading to instability in the control system. The other approach utilizes state feedback control incorporating an adaptive neural network with a forgetting factor. However, this neural network-based method is sensitive to disturbances, making it difficult to achieve stable and efficient on-demand assistance. Furthermore, existing technologies directly use combinations of system position errors, velocity errors, position, or human-robot interaction forces to design controllers, without quantifying the level of robot assistance. This not only makes the controller unsuitable for patients with varying degrees of injury but also complicates the design of "on-demand assistance" controllers to ensure seamless and safe robot assistance.

[0005] Furthermore, unlike position control, different robot motion modes in AAN are accompanied by different robot dynamics and impedances, posing a significant challenge to designing a unified AAN controller with position constraints. The methods mentioned above fail to consider the output position constraints of the rehabilitation robot, potentially leading to robot motion violating the kinematic constraints required during human rehabilitation and causing secondary injury to the patient. Summary of the Invention

[0006] To address the shortcomings and improvement needs of existing technologies, this invention provides an adaptive position-constrained on-demand assisted control method and system for rehabilitation robots, aiming to achieve seamless switching between human-led, robot-led, and human-machine collaborative modes in human-machine interaction systems.

[0007] To achieve the above objectives, according to one aspect of the present invention, an adaptive position-constrained on-demand assisted control method for a rehabilitation robot is provided, comprising:

[0008] Step S1: Collect the current position q and angular velocity of the human-computer interaction system. and human-computer interaction torque τ e ;

[0009] Step S2: Based on the desired trajectory q d The position q and the angular velocity Perform position constraint transformation to obtain the position error transformation amount z of the human-computer interaction system;

[0010] Step S3: Convert the position error z and the human-machine interaction torque τ e By performing a linear combination, we obtain the human motion performance function ψ;

[0011] Step S4: Using the human motion performance function ψ as the input variable, design a continuously differentiable robot assistance level function w(ψ) to characterize the degree of robot assistance to the human body under three strategies; the three strategies include human-dominated mode, human-machine collaborative mode, and robot-dominated mode, with the assistance level increasing sequentially, and the corresponding human motion performance function ψ values ​​range [0, r], respectively. di ), [r di ,r si ), [r si (,+∞), where r di r si These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively.

[0012] Step S5: Use the robot assistance level function w(ψ) as a weight factor of the human-computer interaction system controller, so that the controller can achieve on-demand assistance under different strategies.

[0013] Furthermore, the robot-assisted level function w(ψ) is:

[0014]

[0015] Where, ψ i r represents the human motion performance function with degree of freedom i. di r si These represent the boundary points of the dead zone and the saturation zone, respectively, δ di >0, δ si >0, k1 and k2 represent coefficients, and k1>0 and k2>0 respectively.

[0016] Further, in step S3, the square of the position error conversion amount z and the human-machine interaction torque τ are used. e The linear combination of the squares represents the human motion performance function ψ.

[0017] Further, in step S5, the controller is:

[0018]

[0019] Where τ represents the robot's output torque, The nonlinear function representing the state of a human-computer interaction system, q and These represent the current position and angular velocity of the human-computer interaction system, respectively. This represents the reference angular velocity vector of the human-computer interaction system. yes The derivative with respect to time, This represents the robot's dynamic compensation amount. express The estimate, k represents the relationship between the mass, length, moment of inertia, or degrees of freedom of a robot exoskeleton. s Let represent a positive definite diagonal matrix; s is the designed second sliding manifold, expressed as:

[0020]

[0021] In the formula, Υ is the error transformation matrix. Let λ represent the derivative of the position error transformation z with respect to time, I represent the identity matrix, and λ represent the position error transformation. z This represents a positive definite diagonal matrix.

[0022] Furthermore, the design steps of the controller include:

[0023] Step S501: Substitute the position error conversion amount z into the unconstrained human-computer interaction dynamics model to obtain the first sliding manifold s with position constraints. z ,in,

[0024] Step S502: Use the angular velocity currently output by the human-computer interaction system. First sliding manifold s z The second sliding manifold s is designed with robot-assisted level function.

[0025] Step S503: Replace the angular acceleration in the unconstrained human-computer interaction dynamics model with the second sliding manifold s. A new human-computer interaction dynamics model was obtained;

[0026] Step S504: Based on the new human-computer interaction dynamics model, the Lyapunov stability analysis method is used to obtain the controller.

[0027] Further, step S2 includes: using the desired trajectory q d The current output of the human-computer interaction system is the position q and angular velocity. Using the input as the position constraint, and taking the real-time output of the human knee joint angle from the human-computer interaction system as not exceeding the preset maximum joint angle, the position constraint transformation function is used to perform position constraint transformation to obtain the position error transformation amount z.

[0028] According to a second aspect of the present invention, an adaptive position-constrained on-demand assistance control system for a rehabilitation robot is provided, comprising:

[0029] The data acquisition module is used to acquire the current position q and angular velocity of the human-computer interaction system. and human-computer interaction torque τ e ;

[0030] The position error conversion quantity acquisition module is used to obtain the position error based on the desired trajectory q. d The position q and the angular velocity Perform position constraint transformation to obtain the position error transformation amount z of the human-computer interaction system;

[0031] The human motion performance function acquisition module is used to convert the position error conversion amount z and the human-computer interaction torque τ. e By performing a linear combination, we obtain the human motion performance function ψ;

[0032] The robot-assisted level function design module is used to design a continuously differentiable robot-assisted level function w(ψ) using the human motion performance function ψ as an input variable. This w(ψ) characterizes the degree of robot assistance to the human under three strategies: human-dominated mode, human-machine collaborative mode, and robot-dominated mode, with the degree of assistance increasing sequentially. The corresponding human motion performance function ψ takes values ​​in the range [0, r]. di ), [rdi ,r si ), [r si (,+∞), where r di r si These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively.

[0033] The controller design module is used to use the robot assistance level function w(ψ) as a weight factor of the human-computer interaction system controller, so that the controller can achieve on-demand assistance under different strategies.

[0034] Furthermore, the robot-assisted level function w(ψ) is:

[0035]

[0036] Where, ψ i r represents the human motion performance function with degree of freedom i. di r si These represent the boundary points of the dead zone and the saturation zone, respectively, δ di >0, δ si >0, k1 and k2 represent coefficients, and k1>0 and k2>0 respectively;

[0037] The controller is:

[0038]

[0039] Where τ represents the robot's output torque, The nonlinear function representing the state of a human-computer interaction system, q and These represent the current position and angular velocity of the human-computer interaction system, respectively. This represents the reference angular velocity vector of the human-computer interaction system. yes The derivative with respect to time, This represents the robot's dynamic compensation amount. express The estimate, k represents the relationship between the mass, length, moment of inertia, or degrees of freedom of a robot exoskeleton. s Let represent a positive definite diagonal matrix; s is the designed second sliding manifold, whose expression is:

[0040]

[0041] In the formula, Υ is the error transformation matrix. Let λ represent the derivative of the position error transformation z with respect to time, I represent the identity matrix, and λ represent the position error transformation. z This represents a positive definite diagonal matrix.

[0042] According to a third aspect of the invention, a lower limb rehabilitation assistive robot is provided, comprising: an exoskeleton, and a processor and a memory disposed on the exoskeleton, the memory storing a computer-executable program, which, when executed by the processor, causes the processor to perform the control method as described in any one of the first aspects.

[0043] According to a fourth aspect of the invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the control method as described in any one of the first aspects.

[0044] In summary, the above-described technical solutions conceived in this invention can achieve the following beneficial effects:

[0045] (1) This invention designs a robot assistance level function w(ψ) based on the human motion performance function ψ to characterize the degree of assistance provided by the human-computer interaction system to the human in human-dominated mode, human-computer collaborative mode, and robot-dominated mode. Under the three models, the corresponding human motion performance function ψ is divided into three different regions: dead zone [0, r di ), Activation Zone [r di ,r si ) and saturation region [r si Within these three regions (+∞), the level of assistance increases sequentially. The dead zone and saturation characteristics of the robot's assistance level function w(ψ) ensure that there exists a region allowing free movement during the switching process of the three on-demand assistance strategies, where the position error conversion amount z and the human-machine interaction torque τ are... e When the torque is relatively small (corresponding to human-dominated mode), excessive torque output from the robot joints should be avoided to prevent it from affecting the free movement of the human body; this applies to the position error conversion amount z and the human-machine interaction torque τ. e When the torque is large (corresponding to the robot-dominated mode), excessive torque output is avoided to prevent secondary injury to the patient. By utilizing the globally continuous differentiable characteristics of the robot-assisted horizontal function, a continuous transition between the human-dominated mode and the robot-dominated mode is established, ensuring seamless switching between different strategies during rehabilitation training and improving the continuity and safety of robot assistance.

[0046] (2) Furthermore, based on human movement performance, the present invention designs a robot assistance level function that meets the requirements, with a value between [0-1], which quantifies the degree of robot assistance to humans. It can automatically determine the degree of active participation of the human body in rehabilitation training based on the human body's movement performance, and generate a corresponding degree of assistance, thus laying the foundation for providing personalized rehabilitation training assistance for stroke patients with different motor abilities.

[0047] (3) Furthermore, this invention designs a new human motion performance function based on the position error after position constraint transformation and the human-computer interaction torque, which is used to characterize the human motion performance capability and facilitates the establishment of position constraints for the human-computer interaction system for on-demand auxiliary control. Compared with the prior art, which uses the inverse function of the sum of the squares of the original position error and the human-computer interaction torque to characterize human motion performance, this invention uses a new human motion performance function. As the system tracking error e approaches the set boundary, the corresponding transformed error also approaches infinity, laying the foundation for establishing position constraints for the human-computer system.

[0048] (4) Furthermore, the present invention constructs a novel sliding manifold (second sliding manifold), which includes the robot's position constraints and robot-assisted level function, bridging the gap between on-demand assistance strategy and position constraints; the second sliding manifold replaces the angular acceleration in the unconstrained human-machine interaction dynamics model, and Lyapunov stability analysis method is used to further design a unified controller with position constraints, which reflects the robot's position constraints through position error conversion, and determines the control under different assistance modes simultaneously through only one weight function (i.e., robot-assisted level function), so that the human-machine system can seamlessly switch between three modes to achieve "on-demand assistance" for patients.

[0049] (5) Furthermore, the present invention establishes position constraints by using the real-time output of the human knee joint angle of the human-computer interaction system to determine the error conversion amount after the position constraint transformation, and constructs the subsequent robot-assisted level function and unified controller based on the transformed error conversion amount. This ensures that the motion trajectory of the rehabilitation robot will not violate the preset position constraints throughout the process, thereby improving the safety of the patient's active rehabilitation training.

[0050] In summary, the method of this invention is an AAN human-computer interaction control method for stroke patients in active rehabilitation training that provides seamless assistance and positional constraints. It can use a unified controller to achieve seamless switching of the human-computer interaction system in human-led mode, robot-led mode and human-computer collaboration mode, so as to realize on-demand assistance under three different assistance strategies. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the adaptive position constraint-based on-demand assisted control method for a rehabilitation robot according to the present invention.

[0052] Figure 2 This is a block diagram illustrating the principle of the adaptive position constraint-based on-demand assisted control method for rehabilitation robots according to the present invention.

[0053] Figure 3 The robot-assisted level function w(ψ) designed for embodiments of the present invention i ) and its derivative w′(ψi (Diagram)

[0054] Figure 4 This is a region division diagram of the human-dominated mode, robot-dominated model, and human-machine collaboration model under the robot-assisted level function designed for embodiments of the present invention.

[0055] Figure 5 The robot-assisted level function designed for embodiments of the present invention shows the robot's tracking of the desired trajectory in five cases: {0, 0.25, 0.5, 0.75, 1}.

[0056] Figure 6 The figure shows the experimental results of healthy individuals participating in three modes of the human-computer interaction system controlled by the unified controller designed in this invention.

[0057] Figure 7 This represents the average value of the robot-assisted level function under different human performances. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0059] In this invention, the terms "first," "second," etc., used in the invention and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0060] like Figure 1 , Figure 2 As shown, the adaptive position-constrained on-demand assisted control method for rehabilitation robots of the present invention mainly includes the following steps:

[0061] Step S1: Collect the human-computer interaction torque τ between the user and the rehabilitation robot. e And the position q and angular velocity currently output by the human-computer interaction system.

[0062] Step S2: Based on the desired trajectory q d The current output of the human-computer interaction system is the position q and angular velocity. Perform position constraint transformation to obtain the position error transformation amount z of the human-computer interaction system;

[0063] Step S3: Convert the position error z and the human-machine interaction torque τ e By performing a linear combination, we obtain the human motion performance function ψ;

[0064] Step S4: Design the robot assistance level function w(ψ): Using the human motion performance function ψ as the input variable of the robot assistance level function, design a continuously differentiable robot assistance level function w(ψ) to characterize the degree of robot assistance to the human under three strategies; the three strategies include human-dominated mode, human-machine collaborative mode, and robot-dominated mode, with the degree of assistance increasing sequentially, and the corresponding human motion performance function ψ taking values ​​in the range [0, r], respectively. di ), [r di ,r si ), [r si (,+∞), where r di r si These represent the boundary points of the dead zone and the saturation zone, respectively. Specifically, the degree of assistance of the robot-assisted level function w(ψ) increases sequentially in the human-dominated mode, human-machine collaborative mode, and robot-dominated mode of the human-computer interaction system. The human motion performance function ψ is divided into a dead zone [0, r]. di ), saturation region [r si (,+∞) and the activation region [r] between the dead zone and the saturation zone. di ,r si ), where the dead zone [0, r di ) corresponds to the human-dominated mode in a human-computer interaction system, the saturation region [r si (+∞) corresponds to the region in the robot-dominated mode of the human-computer interaction system, the activation area [r di ,r si The region corresponds to the human-machine collaboration mode; among them, the robot-assisted level function has the property of being globally continuous and differentiable, that is, the human motion performance function ψ is continuously differentiable in the range of [0,+∞).

[0065] Step S5: Design a unified controller: Use the robot-assisted level function as the weight factor of the unified controller of the human-computer interaction system, and use the unified controller for online motion control. Finally, achieve seamless switching of the three strategies (human-led mode, robot-led mode, and human-machine collaborative mode) of the human-computer interaction system through the unified controller to realize on-demand assistance with position constraints.

[0066] Specifically, in step S2, the position constraint transformation is performed based on the real-time output of the human knee joint joint angle from the human-computer interaction system not exceeding the preset maximum range of motion (i.e., the preset maximum joint angle) as the position constraint, to obtain the position error transformation amount z of the human-computer interaction system, wherein the position constraint transformation function z i (t) is:

[0067]

[0068] Among them, ei (t) represents the tracking error of the human-computer interaction system, which is calculated based on the desired trajectory q. d The current output of the human-computer interaction system is the position q and angular velocity. The calculation yields μ i (t) represents the preset performance function, ε i It is a constant, and 0 < ε i <1, where i represents the degrees of freedom of the rehabilitation robot, and t represents the current time; the position constraint transformation function z i (t) Position error conversion amount z under reaction i degrees of freedom i ;

[0069] In the formula, e represents the tracking error of the human-computer interaction system. i (t) and the preset performance function μ i The following conditions are satisfied between (t):

[0070] -ε i μ i (t) <e i (t)<ε i μ i (t)

[0071] Specifically, in step S3, the position error conversion amount z and the human-machine interaction torque τ are used. e The linear combination of their squares represents the human motion performance function ψ:

[0072]

[0073] Where β1 and β2 represent the combination coefficients, n represents the total number of degrees of freedom, and z i τ represents the position error transformation amount under the i-th degree of freedom. ei This represents the human-computer interaction torque in the i-th degree of freedom.

[0074] When the user has sufficient mobility, they can track the desired trajectory without relying on a robot, which is the human-computer interaction torque τ. e Furthermore, the smaller position error conversion value z indicates that ψ is relatively small, meaning the user needs less assistance from the robot and can guide the robot to accurately track the desired trajectory. Conversely, a larger value of ψ means that the human cannot guide the human-machine system to track the desired trajectory and the human needs assistance from the robot system.

[0075] Specifically, in step S4, the robot-assisted level function is designed using the human motion performance function ψ as the input variable. The value of ψ ranges from [0, +∞), representing the human body's motion ability. This invention proposes a function that simultaneously possesses dead zone characteristics and saturation characteristics as the robot-assisted level function w(ψ). i ):

[0076]

[0077] Where, ψ i Let k1 and k2 represent the human motion performance function under degree of freedom i, where k1 and k2 are coefficients, both of which are positive numbers.

[0078] Based on the human-dominated region, robot-dominated region, and human-computer collaboration region in the human-computer interaction system, the human motion performance function ψ is approximately divided into three different regions: the dead zone range [0, r], where the human motion performance function ψ is approximately divided into three different regions. di ), saturation region range [r si [,+∞] and the range between the dead zone and the saturation zone [r di ,r si ):

[0079]

[0080] Where, r di r si These represent the boundary points of the dead zone and the saturation zone, respectively. Their values ​​are related to the patient's own motor ability. δ di and δ si All are positive numbers close to 0, meaning that within the set threshold, δ di and δ si In this embodiment, δ is 0. di and δ si Take 10 -5 In other embodiments, the appropriate option may be selected based on the specific application.

[0081] ψ i ∈[0,r di ) corresponds to the human-dominated region, in which the robot-assisted level function w(ψ) i This indicates the degree of assistance the robot provides to the human body in a human-dominated mode.

[0082] ψ i ∈[r si The range (+∞) corresponds to the robot's dominant region, within which the robot's auxiliary level function w(ψ) is... i This indicates the degree of assistance the robot provides to the human body in a robot-dominated mode.

[0083] ψ i ∈[r di ,r si This corresponds to the human-robot collaboration region, within which the robot's assisted level function w(ψ) is... i This indicates the degree of assistance the robot provides to the human body in a human-machine collaboration mode.

[0084] At the same time, it can be seen that w(ψ)i The robot is globally continuous and differentiable, so the switching between different areas is continuous, enabling the robot to continuously provide users with an appropriate level of assistance.

[0085] Based on the human motion performance function ψ, which is the position error conversion amount z and the human-computer interaction torque τ e In a linear combination, when the person wearing the exoskeleton has sufficiently strong motor ability, without the assistance of the robot, the user guides the human-machine system to follow the desired trajectory with an acceptable tracking error. Regardless of whether the user is willing, any increase in tracking error or human-machine interaction force may indicate a deterioration in human motor performance.

[0086] That is, when the user has good motion performance, the position error conversion amount z and the human-computer interaction torque τ e Smaller (position error conversion amount z and human-computer interaction torque τ) e Within a set first threshold, the linear combination of the parameters ensures that the robot's joints do not output excessive torque that could affect the human's free movement, even within an acceptable positional error transformation amount z. Based on this, the robot-assisted level function designed in this invention has dead-zone characteristics. Even with small tracking errors, the joints of the human-machine interaction system exhibit high compliance, allowing the human to control the behavior of the system.

[0087] When human motor performance is poor, the position error conversion amount z and the human-computer interaction torque τ e Larger (position error conversion amount z and human-computer interaction torque τ) e If the linear combination exceeds a set second threshold, the robot should be able to provide appropriate assistance based on the human's performance. Therefore, the robot's output torque should be limited to prevent excessive output torque from causing secondary injury to the patient. Based on this, the robot assistance level function designed in this invention has saturation characteristics. In this case, the robot dominates the behavior of the human-machine system, and the human-machine interaction torque is regarded as an external disturbance of the system. The first and second thresholds are selected according to the actual application.

[0088] Specifically, in step S5, the step of designing a unified controller by treating the robot's auxiliary level function as a weight factor of the controller includes:

[0089] Step S501: Substitute the position error transformation quantity z into the unconstrained human-robot interaction dynamics model to obtain the human-robot interaction dynamics model with position constraints:

[0090]

[0091] Among them, M z Let M represent the robot's inertia matrix after the position constraint transformation, and M... z =γT M(q)γ, where M(q) represents the inertia matrix of the human-computer interaction system, q represents the current output position of the human-computer interaction system, i.e., the joint angle output by the human-computer system; γ is the error transformation matrix, used to characterize the relationship between the derivative of the tracking error e and the derivative of the position error transformation z of the human-computer interaction system; T Let C be the transpose of γ. z Represents the centrifugal and Coriolis force matrices of the robot after position constraint transformation, where, D represents the robot's centrifugal and Coriolis force matrices before the position constraint transformation; z Let D represent the robot friction coefficient matrix after position constraint transformation, where D z =Υ T DΥ, where D represents the friction force matrix; g z This represents the gravity term after the position constraint transformation, and p r τ represents the product of the derivative of the set position constraints and the tracking error; g(q) represents the gravity term in human-machine dynamics; τ represents the robot's output torque. e The point on the parameter represents the first or second partial derivative of the parameter with respect to time.

[0092] Based on the above human-robot interaction dynamics model with position constraints, the first sliding manifold s with position constraints is obtained. z in, λ z Let represent a positive definite diagonal matrix; where the first sliding manifold is a linear combination of the position error transformation z and its derivative, representing the linear relationship between the position and angular velocity of the transformed human-computer interaction system;

[0093] The unconstrained human-robot interaction dynamics model is as follows:

[0094]

[0095] Step S502: Use the angular velocity currently output by the human-computer interaction system. First sliding manifold and robot-aided level function design for second sliding manifold s:

[0096]

[0097] Where w(ψ) represents the robot-assisted level function, I represents the angular velocity currently output by the human-computer interaction system.

[0098] The second sliding manifold s includes the robot's position constraints and robot-assisted level function, laying the foundation for establishing different human-computer interaction strategies in the future.

[0099] Step S503: Replace the angular acceleration in the unconstrained human-robot interaction dynamics model with the second sliding manifold s. A new human-robot interaction dynamics model is obtained:

[0100]

[0101] in, Let represent the reference angular velocity vector of the human-computer interaction system, and yes The derivative with respect to time, A nonlinear function representing the state of a human-computer interaction system, and This represents the relationship between the mass, length, moment of inertia, or degrees of freedom of the robot's exoskeleton, where D represents the friction matrix.

[0102] Step S504: Design a unified controller for the human-robot interaction system based on the new human-robot interaction dynamics model.

[0103]

[0104] in, This represents the robot's dynamic compensation amount, which is related to the degree of freedom. It compensates for different types of dynamics depending on the different values ​​of the human-assisted level function w(ψ). express The estimate, k s This represents a positive definite diagonal matrix.

[0105] When the degree of freedom is 1 m1 represents the mass of a link with one degree of freedom, l1 represents the length of the link, g represents the acceleration due to gravity, and I1 represents the moment of inertia of the rotating joint.

[0106] Specifically, in the human-dominated mode, the human-computer interaction torque τ e The controller is used to compensate for the gravity term in human-machine dynamics, so that the robot's output torque τ does not affect the human's active movement. That is, the robot's output torque τ is equal to the gravity term g(q) in human-machine dynamics.

[0107] In both robot-dominated and human-robot collaborative modes, a controller is designed based on a novel human-robot interaction dynamics model and Lyapunov stability analysis method, enabling the human-robot interaction system to be passive and thus allowing the trajectory of the human-robot interaction system to track the desired trajectory.

[0108] It can be verified that when the unified controller of the designed human-computer interaction system is substituted into the new human-robot interaction dynamics model, the human-computer interaction system exhibits passivity and the entire system is in a stable state.

[0109] Based on the unified controller of the human-computer interaction system designed above, in the human-dominated mode, w(ψ)=δ di , and δ di When the value is close to 0, the user has sufficient mobility, the human motion performance function ψ is small, and it can guide the robot joints to track the desired trajectory (i.e., the human-computer interaction torque τ). e The tracking error is relatively small (i.e., the position error conversion amount z is relatively small), which is determined by the human-machine interaction torque τ. e The controller drives the human-robot interaction system to move, compensating for the gravity term in human-robot dynamics to ensure that the robot's output torque τ does not affect the human's active movement. The new human-robot interaction dynamics model can be rewritten as:

[0110]

[0111] It can be seen that this formula expresses the movement of the human-machine system driven by human-machine interaction force, and this mode corresponds to the human-dominated mode.

[0112] In robot-dominated mode, w(ψ) approaches 1, indicating insufficient user mobility to guide the robot joints to follow the desired trajectory, resulting in significant tracking errors. In this mode, the human-machine interface controller is a purely passive auxiliary controller with position constraints. This controller treats the force exerted by the user's body on the robot joints as an external disturbance to the human-machine interface system, and forcibly drives the robot to follow the desired trajectory q based on the robot's output torque τ. d .

[0113] In robot-dominated mode, when the human's motor skills are weak, they cannot drive the robot's joints to track the desired trajectory. In this mode, the robot provides sufficient assistance to enable the human-machine system to track the desired trajectory, but there is a certain degree of error. When the human's motor skills are strong, the robot provides less assistance, and the system trajectory can still track the desired trajectory.

[0114] like Figure 3 As shown, the robot's auxiliary level function w(ψ) is set to k1 = 1 and k2 = 5. i ) and w(ψ i The derivative function w′(ψ) is corresponding to ) i (Diagram showing k1 and k2 related to human motor performance; when the human body has strong motor ability, r) di The value of r should be as large as possible; otherwise, siThe values ​​should be kept as small as possible. In other embodiments, different values ​​of k1 and k2 are selected based on human movement performance. It can be seen that the robot-assisted level function w(ψ) i It has global differentiability, dead zone property and saturation property.

[0115] Figure 4 This is a region partitioning diagram for the corresponding human-dominated mode, robot-dominated model, and human-machine collaboration model. The vertical axis represents the human-machine interaction torque τ under the i-th degree of freedom. ei The horizontal axis represents the robot's assisted level function w(ψ). i As can be seen, the human motion performance function ψ divides the human-dominated region, robot-dominated region, and human-machine collaborative region in the human-computer interaction system.

[0116] like Figure 5 As shown, according to the robot-assisted level function w(ψ) i By examining the robot's tracking performance under the five conditions {0, 0.25, 0.5, 0.75, 1}, it can be observed that the robot's tracking accuracy improves with the increase in the level of assistance. That is, the larger w(ψ) is, the better the tracking performance. At the same time, it also ensures that the robot operates within the preset position constraints.

[0117] Experiments were conducted based on the unified controller designed in this invention, and the experiments were carried out in three phases. In the first phase, the subject (user) was asked to imitate a patient with limb weakness, with the exoskeleton driving the person's lower leg movement for 20 seconds; then, the subject actively exerted force to drive the human-machine system's joints to track the desired trajectory for 20 seconds; finally, the subject withdrew the force on the limb and again imitated a patient with limb weakness, with the human-machine system guiding the limb movement for another 20 seconds. The experimental results are shown below. Figure 6 The figure reflects different stages, showing the current output position q of the human-computer interaction system based on the desired trajectory q. d The changes in the incremental air pressure ΔP of the antagonistic pneumatic muscle-driven joint over time, and the robot-assisted level function w(ψ) i The change of ) over time, where ΔP is related to the robot's output torque τ, through w(ψ) i Adjust the stiffness of the drive joint.

[0118] In the first phase, the results showed that the human-computer interaction system exhibited a certain tracking effect and a high level of robot assistance. Due to the periodic variation of the desired trajectory, the tracking error inevitably changes, causing the value of w(ψ) to continuously change. In this process, a smaller tracking error leads to a smaller w(ψ), which in turn reduces tracking performance and increases the tracking error. Conversely, a larger tracking error leads to a larger w(ψ), thus improving tracking performance. If the subject wishes to improve tracking performance and better complete the rehabilitation task, he / she must actively drive the human-computer system to the best of his / her ability, ensuring that the system's trajectory tracks the desired trajectory.

[0119] When the subject actively pushes the human-computer interaction system, the system automatically enters the active phase. It can be observed that initially, even healthy subjects struggle to accurately track the required trajectory with their joints. Therefore, the joint trajectory exhibits significant errors, resulting in a high level of robot assistance. After the subject adapts to the changes in joint angles, the joints track the required trajectory well, leading to a significant decrease in w(ψ) and the robot's output torque, as well as high joint compliance, allowing the subject's limbs to move freely.

[0120] The third stage exhibits a similar phenomenon to the first stage, indicating that the human-computer interaction system can continuously switch between different modes. The continuous variation of w(ψ) enables seamless robot assistance. Furthermore, this invention calculates the average w(ψ) for different stages. Figure 7 As shown, when the subject is in a passive state (i.e., the joints are not actively exerting force), the average w(ψ) is 0.6141, while it is 0.2136 in the active state. The average level of robot assistance decreased by 65.22%, indicating that the average w(ψ) can be used to quantify the robot's assistance to the human body when patients perform rehabilitation tasks.

[0121] This invention designs a robot-assisted level function based on the human body's motion performance function. The dead zone and saturation characteristics of the function divide the human motion performance function ψ into three distinct regions, each corresponding to one of the three assistance strategies of the human-computer interaction system. This provides corresponding robot assistance to the human body. The dead zone and saturation characteristics of the robot-assisted level function ensure that there exists a region allowing free movement during the switching between the three on-demand assistance strategies. This is because even healthy individuals cannot guarantee perfectly precise driving of the human-computer system to track the desired trajectory, considering the position error conversion amount z and the human-computer interaction torque τ. e When the torque is relatively small, the robot joints will not output excessive torque that would affect the free movement of the human body, giving the joints of the human-computer interaction system high compliance; the position error conversion amount z and the human-computer interaction torque τ eWhen the torque is large, the robot's output torque is limited, which avoids causing secondary injury to the patient due to excessive output torque. By utilizing the globally continuous differentiable characteristics of the robot-assisted horizontal function, a continuous transition between human-dominated mode and robot-dominated mode is established, ensuring seamless switching between different modes during rehabilitation training and improving the continuity and safety of robot assistance.

[0122] The robot-assisted level function designed in this invention is based on human motor performance, taking values ​​between [0-1], thus quantifying the degree of robot assistance to the human. Based on the human's motor performance, the function automatically determines the degree of active participation in rehabilitation training and generates a corresponding level of assistance. This lays the foundation for providing personalized rehabilitation training assistance for stroke patients with different motor abilities.

[0123] The positional error transformation value z of the human-computer interaction system reflects the positional constraints of the robot. Based on the positional error transformation value z, a second sliding manifold s is designed. This second sliding manifold s includes the positional constraints of the robot and the robot's auxiliary level function. It can ensure that the motion trajectory of the rehabilitation robot will not violate the pre-set positional constraints throughout the process, thus bridging the gap between the on-demand assistance strategy and the positional constraints and improving the safety of patients' active rehabilitation training.

[0124] A unified controller with position constraints designed based on the second sliding manifold s can simultaneously determine the control under different assistance modes using only a single weight function (i.e., the robot-assisted level function). This allows the human-machine system to seamlessly switch between three modes: human-led mode, robot-led mode, and human-machine collaborative mode, thus achieving "on-demand assistance" for patients.

[0125] This invention also provides an adaptive position-constrained on-demand assisted control system for a rehabilitation robot, mainly comprising:

[0126] The data acquisition module is used to acquire the current position q and angular velocity of the human-computer interaction system. and human-computer interaction torque τ e ;

[0127] The position error conversion quantity acquisition module is used to obtain the position error based on the desired trajectory q. d Position q and angular velocity Perform position constraint transformation to obtain the position error transformation amount z of the human-computer interaction system;

[0128] The human motion performance function acquisition module is used to convert the position error z and the human-computer interaction torque τ. e By performing a linear combination, we obtain the human motion performance function ψ;

[0129] The robot-assisted level function design module is used to design a continuously differentiable robot-assisted level function w(ψ) using the human motion performance function ψ as the input variable. This function characterizes the degree of robot assistance to the human under three strategies: human-dominated mode, human-robot collaborative mode, and robot-dominated mode, with the assistance level increasing sequentially. The corresponding human motion performance function ψ takes values ​​in the range [0, r]. di ), [r di ,r si ), [r si (,+∞), where r di r si These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively.

[0130] The controller design module is used to use the robot's assisted level function w(ψ) as a weight factor for the human-machine interaction system controller, enabling seamless switching between three on-demand assistance strategies with position constraints.

[0131] The robot-assisted level function w(ψ) and the form of the controller are described in the above embodiments.

[0132] For the specific implementation process of each module, please refer to the implementation steps of the adaptive position constraint rehabilitation robot on-demand assisted control method in the above embodiments.

[0133] The present invention also provides a lower limb rehabilitation assistive robot, comprising: an exoskeleton, and a processor and a memory disposed on the exoskeleton, the memory storing a computer-executable program, which, when executed by the processor, causes the processor to perform each implementation step of the adaptive position constraint rehabilitation robot on-demand assistive control method as described in the above embodiments.

[0134] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements each implementation step of the adaptive position constraint rehabilitation robot on-demand assisted control method as described in the above embodiments.

[0135] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for on-demand assisted control of a rehabilitation robot with adaptive position constraints, characterized in that, include: Step S1: Collect the current location of the human-computer interaction system. angular velocity and human-computer interaction torque ; Step S2: Based on the desired trajectory The location and the angular velocity Perform position constraint transformation to obtain the position error transformation amount of the human-computer interaction system. ; Step S3: Convert the position error into a quantity. and the human-computer interaction torque By performing a linear combination, we obtain the human motion performance function. ; Step S4: Using the human motion performance function As input variables, design a continuously differentiable robot-assisted level function. This is used to characterize the degree of robot assistance to the human body under three strategies: human-dominated mode, human-robot collaborative mode, and robot-dominated mode, with the degree of assistance increasing sequentially. The corresponding human motion performance functions are... The value ranges are respectively , , ,in, , These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively. Step S5: Adjust the robot-assisted level function. As a weighting factor in the human-computer interaction system controller, the controller enables on-demand assistance under different strategies; the controller is: in, This indicates the robot's output torque. A nonlinear function representing the state of a human-computer interaction system. and These represent the current position and angular velocity of the human-computer interaction system, respectively. This represents the reference angular velocity vector of the human-computer interaction system. yes The derivative with respect to time, This represents the robot's dynamic compensation amount. express The estimate, This represents the relationship between the mass, length, moment of inertia, or degrees of freedom of a robot exoskeleton. Represents a positive definite diagonal matrix; The expression for the second sliding manifold is: In the formula, Here is the error transformation matrix. Indicates position error conversion amount The derivative with respect to time, Represents the identity matrix. Represents a positive definite diagonal matrix; This represents the first sliding manifold with positional constraints.

2. The control method according to claim 1, characterized in that, The robot-assisted level function for: in, Degrees of freedom The human motion performance function under the following conditions , These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively. , , and Denote the coefficients respectively, and , .

3. The control method according to claim 1, characterized in that, In step S3, the position error conversion amount is used. The square of the sum of the human-computer interaction torques The linear combination of squares characterizes the human motion performance function. .

4. The control method according to claim 1, characterized in that, The design steps for the controller include: Step S501, convert the position error amount Substituting into the unconstrained human-computer interaction dynamics model, we obtain the first sliding manifold with positional constraints. ,in, ; Step S502: Use the angular velocity currently output by the human-computer interaction system. First sliding manifold The second sliding manifold is designed with robot-assisted level function. , ; Step S503: Use the second sliding manifold Replace angular acceleration in unconstrained human-computer interaction dynamics models A new human-computer interaction dynamics model was obtained; Step S504: Based on the new human-computer interaction dynamics model, the Lyapunov stability analysis method is used to obtain the controller.

5. The control method according to claim 4, characterized in that, Step S2 includes: using the desired trajectory The current output position of the human-computer interaction system and angular velocity Using the input as the basis, and taking the real-time output of the human knee joint angle from the human-computer interaction system as not exceeding a preset maximum joint angle as the position constraint, a position constraint transformation function is used to perform position constraint transformation to obtain the position error transformation amount. .

6. An adaptive position-constrained on-demand assistance control system for a rehabilitation robot, characterized in that, include: The data acquisition module is used to acquire the current position of the human-computer interaction system. angular velocity and human-computer interaction torque ; The position error conversion quantity acquisition module is used to obtain the position error based on the desired trajectory. The location and the angular velocity Perform position constraint transformation to obtain the position error transformation amount of the human-computer interaction system. ; The human motion performance function acquisition module is used to convert the position error into a quantity. and the human-computer interaction torque By performing a linear combination, we obtain the human motion performance function. ; The robot-assisted level function design module is used to design the human motion performance function. As input variables, design a continuously differentiable robot-assisted level function. This is used to characterize the degree of robot assistance to the human body under three strategies: human-dominated mode, human-robot collaborative mode, and robot-dominated mode, with the degree of assistance increasing sequentially. The corresponding human motion performance functions are... The value ranges are respectively , , ,in, , These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively. The controller design module is used to implement the robot's assisted level function. As a weighting factor in the human-computer interaction system controller, the controller enables on-demand assistance under different strategies; the controller is: in, This indicates the robot's output torque. A nonlinear function representing the state of a human-computer interaction system. and These represent the current position and angular velocity of the human-computer interaction system, respectively. This represents the reference angular velocity vector of the human-computer interaction system. yes The derivative with respect to time, This represents the robot's dynamic compensation amount. express The estimate, This represents the relationship between the mass, length, moment of inertia, or degrees of freedom of a robot exoskeleton. Represents a positive definite diagonal matrix; The expression for the second sliding manifold is: In the formula, Here is the error transformation matrix. Indicates position error conversion amount The derivative with respect to time, Represents the identity matrix. Represents a positive definite diagonal matrix; This represents the first sliding manifold with positional constraints.

7. The control system according to claim 6, characterized in that, The robot-assisted level function for: in, Degrees of freedom The human motion performance function under the following conditions , These represent the boundary points of the dead zone and the boundary points of the saturation zone, respectively. , , and Denote the coefficients respectively, and , .

8. A lower limb rehabilitation assistive robot, comprising: An exoskeleton, and a processor and a memory disposed on the exoskeleton, the memory storing a computer-executable program, characterized in that, when executed by the processor, the program causes the processor to perform the control method as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the control method as described in any one of claims 1-5.

Images