Operation robot base position optimization method based on working mode analysis

Abstract

The invention relates to the field of surgical robots, and discloses a surgical robot base position optimization method based on working mode analysis, which comprises the following steps: extracting an end effector representative attitude set representing the working mode of a surgeon; a plurality of base positions are randomly generated, for any base position, the robot joint angle corresponding to each representative posture is solved through inverse kinematics, the suitability of the base position to each representative posture of the end effector is evaluated based on the solved joint angle, and comprehensive scoring is carried out; taking each base position parameter and the comprehensive score thereof as an input-output data pair to form a training data set, and training the neural network model to learn a mapping relation between the base position and the score; in an actual operation scene, the trained neural network model is utilized to predict an optimal base position. The optimal base position can be determined based on the specific working mode of a surgeon, and the operability and the working efficiency of the surgical robot are improved to the maximum extent.

Description

Technical Field

[0001] The present invention relates to the technical field of surgical robots, and more specifically, to a method for optimizing the position of a surgical robot base based on working mode analysis. Background Art

[0002] In robot-assisted minimally invasive surgery (RAMIS), the base of the surgical robot is the installation and fixing interface of the robot arm, providing support and fixation for the robot arm. It determines the initial position and direction of the robot arm. Therefore, the position and direction of the base directly affect the reachable range of the robot end effector in the global space. Improper base position may lead to joint restrictions, singularity problems and inaccessible workspace of the operating arm, thus affecting the efficiency and safety of the operation. Therefore, the optimization of the base position of the surgical robot is crucial to the success of the operation. The traditional method relies on the experience of medical staff to set the base position, but this method of manually setting the base position is inefficient and lacks objective quantitative indicators. It is difficult to ensure that the best surgical effect can be achieved in various complex surgical scenarios. In addition, due to improper base position setting, the end effector cannot cover the target surgical area, the robot arm approaches a singular configuration or the operation is interrupted, so the base position needs to be adjusted frequently. In addition, different surgeons have different working modes when performing surgical tasks, including preferred operating techniques and processing styles. These working modes may vary greatly between different surgeons, and the universal base position cannot meet the personalized needs of doctors. Therefore, a method is needed to automatically optimize the position of the surgical robot base according to the surgeon's working mode. Summary of the invention

[0003] The purpose of the present invention is to propose a method for optimizing the base position of a surgical robot based on working mode analysis, so as to determine the optimal base position based on the surgeon's unique working mode, meet the doctor's personalized needs, and maximize the operability and work efficiency of the surgical robot.

[0004] To achieve the above objectives, the present invention proposes a method for optimizing the position of a surgical robot base based on working mode analysis, comprising: By analyzing the end-effector motion data recorded by the surgeon during the operation, a representative set of end-effector postures representing the surgeon's working mode is extracted; Multiple base positions are randomly generated within the feasible range of the base. For any base position, the robot joint angle corresponding to each representative posture at the base position is solved by inverse kinematics. The adaptability of the base position to each representative posture of the end effector is evaluated based on the solved joint angle, and the base position is comprehensively scored. Taking each base position parameter and its corresponding comprehensive score as an input-output data pair to form a training data set, and using the training data set to train a neural network model to learn the mapping relationship between the base position and the score; In actual surgical scenarios, multiple candidate base positions are generated within the feasible range of the base seat, and the trained neural network model is used to predict the comprehensive score of each candidate base position to select the optimal base position.

[0005] Optionally, the step of extracting a representative set of end effector postures representing the surgeon's working mode by analyzing the end effector motion data recorded by the surgeon during the operation includes: In actual surgery, the robot sensor records the posture data of the end effector when the surgeon operates the surgical robot, and the posture data includes the position data and direction data of the end effector; Divide the workspace of the end effector into a cubic grid, where each grid unit is a voxel; Count the number of visits of the end effector in each voxel, select the voxels with a visit number higher than the set threshold as high-frequency voxels, and record the center position of each high-frequency voxel; In each high-frequency voxel, the direction data is clustered using a mean shift clustering algorithm, the optimal bandwidth is determined by maximizing the silhouette coefficient, and the cluster center is taken as the typical direction; The center position of each high-frequency voxel and the typical orientation obtained by clustering are combined to generate a set of representative poses of the end effector.

[0006] Optionally, solving each joint angle of the robot corresponding to each representative posture by inverse kinematics includes: For each base position, converting each end effector representative posture in the global coordinate system in the representative posture set into a local posture in the base coordinate system; A closed inverse kinematics algorithm is used to solve the robot joint angles corresponding to each local posture.

[0007] Optionally, the adaptability of the base position to each representative posture of the end effector is evaluated based on the solved joint angles, and a comprehensive score is given to the base position, including: Calculate the joint margin score corresponding to each representative posture based on the solved key angles; According to the current base position and the angles of each joint of the robot corresponding to the current representative posture, the Jacobian matrix of the robot is determined, and the operability score corresponding to the current representative posture is calculated by Jacobian matrix singular value analysis; A comprehensive score for the current base position is calculated based on the joint margin scores and maneuverability scores of all representative poses.

[0008] Optionally, calculating the joint margin score corresponding to each representative posture based on the solved joint angles includes: According to the allowed angle range of each joint, calculate the median angle of each joint: ,in, q mid,i is the median angle of the i-th joint, q min,i , q max,i are the minimum and maximum angles allowed for the i joints respectively; Calculate the distance between the current actual angle of each joint and its median angle: ,in, q dist,i is the distance between the current angle of the i-th joint and its value, q pose,i is the actual angle of the i-th joint; Normalize the distance between the current angle of each joint and its median value to the range [0, 1]: ,in, m norm,i is the normalized value of the distance between the current angle of the i-th joint and its median value, q dist,i,max For the i The maximum distance that a joint is allowed to deviate from, q dist,i,max = ; Calculate the margin score for each joint: score JM,i =1- m norm,i , where score JM,i Score the margin of the i-th joint; Take the average of the margin scores of all joints to get the final joint margin score corresponding to the current representative posture: , where score JM is the final joint margin score corresponding to the current representative posture, n is the number of robot joints.

[0009] Optionally, calculating the operability score corresponding to each representative posture by Jacobian matrix singular value analysis includes: The Jacobian matrix J Decomposition into linear components Jlinear and the angular velocity component J angular ; Calculate the linear components separately J linear and the angular velocity component J angular The product matrix with its transpose: A linear = J linear J T linear and A angular = J angular J T angular ; Calculating the Matrix A linear The maximum eigenvalue of l max ( A linear ) and the minimum eigenvalue l min ( A linear ), and the matrix A angular The maximum eigenvalue of l max ( A linear ) and the minimum eigenvalue l min ( A linear ); Calculating the Matrix A linear condition number m linear And the matrix A angular condition number m angular ,in: ; Calculate the linear operability score: ; Calculate the angular velocity maneuverability score: ; Calculate the final operability score: score M =score LM + score AM .

[0010] Optionally, it also includes: If the solution of the robot joint angle corresponding to each representative posture solved by inverse kinematics does not exist, or any joint angle solved by inverse kinematics exceeds the corresponding angle allowable range, the corresponding representative posture will be marked as unachievable and excluded from the score calculation; If a solution for the joint angle obtained by inverse kinematics exists and the corresponding joint angle does not exceed the allowed angle range, the corresponding representative posture is marked as achievable and participates in the subsequent scoring calculation.

[0011] Optionally, the comprehensive score of the current base position is calculated by the following formula: ,in, score final An overall score for the base location.

[0012] Optionally, the neural network model includes a multi-layer perceptron model, the multi-layer perceptron model includes an input layer, three hidden layers and an output layer, the activation functions of the three hidden layers are all ReLU; the model training uses mean square error as the loss function; The input of the multi-layer perceptron model is the base position parameters, which include the coordinates of the base and the rotation angle of the base around the vertical axis, and the output is the comprehensive score of the base position.

[0013] Optionally, generating a plurality of candidate base positions in a continuous space of base position parameters, using a trained neural network model to predict a comprehensive score of each candidate base position, and selecting an optimal base position includes: Determine the feasible range of the base position parameters based on the physical limitations of the surgical robot and the actual surgical scenario requirements; Continuously generate all possible candidate base positions within the feasible space of base position parameters according to the set parameter intervals; The trained neural network model is used to predict the comprehensive score of each candidate base position in turn, and the predicted comprehensive scores of all candidate positions are ranked; The base position with the highest comprehensive score is output as the optimal base position.

[0014] The beneficial effects of the present invention are: The present invention analyzes the surgeon's working mode and extracts a representative posture set, so as to accurately identify the doctor's main operating area and direction during the operation, reduce the need for the robot arm to frequently adjust its posture during the operation, and significantly improve the operation efficiency. The present invention quantifies the support degree of the base position for the commonly used postures through inverse kinematics solution and comprehensive scoring mechanism, ensures that the robot arm is away from the joint limit and singularity state during the operation, and reduces the risk caused by the sudden change of the robot arm posture during the operation. The base position and the corresponding comprehensive score are used as the training data of the neural network model. The model captures the complex relationship between the base position and the score through nonlinear mapping. The trained neural network model is used to quickly predict the optimal base position in the actual operation scene, so as to ensure that the robot can cover the commonly used postures of the end effector and avoid joint limits and singularities. The optimized base position can adapt to the doctor's operating habits, so that the doctor can maintain his familiar operating style during the operation, meet the doctor's personalized needs, improve the operability and work efficiency of the robot in performing the operation, thereby improving the fluency and success rate of the operation.

[0015] The system of the present invention has other characteristics and advantages, which will be apparent from the drawings incorporated herein and the following detailed description, or will be described in detail in the drawings incorporated herein and the following detailed description, which together serve to explain the specific principles of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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 in conjunction with the accompanying drawings, in which like reference numerals generally represent like components.

[0017] Figure 1 A flow chart of the steps of a method for optimizing the position of a surgical robot base based on working mode analysis according to the present invention is shown. DETAILED DESCRIPTION

[0018] The present invention will be described in more detail below with reference to the accompanying drawings. Although preferred embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, 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.

[0019] like Figure 1 As shown, a method for optimizing the position of a surgical robot base based on working mode analysis according to the present invention includes: S1: By analyzing the end-effector motion data recorded by the surgeon during surgery, a representative set of end-effector postures representing the surgeon's working mode is extracted; This step mainly analyzes the working mode, extracts the end effector postures commonly used by surgeons, and reduces the computational complexity, including: S101: In an actual operation, recording, by means of a robot sensor, the posture data of the end effector when the surgeon operates the surgical robot, wherein the posture data includes the position data and the direction data of the end effector; In one example, a surgical robot operator (doctor) completes multiple tasks (such as suturing, picking and placing, etc.), and the surgical robot's built-in sensors record the surgeon's end effector posture data during the operation, collecting about 20,000 to 30,000 end effector posture data points. The position data in the posture data is a three-dimensional coordinate, and the direction data is a three-dimensional rotation vector oh =( oh x , oh y , oh z ). Alternatively, the surgeon completes the surgical task on the simulation platform and records the position and orientation data of the end effector.

[0020] S102: Divide the working space of the end effector into a cubic grid, where each grid unit is a voxel; In one example, the workspace is divided into small cubes (voxels), with a typical size of 0.02m on a side, and each voxel is assigned a unique ID to count the frequency of position access by the end effector. The workspace is a three-dimensional spatial area that defines all the positions and directions that the end effector of the surgical robot can reach. In the surgical scene, the workspace includes the area above the operating table, which is the main area for doctors to perform surgical operations.

[0021] S103: Counting the number of visits of the end effector in each voxel, selecting voxels with a number of visits higher than a set threshold as high-frequency voxels, and recording the center position of each high-frequency voxel; In one example, the position data of the end effector during the surgery, including the x, y, and z coordinates, are collected, and for each recorded end effector position, the voxel where it is located is determined. The number of visits of the effector to each voxel is counted, and a fixed number of visits is set as a threshold (for example, the number of visits is greater than 100). Voxels exceeding the threshold are marked as high-frequency voxels, which are used to identify high-frequency visit areas and record their center positions.

[0022] S104: in each high-frequency voxel, clustering the direction data using a mean shift clustering algorithm, determining an optimal bandwidth by maximizing a silhouette coefficient, and taking a cluster center as a typical direction; In one example, within each high-frequency voxel, the orientation data (rotation vector oh ) is used for clustering, the optimal bandwidth is determined by maximizing the silhouette coefficient, and the rotation vector of the cluster center is taken as the typical direction.

[0023] S105: Combining the center position of each high-frequency voxel and the typical direction obtained by clustering to generate a representative posture set of the end effector.

[0024] In one example, the center position of each high-frequency voxel and the corresponding clustering direction are combined into a data pair, and all the data pairs are used as a representative posture set of the end effector.

[0025] Through the above processing, the original data is compressed into a set of key postures, reducing the amount of subsequent calculations.

[0026] S2: Randomly generate multiple base positions within the feasible range of the base. For any base position, solve the robot joint angle corresponding to each representative posture at the base position through inverse kinematics. Based on the solved joint angles, evaluate the adaptability of the base position to each representative posture of the end effector, and give a comprehensive score to the base position. This step specifically includes: S201: randomly generating multiple base positions within a feasible range of the base; In one example, randomly generated within the feasible range of base parameters M =20,000 base positions ( X j , Y j ,Θ j ), covering the horizontal position and rotation angle. The base position is represented by the horizontal coordinate ( X , Y ) and a rotation angle (θ) about the vertical axis (Z axis). The vertical position (Z axis) is usually pre-fixed by the patient's anatomy and therefore does not need to be calculated. X , Y ,Θ) defines the position and orientation of the robot base coordinate system relative to the world coordinate system. The pose of the end effector is determined by its joint angles relative to the base coordinate system. Therefore, the base position directly affects the reach of the end effector in the global workspace.

[0027] S202: For each base position, convert each representative posture of the end effector in the global coordinate system in the representative posture set into a local posture in the base coordinate system; In one example, the transformation matrix from the base coordinate system to the global coordinate system is:

[0028] in, R z (Θ) is the rotation matrix around the Z axis, Z fixed It is a fixed value in the vertical direction (determined by the patient's anatomy).

[0029] The representative posture of the end effector in the global coordinate system Convert to local pose in base coordinate system : .

[0030] S203: Using a closed inverse kinematics algorithm to solve the angles of each joint of the robot corresponding to each local posture.

[0031] In one example, based on the local pose ,The joint angles are directly analyzed using closed-form inverse kinematics using the robot geometric model.

[0032] If the solution of the robot joint angle corresponding to each representative posture solved by inverse kinematics does not exist (such as the target posture exceeds the robot workspace), or any joint angle solved by inverse kinematics exceeds its allowable angle range, the corresponding representative posture will be marked as unattainable and excluded from the score calculation; if the solution of the joint angle solved by inverse kinematics exists and the corresponding joint angle does not exceed the allowable angle range, the corresponding representative posture will be marked as achievable and participate in the subsequent score calculation.

[0033] Specifically, for each base position, inverse kinematics is used to verify whether it can achieve a representative posture. If it cannot be solved (such as exceeding the workspace or mechanical limitations), the posture will be marked as "unachievable" and excluded from the score calculation. Even if an inverse kinematic solution exists, it is still necessary to verify whether the joint angle is within the allowed range. If a joint angle exceeds the limit (such as approaching the limit value), the posture is still considered invalid and will also be marked as "unachievable" and excluded from the score calculation.

[0034] Then, the adaptability of the base position to each representative posture of the end effector is evaluated based on the solved joint angles, and the base position is comprehensively scored. Specifically, the following scoring steps S204-S206 are used to quantify the support degree of the base position for the joint margin and operability, that is, to quantify the adaptability of the base position to the robot's motion performance.

[0035] S204: Calculate the joint margin score corresponding to each representative posture based on the solved key angles, and the specific method includes: According to the allowed angle range of each joint, calculate the median angle of each joint: ,in, q mid,i is the median angle of the i-th joint, q min,i , q max,i are the minimum and maximum angles allowed for the i joints respectively; Calculate the distance between the current actual angle of each joint and its median angle: ,in, q dist,i is the distance between the current angle of the i-th joint and its value, q pose,i is the actual angle of the i-th joint; Normalize the distance between the current angle of each joint and its median value to the range [0, 1]: ,in, m norm,i is the normalized value of the distance between the current angle of the i-th joint and its median value. The larger the value, the closer the joint is to the limit. q dist,i,max For the i The maximum distance that a joint is allowed to deviate from, q dist,i,max = ; Calculate the margin score for each joint: score JM,i =1- m norm,i , where score JM,i Score the margin of the i-th joint; Take the average of the margin scores of all joints to get the final joint margin score corresponding to the current representative posture: , where score JM is the final joint margin score corresponding to the current representative posture, ranging from [0, 1], n is the number of robot joints. score JM A higher score indicates greater joint margin and greater operational safety.

[0036] S205: Determine the Jacobian matrix of the robot according to the current base position and the angles of each joint of the robot corresponding to the current representative posture, and calculate the operability score corresponding to the current representative posture by Jacobian matrix singular value analysis; the specific method includes: The Jacobian matrix J Decomposition into linear components J linear and the angular velocity component J angular ; Calculate the linear components separately J linear and the angular velocity component J angular The product matrix with its transpose: A linear = J linear J T linear and A angular = J angular J T angular ; Calculating the Matrix A linear The maximum eigenvalue of l max ( A linear ) and the minimum eigenvalue l min ( A linear ), and the matrix A angular The maximum eigenvalue of l max ( A linear ) and the minimum eigenvalue l min ( A linear ); Calculating the Matrix A linear condition number m linear And the matrix A angular condition number m angular ,in: ; The larger the condition number, the closer it is to the singular configuration (generally, a condition number greater than 10 is considered close to singularity), and the worse the operability.

[0037] Calculate the linear operability score: ; Calculate the angular velocity maneuverability score: ; Linear operability score LM and angular velocity maneuverability score AM The range is (0, 1]. The closer the score is to 1, the better the operability is, and the closer it is to 0, the higher the singularity risk is.

[0038] Calculate the final operability score: score M =score LM + score AM ; score M The range is (0, 2], which is used to evaluate the overall flexibility of the robot arm. The higher the score, the more suitable it is for complex operations.

[0039] S206: The joint margin scores and operability scores of all achievable representative postures are summed up to obtain a comprehensive score of the current base position, and the calculation formula is: ,in, score final An overall score for the base location.

[0040] Overall Rating score final It reflects the adaptability of a base position to all representative postures of the end effector. The higher the score, the better the motion performance and smoothness of the robot in executing the representative postures of the end effector at the base position, the stronger the adaptability, and the more representative postures that can be achieved.

[0041] Repeat the above steps S202-S206 to complete the calculation of the comprehensive score for each base position.

[0042] S3: taking each base position parameter and its corresponding comprehensive score as an input-output data pair to form a training data set, and using the training data set to train a neural network model to learn a mapping relationship between the base position and the score; In this step, the neural network model includes a multi-layer perceptron model (MLP), which includes an input layer (3 nodes), three hidden layers (48→96→192 nodes) and an output layer (1 node). The activation functions of the three hidden layers are all ReLU; the model training adopts the Adam optimizer (such as learning rate 0.0001, momentum parameter β 1=0.9, β2=0.999), mean square error as the loss function; The input of the multi-layer perceptron model is the base position parameters, which include the coordinates of the base and the rotation angle of the base around the vertical axis, and the output is the comprehensive score of the base position score final .

[0043] S4: In the actual surgical scenario, multiple candidate base positions are generated within the feasible range of the base seat, and the trained neural network model is used to predict the comprehensive score of each candidate base position to select the optimal base position.

[0044] This step specifically includes: S401: Determine the feasible range of the base position parameters according to the physical limitations of the surgical robot and the actual surgical scene requirements. For example, the feasible range of the horizontal position is: X ∈[ X min , X max ], Y ∈[ Y min , Y max ], the feasible range of rotation angle is: Θ∈[Θ min ,Θ max ].

[0045] S402: continuously generating all possible candidate base positions within the feasible space of base position parameters according to the set parameter interval; For example, to set the horizontal position interval: Δ X =0.005 m , Δ Y =0.005 m , rotation angle interval: ΔΘ=0.5°; all possible candidate base positions are generated at intervals within the parameter range, and the total number N is: .

[0046] S403: using the trained neural network model to predict the comprehensive score of each candidate base position in turn, and sorting the predicted comprehensive scores of all candidate positions; Specifically, the parameters of each candidate base position are input into the trained MLP model, the prediction score is output, and the prediction scores of all candidate positions are sorted from high to low.

[0047] S404: Output the base position with the highest comprehensive score as the optimal base position.

[0048] Specifically, the base position with the highest score is output as the optimal base position and recommended to the doctor. Optionally, the top N candidates (such as Top 10) can also be output for reference of intraoperative fine-tuning.

[0049] The core of the method of the present invention is to extract the representative postures used frequently by a specific doctor by analyzing the historical operation data (such as the end effector posture and working mode), and train the MLP model based on these postures. Therefore, the model directly reflects the doctor's operating habits. In the actual application process, for a single doctor and a single surgical task scenario, an MLP model can be trained for the doctor to achieve fully personalized adaptation. The pre-trained MLP model is integrated into the surgical robot operating system as a preoperative planning module. When the doctor performs a specific surgical task, the pre-trained MLP model is used to generate the optimal base position that meets the doctor's operating habits, thereby maximizing the operability and work efficiency of the surgical robot, thereby improving the safety and efficiency of the operation.

[0050] When multiple doctors share a surgical robot and the surgical robot is suitable for a variety of surgical task scenarios, multiple different end-effector representative posture sets can be generated by collecting the end-effector motion data of different doctors in different surgical task scenarios. Each representative posture set corresponds to a surgical task scenario of a doctor. Different training data sets are used to train an MLP model separately for each specific surgical task of each doctor. The pre-trained multiple MLP models are integrated into the surgical robot operating system. When each doctor performs a specific surgical task, the corresponding MLP model is selected to generate the optimal base position that meets the doctor's operating habits.

[0051] In summary, the method of the present invention can generate customized base position suggestions for the unique working modes of different operators, significantly reducing the reliance on traditional experience, and can assist the medical team in quickly determining safe and efficient base positions before surgery, ensuring that postures frequently used during surgery can be performed safely and efficiently. The present invention is applicable to a variety of surgical robot-assisted minimally invasive surgery scenarios, and can significantly reduce surgical risks and improve operational efficiency.

[0052] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes 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 optimizing the position of a surgical robot base based on working mode analysis, characterized in that: include: By analyzing the end-effector motion data recorded by the surgeon during the operation, a representative set of end-effector postures representing the surgeon's working mode is extracted; Multiple base positions are randomly generated within the feasible range of the base. For any base position, the robot joint angle corresponding to each representative posture at the base position is solved by inverse kinematics. The adaptability of the base position to each representative posture of the end effector is evaluated based on the solved joint angle, and the base position is comprehensively scored. Taking each base position parameter and its corresponding comprehensive score as an input-output data pair to form a training data set, and using the training data set to train a neural network model to learn the mapping relationship between the base position and the score; In actual surgical scenarios, multiple candidate base positions are generated within the feasible range of the base seat, and the trained neural network model is used to predict the comprehensive score of each candidate base position to select the optimal base position.

2. The method according to claim 1, characterized in that The method extracts a representative set of end effector postures representing the surgeon's working mode by analyzing the end effector motion data recorded by the surgeon during the operation, including: In actual surgery, the robot sensor records the posture data of the end effector when the surgeon operates the surgical robot, and the posture data includes the position data and direction data of the end effector; Divide the workspace of the end effector into a cubic grid, where each grid unit is a voxel; Count the number of visits of the end effector in each voxel, select the voxels with a visit number higher than the set threshold as high-frequency voxels, and record the center position of each high-frequency voxel; In each high-frequency voxel, the direction data is clustered using a mean shift clustering algorithm, the optimal bandwidth is determined by maximizing the silhouette coefficient, and the cluster center is taken as the typical direction; The center position of each high-frequency voxel and the typical orientation obtained by clustering are combined to generate a set of representative poses of the end effector.

3. The method according to claim 1, characterized in that Solving the robot joint angles corresponding to each representative posture by inverse kinematics includes: For each base position, converting each end effector representative posture in the global coordinate system in the representative posture set into a local posture in the base coordinate system; A closed inverse kinematics algorithm is used to solve the robot joint angles corresponding to each local posture.

4. The method according to claim 1, characterized in that: The adaptability of the base position to each representative posture of the end effector is evaluated based on the solved joint angle, and the base position is comprehensively scored, including: Calculate the joint margin score corresponding to each representative posture based on the solved key angles; According to the current base position and the angles of each joint of the robot corresponding to the current representative posture, the Jacobian matrix of the robot is determined, and the operability score corresponding to the current representative posture is calculated by Jacobian matrix singular value analysis; A comprehensive score for the current base position is calculated based on the joint margin scores and maneuverability scores of all representative poses.

5. The method according to claim 4, characterized in that The calculation of the joint margin score corresponding to each representative posture based on the solved joint angles includes: According to the allowed angle range of each joint, calculate the median angle of each joint: ,in, q mid,i is the median angle of the i-th joint, q min,i , q max,i are the minimum and maximum angles allowed for the i joints respectively; Calculate the distance between the current actual angle of each joint and its median angle: ,in, q dist,i is the distance between the current angle of the i-th joint and its value, q pose,i is the actual current angle of the i-th joint; Normalize the distance between the current angle of each joint and its median value to the range [0, 1]: ,in, μ norm,i is the normalized value of the distance between the current angle of the i-th joint and its median value, q dist,i,max For the i The maximum distance that a joint is allowed to deviate from, q dist,i,max = ; Calculate the margin score for each joint: score JM,i =1- μ norm,i , where score JM,i Score the margin of the i-th joint; Take the average of the margin scores of all joints to get the final joint margin score corresponding to the current representative posture: , where score JM is the final joint margin score corresponding to the current representative posture, n is the number of robot joints.

6. The method according to claim 5, characterized in that The calculation of the operability score corresponding to each representative posture by Jacobian matrix singular value analysis includes: The Jacobian matrix J Decomposition into linear components J linear and the angular velocity component J angular ; Calculate the linear components separately J linear and the angular velocity component J angular The product matrix with its transpose: A linear = J linear J T linear and A angular = J angular J T angular ; Calculating the Matrix A linear The maximum eigenvalue of λ max ( A linear ) and the minimum eigenvalue λ min ( A linear ), and the matrix A angular The maximum eigenvalue of λ max ( A linear ) and the minimum eigenvalue λ min ( A linear ); Calculating the Matrix A linear condition number μ linear And the matrix A angular condition number μ angular ,in: ； Calculate the linear operability score: ; Calculate the angular velocity maneuverability score: ; Calculate the final operability score: score M =score LM + score AM 。 7. The method according to claim 6, characterized in that Also includes: If the solution of the robot joint angle corresponding to each representative posture solved by inverse kinematics does not exist, or any joint angle solved by inverse kinematics exceeds the corresponding angle allowable range, the corresponding representative posture will be marked as unachievable and excluded from the score calculation; If a solution for the joint angle obtained by inverse kinematics exists and the corresponding joint angle does not exceed the allowed angle range, the corresponding representative posture is marked as achievable and participates in the subsequent scoring calculation.

8. The method according to claim 7, characterized in that The comprehensive score of the current base position is calculated by the following formula: ,in, score final An overall score for the base location.

9. The method according to claim 1, characterized in that: The neural network model includes a multi-layer perceptron model, which includes an input layer, three hidden layers and an output layer, and the activation functions of the three hidden layers are all ReLU; the model training uses mean square error as the loss function; The input of the multi-layer perceptron model is the base position parameters, which include the coordinates of the base and the rotation angle of the base around the vertical axis, and the output is the comprehensive score of the base position.

10. The method according to claim 1, characterized in that The method generates a plurality of candidate base positions in the continuous space of base position parameters, predicts the comprehensive score of each candidate base position by using the trained neural network model, and selects the optimal base position, including: Determine the feasible range of the base position parameters based on the physical limitations of the surgical robot and the actual surgical scenario requirements; Continuously generate all possible candidate base positions within the feasible space of base position parameters according to the set parameter intervals; The trained neural network model is used to predict the comprehensive score of each candidate base position in turn, and the predicted comprehensive scores of all candidate positions are ranked; The base position with the highest comprehensive score is output as the optimal base position.

Images