A method for controlling the degrees of freedom of movement of an ophthalmic surgical assistant system

Abstract

The application discloses a motion degree of freedom control method of an ophthalmic auxiliary surgery system, and belongs to the technical field of medical robots. The method is based on a five-degree-of-freedom configuration, and comprises a translation module and an RCM rotation module. First, a multi-stage coordinate system and a kinematics forward solution model are established; the geometric center of the RCM is moved to a safe distance above a target position through visual guidance; during posture adjustment, a weighted pseudo-inverse algorithm is used to solve a translation compensation displacement, so that the position of a needle tip remains constant; an error mapping database of rotation angle and needle tip spatial deviation is further established, a compensation amount is added to a translation instruction at a frequency of not less than 1000 Hz in real time, and drift caused by mechanical error is eliminated; meanwhile, posture and feeding decoupling control is realized. The application also integrates functions such as eyeball follow-up tracking, axial safe needle withdrawal and dynamic assistance teaching. The method realizes high-precision virtual RCM maintenance under limited degrees of freedom, and significantly improves the safety and convenience of operation.

Description

Technical Field

[0001] This invention relates to the medical field, specifically to a method for controlling the degrees of freedom of motion in an ophthalmic auxiliary surgical system. Background Technology

[0002] Ophthalmic surgery, especially posterior segment surgeries such as vitrectomy, retinal reattachment, and epiretinal membrane detachment, places extremely high demands on the precision and safety of surgical instruments. During the procedure, instruments must be inserted into the eye through tiny puncture incisions in the sclera, using these incisions as fulcrums for directional adjustment and axial feed. If the instruments experience lateral wobbling or lateral forces at this fulcrum, it can lead to serious complications such as scleral tearing, vitreous prolapse, or even retinal damage. Therefore, ophthalmic surgical robots must be able to ensure that the surgical instruments perform pure rotational motion around a fixed spatial point, known as the remote center of motion.

[0003] In existing technologies, the implementation of RCM functions is mainly divided into two categories: mechanical implementation and kinematic control implementation. Mechanical RCM mechanisms typically employ structures such as parallelogram linkages, double four-bar linkages, or arc-shaped guide rails, forcing the instrument to rotate around a fixed point through the geometric constraints of rigid components. For example, publicly available technologies disclose RCM devices constructed using parallel mechanisms, which form a virtual rotation center through the coupled motion of multiple sets of linkages. However, mechanical RCM mechanisms have significant drawbacks: firstly, the mechanisms are complex and bulky, easily interfering with surgical microscopes or the patient's face; secondly, adjusting the position of the fixed RCM point is difficult, making the preoperative alignment process cumbersome and time-consuming; and thirdly, the mechanism's rigidity and weight are significant, increasing the burden on the drive system and hindering the achievement of precise force feedback.

[0004] To overcome the limitations of mechanical RCMs, another approach employs redundant-degree-of-freedom robots combined with kinematic control to construct "virtual RCMs." This involves using the coordinated movements of multiple robot joints to mathematically ensure the instrument tip's position is fixed while its direction is adjustable. For example, existing research has disclosed virtual RCM control methods based on six-degree-of-freedom serial robotic arms, using inverse kinematic constraints to rotate the arm's end effector around a preset pivot point. However, such methods typically require at least six degrees of freedom to simultaneously satisfy both position and attitude constraints, leading to high system costs and complex control algorithms. For five-degree-of-freedom configurations (three translational degrees of freedom plus two rotational degrees of freedom), maintaining tip fixation during rotation is an underactuated problem, requiring active compensation from the translational axes. Existing control strategies often employ simple proportional-derivative control, which struggles to eliminate tip drift caused by mechanical assembly errors, non-ideal intersections of rotational axes, etc., especially during high-speed or small-range attitude adjustments, where drift can reach millimeter levels, far exceeding the safety threshold for ophthalmic surgery (typically required to be less than 0.1 mm).

[0005] Furthermore, existing ophthalmic surgical robotic systems generally lack the ability to adapt to the subtle physiological movements of the patient's eyeballs. During surgery, the patient's eyeballs undergo continuous positional and postural changes due to respiration, heartbeat, extraocular muscle tone variations, and involuntary eye movements. If the robot still uses the initially fixed RCM point as the control target, relative displacement and stress will occur between the instrument and the scleral incision, increasing the risk of tissue tearing. Although some studies have proposed dynamic tracking methods based on visual feedback, most of them use simple proportional control and do not consider visual sampling delays and noise, resulting in tracking lag or jitter.

[0006] Furthermore, the existing system's needle withdrawal strategy in emergency situations is relatively simple, usually withdrawing directly along a fixed coordinate axis rather than along the instrument's own axis, which can easily cause secondary damage to the incision. Additionally, when manually adjusting the robot's pose before surgery, the large weight of the mechanism makes operation cumbersome and lacks effective assisted teaching functions.

[0007] In summary, how to achieve high-precision, dynamically adaptive virtual RCM control with multiple safety protection mechanisms on a low-cost, compact five-DOF ophthalmic surgical robot platform is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention provides a method for controlling the degrees of freedom of motion in an ophthalmic auxiliary surgical system.

[0009] This invention is achieved through the following technical solution:

[0010] The present invention provides a method for controlling the degrees of freedom of motion in an ophthalmic auxiliary surgical system, the ophthalmic auxiliary surgical system comprising a translation module, an RCM rotation module mounted at the end of the translation module, and a central control unit;

[0011] The translation module includes a trolley X-axis, a trolley Y-axis, and a trolley Z-axis arranged along three-dimensional orthogonal directions in space. Each axis is driven by a linear drive mechanism and equipped with a position feedback unit.

[0012] The RCM rotating module includes an RCM horizontal rotating joint that realizes horizontal yaw rotation around a first rotating axis and an RCM vertical ring mechanism that realizes vertical pitch rotation around a second rotating axis.

[0013] The central control unit is electrically connected to the translation module and the RCM rotation module, and is used to execute the following control steps:

[0014] S1. Multi-coordinate system modeling and kinematic mapping construction;

[0015] Establish a multi-level coordinate system including a base coordinate system, a translation end coordinate system, an RCM geometric coordinate system, and a tool coordinate system, wherein:

[0016] The base coordinate system is fixedly set on the equipment base, and the Z-axis direction is consistent with the Z-axis guide rail direction of the trolley.

[0017] The translation end coordinate system is set at the equivalent end reference point formed by the superposition of the X-axis, Y-axis and Z-axis motion of the trolley. The equivalent end reference point is determined by the combination of the three-axis displacements in the base coordinate system.

[0018] The RCM geometric coordinate system takes the theoretical spatial intersection of the first rotation axis and the second rotation axis as its origin;

[0019] The tool coordinate system is set at the tip of the surgical needle;

[0020] Based on the above coordinate system, the forward kinematics model is constructed using the DH parameter method:

[0021]

[0022] in, , , For the displacement variable of the translation module, , For the rotational joint angle variable, The fixed structural length from the origin of the RCM geometric coordinate system to the needle tip;

[0023] S2. Visual-guided spatial coarse positioning control;

[0024] The central control unit acquires the position data of the needle insertion point in the eyeball in the image coordinate system through a vision sensor, and converts the position data into the target position in the base coordinate system through a pre-calibrated hand-eye transformation matrix. ;

[0025] Based on the spatial deviation between the current RCM geometric coordinate system origin position and the target position, generate three-axis displacement control commands for the translation module;

[0026] The translation module is driven to move along a continuous trajectory, so that the origin of the RCM geometric coordinate system reaches a reference position located 20mm to 80mm above the target position along the Z-axis.

[0027] The continuous trajectory is generated by an S-curve velocity planning algorithm, and the maximum velocity, acceleration and jerk are constrained to suppress motion impact.

[0028] S3. RCM centering control based on inverse kinematics constraints;

[0029] Upon receiving attitude adjustment input, the central control unit calculates the unit vector of the needle tip direction based on the current rotation angle;

[0030] The position of the needle tip in the base coordinate system is constrained to a fixed point. Establish a position error model;

[0031] The mapping relationship between translation variables and needle tip position is established based on the Jacobian matrix, and the translation compensation displacement is solved by the weighted pseudo-inverse algorithm.

[0032] The weighted pseudo-inverse solution satisfies the minimum translational displacement norm condition to reduce the mechanism disturbance caused by multi-axis linkage, thereby maintaining the stability of the needle tip spatial position during rotation.

[0033] S4. Virtual RCM dynamic compensation control based on error mapping;

[0034] To address the needle tip spatial offset caused by the non-ideal intersection of the rotation axes, the central control unit executes the following control process:

[0035] An error mapping database is established to characterize the functional mapping relationship between rotation angle parameters and needle tip spatial deviation, wherein the rotation angle parameters include the angle of the RCM horizontal rotation joint and the angle of the RCM vertical ring mechanism.

[0036] The database is rotated by an angle. , As input variables, the needle tip spatial deviation vector For output variables;

[0037] The error mapping database is obtained by pre-calibrating the spatial position deviation of the tip of the RCM rotating module under different combinations of rotation angles. The calibration process establishes the correspondence between the rotation angle and the spatial deviation based on actual measurement data.

[0038] During system operation, the current rotation angle is acquired at a sampling frequency of not less than 1000Hz;

[0039] Based on the rotation angle, the current spatial deviation is calculated in the error mapping database using an interpolation algorithm;

[0040] The spatial deviation is added as a compensation amount to the translation module control command:

[0041] ;

[0042] The needle tip offset is offset by the compensation displacement of the translation module, so that the needle tip can maintain a stable virtual teleoperation center in space.

[0043] S5. Operation control based on attitude and feed decoupling;

[0044] The central control unit decomposes the operation input into attitude adjustment components and axial feed components;

[0045] During the attitude adjustment process, compensation control is performed in steps S3 and S4.

[0046] During axial feed, the rotary joint angle remains constant, and a translational speed command is generated based on the unit vector of the needle tip direction. The surgical needle is made to perform a linear feed motion along its own axis.

[0047] The "motion freedom control" in multi-degree-of-freedom medical robots essentially involves the problem of precise constraint of spatial pose, especially in ophthalmic surgery scenarios, where surgical instruments need to adjust their posture around a fixed spatial point, which is the remote center of motion (RCM). Since it is difficult for real mechanical structures to achieve ideal multi-axis convergence, it is necessary to construct a "virtual RCM" at the motion level through control algorithms, thereby achieving mathematical stability of the rotation center.

[0048] The translation module provides positional freedom in three-dimensional space, essentially functioning as a linear displacement actuator in a Cartesian coordinate system. Through independent drive in the X, Y, and Z directions, the end effector can be moved to any reachable spatial position. Unlike traditional serial robotic arms, this type of structure offers better rigidity and positioning accuracy, but lacks attitude adjustment capabilities, thus requiring its use in conjunction with a rotation module.

[0049] The RCM rotation module achieves attitude adjustment through two mutually orthogonal rotational degrees of freedom. The horizontal yaw rotational joint changes the azimuth angle of the instrument in the horizontal plane, while the vertical ring mechanism achieves pitch motion through a gimbal-like structure. The combination of the two allows the surgical needle to be spatially oriented around a theoretical RCM point.

[0050] In coordinate modeling, the purpose of introducing multiple coordinate systems is to decompose complex mechanical motion into multiple simple rigid body transformations. The base coordinate system serves as a global reference system, used to uniformly describe the position of all components in the system; the translation end coordinate system reflects the position state of the translation module output; the RCM geometric coordinate system describes the position of the rotation center; and the tool coordinate system is directly related to the surgical needle tip and is the ultimate embodiment of control precision.

[0051] By establishing a kinematic model using the DH parameter method, the system can map joint space variables (displacement and angle) to position coordinates in Cartesian space. The given forward kinematics expression essentially describes the combination of "translation, rotation, and fixed length," where the order of the rotation matrices reflects the series relationship of the actual mechanical structure.

[0052] The coarse localization step guided by vision addresses the "global localization problem," that is, moving the system from its initial position to the vicinity of the surgical area. The vision system determines the needle insertion point by recognizing anatomical features of the eyeball (such as the iris and sclera structures), and then unifies the image space and mechanical space through hand-eye calibration. The hand-eye calibration matrix here is essentially a rigid body transformation matrix used to describe the relationship between the camera coordinate system and the robot's base coordinate system.

[0053] Reverse kinematics (RCM) alignment control is one of the key steps in the system. Its core lies in treating the needle tip position as a constraint point and solving for it using inverse kinematics to ensure that the point remains stationary during rotation. Since the system has redundant degrees of freedom (3 translational degrees of freedom are used to compensate for position changes caused by 2 rotational degrees of freedom), it is necessary to select the optimal solution using optimization methods (such as least squares or weighted pseudo-inverse) to avoid unnecessary mechanism motion.

[0054] Virtual RCM dynamic compensation control further considers non-ideal factors in actual mechanical systems, such as incomplete intersection of rotation axes, assembly errors, and structural deformation. These errors can cause minute displacements at the needle tip during pure rotational motion. By establishing a mapping relationship between angle and spatial deviation and compensating in real time during operation, mechanical errors can be "eliminated" at the control level, thereby enabling the system to exhibit ideal RCM characteristics.

[0055] In decoupled control, separating attitude adjustment from axial feed reduces control complexity and improves operational safety. The attitude adjustment phase focuses primarily on directional control, while the feed phase executes linear motion along the needle axis, thus avoiding tissue damage caused by lateral shearing forces. This decoupling method is of great significance in minimally invasive surgical robots.

[0056] Furthermore, the specific steps for establishing the multi-level coordinate system mapping model in step S1 include:

[0057] The base coordinate system is established at the center of the fixed base of the ophthalmic auxiliary surgery system, with its Z-axis parallel to the guide rail direction of the Z-axis of the trolley;

[0058] The translation end coordinate system is established at the intersection point of the three axes of the translation module, and the translation transformation relationship between the base coordinate system and the translation end coordinate system is defined by a homogeneous transformation matrix.

[0059] The rotation axis of the RCM horizontal rotary joint is defined as the first rotation axis, and the rotation axis of the RCM vertical ring mechanism is defined as the second rotation axis. The RCM geometric coordinate system is established based on the theoretical intersection of the first rotation axis and the second rotation axis.

[0060] During the establishment of the RCM geometric coordinate system, a three-dimensional spatial offset parameter is introduced. It is used to describe the actual intersection point offset of the rotation axis caused by mechanical assembly error, and the offset parameter is embedded in the pose transformation matrix from the translation end coordinate system to the RCM geometric coordinate system;

[0061] The tool coordinate system is established at the position of the surgical needle tip, and the position of the needle tip relative to the origin of the RCM geometric coordinate system is defined as a fixed pendulum length vector; the central control unit constructs a rotation transformation matrix sequence by reading the encoder values ​​of the RCM rotation module in real time, thereby realizing the forward kinematic mapping calculation from the tool coordinate system to the base coordinate system.

[0062] Establishing a multi-level coordinate system is a fundamental task in robot control systems. Its purpose is to achieve unified modeling of complex systems by describing the motion relationships of different components in layers. The base coordinate system is usually selected on the fixed base of the device, which serves as a global reference frame, and all other coordinate systems are related to it through relative transformations.

[0063] The translation end coordinate system is used to describe the position state of the translation module output. In actual engineering, since the three axes are usually in series, there is no strictly defined "physical intersection point". Therefore, the concept of "equivalent end point" is used to combine the displacements of the three axes into a virtual reference point, which facilitates unified modeling and control.

[0064] The RCM geometric coordinate system is established based on the theoretical intersection of the two rotation axes. Ideally, this intersection is the teleoperation center. However, in actual mechanical systems, due to machining and assembly errors, the two rotation axes often have slight offsets, thus requiring the introduction of offset parameters. The offset is modeled. These offset parameters can be obtained through calibration experiments and compensated for in the kinematic model, thereby improving the system accuracy.

[0065] The tool coordinate system is directly linked to the surgical needle tip, and its positional accuracy directly determines the safety of the surgical procedure. By defining the needle tip position as a fixed pendulum length vector relative to the RCM coordinate system, rotational motion can be transformed into a simple spatial coordinate transformation, thereby simplifying the control model.

[0066] Real-time reading of encoder data enables the system to acquire the current joint angle and achieve a step-by-step mapping from the tool coordinate system to the base coordinate system by constructing a sequence of rotation matrices. This process is essentially a chain product of rigid body transformations, the result of which is used to update the position of the needle tip in space in real time, providing basic data for subsequent control.

[0067] Furthermore, the specific steps for performing vision-guided coarse spatial positioning control in step S2 include:

[0068] The central control unit acquires image signals of the patient's eyes through a visual sensor integrated into the surgical microscope, and extracts the needle insertion site features using the following processing method:

[0069] Iris edge contours were extracted using the Canny edge detection operator;

[0070] The iris center position is determined by performing a circle fitting on the iris edge points using the least squares method.

[0071] Determine the spatial location of the needle insertion point by combining a pre-designed anatomical model;

[0072] Depth information of the needle insertion point is obtained through binocular vision or structured light depth sensing.

[0073] The pixel coordinates in the image coordinate system are converted into three-dimensional physical coordinates in the base coordinate system using a preset hand-eye calibration matrix.

[0074] The central control unit generates the motion trajectory of the translation module using a seven-segment lifting speed curve planning algorithm based on the deviation between the current position and the target position. The maximum acceleration and jerk are preset according to the dynamic response characteristics of each axis drive system to suppress the inertial impact during the displacement stopping phase.

[0075] The visual perception module plays a crucial role in ophthalmic surgical robots, providing localization and feedback. Its core task is to extract stable and anatomically significant feature points from complex biological tissue images. The iris, due to its clear edges and high contrast, is often used as a localization reference; therefore, using edge detection operators (such as Canny) to extract contour information exhibits good robustness.

[0076] After obtaining the edge points, the geometric center of the iris can be estimated by performing circle fitting using the least squares method. This center position usually corresponds to the optical axis of the eyeball and can be used as a reference for spatial positioning. Combined with a pre-set anatomical model, the location of the needle insertion point can be further deduced.

[0077] Acquiring depth information is crucial for 3D localization. Using binocular vision or structured light methods, the depth coordinates of a target point in space can be obtained, enabling a mapping from a 2D image to 3D space. The hand-eye calibration matrix is ​​then used to transform these 3D coordinates to the robot's base coordinate system, allowing the visual information to be directly used in the control system.

[0078] The trajectory planning section employs a seven-segment velocity curve (i.e., an S-curve), characterized by the introduction of jerk constraints during acceleration and deceleration phases, resulting in smoother velocity changes. This method effectively reduces mechanical shock, minimizes vibration, and improves the system's positioning accuracy and stability. In medical settings, this smooth control is particularly crucial, as any sudden mechanical shock could pose a risk to the patient.

[0079] Furthermore, trajectory planning also needs to consider the dynamic performance of the drive system, such as maximum acceleration, maximum speed, and response delay. By setting these parameters appropriately, motion efficiency can be improved while ensuring safety.

[0080] Furthermore, the implementation of RCM centering control in step S3 includes:

[0081] The central control unit acquires the rotation angle change from the interactive input in real time. And calculate the unit vector of the needle tip direction based on the current angle;

[0082] With the spatial position of the needle tip set as a constraint constant, the position error equation is established: ;in, The Jacobian matrix for the translation module with respect to the tip position;

[0083] The compensation displacement is solved using the weighted pseudo-inverse method: The weight matrix is ​​used for weighting. It is a diagonal matrix, and its diagonal elements are composed of the reciprocals of the maximum velocities of each translation axis;

[0084] The central control unit selects the solution that minimizes the sum of the squares of the translational displacements as the execution solution.

[0085] RCM centering control is essentially a constrained kinematics problem with redundant degrees of freedom. In this system, although the mechanism as a whole has 5 degrees of freedom, when executing the constraint of "fixed needle tip spatial position," the 3 degrees of freedom of the translation module compensate for the disturbances caused by rotation. Jacobian matrix. This describes the linear mapping relationship between a small displacement of the translation axis and the change in the position of the needle tip; its physical meaning can be understood as a "sensitivity matrix." By introducing a weighted pseudo-inverse solution, the distribution of execution variables can be optimized while satisfying position constraints, making the system tend to choose the "lowest cost" motion path.

[0086] weight matrix The introduction of this approach allows for the adjustment of motion priorities for different axes based on the actual performance of the mechanism. For example, axes with slower dynamic responses or weaker structural rigidity can be assigned greater weights to reduce their involvement, thereby improving the overall system stability. Furthermore, this solution method exhibits good numerical stability and avoids solution divergence near singular configurations.

[0087] Furthermore, the specific steps for performing virtual RCM dynamic compensation control in step S4 include:

[0088] An error mapping database is established through offline calibration. The database adopts a two-dimensional discrete grid structure, and its input variable is the rotation angle. , The output variable is the needle tip spatial deviation vector. ;

[0089] The interval between the angle grid divisions is no greater than 1°;

[0090] During online operation, the central control unit collects the current rotation angle at a frequency of no less than 1000Hz and uses a trilinear interpolation algorithm to calculate the corresponding spatial deviation vector in the error mapping database.

[0091] The central control unit acquires the vibration signal output by the inertial measurement unit and extracts the mechanical vibration component through high-pass filtering;

[0092] The spatial deviation vector and the vibration signal are fused to generate a comprehensive compensation amount, which is then superimposed as a feedforward term into the position control command of the translation module.

[0093] The compensation signal is processed by a second-order Butterworth low-pass filter before being output.

[0094] Virtual RCM dynamic compensation control, at its core, transforms mechanical errors from "uncontrollable disturbances" into "predictable models." The establishment of the error mapping database is essentially a process of discrete sampling and function approximation of the non-ideal geometric characteristics of the mechanism. By performing gridded sampling within the rotation angle space, the complex spatial error field can be transformed into a queryable data structure. In the online phase, interpolation algorithms (such as trilinear interpolation) can achieve error estimation in continuous space while ensuring computational efficiency.

[0095] Introducing vibration signals acquired by an inertial measurement unit (IMU) expands the compensation control from purely static error correction to dynamic disturbance suppression. High-pass filtering extracts high-frequency components from structural vibrations, while low-pass filtering suppresses noise and discontinuities in the compensation signal, achieving a balance between response speed and control smoothness. This control framework of "offline modeling, online correction, and signal fusion" enables the virtual RCM point to maintain sub-millimeter level or even higher stability under complex real-world conditions.

[0096] Furthermore, the method also includes an eye-tracking step based on visual feedback, comprising:

[0097] The surgical field images are continuously acquired using a visual sensor, and a template matching algorithm is used to track the vascular branches or iris texture features on the surface of the eyeball and calculate the pixel displacement in the image.

[0098] The pixel displacement is converted into a physical displacement deviation in the base coordinate system using a hand-eye calibration matrix;

[0099] The state of the displacement deviation is estimated using a discrete linear Kalman filter algorithm, and the state variables include position and velocity components.

[0100] The predicted displacement increment is superimposed on the target position of the virtual teleoperation center and the translation module is driven to perform follow-up control.

[0101] The attitude of the RCM rotating module is adjusted synchronously during the follow-up process.

[0102] Eye-tracking control is a typical application of visual servo control in the field of minimally invasive surgery.

[0103] By continuously tracking the surface texture features of the eyeball (such as blood vessel branches or iris texture), the displacement information of the target area in the image plane can be obtained in real time. The role of the hand-eye calibration matrix is ​​to establish a rigid transformation relationship between the camera coordinate system and the robot base coordinate system, thereby realizing the mapping from two-dimensional pixel space to three-dimensional physical space.

[0104] The introduction of Kalman filtering enables the system to make optimal estimates of the target motion state even in the presence of measurement noise and processing delays. Its prediction-update mechanism can effectively improve the foresight of the control system. At the control execution level, the predicted displacement is superimposed on the RCM target point, which is equivalent to "moving" the virtual fulcrum in real time, so that the robot always operates around the new needle insertion position, thereby avoiding additional stress caused by eye movements. Compared with the traditional fixed RCM structure, this dynamic RCM concept significantly improves the system's adaptability to physiological movements.

[0105] Furthermore, the method also includes a one-button axial safety needle retraction control step, comprising:

[0106] The central control unit records the current pose state vector in real time;

[0107] When an emergency withdrawal signal is received, the rotation angle of the RCM rotating module is locked.

[0108] Calculate the unit vector of the needle tip direction based on the current rotation angle:

[0109] ;

[0110] The unit vector is decomposed into the directions of each axis of the translation module, and a three-axis synchronous speed control command is generated;

[0111] The motor current change rate is monitored in real time during the needle retraction process, and deceleration control is triggered when the current exceeds a preset threshold.

[0112] The one-button axial safety needle retraction control reflects the design priority of safety under extreme working conditions.

[0113] In this control mode, by locking the rotational degree of freedom, the multi-degree-of-freedom system is reduced to a single-degree-of-freedom motion problem along the tool axis, thus greatly simplifying the control strategy. The calculation of the unit vector of the needle tip direction is derived from the current attitude angle, which is essentially the transformation of rotational space information into a direction vector in Cartesian space. By decomposing this vector into each translation axis, multi-axis coordinated linear motion can be achieved.

[0114] This method avoids the curve avoidance problem that may occur in traditional path planning, thereby reducing the risk of lateral shearing of surrounding tissues. Simultaneously, by monitoring the rate of change of motor current, it can indirectly reflect changes in the resistance experienced by the needle tip, thus allowing for timely speed adjustments or even stopping of movement in case of abnormalities, achieving a force feedback-based safety protection mechanism.

[0115] Furthermore, the method also includes a dynamic-assisted teaching step, comprising:

[0116] Establish a system dynamics model:

[0117] ;

[0118] The inertial and centrifugal terms are ignored.

[0119] The central control unit calculates and outputs the gravity compensation torque based on the current posture.

[0120] Static friction compensation is added based on the direction of motor motion;

[0121] External forces are estimated through current feedback, and the external forces are mapped to target velocities through admittance control.

[0122] Dynamics-assisted teaching is an important form of human-machine collaborative control.

[0123] Ignoring inertia and centrifugal terms under low-speed operating conditions simplifies the dynamic model to a static equilibrium model, a treatment with good engineering applicability. Gravity compensation counteracts the mechanism's own weight, allowing the operator to move the system without having to overcome gravity. Friction compensation further eliminates static frictional resistance between the guide rail and transmission components, thus preventing "jamming."

[0124] Estimating external force using motor current is a method that does not require external force sensors. Its core lies in establishing a mapping relationship between current and output torque. Admittance control strategy defines the system's "compliance" with external forces, that is, the dynamic relationship parameters between input force and output motion. By reasonably setting the admittance parameters, the system can exhibit characteristics similar to "low damping and low stiffness," thereby achieving a smooth human-machine interaction experience.

[0125] Furthermore, the method also includes a virtual space constraint control step, comprising:

[0126] Establish a geometric model of the restricted area and define a distance function between the needle tip position and the boundary of the restricted area;

[0127] When the distance is less than a preset threshold, the motion command is attenuated by adjusting the control gain;

[0128] When the needle tip reaches the boundary, the speed command in the corresponding direction is limited to zero, and a tactile feedback signal is output.

[0129] Virtual space constraint control is a software-based method for constructing security boundaries.

[0130] By defining a geometric model of the restricted area in the control space, anatomical information can be directly incorporated into the control strategy. A distance function quantifies the spatial relationship between the current needle tip position and the danger zone, and its value serves as the basis for control adjustments. Approaching the boundary, a "soft constraint" is achieved by continuously attenuating the control gain, allowing the operator to gradually experience increased resistance as they approach the danger zone. Upon touching the boundary, a "hard constraint" is achieved by setting the speed command to zero, preventing further intrusion.

[0131] By combining haptic feedback devices, this constraint information can be intuitively transmitted to the operator, thus forming a closed-loop human-machine perception system. This virtual constraint mechanism is highly flexible and can be dynamically adjusted according to different surgical needs.

[0132] Furthermore, the central control unit adopts a multi-core parallel computing architecture, wherein:

[0133] Different computing cores perform motion control, vision processing, and security monitoring tasks respectively;

[0134] Each computing core interacts with other computing cores via a shared bus and a double buffering mechanism;

[0135] The system's end-to-end control delay is no greater than a preset threshold;

[0136] The system's operating status is monitored through a watchdog timer, and security protection controls are executed in case of abnormalities.

[0137] The multi-core parallel computing architecture reflects a comprehensive consideration of system real-time performance and security.

[0138] Assigning different functional modules to independent computing cores can effectively avoid resource contention between tasks and improve overall operating efficiency. Motion control tasks typically require high-frequency execution to ensure trajectory accuracy, while vision processing depends on the image sampling period. The two differ significantly in time scale, and a multi-core architecture can achieve decoupled operation.

[0139] The double buffering mechanism aims to avoid data read / write conflicts, ensuring deterministic latency in data exchange between modules. Watchdog timers and heartbeat detection mechanisms are typical examples of safety redundancy designs. By monitoring the system's operational status, they trigger protective measures upon detecting anomalies, thereby preventing the risk of loss of control. In medical settings, such designs are crucial for ensuring patient safety.

[0140] The beneficial effects of this invention are as follows:

[0141] First, this invention achieves high-precision virtual RCM functionality with only five degrees of freedom by decoupling and coupling the translation and rotation modules for coordinated control. Utilizing a weighted pseudo-inverse algorithm to solve for translational compensation displacement effectively suppresses mechanism disturbances caused by multi-axis linkage, keeping the spatial drift of the surgical needle tip within tens of micrometers during attitude adjustment. This significantly surpasses the control precision of existing five-degree-of-freedom configurations and meets the stringent requirements for fulcrum stability in ophthalmic surgery.

[0142] Secondly, this invention introduces a dynamic compensation mechanism based on an error mapping database. It obtains the functional relationship between the rotation angle and the spatial deviation of the needle tip through offline calibration, and performs real-time interpolation and compensation at a frequency of no less than 1000Hz during online operation. This mechanism effectively eliminates non-ideal RCM drift caused by machining tolerances, assembly errors, and structural deformation, reducing the dependence on the precision of mechanical components and thus lowering manufacturing costs while ensuring performance.

[0143] Third, the visual feedback-based eye-tracking control provided by this invention uses a Kalman filter to estimate and predict eye movements, enabling the virtual RCM point to be "attached" to the needle insertion point during physiological movements in real time. This function effectively avoids scleral incision stress caused by micro-movements of the eyeball, significantly improves the compatibility between instruments and tissues during surgery, and reduces the risk of iatrogenic tissue damage.

[0144] Fourth, the one-button axial safety needle withdrawal control of this invention ensures that the surgical needle withdraws strictly along its own axis by locking the rotary joint and generating a translational speed command along the current needle tip axis unit vector. This mechanism avoids the lateral shearing force that may be generated by traditional needle withdrawal methods. At the same time, combined with real-time monitoring of the motor current change rate, it can detect abnormal resistance and automatically decelerate, providing a reliable guarantee for safe evacuation in emergency situations.

[0145] Fifth, the dynamic-assisted teaching control of this invention maps external forces into target velocities through gravity compensation and static friction feedforward, combined with first-order admittance control, allowing the operator to experience a "zero-gravity floating" feel when manually dragging the robotic arm. This function greatly simplifies the preoperative alignment and posture adjustment process, reduces the operator's physical burden, and improves the convenience of clinical use.

[0146] Sixth, the virtual space constraint control constructed in this invention can form a safety boundary at the software level by defining a geometric model of the restricted area and implementing distance-based adaptive gain attenuation. When the instrument approaches the danger zone, the operator can feel the gradually increasing virtual resistance. When it reaches the boundary, the movement is completely blocked and accompanied by a tactile warning, thereby effectively preventing the instrument from accidentally entering key anatomical structures such as the lens and retina, and improving the overall safety of the surgery.

[0147] Seventh, this invention employs a multi-core parallel computing architecture, allocating motion control, vision processing, and safety monitoring to independent computing cores. Combined with double-buffered data interaction and a watchdog timer, it achieves real-time response capability with end-to-end control latency of less than 1 millisecond. This architecture ensures the synchronization of high-frequency compensation control and visual tracking, while also possessing hardware-level safety redundancy, providing a solid foundation for stable operation in complex surgical environments. Attached Figure Description

[0148] Figure 1 : A flowchart of the process of this invention. Detailed Implementation

[0149] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0150] Example: Figure 1As shown, this embodiment of the invention provides a method for controlling the motion degrees of freedom of an ophthalmic assisted surgical system. This method is based on a five-degree-of-freedom ophthalmic surgical robot system, which includes a translation module, an RCM rotation module mounted at the end of the translation module, and a central control unit.

[0151] 1. System Hardware Configuration

[0152] The translation module is a Cartesian coordinate linear motion platform, comprising a trolley X-axis, a trolley Y-axis, and a trolley Z-axis arranged along three-dimensional orthogonal directions in space. Each axis is driven by a high-precision linear motor or ball screw pair and equipped with a position feedback unit such as an optical or magnetic scale to achieve sub-millimeter positioning accuracy.

[0153] The RCM rotary module is mounted on the end of the translation module, i.e., the Z-axis of the trolley. It includes an RCM horizontal rotary joint that enables horizontal yaw rotation around a first rotation axis, which is a vertical axis, and an RCM vertical ring mechanism that enables vertical pitch rotation around a second rotation axis, which is a horizontal axis. The vertical ring mechanism typically adopts an arc-shaped guide rail structure, and its end is used to hold surgical needles or surgical instruments.

[0154] The central control unit is a multi-axis motion controller, which is electrically connected to the drivers and sensors of the translation module and the RCM rotation module, and is used to execute the multi-level, high-precision control steps described in this invention.

[0155] 2. Multi-coordinate system modeling and kinematic mapping

[0156] To achieve precise control over the needle tip's position, it is first necessary to establish a systematic kinematic model.

[0157] 2.1 Establishing the coordinate system

[0158] This invention establishes a four-level coordinate system. The base coordinate system is fixed at the center of the equipment base or trolley base, with its Z-axis direction aligned with the Z-axis guide rail direction of the trolley, serving as a global reference system. The translation end coordinate system is set at the equivalent end reference point formed by the superposition of the X, Y, and Z axis movements of the trolley. In actual physical structures, the mechanical intersection points of the three-axis sliders may not coincide. This invention combines the three-axis displacements in the base coordinate system using forward kinematics to define a virtual equivalent end point, serving as the output end of the translation module. The RCM geometric coordinate system takes the theoretical spatial intersection of the first and second rotation axes as its origin; ideally, this point is the teleoperation center required for surgery. The origin of the tool coordinate system is set at the tip of the surgical needle.

[0159] 2.2 Homogeneous Transformation Considering Assembly Errors

[0160] Considering machining and assembly errors, the actual axes of the two rotation axes may not intersect ideally. Therefore, a three-dimensional spatial offset parameter is introduced during the transformation from the translation end coordinate system to the RCM geometric coordinate system. This offset parameter describes the deviation between the theoretical and actual intersection points. It can be obtained through calibration using external measuring equipment such as a laser tracker.

[0161] 2.3 Forward Kinematics Model

[0162] Based on the above coordinate system, the forward kinematics model is constructed using the DH parameter method. For a given translational displacement... , , Horizontal rotation angle and pitch angle The position of the surgical needle tip in the base coordinate system for:

[0163] ;

[0164] in The pendulum length is the fixed structural length from the origin of the RCM geometric coordinate system to the tip of the needle. and This is the standard rotation matrix. The formula clearly expresses the sequential kinematic relationship of translation, rotation, and re-translation.

[0165] 3. Global coarse positioning and automatic RCM calibration

[0166] This step aims to safely and quickly move the robot from any initial position to the vicinity of the surgical area, preparing for subsequent delicate procedures.

[0167] 3.1 Vision-based target localization

[0168] The central control unit acquires images of the patient's eyes using a vision sensor integrated into the surgical microscope. The following image processing workflow is employed to extract the location of the needle insertion point in the eyeball: First, the iris edge contour is extracted using the Canny edge detection operator. Then, a circle is fitted to the iris edge points using the least squares method to determine the iris center position. Next, combined with a pre-defined eyeball anatomical model, or through binocular vision or structured light depth sensing, the three-dimensional spatial position of the needle insertion point in the camera coordinate system is determined. Finally, using a pre-calibrated hand-eye calibration matrix—a rigid body transformation matrix from the camera coordinate system to the robot's base coordinate system—the pixel coordinates in the image coordinate system are converted into the three-dimensional physical target position in the base coordinate system. .

[0169] 3.2 S-curve speed planning and determination of safety distance

[0170] Obtain target location Then, the controller adjusts the position of the origin of the current RCM geometric coordinate system according to... The spatial deviation between them generates a three-axis displacement control command for the translation module. The translation module is driven to move along a continuous trajectory, so that the origin of the RCM geometric coordinate system reaches a reference position located directly above the target position and within a safe distance along the Z-axis.

[0171] In this embodiment, the reference position is preferably defined as being located at the target position. The range is 20 mm to 80 mm along the upper Z-axis. The lower limit of 20 mm is determined based on surgical safety: if the distance is too close (less than 20 mm), during high-speed coarse positioning, the end effector or RCM module may accidentally collide with the surface of the eyeball or surrounding facial tissue, significantly increasing the risk. The upper limit of 80 mm is determined based on the positioning accuracy and motion efficiency of visual guidance: on the one hand, beyond 80 mm, the field of view and depth measurement accuracy of the visual sensor decrease, which may lead to... On the one hand, excessively long travel distances can amplify measurement errors; on the other hand, excessively long travel distances can reduce positioning efficiency. Therefore, 20 mm to 80 mm is a relatively optimal safe distance range obtained through a limited number of routine experiments, taking into account the anatomical space limitations of ophthalmic surgery, system positioning accuracy, and movement efficiency. Within this range, absolute safety can be guaranteed while providing a sufficiently accurate starting position for subsequent RCM centering control. In other embodiments of the invention, this range can also be adjusted to 25 mm to 60 mm depending on the specific surgical approach or instrument size.

[0172] 3.3 Generation of S-curve Trajectory

[0173] To ensure smooth motion and suppress shocks, a seven-segment S-curve velocity planning algorithm is used to generate a continuous trajectory. The controller presets the maximum speed based on the dynamic response characteristics of each axis drive system. Maximum acceleration and maximum jerk By constraining the jerk, the velocity curve changes continuously, effectively suppressing the inertial impact during the start and stop phases of motion, and avoiding potential risks to patients caused by mechanical vibration.

[0174] 4. RCM centering control

[0175] After coarse localization, the system enters RCM centering mode. The core of this mode is that when the doctor inputs posture adjustment commands via the main hand or joystick... and At that time, the system automatically calculates the compensation displacement of the translation module, so that the position of the needle tip in space remains constant, as if it were orbiting a fixed point. Rotate.

[0176] 4.1 Position Error Modeling

[0177] The controller acquires the current unit vector of the needle tip direction in real time and constrains the position of the needle tip in the base coordinate system to a fixed point. When the rotary joint undergoes a slight rotation, the ideal position of the needle tip should remain unchanged, but due to the kinematic relationship of the mechanism, a slight offset will occur. Establish the position error equation:

[0178] ;

[0179] in This is the deviation between the theoretical and actual positions of the needle tip caused by rotational motion. For the displacement of each axis of the translation module , , The Jacobian matrix for the needle tip position, The translational compensation displacement is to be determined. Jacobian matrix. The specific form can be obtained by taking the partial derivative of the forward kinematics formula. The first three columns correspond to the unit mapping of the translation module, and the last two columns correspond to the influence of rotational motion on the tip position. In RCM centering mode, we only solve for the translation compensation, so J degenerates into a 3×3 submatrix with diagonal elements close to 1 and off-diagonal elements depending on the current attitude angle.

[0180] 4.2 Solving for the optimal solution based on weighted pseudoinverse

[0181] Since three translational degrees of freedom are used to compensate for the deviations caused by the two rotational degrees of freedom, the system is redundantly driven. To select the optimal solution and suppress mechanism disturbances, this invention employs a weighted pseudo-inverse algorithm to solve for the compensation displacement:

[0182] ;

[0183] in It is a 3×3 diagonal weight matrix, with its diagonal elements The weights correspond to the X, Y, and Z axes, respectively.

[0184] In this invention, the weight matrix The diagonal elements are set to the reciprocal of the maximum velocity along each translation axis, i.e. This design differs fundamentally from the conventional identity matrix weighting, i.e., the least squares solution, and significantly solves the problem of suppressing structural perturbations in ophthalmic surgery. The least squares solution minimizes only the L2 norm of the displacement vector. This may cause the calculated compensation amount to rely excessively on certain axes with slow dynamic response or dead zones, thereby causing motion jamming or shock.

[0185] The weights used in this invention make... Axis with larger dimensions, i.e., better dynamic performance, receives greater weight and is used preferentially in the solution process; while Axis weights that are smaller, meaning those with limited dynamic performance, are assigned smaller amounts of motion. In RCM alignment scenarios requiring high-frequency, minute compensation, this capability-based allocation strategy avoids sending compensation commands to axes that cannot respond quickly, thus effectively suppressing mechanism disturbances and vibrations caused by motion incoordination during multi-axis linkage. Experimental data shows that, at the same attitude adjustment speed, the system using the weighted pseudo-inverse method of this invention reduces the end-effector vibration amplitude by approximately 40% compared to the system using identity matrix weights, and also exhibits smaller peak motor current and smoother motion, which is crucial for ophthalmic surgery requiring extreme stability.

[0186] 5. Virtual RCM dynamic compensation control

[0187] In actual mechanical systems, due to machining and assembly errors, the rotation axes of the RCM horizontal rotary joint and the vertical ring mechanism often cannot achieve ideal spatial intersection. This results in a small but not negligible spatial drift at the surgical needle tip when simply adjusting the posture, even with compensation according to the ideal kinematic model in step S3. To eliminate this drift, this invention introduces real-time dynamic compensation control based on an error mapping database.

[0188] 5.1 Offline calibration method for error mapping database

[0189] The establishment of this database is the core of compensation control, and its purpose is to obtain the needle tip spatial deviation vector under different combinations of rotation angles. The precise values ​​are obtained. The calibration process is performed before the system leaves the factory or during routine maintenance, and the specific steps are as follows.

[0190] First, the system is placed in the measurement field of a high-precision external measuring device. This device can be a laser tracker or an optical coordinate measuring machine. A specially designed target sphere is fixed at the tip of the surgical needle for the external measuring device to track the three-dimensional spatial coordinates of the needle tip in real time. To convert the data measured by the external measuring device to the robot's base coordinate system, coordinate system alignment calibration needs to be performed. Specifically, reflective target spheres are fixed at multiple known locations on the robot base. The spatial coordinates of these target spheres are measured using the external measuring device, and their nominal coordinates in the robot's base coordinate system are recorded. The rigid body transformation matrix between the two coordinate systems is calculated using singular value decomposition. After alignment, all readings from the external measuring device can be directly expressed as coordinates in the base coordinate system.

[0191] Then, the two rotary joints of the RCM rotary module are driven to a series of preset angle combinations. The angle grid intervals should be sufficiently fine, for example... and A point was taken every 1 degree within the range of -45 degrees to +45 degrees, forming a total of 91×91 grid nodes. At each node, an external measuring device recorded the actual spatial coordinates of the needle tip. To ensure that the translation module does not introduce additional motion during the calibration process, the translation module should be locked in a fixed reference position, for example, by placing the origin of the RCM geometric coordinate system at the center of the measurement field.

[0192] At the same time, the central control unit calculates the theoretical coordinates of the needle tip under this angle combination based on the ideal kinematic model. The calculation of these theoretical coordinates assumes ideal intersection of rotation axes and no other geometric errors, and uses the same translation module as during calibration to lock the position. For each mesh node, the tip spatial deviation vector... The deviation vector contains three components. This represents the deviation of the actual needle tip from the theoretical needle tip caused by mechanical errors.

[0193] To further improve calibration accuracy, multiple measurements can be taken at each grid node and the average value taken to eliminate random measurement noise. The final discrete data points obtained from the calibration form a two-dimensional lookup table, with the rotation angle as the input. and The output is a three-dimensional deviation vector. If the grid interval is 1 degree, the entire database contains approximately 8281 data points. This database is stored in the non-volatile memory of the central control unit for online runtime access.

[0194] 5.2 Online Dynamic Compensation Logic

[0195] After the system enters surgical mode, the central control unit reads the encoder values ​​of the RCM rotation module in real time at a sampling frequency of no less than 1000 Hz to obtain the current rotation angle. and Since the actual angle may fall between discrete grid nodes, the controller uses a trilinear interpolation algorithm to quickly calculate the spatial deviation vector corresponding to the current angle in the error mapping database. The trilinear interpolation is implemented on a two-dimensional grid as follows: First, find the four adjacent grid nodes containing the current angle point, and then perform linear weighting in both dimensions to obtain a smooth and continuous bias estimate.

[0196] After obtaining the current deviation, the controller adds it as a compensation to the position control command of the translation module. Specifically, the target position that the translation module was originally expected to reach... , , After correction, it becomes:

[0197] ;

[0198] ;

[0199] ;

[0200] The correction command is fed into the servo control loop of the translation module in the form of position feedforward. With a sampling frequency as high as 1000 Hz, the translation module can respond to millisecond-level needle tip drift, offsetting rotational offsets in real time through micrometer-level minute movements. The final result is that, during pure attitude adjustment, the surgical needle tip maintains an extremely stable virtual teleoperation center in space, with its drift controlled within tens of micrometers.

[0201] To further enhance the robustness of the compensation, an inertial measurement unit (IMU) can be integrated into the system. This unit, installed near the RCM rotating module, detects high-frequency vibration signals from the mechanical structure. The central control unit reads the output of the IMU, extracts the mechanical flutter component through a high-pass filter with a cutoff frequency of 30 Hz, and then adds this component to the spatial deviation vector obtained from the error mapping database. The resulting comprehensive compensation is processed by a second-order Butterworth low-pass filter with a cutoff frequency set to 200 Hz to filter out high-frequency noise in the control signal before being superimposed onto the translation module's commands. This compensation strategy, combining static and dynamic errors, ensures that the virtual RCM point remains highly stable even under external disturbances.

[0202] 6. Decoupled operation control of attitude and feed

[0203] During delicate surgical procedures, this invention decomposes the surgeon's input into two independent components: an attitude adjustment component and an axial feed component. This decoupled control reduces operational complexity and improves safety.

[0204] When the doctor adjusts their posture, that is, changes the pointing angle of the surgical needle via a joystick, the central control unit executes the RCM centering control in step S3 and the virtual RCM dynamic compensation control in step S4. At this time, the translation module maintains the spatial position of the needle tip, while the rotation module is responsible for changing the direction of the needle tip. The two movements are tightly coupled at the control algorithm level, but their functions are clearly distinguished.

[0205] When the surgeon performs axial feed operations, such as pushing the surgical needle forward along its own axis into the eye tissue, the central control unit maintains the angle of the rotation joint constant. The controller operates based on the current needle tip direction unit vector. , , and the feed rate given by the doctor Generate speed commands for the three axes of the translation module:

[0206] ;

[0207] ;

[0208] ;

[0209] This speed command ensures that the surgical needle moves strictly along its own axis in a straight line, avoiding damage to the scleral incision caused by lateral shearing forces. Because the rotary joint is locked during feeding, the system's kinematic model is simplified to a single-degree-of-freedom linear motion, significantly improving control robustness.

[0210] 7. Eye-tracking control based on visual feedback

[0211] During surgery, the patient's eyeballs may change position due to breathing, heartbeat, or involuntary micro-movements. If the virtual RCM point remains absolutely stationary, eyeball movement will cause relative displacement between the surgical needle and the scleral incision, potentially leading to tissue tearing. Therefore, this invention also provides an eyeball tracking control method.

[0212] The central control unit continuously acquires surgical field images via a visual sensor. Stable features on the ocular surface, such as iris texture, vascular branches, or pre-placed scleral markers, are tracked using template matching algorithms or optical flow methods. Motion vectors in pixel coordinates are obtained by calculating the displacement of feature points in two consecutive frames. , .

[0213] Using a pre-calibrated hand-eye transformation matrix, the pixel displacement is converted into a physical displacement deviation in the base coordinate system. This deviation represents the actual amount of eye movement.

[0214] Considering the noise and delay inherent in visual sampling, the controller employs a discrete linear Kalman filter to estimate the displacement deviation. The state variables of the Kalman filter are defined as follows: ,in For positional components, Let be the velocity component. The state transition equation of the system is: ,in Add time step to identity matrix The velocity integral term is formed; the observation equation is ,in To select the observation matrix for the position components, This refers to the displacement deviation obtained from image processing. The process noise covariance matrix Q and the observation noise covariance matrix R are preset based on the system's dynamic characteristics and sensor noise levels. For example, the process noise corresponding to the diagonal elements of Q is set to 0.01 square millimeters, the velocity process noise is set to 1 square millimeter per square second, and R is set to 0.1 square millimeters. Through the Kalman filter prediction-update mechanism, the system can estimate the optimal position of the eyeball at the current moment and predict the possible position at the next moment.

[0215] The filtered positional deviation is superimposed on the target position of the virtual teleoperation center as a compensation. In other words, the originally fixed RCM point now moves in real time with eye movements. The controller drives the three-axis translation module to perform follow-up motion, while simultaneously fine-tuning the attitude of the RCM rotation module as needed to ensure that the relative pose between the surgical needle tip and the eye incision remains consistent. This dynamic RCM control makes the surgical needle appear glued to the eyeball, greatly reducing the risk of tissue damage.

[0216] 8. One-button axial safety needle retraction control

[0217] In case of emergencies during surgery, such as sudden patient movement, abnormally high intraocular pressure, or other complications, the system needs to quickly and safely withdraw the surgical needle from the eyeball. This invention provides a one-button axial safe needle withdrawal function.

[0218] The system normally records the current pose vector in real time, including the angles of each joint and the position of the translation module. When the operator presses the emergency retraction button or the system triggers the automatic protection signal, the central control unit immediately locks the two rotary joints of the RCM rotary module, keeping their angles unchanged.

[0219] The controller adjusts according to the current horizontal rotation angle. and pitch angle Calculate the axial unit vector of the surgical needle tip. For the Yaw-Pitch rotation sequence used in this invention, the formula for calculating this unit vector is:

[0220] ;

[0221] ;

[0222] ;

[0223] Then decompose the unit vector into the X, Y, and Z axes of the translation module. Set a safe needle retraction speed. For example, if the command speed is 5 millimeters per second, then the command speeds for the three axes are respectively... The translation module executes multi-axis synchronous linear motion according to the speed command, causing the surgical needle to exit the eyeball at a uniform speed along its own axis.

[0224] Throughout the needle withdrawal process, the controller also monitors the current change rate of each drive motor in real time. The current change rate can indirectly reflect the resistance encountered by the needle tip. If the current change rate exceeds a preset safety threshold, such as 5 times the rated current change rate per second, it indicates that tissue jamming or other abnormal resistance may be encountered during needle withdrawal. The controller will immediately trigger deceleration control, linearly reducing the needle withdrawal speed to zero, awaiting further instructions from the operator. This force feedback-based protection mechanism further ensures the safety of the needle withdrawal process.

[0225] 9. Dynamic-assisted teaching control

[0226] During preoperative preparation or equipment adjustment, the operator may need to manually drag the robotic arm to a specific position. Since the translation and rotation modules have a certain weight, direct dragging is quite strenuous. This invention provides a dynamic-assisted teaching method that makes the system appear to float in zero gravity.

[0227] The central control unit establishes a simplified dynamic model of the system. In slow-speed drag mode, the speed is typically below 10 mm / s, and the acceleration is very small. Therefore, the inertial and centrifugal force terms can be ignored, and the model is simplified to a gravity term. With friction term The controller, based on the current angles and angular velocities of each joint, uses a pre-calculated gravity compensation torque table based on the Lagrange equation or the Newton-Euler method to obtain the static support torque required for each motor in real time. Simultaneously, a constant static friction compensation torque is superimposed according to the motor's direction of motion; its magnitude is determined by experimentally measured Coulomb friction force, and its direction is opposite to the direction of motion.

[0228] To convert the operator's external force into system motion, the controller employs an admittance control strategy. Admittance control essentially defines the dynamic relationship between external force and motion, typically represented as a second-order or first-order system. In this embodiment, a first-order admittance relationship is used: desired velocity... external forces Proportional, that is ,in This is the admittance coefficient matrix, where the diagonal elements correspond to the proportional coefficients of each axis, for example, 0.1 mm / s / Newton. Specifically, the system reads the current feedback from the motor driver in real time and estimates the external force through the linear relationship between current and torque. ,in Let be the motor torque constant. Substituting the estimated external force into the admittance relationship, the desired speed is obtained, and then the motor is driven by the speed controller. After compensation for gravity and friction, the operator only needs to apply a very small pushing force to easily move the entire robotic arm, resulting in a light and smooth feel.

[0229] 10. Virtual Space Constraint Control

[0230] To prevent the surgical needle tip from accidentally entering dangerous areas, such as touching the lens or retina, the present invention also provides a virtual space constraint control function.

[0231] First, a geometric model of the restricted area is established in the central control unit. This model can be a sphere, cylinder, or an irregular shape reconstructed from the patient's preoperative imaging data, used to describe the tissue structures that need to be protected. The controller calculates in real time the minimum distance d_safe between the current surgical needle tip position and the boundary of the restricted area.

[0232] When the distance is less than the preset warning threshold When the distance is less than 1 mm, for example, the controller begins to attenuate the motion command. Attenuation gain. Defined as and The ratio, i.e. The operator's desired speed command. After attenuation, it becomes The operator will feel a gradually increasing virtual resistance, indicating that they are approaching a danger zone.

[0233] When the tip of the needle reaches the boundary of the restricted area, that is attenuation gain When the speed reaches zero, the controller forcibly limits the speed command in the corresponding direction to zero, preventing the needle tip from penetrating further. Simultaneously, the system outputs tactile feedback signals to the operator, such as generating a 200 Hz vibration through the main hand, to enhance the warning effect. This virtual constraint mechanism is entirely implemented in software, requiring no modification to the hardware structure, and possesses strong flexibility and scalability.

[0234] 11. Multi-core parallel control architecture

[0235] To ensure the real-time and reliable execution of all the aforementioned control tasks, the central control unit employs a multi-core parallel computing architecture. Specifically, three independent computing cores are allocated for motion control, vision processing, and safety monitoring, respectively. The motion control core runs high-frequency position loop, velocity loop, and current loop algorithms, typically at frequencies ranging from 4 kHz to 20 kHz. The vision processing core is responsible for image acquisition, feature extraction, and coordinate transformation, operating at a frequency consistent with the camera frame rate, such as 60 Hz or 100 Hz. The safety monitoring core independently runs a watchdog timer and various safety logics, monitoring the system status at a frequency of no less than 1 kHz.

[0236] Data exchange between the computing cores is achieved through a shared bus and a double-buffering mechanism. The double-buffering mechanism ensures that when one core writes new data, the other always reads a complete and consistent data block, avoiding data conflicts. The end-to-end control latency of the entire system—the total time from sensor acquisition to control command output to the driver—is controlled to within 1 millisecond. A watchdog timer continuously monitors the system's heartbeat; if any core times out or the program crashes, hardware-level safety protection is immediately triggered, including cutting off the motor power supply and activating the electromagnetic brake, ensuring the system can safely shut down in abnormal situations.

[0237] 12. Overall beneficial effects

[0238] In summary, the motion freedom control method for the ophthalmic assisted surgical system proposed in this invention achieves high-precision virtual RCM control in a five-degree-of-freedom configuration by establishing a multi-level coordinate system and a precise kinematic model. Its core innovation lies in utilizing the three degrees of freedom of the translation module to actively compensate for the geometric errors and dynamic offsets of the rotation module, enabling the system to achieve superior remote operation center stability compared to traditional structures without relying on high-precision machining. Simultaneously, it integrates multiple high-order control functions such as visual follow-up, safe needle withdrawal, dynamic teaching, and virtual constraints, comprehensively meeting the stringent requirements of ophthalmic surgery for safety, precision, and ease of operation. This method is particularly suitable for procedures such as vitrectomy and retinal surgery, which have extremely high requirements for scleral incision protection, and has significant clinical application value.

[0239] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art can make various improvements and modifications without departing from the spirit and principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for controlling the degrees of freedom of motion in an ophthalmic auxiliary surgical system, characterized in that, The ophthalmic assisted surgical system includes a translation module, an RCM rotation module installed at the end of the translation module, and a central control unit; The translation module includes a trolley X-axis, a trolley Y-axis, and a trolley Z-axis arranged along three-dimensional orthogonal directions in space. Each axis is driven by a linear drive mechanism and equipped with a position feedback unit. The RCM rotating module includes an RCM horizontal rotating joint that realizes horizontal yaw rotation around a first rotating axis and an RCM vertical ring mechanism that realizes vertical pitch rotation around a second rotating axis. The central control unit is electrically connected to the translation module and the RCM rotation module, and is used to execute the following control steps: S1. Multi-coordinate system modeling and kinematic mapping construction; Establish a multi-level coordinate system including a base coordinate system, a translation end coordinate system, an RCM geometric coordinate system, and a tool coordinate system, wherein: The base coordinate system is fixedly set on the equipment base, and the Z-axis direction is consistent with the Z-axis guide rail direction of the trolley. The translation end coordinate system is set at the equivalent end reference point formed by the superposition of the X-axis, Y-axis and Z-axis motion of the trolley. The equivalent end reference point is determined by the combination of the three-axis displacements in the base coordinate system. The RCM geometric coordinate system takes the theoretical spatial intersection of the first rotation axis and the second rotation axis as its origin; The tool coordinate system is set at the tip of the surgical needle; Based on the above coordinate system, the forward kinematics model is constructed using the DH parameter method: in, , , For the displacement variable of the translation module, , For the rotational joint angle variable, The fixed structural length from the origin of the RCM geometric coordinate system to the needle tip; S2. Visual-guided spatial coarse positioning control; The central control unit acquires the position data of the needle insertion point in the eyeball in the image coordinate system through a vision sensor, and converts the position data into the target position in the base coordinate system through a pre-calibrated hand-eye transformation matrix. ; Based on the spatial deviation between the current RCM geometric coordinate system origin position and the target position, generate three-axis displacement control commands for the translation module; The translation module is driven to move along a continuous trajectory, so that the origin of the RCM geometric coordinate system reaches a reference position located 20mm to 80mm above the target position along the Z-axis. The continuous trajectory is generated by an S-curve velocity planning algorithm, and the maximum velocity, acceleration, and jerk are constrained to suppress motion impact. S3. RCM centering control based on inverse kinematics constraints; Upon receiving attitude adjustment input, the central control unit calculates the unit vector of the needle tip direction based on the current rotation angle; The position of the needle tip in the base coordinate system is constrained to a fixed point. Establish a position error model; The mapping relationship between translation variables and needle tip position is established based on the Jacobian matrix, and the translation compensation displacement is solved by the weighted pseudo-inverse algorithm. The weighted pseudo-inverse solution satisfies the minimum translational displacement norm condition to reduce the mechanism disturbance caused by multi-axis linkage, thereby maintaining the stability of the needle tip spatial position during rotation. S4. Virtual RCM dynamic compensation control based on error mapping; To address the needle tip spatial offset caused by the non-ideal intersection of the rotation axes, the central control unit executes the following control process: An error mapping database is established to characterize the functional mapping relationship between rotation angle parameters and needle tip spatial deviation, wherein the rotation angle parameters include the angle of the RCM horizontal rotation joint and the angle of the RCM vertical ring mechanism. The database is rotated by an angle. , As input variables, the needle tip spatial deviation vector For output variables; The error mapping database is obtained by pre-calibrating the spatial position deviation of the tip of the RCM rotating module under different combinations of rotation angles. The calibration process establishes the correspondence between the rotation angle and the spatial deviation based on actual measurement data. During system operation, the current rotation angle is acquired at a sampling frequency of not less than 1000Hz; Based on the rotation angle, the current spatial deviation is calculated in the error mapping database using an interpolation algorithm; The spatial deviation is added as a compensation amount to the translation module control command: ； The needle tip offset is offset by the compensation displacement of the translation module, so that the needle tip can maintain a stable virtual teleoperation center in space. S5. Operation control based on attitude and feed decoupling; The central control unit decomposes the operation input into attitude adjustment components and axial feed components; During the attitude adjustment process, compensation control is performed in steps S3 and S4. During axial feed, the rotary joint angle remains constant, and a translational speed command is generated based on the unit vector of the needle tip direction. The surgical needle is made to perform a linear feed motion along its own axis.

2. The method for controlling the degrees of freedom of motion in an ophthalmic auxiliary surgical system according to claim 1, characterized in that: The specific steps for establishing the multi-level coordinate system mapping model in step S1 include: The base coordinate system is established at the center of the fixed base of the ophthalmic auxiliary surgery system, with its Z-axis parallel to the guide rail direction of the Z-axis of the trolley; The translation end coordinate system is established at the intersection point of the three axes of the translation module, and the translation transformation relationship between the base coordinate system and the translation end coordinate system is defined by a homogeneous transformation matrix. The rotation axis of the RCM horizontal rotary joint is defined as the first rotation axis, and the rotation axis of the RCM vertical ring mechanism is defined as the second rotation axis. The RCM geometric coordinate system is established based on the theoretical intersection of the first rotation axis and the second rotation axis. During the establishment of the RCM geometric coordinate system, a three-dimensional spatial offset parameter is introduced. It is used to describe the actual intersection point offset of the rotation axis caused by mechanical assembly error, and the offset parameter is embedded in the pose transformation matrix from the translation end coordinate system to the RCM geometric coordinate system; The tool coordinate system is established at the position of the surgical needle tip, and the position of the needle tip relative to the origin of the RCM geometric coordinate system is defined as a fixed pendulum length vector; the central control unit constructs a rotation transformation matrix sequence by reading the encoder values ​​of the RCM rotation module in real time, thereby realizing the forward kinematic mapping calculation from the tool coordinate system to the base coordinate system.

3. The method according to claim 1, characterized in that: The specific steps for performing vision-guided coarse spatial positioning control in step S2 include: The central control unit acquires image signals of the patient's eyes through a visual sensor integrated into the surgical microscope, and extracts the needle insertion site features using the following processing method: Iris edge contours were extracted using the Canny edge detection operator; The iris center position is determined by performing a circle fitting on the iris edge points using the least squares method. Determine the spatial location of the needle insertion point by combining a pre-designed anatomical model; Depth information of the needle insertion point is obtained through binocular vision or structured light depth sensing. The pixel coordinates in the image coordinate system are converted into three-dimensional physical coordinates in the base coordinate system using a preset hand-eye calibration matrix. The central control unit generates the motion trajectory of the translation module using a seven-segment lifting speed curve planning algorithm based on the deviation between the current position and the target position. The maximum acceleration and jerk are preset according to the dynamic response characteristics of each axis drive system to suppress the inertial impact during the displacement stopping phase.

4. The method according to claim 1, characterized in that: The implementation of RCM centering control in step S3 includes: The central control unit acquires the rotation angle change from the interactive input in real time. And calculate the unit vector of the needle tip direction based on the current angle; With the spatial position of the needle tip set as a constraint constant, the position error equation is established: ;in, This is the Jacobian matrix for the translation module relative to the tip position; The compensation displacement is solved using the weighted pseudo-inverse method: The weight matrix is ​​used for weighting. It is a diagonal matrix, and its diagonal elements are composed of the reciprocals of the maximum velocities of each translation axis; The central control unit selects the solution that minimizes the sum of the squares of the translational displacements as the execution solution.

5. The method according to claim 1, characterized in that: The specific steps for performing virtual RCM dynamic compensation control in step S4 include: An error mapping database is established through offline calibration. The database adopts a two-dimensional discrete grid structure, and its input variable is the rotation angle. , The output variable is the needle tip spatial deviation vector. ; The interval between the angle grid divisions is no greater than 1°; During online operation, the central control unit collects the current rotation angle at a frequency of no less than 1000Hz and uses a trilinear interpolation algorithm to calculate the corresponding spatial deviation vector in the error mapping database. The central control unit acquires the vibration signal output by the inertial measurement unit and extracts the mechanical vibration component through high-pass filtering; The spatial deviation vector and the vibration signal are fused to generate a comprehensive compensation amount, which is then superimposed as a feedforward term into the position control command of the translation module. The compensation signal is processed by a second-order Butterworth low-pass filter before being output.

6. The method according to claim 1, characterized in that: The method further includes an eye-tracking step based on visual feedback, including: The surgical field images are continuously acquired using a visual sensor, and a template matching algorithm is used to track the vascular branches or iris texture features on the surface of the eyeball and calculate the pixel displacement in the image. The pixel displacement is converted into a physical displacement deviation in the base coordinate system using a hand-eye calibration matrix; The state of the displacement deviation is estimated using a discrete linear Kalman filter algorithm, and the state variables include position and velocity components. The predicted displacement increment is superimposed on the target position of the virtual teleoperation center and the translation module is driven to perform follow-up control. The attitude of the RCM rotating module is adjusted synchronously during the follow-up process.

7. The method according to claim 1, characterized in that: The method further includes a one-click axial safety needle retraction control step, including: The central control unit records the current pose state vector in real time; When an emergency withdrawal signal is received, the rotation angle of the RCM rotating module is locked. Calculate the unit vector of the needle tip direction based on the current rotation angle: ； The unit vector is decomposed into the directions of each axis of the translation module, and a three-axis synchronous speed control command is generated; The motor current change rate is monitored in real time during the needle retraction process, and deceleration control is triggered when the current exceeds a preset threshold.

8. The method according to claim 1, characterized in that: The method also includes a dynamic-assisted teaching step, including: Establish a system dynamics model: ； The inertial and centrifugal terms are ignored. The central control unit calculates and outputs the gravity compensation torque based on the current posture. Static friction compensation is added based on the direction of motor motion; External forces are estimated through current feedback, and the external forces are mapped to target velocities through admittance control.

9. The method according to claim 1, characterized in that: The method further includes a virtual space constraint control step, comprising: Establish a geometric model of the restricted area and define a distance function between the needle tip position and the boundary of the restricted area; When the distance is less than a preset threshold, the motion command is attenuated by adjusting the control gain; When the needle tip reaches the boundary, the speed command in the corresponding direction is limited to zero, and a tactile feedback signal is output.

10. The method according to claim 1, characterized in that: The central control unit adopts a multi-core parallel computing architecture, wherein: Different computing cores perform motion control, vision processing, and security monitoring tasks respectively; Each computing core interacts with other computing cores via a shared bus and a double buffering mechanism; The system's end-to-end control delay is no greater than a preset threshold; The system's operating status is monitored through a watchdog timer, and security protection controls are executed in case of abnormalities.

Images