Teleoperation robot auxiliary system based on MYO bracelet and virtual clamp and control method

Abstract

The invention discloses a teleoperation robot auxiliary system based on an MYO bracelet and a virtual clamp and a control method. The system comprises a teleoperation module, a track learning module, an arm rigidity extraction module and a virtual clamp module. The arm stiffness extraction module obtains forearm electromyographic signals and muscle stiffness information based on the MYO bracelet; the teleoperation module is used for realizing robot teleoperation of space matching between the master end and the slave end based on TouchX equipment; the trajectory learning module obtains muscle stiffness information, robot state information, task information and a learning expert trajectory data set, and generates an optimal reference trajectory; the virtual fixture module generates auxiliary virtual fixture guiding force based on the optimal reference trajectory and the robot state, operation tremor suppression and environment interference compensation are achieved through adaptive adjustment of trajectory gravitation weight and muscle stiffness value, and the operation tremor suppression and environment interference compensation are fed back to an operator through Touch X equipment. According to the invention, the precision and efficiency of teleoperation can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of teleoperation robots, and particularly relates to a teleoperation robot assistance system and control method based on a MYO bracelet and virtual fixtures. Background Art

[0002] With the development of fields such as industrial automation, precision assembly, and dangerous operations, robots have become important tools to replace humans. Especially in complex and high-risk working environments, teleoperation robots can help operators remotely complete high-precision tasks. However, traditional teleoperation systems usually rely on physical input devices such as joysticks or buttons. This method has problems such as large operation errors, a steep learning curve, low precision, being easily affected by hand tremors and environmental interference, and lacking real-time feedback, making it difficult to adapt to changes in dynamic environments.

[0003] Currently, for problems such as environmental interference, operator inexperience, and hand tremors during the teleoperation of robots, this makes it difficult for teleoperation robots to complete tasks with high precision requirements. And when there are obstacles in the environment, collisions are likely to occur due to perturbations.

[0004] For example, in the teleoperation robot grinding and cutting integrated processing system and method based on virtual fixtures with the patent publication number CN113386142A, the system obtains the residual features of the casting and the relative pose of the end effector of the slave robot through a vision system, generates a virtual fixture suitable for processing the residual features, and constrains and guides the end movement of the active manipulation robot through the artificial repulsive force field applied by the virtual fixture, so as to coordinate the movement of the slave processing robot to complete the processing task. However, it relies on a relatively complex vision system and virtual fixture control, requires high computing resources to process the adjustment of the virtual fixture and force field control, and does not provide feedback to the operator to guide the operator's teleoperation.

[0005] For example, in a teleoperation robot grinding method based on force sense virtual fixture assistance with the patent publication number CN115284317A, this method creates a virtual fixture suitable for different casting grinding requirements through a virtual environment display module using CHAI3D software. After the virtual fixture is created, its pose is adjusted to ensure that there is no collision between the virtual fixture and the end of the master hand, and the operator is guided to operate by calculating the feedback force. However, this method does not identify the operator's intention and does not handle sudden external interferences. Summary of the Invention

[0006] To overcome the defects and deficiencies of the prior art, the present invention provides a teleoperation robot assistance system and control method based on a MYO bracelet and a virtual fixture. The present invention collects expert trajectories, establishes a trajectory library based on the expert trajectories, and performs teleoperation on the robot based on a TouchX device. At the beginning of the teleoperation, the MYO bracelet is used to obtain the muscle activity and stiffness of the operator to represent human intentions, and the muscle stiffness information, robot state information, and task information are input into the trajectory learning module together. The dynamic system in the trajectory learning module will generate an optimal reference trajectory. During the teleoperation process, the virtual fixture module will combine the optimal reference trajectory and the robot state to generate an auxiliary virtual fixture guiding force, which is fed back to the operator through the TouchX device to pull the operator's hand for teleoperation. The present invention can effectively improve the accuracy and efficiency of teleoperation.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] The present invention provides a teleoperation robot assistance system based on a MYO bracelet and a virtual fixture, including: a teleoperation module, a trajectory learning module, an arm stiffness extraction module, and a virtual fixture module;

[0009] The arm stiffness extraction module obtains forearm electromyogram signals and muscle stiffness information based on the MYO bracelet;

[0010] The teleoperation module realizes the teleoperation of the master-slave end space-matched robot based on the TouchX device;

[0011] The trajectory learning module obtains muscle stiffness information, robot state information, and task information, learns the expert trajectory data set, and generates an optimal reference trajectory;

[0012] The virtual fixture module generates an auxiliary virtual fixture guiding force based on the optimal reference trajectory and the robot state, and feeds it back to the operator through the TouchX device.

[0013] The present invention also provides a control method for a teleoperation robot assistance system based on a MYO bracelet and a virtual fixture, including the following steps:

[0014] Collect expert trajectory data and environmental obstacle information, complete the robot task based on the teleoperation module, and generate a demonstration data set;

[0015] Construct a trajectory learning dynamic system model in the trajectory learning module and train it based on the expert trajectory data set;

[0016] Remote control the slave robot based on the TouchX device to perform master-slave end space matching;

[0017] Obtain forearm electromyogram signals and muscle stiffness information based on the MYO bracelet;

[0018] Input the muscle stiffness information, robot state information, and task information into the trained trajectory learning dynamic system model to obtain the optimal reference trajectory;

[0019] Calculate the virtual fixture guiding force value based on the current real-time position of the robot's end, the optimal reference trajectory, and the muscle stiffness information.

[0020] As a preferred technical solution, the demonstration dataset is expressed as:

[0021] where x t,n is the position state of the robot at time t in the nth demonstration, is the velocity state, x obj is the obstacle information in the environment, A t,n is the expert arm stiffness value.

[0022] As a preferred technical solution, the steps for generating the demonstration dataset include:

[0023] The position state x t,n and velocity state of the robot are obtained by remotely operating the slave robot through the master haptic device;

[0024] Calibrate the coordinate systems of the master and slave robots, define the initial pose X l0 =[x l0 , y l0 , z l0 of the master haptic device and the initial pose X f0 =[x f0 , y f0 , z f0 of the slave manipulator. By constructing the scaling matrix S = diag(S x , S y , S z ), make the master-slave displacements satisfy X f = X f0 + S·(X l - X l0 ), perform real-time pose mapping on the master and slave robots, collect the real-time pose X l (t) = [x l (t), y l (t), z l (t)] of the haptic device, and calculate the target pose of the slave end and send it to the robot for execution;

[0025]

[0026] Solve for the joint angles through inverse kinematics and send them to the slave manipulator to obtain the position state x t,n, the time derivative of the position is obtained to get the velocity state

[0027] Obtain the electromyogram signal EMG collected from the i-th channel in the MYO bracelet i , calculate the sum of the amplitudes of the electromyogram signals as:

[0028]

[0029] Perform sliding window filtering on the sum of the amplitudes of the electromyogram signals to obtain the envelope E:

[0030]

[0031] where, w win is the sliding window length;

[0032] The muscle stiffness value of the expert's arm is expressed as:

[0033]

[0034] where, α is the non-linear coefficient.

[0035] As a preferred technical solution, construct the trajectory learning dynamic system model in the trajectory learning module, specifically expressed as:

[0036]

[0037] where, represents the energy function, represents the modulation matrix, x represents the robot state input, and z represents the vector composed of the robot state input, obstacle information, the energy value at state x, and the muscle stiffness A.

[0038] As a preferred technical solution, the energy function is specifically expressed as:

[0039]

[0040] where, the energy function V(x) is composed of the learnable neural network P1(x) and the quadratic form P2(x). The P1(x) term is the weighted sum of the weight parameter w and the feature extraction function f k (x), g(x) maps the input from the dx dimension to the dh dimension, c(x) performs non-linear feature expansion on the state x, and the parameters a k , b k are the feature parameters, and the activation function is the tanh function, and the parameter κ is the quadratic term coefficient;

[0041] The vector z is expressed as:

[0042]

[0043] Among them, represents the trajectory learning dynamic system model, and x obj represents the obstacle information, represents the energy value at state x, and A represents the muscle stiffness;

[0044] Multiply the vector z by a degree factor according to the distance between the current end position and the obstacle and the degree of arm muscle tension:

[0045]

[0046]

[0047] Among them, the parameters α1 and α2 determine the smoothness of the tracking and obstacle avoidance switching process, β represents the safety margin, and d represents the distance from the current robot position to the obstacle;

[0048] The vector z is input into a three-layer fully connected neural network, and the output is a six-dimensional vector:

[0049] n = Net(z; Θ N )

[0050] Based on the six-dimensional vector n, a lower triangular matrix is formed, which is expressed as:

[0051]

[0052] Based on the Cholesky formula, a positive definite matrix is formed

[0053] As a preferred technical solution, the slave robot is remotely controlled based on the TouchX device to perform master-slave end space matching:

[0054]

[0055] Among them, X l0 、X f0 are the end Cartesian position coordinates of Touch X and the slave robot at the initial state, X l (t) is the real-time position of the master robot, and X f (t) is the real-time position of the slave robot;

[0056] The real-time position of the master robot is obtained, and the expected position of the end of the slave robot is obtained through matching calculation and sent to the robot for execution to achieve teleoperation.

[0057] As a preferred technical solution, according to the current end real-time position, the optimal reference trajectory and the muscle stiffness information of the robot, the virtual fixture guiding force value is calculated, which is specifically expressed as:

[0058]

[0059] Among them, A max and A min are the set muscle stiffness amplitudes, k is the manually adjusted distance weight factor, and x c,best is the optimal reference trajectory is the point in B that is closest to the real-time position X of the current robot end.

[0060] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0061] (1) By designing a virtual fixture module, the present invention can provide different degrees of guiding force assistance to the operator according to information such as the robot state, task objectives, and the operator's muscle state, so as to improve the operator's accuracy.

[0062] (2) The present invention designs a trajectory learning module based on an autonomous dynamic system according to the operator's arm stiffness information, task information, and environmental information, and comprehensively considers various information to generate an optimal guidance trajectory, providing assistance for unskilled teleoperation operators or providing higher-precision guidance for skilled operators.

[0063] (3) The present invention collects robot information and human intentions through the MYO bracelet and the TouchX device, and adjusts the parameters of the virtual fixture module in real time to provide different degrees of assistance to the operator under different circumstances of the teleoperation task. Description of the Drawings

[0064] Figure 1 is a schematic diagram of the implementation architecture of the teleoperation robot assistance system based on the MYO bracelet and the virtual fixture of the present invention;

[0065] Figure 2 is a schematic diagram of the flow of the control method of the teleoperation robot assistance system based on the MYO bracelet and the virtual fixture of the present invention. Detailed Embodiments

[0066] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present 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 only used to explain the present invention and are not used to limit the present invention.

[0067] Embodiment

[0068] As Figure 1 shown, this embodiment provides a teleoperation robot assistance system based on the MYO bracelet and the virtual fixture, including: a teleoperation module, a trajectory learning module, an arm stiffness extraction module, and a virtual fixture module;

[0069] Among them, the design purpose of the teleoperation module is to enable remote control of the robot through an external device, so that the robot can complete tasks in dangerous environments or environments where humans cannot enter. In this embodiment, the teleoperation module is built through a TouchX device. First, the operation spaces of the robot and the teleoperation device are paired through master-slave space matching to ensure that the posture of the teleoperation device end matches the end of the robot, and the robot can move within a reasonable range when the operator remotely controls the robot. Then, the control fluency and flexibility of the operator's control of the robot during teleoperation are adjusted through a proportional adjustment factor to ensure the sensitivity and followability of teleoperation. The TouchX device can remotely control the robot through a handheld joystick, and the built-in gyroscope can map and match the end of the joystick with the end posture of the robot;

[0070] The trajectory learning module is used to model and generalize the expert trajectories collected and constructed. In the teleoperation system, the trajectory learning module extracts, learns, and models the trajectories of some expert teleoperations to complete tasks. Then, when an untrained operator performs a teleoperation task, the optimal reference trajectory is automatically generated according to the task information, and guidance is generated at the teleoperation device end according to the optimal reference trajectory and the robot state to guide the operator to complete the task more stably and accurately; the trajectory learning module is implemented through an autonomous dynamic system and has multiple dynamically adjustable parameters, including the number of layers of the neural network, the number of neurons in the hidden layer, the characteristic parameters in the energy function, etc.;

[0071] The arm stiffness extraction module extracts the operator's arm stiffness information and muscle activity information. In the teleoperation system, the operator wears a MYO bracelet on the arm to obtain the electromyogram signal of the operator's forearm. The muscle stiffness value is obtained through sliding window filtering and nonlinear transformation. By processing the information, the operator's intention can be recognized. According to the tension degree of the operator's arm muscles, the trajectory attention degree of the trajectory learning module and the feedback degree of the virtual fixture guiding force are adjusted, so as to improve the intelligence and flexibility of the teleoperation system and also improve the operator's teleoperation experience. The MYO bracelet collects the surface electromyogram signal of the operator's arm through 8 sensing channels and transmits the data to the computer at the operation end in real time through Bluetooth wirelessly. The received electromyogram signal is filtered and finally converted into the operator's muscle activity data through an algorithm;

[0072] The virtual fixture module obtains the robot state and the optimal reference trajectory in real time, searches for the current optimal target point of the robot on the optimal reference trajectory, constructs the resultant gravitational force of the optimal target point and the task completion point on the robot end, and feeds it back to the teleoperation device end to guide the operator to remotely control the robot. The virtual fixture module includes multiple dynamically adjustable parameters, including the gravitational weight of the task completion point on the robot, the gravitational weight of the trajectory optimal point on the robot, and the attention weight dynamically adjusted according to the tension degree of the operator's arm muscles.

[0073] As shown Figure 2 in the figure, this embodiment also provides a control method for a teleoperation robot assistance system based on a MYO bracelet and a virtual fixture. The operator side is a user holding a Touch X end joystick and a MYO bracelet. The Touch X end joystick is used to remotely control the slave robot, and the MYO bracelet is used to collect the operator's muscle activity to obtain the arm stiffness value and obtain the operator's intention. The operator side and the robot side are connected through a network. The specific steps include:

[0074] S1: Collect expert trajectory data and environmental obstacle information. The human expert completes the robot task through teleoperation teaching to generate a demonstration dataset where x t,n is the position state of the robot at time t in the nth demonstration, is the velocity state, x obj is the obstacle information in the environment (manually input), A t,n is the expert arm stiffness value;

[0075] The robot state data x t,n , is obtained by using the master-side haptic device to remotely operate the slave robot. First, the coordinate systems of the master and slave robots are calibrated. Define the initial pose X l0 =[x l0 ,y l0 ,z l0 of the master-side haptic device and the initial pose X f0 =[x f0 ,y f0 ,z f0 of the slave manipulator. By designing an appropriate scaling matrix S = diag(S x ,S y ,S z ), make the master-slave displacement satisfy X f =X f0 +S·(X l -X l0 ), and then perform real-time pose mapping on the master and slave robots. Collect the real-time pose X l (t)=[x l (t),y l (t),z l (t)] of the haptic device, calculate the target pose of the slave end and send it to the robot for execution;

[0076]

[0077] And solve the joint angles through inverse kinematics and send them to the slave manipulator. Then, the position data x t,n, taking the time derivative of the position gives

[0078] To calculate the arm stiffness value, it is necessary to first obtain the 8-channel raw electromyogram signals EMG collected by the MYO bracelet i , EMG i represents the electromyogram signal collected by the i-th channel of the MYO bracelet, and k is the current sampling time. Then the sum S of the amplitudes of the 8 raw electromyogram signals collected by the MYO bracelet at a sampling time can be expressed as:

[0079]

[0080] Define E as the envelope of the sum of the amplitudes of the 8 raw electromyogram signals, and w win is the sliding window length. Performing sliding window filtering on the sum S of the amplitudes of the 8 raw electromyogram signals can obtain its envelope E:

[0081]

[0082] Define A as the muscle stiffness and α as the non-linear coefficient. Transforming the envelope E of the sum of the amplitudes of the 8 electromyogram signals can obtain the muscle stiffness A:

[0083]

[0084] S2: After obtaining the expert dataset, the trajectory learning dynamic system model in the trajectory learning module is constructed as:

[0085]

[0086] where the energy function is calculated by the following formula:

[0087]

[0088] Among them, the energy function V(x) is mainly composed of the learnable neural network P1(x) and the quadratic form P2(x). The P1(x) term is the weighted sum of the weight parameter w and the feature extraction function f k (x), g(x) maps the input from the dx dimension to the dh dimension, c(x) performs non-linear feature expansion on x, enabling the system to capture complex non-linear relationships, and the parameters a k , b k are feature parameters, and the activation function is is the tanh function. The quadratic form P2(x) ensures the radial unboundedness of the energy function V(x), and the parameter κ is the quadratic term coefficient;

[0089] Modulation matrix is calculated as follows. First, the robot state input x and Obstacle information x obj and the energy value at state x and muscle stiffness A are combined into a vector z:

[0090]

[0091] A scaling factor needs to be multiplied by vector z according to the distance between the current end - effector position and the obstacle and the degree of arm muscle tension:

[0092]

[0093] It is considered that the parameters α1 and α2 determine the smoothness of the tracking and obstacle - avoidance switching process, β represents the safety margin, and d represents the distance from the current robot position to the obstacle. According to different inputs of z, z can be divided into the following situations: when the distance to the obstacle is far, the value of α1 is large and the value of α2 is small. At this time, the input pays more attention to the position x, that is, on the task execution trajectory. When the distance is close, the value of α1 is small and the value of α2 is large. At this time, the input pays more attention to the obstacle - avoidance part. And the value of the arm stiffness A is used as a complete input. When the value is large, it means that the arm is tense, which may be due to encountering an obstacle or needing to complete a high - precision operation without encountering an obstacle. When the value is small, it means a simple teleoperation following task. z will be used as the input vector of a neural network. The parameters α1, α2 and the human arm stiffness value A will affect the activation of different neurons in the neural network, and also represent different states encountered during the teleoperation process. Input z into a three - layer fully - connected neural network. The number of neurons in the hidden layer of the neural network is 80, the activation function is the tanh function, and the output is a six - dimensional vector n:

[0094] n = Net(z; Θ N )

[0095] Use the six - dimensional vector n to form a lower - triangular matrix N, where the values of matrix N are:

[0096]

[0097] Then use the Cholesky formula to form a positive - definite matrix M:

[0098]

[0099] Combining M and V forms a trajectory learning system:

[0100]

[0101] Use the expert data set for parameter optimization and training. By solving the following loss function for the optimization problem, the parameters of the model can be obtained:

[0102]

[0103] After obtaining the optimal model parameters, the model can be constructed and the expert trajectory can be generalized:

[0104] S3: Start using the master robot's haptic device TouchX to remotely control the slave robot and perform master-slave spatial matching;

[0105]

[0106] where X l0 and X f0 are the end Cartesian position coordinates of Touch X and the slave robot at the initial state, X l (t) is the real-time position of the master robot, and X f (t) is the real-time position of the slave robot. The real-time position of the master robot is obtained and then the expected position of the end of the slave robot is obtained through matching calculation, and then sent to the robot for execution to achieve teleoperation;

[0107] S4: Obtain the operator's arm EMG signal through the MYO device on the operator's arm and obtain the operator's arm stiffness value A;

[0108] S5: According to the task situation, input the task starting point information and the operator's arm stiffness A into the trained trajectory learning dynamic system model to obtain the optimal reference trajectory

[0109] S6: According to the current end real-time position X B of the robot, the optimal reference trajectory and the operator's arm stiffness value A, calculate the virtual fixture guiding force value through the following formula:

[0110]

[0111] where A max and A min are the set muscle stiffness amplitudes, k is the distance weight factor adjusted by the user, and x c,best is the point on the optimal reference trajectory closest to the current end real-time position X B of the robot. When the value of the operator's arm stiffness value A is larger, it means that it is in a situation where the task requires fine alignment or an emergency situation where an obstacle is encountered. At this time, the value of the query sentence feedback force F will be increased, so that the virtual fixture can better guide the operator to perform teleoperation and effectively reduce the hand tremor problem of the operator.

[0112] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A teleoperation robot-assisted system based on a MYO bracelet and a virtual fixture, characterized in that, It includes: A teleoperation module, a trajectory learning module, an arm stiffness extraction module, and a virtual fixture module; The arm stiffness extraction module obtains forearm electromyogram signals and muscle stiffness information based on a MYO bracelet; The teleoperation module realizes master-slave spatial matching robot teleoperation based on a TouchX device; The trajectory learning module obtains muscle stiffness information, robot state information, and task information, learns an expert trajectory data set, and generates an optimal reference trajectory; The virtual fixture module generates an auxiliary virtual fixture guiding force based on the optimal reference trajectory and the robot state, and feeds it back to the operator through the TouchX device.

2. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 1, characterized in that, It includes the following steps: Collect expert trajectory data and environmental obstacle information, complete robot tasks based on the teleoperation module, and generate a demonstration data set; Construct a trajectory learning dynamic system model in the trajectory learning module and train it based on the expert trajectory data set; Remote control the slave robot based on the TouchX device to perform master-slave spatial matching; Obtain forearm electromyogram signals and muscle stiffness information based on the MYO bracelet; Input the muscle stiffness information, robot state information, and task information into the trained trajectory learning dynamic system model to obtain an optimal reference trajectory; Calculate the virtual fixture guiding force value according to the current real-time position of the robot end, the optimal reference trajectory, and the muscle stiffness information.

3. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 2, characterized in that, The demonstration data set is represented as: where x t,n is the position state of the robot at time t in the nth demonstration, is the velocity state, x obj is the obstacle information in the environment, A t,n is the stiffness value of the expert's arm.

4. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 3, characterized in that, The steps for generating the demonstration data set include: The position state x of the robot t,n and the speed state are obtained from the slave robot through teleoperation by the master haptic device; Calibrate the coordinate systems of the master-slave robots, and define the initial pose \(X\) of the master haptic device l0 =\([x l0 ,y l0 ,z l0 \) and the initial pose \(X\) of the slave manipulator f0 =\([x f0 ,y f0 ,z f0 \). By constructing a scaling matrix \(S = diag(S x ,S y ,S z \)), make the master-slave displacements satisfy \(X f =X f0 +S\cdot(X l -X l0 \)), perform real-time pose mapping on the master-slave robots, collect the real-time pose \(X l (t)=[x l (t),y l (t),z l (t)]\), calculate the target pose of the slave end and send it to the robot for execution; Solve for the joint angles through inverse kinematics and send them to the slave manipulator to obtain the position state x of the robot performing the task t ,n , take the time derivative of the position to obtain the velocity state Obtain the electromyogram signal EMG collected from the i-th channel in the MYO bracelet i , and calculate the sum of the amplitudes of the electromyogram signals as follows: Perform sliding window filtering on the sum of the electromyogram signal amplitudes to obtain an envelope E: where w win is the sliding window length; The muscle stiffness value of the expert arm is expressed as: where α is a non-linear coefficient.

5. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 2, characterized in that Construct a trajectory learning dynamic system model in the trajectory learning module, which is specifically expressed as: Among them, represents the energy function, represents the modulation matrix, x represents the robot state input, and z represents a vector composed of the robot state input, obstacle information, the energy value at state x, and the muscle stiffness A.

6. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 5, characterized in that, Energy function Specifically expressed as: Among them, the energy function V(x) is composed of a learnable neural network P1(x) and a quadratic form P2(x). The P1(x) term is the weighted sum of the weight parameter w and the feature extraction function f k (x). g(x) maps the input from the dx dimension to the dh dimension, c(x) performs a non-linear feature expansion on the state x, and the parameters a k , b k are feature parameters, and the activation function is the tanh function, and the parameter κ is the quadratic term coefficient; The vector z is expressed as: Among them, represents the trajectory learning dynamic system model, and x obj represents obstacle information, represents the energy value at state x, and A represents muscle stiffness; Multiply the vector z by a degree factor according to the distance between the current end position and the obstacle and the degree of arm muscle tension: where the parameters α1 and α2 determine the smoothness of the tracking and obstacle avoidance switching process, β represents the safety margin, and d represents the distance from the current robot position to the obstacle; The vector z is input into a three-layer fully connected neural network, and the output is a six-dimensional vector: n = Net(z; Θ N ) Based on the six-dimensional vector n, form a lower triangular matrix, which is expressed as: Construct a positive definite matrix based on the Cholesky formula 7. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 2, characterized in that, Remote control the slave robot based on the TouchX device to perform master-slave spatial matching: Among them, X l0 , X f0 are the Cartesian position coordinates of Touch X and the end of the slave robot at the initial state, and X l (t) is the real-time position of the master robot, and X f (t) is the real-time position of the slave robot; Obtain the real-time position of the master robot in real time, obtain the expected position of the slave robot end through matching calculation, and send it to the robot for execution to realize teleoperation.

8. The control method of the teleoperation robot-assisted system based on the MYO bracelet and the virtual fixture according to claim 2, wherein Calculate the virtual fixture guiding force value according to the current real-time position of the robot end, the optimal reference trajectory, and the muscle stiffness information, which is specifically expressed as: Among them, A max and A min are the set muscle stiffness amplitudes, k is the artificially adjusted distance weight factor, and x c,best is the optimal reference trajectory the point in it that is closest to the real-time position X B of the current robot end effector.

Images