A method for estimating patient lateral force during surgical robot operation

Abstract

The present invention relates to the field of surgical robot control technology and discloses a method for estimating the patient-side force during surgical robot surgery, comprising: constructing a dynamic model of the patient-side robotic arm of the surgical robot in free space; inserting the patient-side robotic arm into the patient's abdominal wall through a cannula to collect joint data, and simultaneously predicting the free-space torque of each joint at the current moment through the dynamic model; using the collected position, velocity data, and free-space torque of each joint as input data, and using the residual between the measured torque and free-space torque of each joint at each moment as label data, and training a neural network model for each joint separately; and calculating the contact force between the end effector and the patient's tissue based on the current measured torque of each joint, the free-space torque, and the predicted additional torque of each joint. The present invention can accurately estimate the interaction force between the surgical robot end effector and the tissue, thereby improving the safety and accuracy of the surgery.

Description

Technical Field

[0001] The present invention relates to the technical field of surgical robot control, and more particularly, to a method for estimating the lateral force of a patient during a surgical robot operation. Background Art

[0002] Surgical robots are widely used in modern medical technology for their ability to improve surgical precision and safety. These robots assist surgeons in performing complex surgical procedures, particularly in minimally invasive surgeries. However, existing surgical robots face significant technical challenges in providing tactile feedback, limiting the intuitiveness and safety of these procedures.

[0003] When surgical robots (such as the da Vinci system) assist in minimally invasive surgery, doctors need to accurately sense the interaction force (such as resistance during cutting and suturing) between the end effector (including surgical tools) and the patient's tissue (such as body tissue in the abdominal cavity) to provide tactile feedback, thereby improving surgical safety and accuracy. In the existing technology, the tactile feedback system of surgical robots mainly relies on end effector force sensors or model-based dynamic calculations. Although the force sensor method can directly measure the interaction force between the instrument and the tissue, it has problems such as high cost, difficulty in sterilization, and difficulty in integration in narrow surgical spaces. Model-based dynamic calculation methods estimate the interaction force by establishing a dynamic model of the robot, but these models usually require a large amount of experimental data for calibration, and are sensitive to changes in the robot structure and environment, making it difficult to adapt to different surgical conditions and instruments.

[0004] In addition, existing patient-side force estimation algorithms are deficient in processing the interaction forces between the robot and the cannula. The cannula is a necessary channel for the surgical robot to enter the patient's body. Usually in abdominal surgery, the cannula is inserted into the patient's body wall through a small incision, providing a channel for surgical instruments to enter the body. During the operation, the patient-side robotic arm is inserted into the patient's abdominal cavity or other surgical site through the cannula, and the surgical tools installed at its end perform the surgical operation. However, the friction and other interaction forces generated by the contact between the cannula and the patient-side robotic arm will have a significant impact on the estimation of the interaction force between the end effector and the tissue. For example, the friction and resistance of the cannula act on the instrument shaft, which is transmitted through the joints of the robotic arm, and ultimately affects the motion accuracy and force feedback of the end effector. Existing technologies find it difficult to accurately compensate for these additional torques, resulting in a decrease in the accuracy of tactile feedback between the end surgical tool and the tissue, which in turn affects the accuracy and safety of the doctor's operation. Summary of the Invention

[0005] The purpose of the present invention is to propose a method for estimating the patient's lateral force during surgical robot surgery, so as to accurately estimate the interaction force between the surgical robot end effector and tissue, improve the safety and accuracy of the surgery, and provide surgeons with more reliable tactile feedback.

[0006] To achieve the above objectives, the present invention proposes a method for estimating the lateral force of a patient during a surgical robot operation, comprising:

[0007] Constructing a dynamic model of the patient-side robotic arm of the surgical robot in free space, wherein the dynamic model is used to predict the free-space torque of the joint based on the position, velocity, and acceleration of the joint;

[0008] Inserting the patient-side robotic arm through the cannula into the patient's abdominal wall, collecting joint data including the interaction between the patient-side robotic arm and the cannula, including the position, velocity, and measured torque of each joint; and predicting the free space torque of each joint at the current moment using the dynamic model;

[0009] The collected position and velocity data of each joint and the free-space torque of each joint are used as input data, and the residual between the measured torque of each joint and the free-space torque at each moment is used as label data to form a training dataset. A neural network model is trained for each joint using the training dataset to predict the additional torque of each joint caused by the interaction between the patient-side robotic arm and the cannula.

[0010] When the end effector contacts the patient's tissue during surgery, the contact force between the end effector and the patient's tissue is calculated based on the current measured torque of each joint, the free space torque of each joint predicted by the dynamic model, and the additional torque of each joint predicted by the neural network model.

[0011] Optionally, constructing a dynamic model of the patient-side robotic arm of the surgical robot includes:

[0012] The kinematic model of the patient-side robotic arm was established based on the improved DH parameter method. The coordinate system of each joint was defined and the position relationship between adjacent joints was expressed using a homogeneous transformation matrix.

[0013] Based on the kinematic model, a dynamic model including connecting rod inertia, gravity, joint friction and spring stiffness is constructed using the Euler-Lagrange equation;

[0014] Dynamic parameters of the dynamic model are identified by combining the optimized excitation trajectory.

[0015] Optionally, the identifying dynamic parameters of the dynamic model by combining the optimized excitation trajectory includes:

[0016] Based on the Fourier series, a periodic excitation trajectory that can fully stimulate the dynamic characteristics of the robot is generated. Based on the upper and lower limits of joint position, velocity, and acceleration, a convex optimization method is used to optimize the parameters of the excitation trajectory to ensure that the robot moves within a safe range.

[0017] Control the patient-side robotic arm to run the optimized excitation trajectory in free space and collect joint position, velocity, acceleration and measured joint torque data;

[0018] Construct a regression matrix based on the collected joint position, velocity, and acceleration. Each row of the regression matrix corresponds to a sampling point containing joint position, velocity, and acceleration data.

[0019] Perform QR decomposition on the regression matrix to obtain the basis parameter regression matrix;

[0020] Using the basis parameter regression matrix and the measured joint torque data, the weighted least squares method is used to solve the basis parameters; the basis parameters are linear combinations of dynamic parameters, which are the minimum parameter set to describe the dynamics of the manipulator;

[0021] The dynamic parameters include connecting rod length, connecting rod mass, connecting rod center of mass position, connecting rod inertia tensor, joint friction coefficient and spring stiffness.

[0022] Optionally, the objective function of solving the basis parameters using the weighted least squares method is:

[0023]

[0024] in, is the basis parameter regression matrix, δ b is the basis parameter vector, is the measured joint torque at the i-th sampling point, m is the number of sampling points, is the weight coefficient, which is used to balance the magnitude differences of each joint. .

[0025] Optionally, the neural network model is a multi-layer perceptron trained independently for each joint, the multi-layer perceptron includes an input layer, a hidden layer and an output layer, the hidden layer includes two fully connected layers, and the activation function is ReLU.

[0026] Optionally, the input data of the neural network model specifically include: a joint position sequence and a velocity sequence within a historical time window, and a free space torque of the current joint predicted by the dynamics model at the current moment;

[0027] The output of the neural network model is the compensation torque of the current joint.

[0028] Optionally, when training the neural network model, mean square error is used as a loss function, and model parameters that minimize the loss function are found as training targets.

[0029] Optionally, the residual calculation formula between the measured torque of each joint and the free space torque is:

[0030]

[0031] in, is the vector of the measured torque of each joint, including the torque actually measured by each joint motor, is the vector of the free space torque of each joint, including the torque of each joint in free space motion predicted by the dynamic model, , is the basis parameter regression matrix, is the basis parameter vector.

[0032] Optionally, the method of calculating the contact force between the end effector and the patient tissue includes:

[0033] The position, velocity and acceleration of each joint collected in real time are input into the dynamic model, and the free space torque of each joint is predicted by the dynamic model;

[0034] Inputting the real-time collected joint position and velocity data and the free space torque predicted by the dynamic model into the neural network model and outputting the predicted joint additional torque;

[0035] The joint free space torque is compensated by the additional joint torque output by the neural network model;

[0036] Based on the current measured torque of each joint and the compensated joint torque, combined with the inverse matrix of the spatial Jacobian matrix, the Cartesian force of the end effector is calculated.

[0037] Alternatively, the formula for calculating the Cartesian force of the end effector is:

[0038]

[0039] in, is the contact force vector between the end effector and the patient tissue, J is the spatial Jacobian matrix at the current joint position, is the measured torque vector of each joint at the current moment, is the free space torque vector of each joint predicted by the dynamic model, is the compensation torque vector output by the neural network model.

[0040] The beneficial effects of the present invention are:

[0041] The present invention first constructs a dynamic model of the patient-side robotic arm in free space (non-contact). The dynamic model predicts the free-space torques of the joints. The patient-side robotic arm is then inserted into the patient through a cannula and joint data collected during the insertion process, including joint positions, velocities, and measured joint torques. Combined with the joint data collected during the interaction between the robotic arm and the cannula and the free-space joint torques calculated by the dynamic model, a neural network model is used to learn the additional torque generated by the interaction between the cannula and the patient-side robotic arm to compensate for the torque predicted by the dynamic model. The contact force between the end effector and the patient's tissue is then estimated based on the current measured torques of each joint, the free-space torques predicted by the dynamic model, and the additional torques predicted by the neural network model. This accurately estimates the interaction forces between the surgical robot end effector and the tissue, providing precise tactile feedback for the surgical robot and improving surgical safety and accuracy. Furthermore, by combining the dynamic model with a neural network learning approach, the present invention reduces the requirements for training data coverage compared to pure learning methods while maintaining force estimation accuracy, thereby improving adaptability to diverse surgical scenarios and operating conditions.

[0042] Furthermore, the present invention completes dynamic parameter identification by optimizing the excitation trajectory and extracting basis parameters through QR decomposition, thereby improving the robustness of the model; adopts a small-scale neural network model for joint training, requiring only a small amount of intraoperative data to compensate for casing friction in real time, thus realizing lightweight neural network compensation; based on the spatial Jacobian matrix, the joint torque is efficiently converted into the end Cartesian force, realizing real-time force mapping, ensuring the real-time nature of intraoperative tactile feedback, and improving the intuitiveness and safety of the doctor's surgery.

[0043] The method of the present invention has other features and advantages that will be apparent from or will be described in detail in the accompanying drawings and subsequent detailed description incorporated herein, which together serve to explain the specific principles of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and other objects, features and advantages of the present invention will become more apparent through a more detailed description of exemplary embodiments of the present invention with reference to the accompanying drawings, in which like reference numerals generally represent like components.

[0045] Figure 1 A flow chart of a method for estimating patient lateral force during surgical robot operation according to the present invention is shown. DETAILED DESCRIPTION

[0046] The present invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention may be implemented in various forms and is not limited to the embodiments set forth herein. Rather, these embodiments are provided to make the present invention more thorough and complete and to fully convey the scope of the present invention to those skilled in the art. Example

[0047] like Figure 1 As shown, this embodiment provides a method for estimating the lateral force of a patient during surgery by a surgical robot, comprising:

[0048] S1: Construct a dynamic model of the patient-side robotic arm of the surgical robot in free space. The dynamic model is used to predict the free-space torque of the joint based on the position, velocity, and acceleration of the joint.

[0049] This step specifically includes:

[0050] S101: Based on the improved DH parameter method, the kinematic model of the patient-side robotic arm is established, the coordinate system of each joint is defined, and the position relationship between adjacent joints is expressed using a homogeneous transformation matrix;

[0051] Specifically, the kinematic model describes the relationship between the positions of each robot joint and the pose (position and attitude) of the end effector, without involving forces or mass. Its core is the coordinate transformation and Jacobian matrix. Based on the structural characteristics of the patient-side robotic arm, the coordinate system for each joint is defined. This is typically accomplished using Denavit-Hartenberg (DH) parameters. This method utilizes a modified DH parameterization that better adapts to the specific structure of the patient-side robotic arm.

[0052] In one example, the improved DH parameter method is used to transform each joint i The coordinate system is defined by the following four parameters:

[0053] adjacent joints i −1 to i Homogeneous transformation matrix of coordinate system for:

[0054]

[0055] Among them, Rot( z , θ i ) is the rotation angle θ of the current joint around the z-axis of the previous joint i The rotation matrix, Trans( z , d i ) is the translation distance d of the current joint along the z-axis of the previous joint i The translation matrix, Trans(x , a i ) is the translation distance a of the link along the current joint x-axis i The translation matrix, Rot( x , α i ) is the rotation angle of the link around the current joint x-axis α i The rotation matrix of .

[0056] The expanded matrix form is:

[0057]

[0058] Then, the kinematic chain of the manipulator is constructed based on the homogeneous transformation matrix. By calculating the homogeneous transformation matrix layer by layer, the precise position and posture of the end effector are finally obtained. For example, from the base (joint 0) to the end effector (joint n )'s overall transformation matrix for:

[0059]

[0060] From the matrix Extract the translation part to get the end position (such as the fourth column) from the matrix Extract the rotation matrix (such as the first three columns) to get the end pose.

[0061] Then, the mapping relationship between joint velocity and end velocity is described by the Jacobian matrix:

[0062]

[0063] in, are the terminal linear velocity and angular velocity vectors, is the spatial Jacobian matrix, which is composed of the rotation quantities of each joint. is the joint position vector, is the joint velocity (angular velocity) vector.

[0064] The improved DH method is more suitable for serial robots, with more intuitive parameter definition and smaller calculation errors. The improved DH parameters and homogeneous transformation matrix ensure high accuracy of end-point pose calculation.

[0065] S102: Based on the kinematic model, a dynamic model including connecting rod inertia, gravity, joint friction, and spring stiffness is constructed using the Euler-Lagrange equation;

[0066] Specifically, this step builds an accurate model of the robot's own dynamics (excluding external interference). The dynamic model describes the joint torque and motion (joint position , joint speed , joint acceleration ), involving physical effects such as inertia, gravity, and friction. This method uses the Euler-Lagrange equations to derive the torque equation that includes inertia, gravity, friction, and spring forces.

[0067] The Euler-Lagrange equation is:

[0068]

[0069] Where L is the Lagrangian, K is the total kinetic energy of the system, P is the total potential energy of the system.

[0070] The kinetic equation is:

[0071]

[0072] After expansion, the nonlinear dynamic equation is:

[0073]

[0074] in, is the joint torque (joint motor torque), is the joint position vector, is the joint velocity vector, is the joint acceleration vector, is the inertia matrix, which describes the mass distribution and inertia characteristics of the robot, is the Coriolis force matrix, which includes Coriolis force and centrifugal force terms; is the gravity term, which is calculated from the gravitational potential energy gradient, is the friction term, usually modeled as a combination of viscous friction and Coulomb friction.

[0075] The above nonlinear dynamic equation can be converted into a linear parameterized form:

[0076]

[0077] in, δ is a dynamic parameter, is a regression matrix, each row corresponds to a kinetic equation at a time point, and each column corresponds to a dynamic parameter.

[0078] The conversion method is: first define the dynamic parameters δ , including all physical quantities that affect dynamics, dynamic parameters δ Including connecting rod mass, center of mass position, inertia tensor, friction coefficient, etc. Then each nonlinear term in the nonlinear dynamic equation is expanded into a linear combination of dynamic parameters, and each coefficient after expansion is extracted as a regression matrix WColumns, construct regression matrix W .

[0079] Reorganize the kinetic equations into dynamic parameters δ It is a linear combination of , which is convenient for subsequent parameter identification.

[0080] S103: Dynamic parameter identification of the dynamic model is performed by combining the optimized excitation trajectory, as follows:

[0081] S1031: Generate a periodic excitation trajectory based on the Fourier series that can fully stimulate the dynamic characteristics of the robot, and optimize the parameters of the excitation trajectory using a convex optimization method based on the constraints of the joint position, velocity, and acceleration upper and lower limits to ensure that the robot moves within a safe range;

[0082] Specifically, Fourier series is used to generate a periodic trajectory that can fully stimulate the dynamic characteristics of the robot. Preferably, the goal is to minimize the regression matrix W The condition number of is expressed as:

[0083]

[0084] in, express j The change of joint coordinates over time, that is, the function of the position of the joint over time; express j The offset or initial position of the joint coordinates is the reference starting point of the trajectory; Indicates the k The amplitude coefficient of the sine term of the subharmonic determines the contribution of the sine part of the harmonic to the trajectory; Indicates the k The amplitude coefficient of the cosine term of the subharmonic determines the contribution of the cosine part of the harmonic to the trajectory; k Indicates the number of harmonics; ω represents the fundamental frequency, which determines the periodic change rate of the trajectory; t is a time variable, indicating the change of trajectory over time; nH is the total number of harmonics.

[0085] Among them, the fundamental frequency ω Total harmonics nH The selection needs to be optimized, with the preferred fundamental frequency being 0.18 Hz and the harmonic order being 6.

[0086] Then, based on the constraints of joint position, velocity and acceleration upper and lower limits, a convex optimization method (such as the convex optimization tool CVXPY package combined with the SCS solver) is used to optimize the amplitude coefficient of the excitation trajectory. and , to ensure that the robot moves within a safe range.

[0087] S1032: Control the patient-side robotic arm to run the optimized excitation trajectory in free space (without external contact) and collect joint position, velocity, acceleration, and joint measured torque data;

[0088] The joint positions, velocities, and accelerations are measured by the robot's built-in encoders or other position / velocity sensors. The measured joint torques are measured by torque sensors at the robot's joints. This data is collected and transmitted by the robot's operating system for subsequent dynamic parameter identification and torque estimation. This step drives the robotic arm using the excitation trajectory, recording the joint positions, velocities, accelerations, and measured torques at each moment.

[0089] S1033: Constructing a regression matrix based on the collected joint positions, velocities, and accelerations, where each row of the regression matrix corresponds to a sampling point containing joint position, velocity, and acceleration data;

[0090] Specifically, construct the regression matrix W , based on the kinetic equation , each row corresponds to a sampling point, and the elements are calculated from the joint position, velocity, and acceleration, describing the contribution of each dynamic parameter to the torque.

[0091] S1034: Perform QR decomposition on the regression matrix to obtain a basis parameter regression matrix;

[0092] Specifically, for the regression matrix W Perform column pivot QR decomposition:

[0093]

[0094] in, P is the permutation matrix, Q is an orthogonal matrix, R Indicates independent columns for the nonzero diagonal elements of an upper triangular matrix.

[0095] Select the basis parameters, select b Linearly independent columns constitute the basis parameter regression matrix ,in, P b is the permutation matrix. Basis parameters δ b is the original parameter δ A linear combination of , is the minimum set of parameters that describe the system, where T represents the transpose.

[0096] Redundant parameters are eliminated through QR decomposition to reduce the complexity of identification.

[0097] S1035: Using the basis parameter regression matrix and the measured joint torque data, the weighted least squares method is used to solve the basis parameters; the basis parameters are linear combinations of dynamic parameters, which are the minimum parameter set to describe the dynamics of the manipulator;

[0098] The objective function of solving the basis parameters using the weighted least squares method is:

[0099]

[0100] in, is the basis parameter regression matrix, δ b is the basis parameter vector, is the measured joint torque at the i-th sampling point, m is the number of sampling points, is the weight coefficient, which is used to balance the magnitude differences of each joint. .

[0101] Fitting dynamic parameters by least squares method δ b , so that the model predicts the torque As close to the measured torque as possible ,in .

[0102] The above-mentioned dynamic modeling process is already known in the art, and the specific details will not be repeated here.

[0103] The robot's own dynamic torque is accurately separated through this step S1, providing a benchmark for subsequent compensation.

[0104] S2: Inserting the patient-side robotic arm through the cannula into the patient's abdominal wall, during which joint data including the interaction between the patient-side robotic arm and the cannula are collected, including the position, velocity, and measured torque of each joint; and simultaneously, the free space torque of each joint at the current moment is predicted using the dynamic model.

[0105] Specifically, the patient-side robotic arm collects joint status and torque data in real time when the cannula is inserted and passes through the abdominal phantom. During the insertion process, only the patient-side robotic arm contacts the cannula, and the end effector does not contact the tissue in the patient's body, thereby ensuring that the collected joint status and torque data only contain data generated by the interaction between the patient-side robotic arm and the collar. The process of inserting the patient-side robotic arm through the cannula into the patient's abdominal wall usually takes about 4 minutes (about 48,000 data points). The preferred sampling rate is 200Hz (window size 5), which can generate about 9,600 training samples. While collecting the above data, the dynamic model is used according to the actual torque and position of each joint collected at the current moment. ,speed and acceleration Real-time prediction of the free space torque of each joint at the current moment , used for subsequent training of neural network models.

[0106] S3: Using the collected position and velocity data of each joint and the free space torque of each joint as input data, and the residual between the measured torque of each joint and the free space torque at each moment as label data, a training dataset is formed. A neural network model is trained for each joint using the training dataset to predict the additional torque of each joint caused by the interaction between the patient-side robotic arm and the cannula.

[0107] Specifically, in this step, the free space prediction torque is calculated using the dynamic model , the residual torque calculation formula between the measured torque of the joint and the free space torque is:

[0108] in, is the vector of the joint measured torque, including the torque actually measured by each joint motor, is the vector of the free space torque of each joint, including the torque of each joint in free space motion predicted by the dynamic model, , is the basis parameter regression matrix, is the basis parameter vector.

[0109] The neural network model is a multi-layer perceptron (MLP) trained independently for each joint. The multi-layer perceptron includes an input layer, a hidden layer, and an output layer. The hidden layer includes two fully connected layers, and the activation function is ReLU.

[0110] The input data of the neural network model specifically includes: the joint position sequence within the historical time window (such as the window size is 5) and velocity sequence , and the free space torque of the current joint predicted by the dynamic model at the current moment The output of the neural network model is the predicted compensation torque of the current joint .

[0111] When the neural network model is trained, the mean square error is used as the loss function: , and find the model parameters that minimize the loss function as the training target.

[0112] By training a neural network model for each joint separately, the patient-specific sleeve friction pattern is learned, and the additional torque of each joint caused by the interaction between the patient-side robotic arm and the sleeve is predicted to compensate for the free space joint torque predicted by the dynamic model, thereby dynamically correcting the dynamic model prediction error.

[0113] S4: When the end effector contacts the patient tissue during surgery, the contact force between the end effector and the patient tissue is calculated based on the current measured torque of each joint, the free space torque of each joint predicted by the dynamic model, and the additional torque of each joint predicted by the neural network model.

[0114] In this step, the method for calculating the contact force between the end effector and the patient tissue includes:

[0115] The position, velocity and acceleration of each joint collected in real time are input into the dynamic model, and the free space torque of each joint is predicted by the dynamic model;

[0116] Inputting the real-time collected joint position and velocity data and the free space torque predicted by the dynamic model into the neural network model and outputting the predicted joint additional torque;

[0117] The joint free space torque is compensated by the additional joint torque output by the neural network model;

[0118] Based on the current measured torque of each joint and the compensated joint torque, combined with the inverse matrix of the spatial Jacobian matrix, the Cartesian force of the end effector is calculated.

[0119] The formula for calculating the Cartesian force of the end effector is:

[0120]

[0121] in, is the contact force vector between the end effector and the patient's tissue at the current moment (including force and torque in the XYZ directions), J is the spatial Jacobian matrix at the current joint position, is the measured torque vector of each joint at the current moment, is the free space torque vector of each joint predicted by the dynamic model, is the compensation torque vector output by the neural network model.

[0122] It should be noted that the measured torque The actual torque of each joint during the actual surgical action, including the torque required to maintain the basic posture (corresponding to the torque required to maintain the posture without external interference, that is, the free space torque of the joint ), resisting the casing interaction torque caused by casing friction, etc. (corresponding to the compensation torque predicted by the neural network model ) and the torque generated by the interaction between the end effector and the patient's tissue, so This is the torque generated by the interaction between the end effector and the patient's tissue, which can be converted into a force in Cartesian space through the inverse matrix of the spatial Jacobian matrix.

[0123] The accurate contact force between the end effector and tissue is calculated to provide intraoperative tactile feedback, thereby improving the doctor's surgical intuition and safety.

[0124] The training time of the small-scale neural network model used in this method is typically less than 20 seconds, thus supporting rapid intraoperative deployment. The training data collected by this method is based on specific patient data, and the trained neural network model can be applied to torque compensation during surgery on specific patients. Due to differences in physiological structure, tissue hardness, body wall thickness, etc., these differences can affect the interaction force between the cannula and the robotic arm. Therefore, for different patients, a brief data collection process (e.g., about 4 minutes) can be performed before each surgery to collect specific data for the current patient. The collected data can then be used to quickly retrain the neural network model to adapt to the current patient and surgical conditions. During surgery, the retrained model is used to estimate in real time the additional joint torque caused by the interaction force between the robotic arm and the cannula and compensate for the free-space joint torque, thereby improving the accuracy of the end-effector force estimation.

[0125] While various embodiments of the present invention have been described above, the above description is intended to be illustrative, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for estimating the lateral force of a patient during surgical robot surgery, characterized in that: include: Constructing a dynamic model of the patient-side robotic arm of the surgical robot in free space, wherein the dynamic model is used to predict the free-space torque of the joint based on the position, velocity, and acceleration of the joint; Inserting the patient-side robotic arm through the cannula into the patient's abdominal wall, collecting joint data including the interaction between the patient-side robotic arm and the cannula, including the position, velocity, and measured torque of each joint; and predicting the free space torque of each joint at the current moment using the dynamic model; The collected position and velocity data of each joint and the free-space torque of each joint are used as input data, and the residual between the measured torque of each joint and the free-space torque at each moment is used as label data to form a training dataset. A neural network model is trained for each joint using the training dataset to predict the additional torque of each joint caused by the interaction between the patient-side robotic arm and the cannula. When the end effector contacts the patient's tissue during surgery, the contact force between the end effector and the patient's tissue is calculated based on the current measured torque of each joint, the free space torque of each joint predicted by the dynamic model, and the additional torque of each joint predicted by the neural network model.

2. The method for estimating the lateral force of a patient during surgery with a surgical robot according to claim 1, wherein: The dynamic model of the patient-side robotic arm of the surgical robot is constructed as follows: The kinematic model of the patient-side robotic arm was established based on the improved DH parameter method. The coordinate system of each joint was defined and the position relationship between adjacent joints was expressed using a homogeneous transformation matrix. Based on the kinematic model, a dynamic model including connecting rod inertia, gravity, joint friction and spring stiffness is constructed using the Euler-Lagrange equation; Dynamic parameters of the dynamic model are identified by combining the optimized excitation trajectory.

3. The method for estimating the lateral force of a patient during surgery with a surgical robot according to claim 2, wherein: The identifying of dynamic parameters of the dynamic model by combining the optimized excitation trajectory includes: Based on the Fourier series, a periodic excitation trajectory that can fully stimulate the dynamic characteristics of the robot is generated. Based on the upper and lower limits of joint position, velocity, and acceleration, a convex optimization method is used to optimize the parameters of the excitation trajectory to ensure that the robot moves within a safe range. Control the patient-side robotic arm to run the optimized excitation trajectory in free space and collect joint position, velocity, acceleration and measured joint torque data; A regression matrix is ​​constructed based on the collected joint position, velocity, and acceleration data. Each row of the regression matrix corresponds to a sampling point containing joint position, velocity, and acceleration data. Perform QR decomposition on the regression matrix to obtain the basis parameter regression matrix; Using the basis parameter regression matrix and the measured joint torque data, the weighted least squares method is used to solve the basis parameters; the basis parameters are linear combinations of dynamic parameters, which are the minimum parameter set to describe the dynamics of the manipulator; The dynamic parameters include connecting rod length, connecting rod mass, connecting rod center of mass position, connecting rod inertia tensor, joint friction coefficient and spring stiffness.

4. The method for estimating the patient's lateral force during surgical robot surgery according to claim 3, characterized in that: The objective function of solving the basis parameters using the weighted least squares method is: ， in, is the basis parameter regression matrix, δ b is the basis parameter vector, is the measured joint torque at the i-th sampling point, m is the number of sampling points, is the weight coefficient, which is used to balance the magnitude differences of each joint. .

5. The method for estimating the patient's lateral force during surgical robot surgery according to claim 1, characterized in that: The neural network model is a multi-layer perceptron trained independently for each joint. The multi-layer perceptron includes an input layer, a hidden layer, and an output layer. The hidden layer includes two fully connected layers, and the activation function is ReLU.

6. The method for estimating the patient's lateral force during surgical robot surgery according to claim 5, characterized in that: The input data of the neural network model specifically includes: a joint position sequence and a velocity sequence within a historical time window, and a free space torque of the current joint predicted by the dynamics model at the current moment; The output of the neural network model is the compensation torque of the current joint.

7. The method for estimating the patient's lateral force during surgical robot surgery according to claim 6, characterized in that: When the neural network model is trained, mean square error is used as the loss function, and the model parameters that minimize the loss function are found as the training target.

8. The method for estimating the patient's lateral force during surgical robot surgery according to claim 4, characterized in that: The residual calculation formula between the measured torque of each joint and the free space torque is: ， in, is the vector of the measured torque of each joint, including the torque actually measured by each joint motor, is the vector of the free space torque of each joint, including the torque of each joint in free space motion predicted by the dynamic model, , is the basis parameter regression matrix, is the basis parameter vector.

9. The method for estimating the lateral force of a patient during surgery by a surgical robot according to claim 8, wherein: Methods for calculating the contact force between the end effector and the patient tissue include: The position, velocity and acceleration of each joint collected in real time are input into the dynamic model, and the free space torque of each joint is predicted by the dynamic model; Inputting the real-time collected joint position and velocity data and the free space torque predicted by the dynamic model into the neural network model and outputting the predicted joint additional torque; The joint free space torque is compensated by the additional joint torque output by the neural network model; Based on the current measured torque of each joint and the compensated joint torque, combined with the inverse matrix of the spatial Jacobian matrix, the Cartesian force of the end effector is calculated.

10. The method for estimating the patient's lateral force during surgical robot surgery according to claim 9, characterized in that: The formula for calculating the Cartesian force of the end effector is: ， in, is the contact force vector between the end effector and the patient tissue, J is the spatial Jacobian matrix at the current joint position, T represents transpose, is the measured torque vector of each joint at the current moment, is the free space torque vector of each joint predicted by the dynamic model, is the compensation torque vector output by the neural network model.

Images