Thoracolaparoscopic surgical instrument microtremor intelligent compensation and control system and method

Abstract

The present application relates to the technical field of medical instrument automation and intelligent control, and particularly relates to a thoracolaparoscopic surgery instrument micro-tremor intelligent compensation and control system and method, the method comprising synchronously collecting inertial sensor data and a laparoscope visual depth image data of a mechanical arm end, using a nonlinear kinematic model to decouple end motion data motion intention, separating active control intention motion data from physiological high-frequency tremor data, dynamically adjusting compensation control parameters according to a shortest safe distance from the end of the instrument to a cavity wall or surrounding tissue monitored in real time by the depth image during surgery to balance tremor suppression and operation safety. More than 82% tremor suppression rate is achieved, which is much higher than traditional low-pass filtering and Kalman filtering, and tremor suppression effect is significantly improved; and a pure data-driven intention decoupling scheme based on a causal optimal state observer realizes full-cycle non-sensitization operation without increasing operation burden.

Description

Technical Field

[0001] This invention relates to the field of medical device automation and intelligent control technology, and in particular to an intelligent compensation and control system and method for micro-vibration of thoracoscopic and laparoscopic surgical instruments. Background Technology

[0002] In thoracoscopic and laparoscopic minimally invasive surgery, surgeons need to operate slender rod-shaped surgical instruments within confined spaces and severely obstructed views, such as the thoracic cavity and deep pelvis. Due to the size limitations of the minimally invasive access channels, these instruments typically have a very high length-to-diameter ratio, resulting in relatively weak overall rigidity and poor resistance to bending deformation. In actual surgical procedures, the following three progressively layered technical problems exist:

[0003] Level 1: The insufficient rigidity of slender rod-shaped instruments amplifies physiological tremors step by step. Microscopic, high-frequency tremors in the surgeon's hand (typically within the 8-12Hz range) are amplified through the lever arm effect of slender rod-shaped instruments, resulting in large-amplitude, unintended tremors at the instrument's end effector. If these end effector tremors cannot be accurately identified and suppressed, they can easily damage surrounding vital blood vessels and fragile organs during delicate cutting, suturing, or dissection, significantly increasing surgical risks.

[0004] The second level: Endoscopic visual systems inherently suffer from transmission and processing time lags. The acquisition, encoding, transmission, and processing of endoscopic visual images objectively involve a fixed time delay, causing the position of the instruments seen by the operator on the display terminal to lag behind the actual physical movement of the instruments. This means that the surgeon's operational decisions based on delayed visual information are inherently time-biased, further exacerbating the risk of misoperation within narrow cavities.

[0005] The third level: Current technologies do not jointly physical model the visual time delay and the nonlinear deformation of the instruments. Current tremor suppression solutions mainly include the following technical approaches: First, signal processing methods based on low-pass filters or Kalman filters, which eliminate high-frequency tremor components by filtering the master hand control signals in the frequency domain. However, these methods do not consider the mechanical characteristics of the instruments themselves, and the filtering delay is further superimposed on the visual time delay. Second, tremor modeling methods based on offline neural networks, such as fuzzy wavelet neural networks (FWNN) or least squares support vector machines, generate inverse compensation signals by offline modeling of tremor behavior. However, offline models cannot adapt to the real-time changes and nonlinear deformation of the instrument configuration during surgery.

[0006] Chinese patent CN115607297B discloses a master-slave surgical robot control system and method for tremor suppression, which separates and compensates for master hand tremor signals and misoperation signals through a filtering module, an adaptive extraction module, and a feedforward control module. However, this scheme only performs signal-level filtering processing, does not consider the systematic impact of visual transmission delay on the control loop, and does not establish a time-varying stiffness model for the nonlinear deformation of slender instruments. Therefore, the compensation accuracy is insufficient in confined space operations in deep body cavities.

[0007] Chinese patent CN101785704A discloses an adaptive filtering device for a master-slave minimally invasive surgical robot system, which uses a fuzzy wavelet neural network to offline model the operator's hand tremor behavior and generate a reverse compensation signal. However, this solution also suffers from the following drawbacks: it does not consider control delays caused by visual time lag; it does not model the time-varying stiffness of the instrument's flexible deformation, making it impossible for the offline model to track changes in the instrument's configuration in real time during surgery; and it lacks a spatial constraint protection mechanism, failing to dynamically adjust the compensation strategy to ensure operational safety when the instrument tip approaches the cavity wall or organs.

[0008] In summary, existing technologies have failed to simultaneously address the systemic impact of visual time lag and the physical modeling of time-varying nonlinear deformation of slender instruments, nor have they provided an intraoperative spatial constraint protection mechanism. There is an urgent need for a technical solution that can jointly model the above factors physically and achieve precise tremor compensation. Summary of the Invention

[0009] The purpose of this invention is to provide an intelligent compensation and control system and method for minute vibrations of thoracoscopic and laparoscopic surgical instruments, in order to solve the problems mentioned in the background art. The specific technical solution is as follows:

[0010] The first objective of this invention is to provide a method for intelligent compensation and control of minute vibrations in thoracoscopic and laparoscopic surgical instruments, comprising the following steps:

[0011] S1: Synchronously acquire inertial sensing data and endoscopic visual depth image data from the end effector of the robotic arm;

[0012] S2: Based on a preset visual time delay constant The cantilever length L(t) of the device entering the card sleeve is calculated in real time by the positive kinematics of the robotic arm. A nonlinear kinematic model containing the time-varying stiffness matrix K(t) is constructed to describe the actual spatial position of the device end under the combined action of time delay and flexible deformation.

[0013] S3: The nonlinear kinematic model is used to decouple the end-effector motion data from the motion intent, separating the active control intention motion data from the physiological high-frequency tremor data;

[0014] S4: Based on the separated high-frequency tremor data and the visual time delay constant The vibration displacement is predicted in advance by feedforward, and the dynamic reverse compensation torque is calculated in real time by sliding mode control algorithm and output to the servo motor of the robotic arm joint to counteract the vibration displacement.

[0015] S5: The shortest safe distance from the instrument tip to the cavity wall or surrounding tissue, as monitored in real time based on intraoperative depth images. The compensation control parameters are dynamically adjusted.

[0016] Through visual time delay constant The joint physical modeling of the time-varying stiffness matrix K(t) completely solves the shortcomings of existing technologies that only perform signal filtering without considering the mechanical essence, achieving a tremor suppression rate of over 82% in deep body cavities, which is significantly higher than traditional low-pass filtering and Kalman filtering.

[0017] Preferably, in step S3, a third-order linear state-space model is constructed with the device end position, velocity, and acceleration as state variables. Jerk is used as process noise, and its variance weighting matrix is ​​set to penalize large Jerk values. A smoothed operational intention baseline is estimated in real time using only historical observation data through a Kalman filter framework as the active control intention motion data. The reconstructed actual motion trajectory is subtracted from the active control intention motion data to obtain a zero-mean residual signal as the physiological high-frequency tremor data.

[0018] Preferably, in step S3, the motion intention decoupling further includes a fatigue time-frequency adaptive adjustment step: performing wavelet packet transform on the physiological high-frequency tremor data to dynamically identify the center frequency of the frequency band with the highest tremor energy proportion; when the center frequency is detected to drift towards a lower frequency direction due to long-term operational fatigue and exceed a preset variance threshold, the filter cutoff frequency is automatically lowered to broaden the tremor capture bandwidth; when the center frequency is detected to rise towards a higher frequency direction and exceed a preset recovery threshold, the filter cutoff frequency is automatically raised to restore the standard bandwidth, ensuring that when the tremor frequency shifts downward and the amplitude amplifies due to the operator's physical exertion during long-term surgery, low-frequency large-amplitude jitter can still be completely eliminated, while avoiding unnecessary control noise introduced by excessively wide bandwidth after normal recovery.

[0019] Preferably, in step S2, the nonlinear kinematic model adopts an incremental form, and its expression is:

[0020] ;

[0021] in, Let J(q) be the total time delay constant for visual transmission and image processing, J(q) be the Jacobian matrix of the robot arm joint space, q be the joint angle, and q̇ be the joint angular velocity. The equivalent external force at the end measured by the inertial sensor, K(t) is the time-varying stiffness matrix; the incremental integration interval is from To t, s < hour Undefined problems The actual spatial position of the end effector with deformation and time delay compensation is physically defined as: the position at the previous moment + the time-delay-corrected increment of joint motion + the increment of flexible deformation displacement caused by external force. This deformation increment only reflects the new deformation caused by changes in external force within the current time step Δt, not the absolute deformation. When the external force is constant... =0, the deformation increment is zero, and the end position does not drift; a non-zero deformation increment is only generated when the external force changes, which mathematically ensures that the model is defined at any time and is suitable for real-time recursive calculation.

[0022] Preferably, the solution method for the time-varying stiffness matrix K(t) is as follows: the tamper point corresponds to the remote motion center of the robotic arm, which is a fixed three-dimensional coordinate point known through preoperative calibration in the robot base coordinate system; the system obtains the angles of each joint in real time through the joint encoder of the robotic arm, and calculates the three-dimensional spatial distance of the instrument end relative to the remote motion center in real time by combining forward kinematics (FK), which is the cantilever length L(t); at the same time, the cantilever length L(t) is recursively estimated based on the speed of the instrument entering and exiting the tamper sleeve to compensate for the visual time delay of the depth image itself, realize the real-time online update of the stiffness coefficient, avoid the defect of offline modeling that cannot track the changes in instrument configuration during surgery, and compensate for the delay error of the depth image itself through recursive estimation.

[0023] Preferably, in step S4, the dynamic reverse compensation torque The calculation is based on the sliding mode control algorithm, and the formula is as follows:

[0024] ;

[0025] in, The dominant angular frequency of the flutter extracted through time-frequency analysis The phase of the flutter displacement signal is rotated in the future direction by a phase lead compensator. · The time-delay compensation advance prediction value obtained from radians The time derivative of the flutter displacement signal is used to predict the flutter velocity signal after time delay compensation. and These are the proportional gain coefficient and the derivative gain coefficient of the controller, respectively. Let λ be the saturation function of the boundary layer thickness Φ, and λ be the sliding mode switching gain, whose values ​​satisfy... d(t) is the upper bound of the system modeling error and external disturbance, and the first term is the PD feedforward term, based on The first term is to predict the residual of the tremor displacement in advance; the second term is the sliding mode switching term, which uses the saturation function sat(S / Φ) of the boundary layer thickness Φ to replace the traditional sign function sgn(S). Linear control is used inside the boundary layer and switching control is used outside the boundary layer. While maintaining the strong robustness of sliding mode control, it effectively suppresses high-frequency jitter and avoids the sawtooth fluctuation of the compensation torque from causing secondary interference to the delicate operation of the surgery.

[0026] Preferably, in step S5, the spatial constraint protection adopts an anisotropic damping strategy: when the shortest safe distance When the resistance is below the safety threshold, normal damping is maintained in the direction parallel to the direction of the doctor's intended movement, while the differential gain coefficient is exponentially amplified in the direction of the tremor disturbance orthogonal to that direction. By selectively applying high damping, the instrument tip can effectively prevent accidental contact with the cavity wall or important tissues without affecting the doctor's normal operating feel, thus achieving an intelligent balance between tremor suppression and operational safety.

[0027] Preferably, the control method further includes a patient physiological motion feedforward compensation step: extracting the periodic displacement of the cavity wall caused by the patient's breathing and / or heartbeat from the depth image data, establishing a periodic motion prediction model, and superimposing the predicted displacement as a feedforward quantity into the inverse compensation torque in step S4 to counteract the disturbance of the device end effector caused by the patient's physiological motion. This solves the device-tissue relative motion error caused by the patient's spontaneous physiological motion such as breathing and heartbeat, further reducing the risk of accidental injury.

[0028] Preferably, when the instrument tip contacts a tissue surface with known stiffness or identifiable characteristics, the parameters in the time-varying stiffness matrix K(t) are calibrated online based on the relationship between deformation feedback and contact force. The model parameters are continuously corrected using real tissue contact mechanics information during the operation, ensuring that the accuracy of K(t) does not drift due to factors such as instrument wear and temperature changes throughout the entire surgical cycle.

[0029] The second objective of this invention is to provide an intelligent compensation and control system for minute vibrations of thoracoscopic and laparoscopic surgical instruments, comprising: a multi-source sensing module for synchronously acquiring inertial sensing data from the end effector of the robotic arm and laparoscopic visual depth image data;

[0030] The nonlinear kinematics modeling module is used for modeling based on a preset visual time delay constant. The cantilever length L(t) of the device entering the card sleeve is calculated in real time by the positive kinematics of the robotic arm. A nonlinear kinematic model containing the time-varying stiffness matrix K(t) is constructed to describe the actual spatial position of the device end under the combined action of time delay and flexible deformation.

[0031] The motion intention decoupling module is used to decouple the end motion data from the nonlinear kinematic model, separating the active control intention motion data from the physiological high-frequency tremor data.

[0032] An adaptive torque compensation module is used to compensate for high-frequency vibration data and the visual time delay constant. The vibration displacement is predicted in advance by feedforward, and the dynamic reverse compensation torque is calculated in real time by sliding mode control algorithm and output to the servo motor of the robotic arm joint to counteract the vibration displacement.

[0033] The dynamic closed-loop fine-tuning module is used to monitor the shortest safe distance between the instrument tip and the cavity wall or surrounding tissue in real time based on the depth images described during the procedure. The compensation control parameters are dynamically adjusted.

[0034] Preferably, the motion intent decoupling module integrates: a minimum jerk optimizer, used to construct a minimum jerk target functional within a sliding time window to optimize and fit the actual motion trajectory to extract the operation intent baseline; a wavelet packet transformer, used to perform time-frequency decomposition on high-frequency tremor data and dynamically identify the center frequency of the frequency band with the highest tremor energy proportion; and a bidirectional adaptive frequency band adjuster, used to automatically lower or raise the filter cutoff frequency according to the monitored center frequency drift direction, thereby achieving integrated processing of intent decoupling and fatigue tracking without the need for additional physiological sensors.

[0035] It has the following beneficial effects:

[0036] 1. This invention utilizes visual time lag The joint physical modeling with time-varying stiffness K(t) achieves a flutter suppression rate of over 82%, which is far higher than traditional low-pass filtering and Kalman filtering, significantly improving the flutter suppression effect. Furthermore, the pure data-driven intent decoupling scheme based on the causal optimal state observer enables seamless operation throughout the entire cycle without increasing the operational burden.

[0037] 2. Wavelet packet transform and bidirectional adaptive frequency band adjustment mechanism ensure that the tremor suppression effect does not decay throughout the long surgical process. The anisotropic damping strategy applies high damping in the tremor suppression direction and maintains normal damping in the intended motion direction, achieving an intelligent balance between accuracy and safety. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a method for intelligent compensation and control of micro-vibrations in thoracoscopic and laparoscopic surgical instruments according to the present invention. Detailed Implementation

[0040] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a further detailed explanation of the intelligent compensation and control system and method for micro-vibrations in thoracoscopic and laparoscopic surgical instruments proposed in this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the accompanying drawings are all in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of this invention.

[0041] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0042] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the illustrations only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the state, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0043] Example 1: Joint Modeling of Time-Delayed Deformation and Adaptive Torque Compensation

[0044] This embodiment details the visual time delay-based approach. The joint physical modeling of time-varying stiffness K(t) and the technical implementation of sliding mode adaptive torque compensation.

[0045] 1.1 System Hardware Composition

[0046] The system hardware mainly includes the following components:

[0047] (1) A six-axis inertial sensor (IMU) is placed at the base of the surgical instrument, i.e., at the proximal end before entering the trocar cannula, to synchronously acquire triaxial acceleration and triaxial angular velocity signals at a sampling frequency of not less than 200Hz. The IMU is placed away from the part of the instrument that enters the trocar cannula to avoid affecting aseptic operation, while ensuring that it can sensitively capture the original motion information of the operator's hand.

[0048] (2) A binocular depth camera is placed at the distal end of the surgical field to acquire depth image data within the laparoscopic field of view at a frame rate of not less than 60Hz. The binocular depth camera is fixed on a special bracket on the operating table, and its field of view covers the entire surgical operation area, and can simultaneously capture the three-dimensional spatial position of the instrument tip and the depth information of the cavity wall.

[0049] (3) Hardware synchronization signal generator sends a unified clock pulse signal to the IMU and depth camera to ensure that the timestamps of inertial sensing data and depth image data are aligned at the microsecond level, providing an accurate time reference for time delay modeling.

[0050] 1.2 Visual time delay constant Measurement

[0051] Visual time delay constant The determination was performed through preoperative calibration experiments. (Reference) Figure 1 The specific steps are as follows:

[0052] (1) Place a high-contrast marker with a known precise location in the calibration environment. The physical location of the marker is precisely determined by a laser rangefinder and used as the true reference.

[0053] (2) Start the complete image acquisition-transmission-processing-display link: the binocular depth camera acquires the marked point image → transmits it to the image processing unit through a dedicated image transmission cable → the image processing unit performs depth calculation and three-dimensional reconstruction → the reconstruction result is superimposed and displayed on the surgical monitor.

[0054] (3) Simultaneously capture the actual motion of the marker points and the displayed image on the surgical monitor using a high-speed camera (1000fps). Compare the actual motion time of the marker points with the corresponding motion time in the displayed image frame by frame. The time difference between the two is the end-to-end visual time delay. Repeat the measurement 50 times and take the average value to obtain the visual time delay constant. The calibration value is typically in the range of 30-80ms.

[0055] (4) The calibration results The values ​​are written to the system configuration file and used as fixed parameters for the nonlinear kinematic model.

[0056] 1.3 Incremental Form of Nonlinear Kinematic Model

[0057] This embodiment uses an incremental nonlinear kinematic model, the mathematical expression of which is:

[0058] ;

[0059] Among them, the first item The estimated end position of the previous recursive time step is used as the starting point for the current recursion. At the initial time (t=0). It is determined by combining the initial geometry of the instrument with the initial measurements from the binocular depth camera.

[0060] Second item : via visual time lag Corrected joint movement increments. Because the visual image seen by the doctor at time s reflects... The actual state at any given time, therefore the actual contribution of joint motion to the end-effector position needs to be determined using the time-delayed Jacobian matrix. Let's calculate. The integration interval is... Where Δt is the control period (usually 5-10 ms), ensuring that the integration interval is always positive, thus avoiding the limitations of traditional integration methods where s < 0. hour Undefined mathematical problems.

[0061] Third item :external force When applied to the end of an instrument, the increase in deformation within the current time step is caused by the flexible deformation of the slender instrument. The stiffness is indirectly calculated using Newton's second law from the IMU acceleration signal and the known mass of the instrument; K(t) is the time-varying stiffness matrix, which physically represents the reciprocal of the deformation displacement caused by a unit force under the current configuration of the instrument. This term takes the difference between the deformation displacement at the current moment and the previous moment, rather than the absolute deformation. When the external force is constant, =0, the deformation increment is zero, and the end position does not drift; a non-zero deformation increment is only generated when the external force changes, which mathematically guarantees the stability of the recursive model.

[0062] 1.4 Solving for the Time-Varying Stiffness Matrix K(t)

[0063] The time-varying stiffness matrix K(t) is solved in real time based on the bending deformation equation of the cantilever beam. The specific steps are as follows:

[0064] (1) The cantilever length L(t) of the instrument entering the puncture tube is calculated in real time using the forward kinematics of the robotic arm. The puncture point corresponds to the remote center of motion (RCM) of the robotic arm, which is a fixed three-dimensional coordinate point known through preoperative calibration in the robot base coordinate system. The system obtains the angles of each joint in real time through the joint encoder of the robotic arm, and calculates the three-dimensional spatial distance of the instrument end relative to the RCM in real time by combining forward kinematics (FK). This distance is the cantilever length L(t). Depth image data is used to verify the actual deformation deviation of the instrument end from the visual end and is fused with inertial sensor data to improve the overall estimation accuracy; it is not directly used to measure L(t).

[0065] (2) The slender surgical instrument is modeled as a uniform cross-section Euler-Bernoulli cantilever beam, whose bending stiffness EI is related to the material's elastic modulus E and the moment of inertia I of the cross section. For a circular cross-section instrument (diameter d), the moment of inertia of the cross section is... .

[0066] (3) According to the bending deformation equation of the cantilever beam, under the action of the lateral force F at the end, the deformation displacement at the end is Therefore, the lateral stiffness coefficient The axial stiffness coefficient is determined based on the elastic modulus and cross-sectional area of ​​the instrument material. , where A is the cross-sectional area.

[0067] (4) Combine the stiffness coefficients of the three orthogonal directions to construct the time-varying stiffness matrix. Note that in actual surgical procedures, due to the constantly changing directions of gravity and tissue reaction forces, the K(t) matrix may need to undergo real-time rotation and transformation between the world coordinate system and the instrument local coordinate system.

[0068] (5) To compensate for the depth image itself Delay, and recursively estimate L(t): ,in The linear velocity of the instrument entering and exiting the tamper sleeve is obtained by integrating the IMU acceleration signal. Substitute L(t) into the bending deformation equation to calculate K(t), thereby eliminating the influence of depth image time delay on stiffness estimation.

[0069] (6) To achieve accurate stiffness modeling for different surgical instruments, the system integrates an instrument identification unit (such as an RFID reader or digital chip interface) at the instrument interface of the robotic arm end effector. This unit automatically reads the unique model code of the currently connected instrument during instrument assembly. The system has a pre-installed instrument stiffness parameter database, which stores the equivalent elastic modulus of each instrument model as calibrated at the factory. Equivalent cross section moment of inertia And geometric dimensional parameters. For non-solid cylindrical instruments with complex internal structures (such as built-in wire drives, hollow tubing), and These are equivalent parameters pre-determined through standard cantilever beam bending tests, rather than based on simple geometric formulas (such as...). The theoretical calculated value. After identifying the instrument model, the system automatically retrieves the corresponding parameters from the database and substitutes them into the real-time calculation of K(t).

[0070] 1.5 Sliding Mode Controller Design

[0071] This embodiment uses a sliding mode controller (SMC) based on a saturation function to calculate the dynamic reverse compensation torque. The mathematical expression of the controller is:

[0072] ;

[0073] in:

[0074] The dominant angular frequency of the flutter extracted based on wavelet packet transform. The phase of the flutter displacement signal is rotated in the future direction by a phase lead compensator. · The time-delay compensation advance prediction value obtained from the radian. Specifically: the dominant frequency component of the flutter signal is extracted based on wavelet packet transform, and its angular frequency is set to be... The frequency domain representation of the flutter displacement is obtained by using a phase lead compensator. · The radian value is used to obtain a lead prediction signal with amplitude fidelity and phase alignment. The implementation method employs a phase-locked loop (PLL) based approach to track the dominant frequency in real time and adjust the phase lead angle online: the PLL uses the current dominant frequency component of the flutter signal as a reference input, and through a phase detector, loop filter, and numerically controlled oscillator, it performs closed-loop tracking and outputs a real-time estimate of the current dominant frequency's angular frequency. Phase lead compensator according to and The product of these factors generates a phase rotation factor, thus achieving phase lead compensation. This scheme avoids the problems associated with linear extrapolation in traditional Taylor expansions. The problem of amplitude surge and phase disorder that occurs near the tremor period.

[0075] and These are the proportional gain and the derivative gain, respectively. In this embodiment, The value ranges from 50 to 150 N·m / m. The value ranges from 5 to 20 N·m·s / m, and the specific value is determined by parameter tuning before the operation based on the type of instrument and the surgical scenario.

[0076] The saturation function of the boundary layer thickness Φ is defined as:

[0077] When |S|≤Φ = (Linear control within the boundary layer);

[0078] When |S|>Φ =sgn(S) (Boundary layer external switch control).

[0079] Where S is the sliding surface function. (c is the slope coefficient of the sliding surface); The boundary layer thickness is taken as [value missing] in this embodiment. =0.05. Replacing the traditional sign function sgn(S) with a saturation function can eliminate high-frequency chattering of the control torque, resulting in a smooth transition within the boundary layer rather than a hard switch.

[0080] λ is the sliding mode switching gain, and its value must satisfy the following conditions: , where d(t) is the upper bound of the system modeling error and external disturbance.

[0081] 1.6 Simulation Verification Data

[0082] To verify the technical effectiveness of this embodiment, the indicators in Table 1 were used for evaluation in a comparative simulation with existing technologies.

[0083] Table 1 Simulation Quantitative Index Evaluation Table

[0084] Traditional low-pass filter (Butterworth 4th order, cutoff frequency 6Hz) Approximately 45% Approximately 25ms The high-frequency components of the intended motion are attenuated. Kalman filtering (linear model) Approximately 58% Approximately 15ms Model linearization error causes distortion of some intended signals. This plan ( +K(t) joint modeling +SMC) Approximately 82% After compensation, the equivalent is 0ms Intended movement is fully preserved, with only the directional component of tremor suppressed.

[0085] Expected Results Analysis: Traditional low-pass filtering attenuates both the vibration frequency band (8-12Hz) and the intended motion signal, resulting in limited suppression and introducing phase delay. While Kalman filtering can improve performance to some extent, its linearized model fails to capture the nonlinear stiffness variations of slender instruments. This proposed solution... By predicting and eliminating phase delay in advance, eliminating nonlinear deformation error through real-time modeling of K(t), and applying anisotropic compensation only to the direction of tremor, the tremor suppression rate and operational fidelity are significantly better than existing technologies.

[0086] Example 2: Non-intrusive intent decoupling and bidirectional fatigue tracking

[0087] This embodiment details a pure data-driven intent decoupling scheme based on minimal Jerk and a bidirectional adaptive fatigue tracking mechanism based on wavelet packet transform.

[0088] 2.1 Mathematical expression of minimal Jerk intention decoupling

[0089] This embodiment employs the Minimum Jerk model, based on the prior law of smoothness in natural human movement, for extracting operational intent. The fundamental law of human voluntary movement is that when the nervous system performs point-to-point voluntary movement, it tends to minimize the square integral of the jerk (the derivative of acceleration) of the movement trajectory. This law has been confirmed by classic research in the field of neuromotor control (Flash & Hogan, 1985).

[0090] The classical theoretical form of this law is as follows: For a one-dimensional trajectory x(t), the minimum Jerk objective functional is defined as:

[0091] ;

[0092] However, this classical form requires known position and velocity boundary conditions at the terminal time T, which cannot be satisfied in real-time surgical scenarios. Therefore, this embodiment designs a real-time causal optimal state observer based on the minimum jerk criterion, rather than directly solving the two-point boundary value problem (BVP). Specifically, a third-order linear state-space model is constructed with the instrument end-effector position, velocity, and acceleration as state variables. Jerk is treated as process noise, and its variance weighting matrix is ​​set to penalize large Jerk values. Using a Kalman filter framework, the smoothed intention trajectory is recursively estimated in real time at the current time, relying only on historical observation data (without future terminal conditions). The equivalent physical meaning of this observer is: under the criterion of minimizing the expected jerk, causally optimal smoothing is performed on the original motion signal containing tremors to extract the intention baseline that conforms to the prior law of smoothing natural human motion.

[0093] Actual measurement trajectory With the baseline of smoothing intention Making a difference:

[0094] ;

[0095] The obtained zero-mean residual signal This refers to physiological high-frequency tremor data.

[0096] 2.2 Basis for Selecting the Length of the Sliding Time Window

[0097] The sliding time window length was set to 100ms (correcting the 50ms setting in the earlier solution), and the selection criteria are as follows:

[0098] (1) The dominant frequency range of physiological tremor is usually 8-12Hz, and the corresponding tremor period is 83-125ms. A 100ms window can cover at least one complete tremor period, ensuring that the window contains sufficient tremor signal features for frequency domain analysis.

[0099] (2) Frequency domain resolution This resolution can effectively distinguish the vibration frequency components around 8Hz and 12Hz, meeting the accuracy requirements for vibration dominant frequency identification.

[0100] (3) If the window length is too short (e.g., 50ms), the frequency domain resolution will be low. It cannot effectively distinguish between 8Hz and 12Hz jitter components; if the window length is too long (e.g., 200ms), the response delay to changes in operator intent will be too large, affecting real-time performance. 100ms achieves the optimal balance between frequency resolution and time-domain response speed.

[0101] (4) Considering that the IMU sampling frequency is 200Hz and the 100ms window corresponds to 20 sampling points, it just meets the requirement of the Kalman filter state observer to obtain sufficient observation data in each recursive cycle, thus ensuring the statistical reliability of the smooth estimation of the intention trajectory.

[0102] 2.3 Wavelet Packet Transform and Spectral Centroid Tracking

[0103] For the separated flutter residual signal Perform three-level wavelet packet decomposition:

[0104] (1) The Daubechies4 (db4) wavelet basis function is selected. This wavelet basis has good tight support and smoothness, and is suitable for time-frequency analysis of non-stationary biological signals such as physiological tremor signals.

[0105] (2) After the three-layer wavelet packet decomposition, eight sub-bands are obtained (divided into 12.5Hz intervals from 0Hz to 100Hz), covering the complete frequency domain range of the tremor signal.

[0106] (3) Calculate the signal energy for each sub-band. ,in The wavelet packet coefficients of the i-th sub-band are used to reconstruct the signal.

[0107] (4) Calculate the spectral centroid, which is the center frequency of the band with the highest proportion of tremor energy: ,in The center frequency of each sub-band. The energy of each sub-band.

[0108] (5) Continuous tracking The time series is analyzed, and its sliding variance is calculated. When the variance continuously increases within a preset time window (e.g., 30 consecutive seconds) and exceeds a variance threshold, the variance is considered. At that time, it was determined that the dominant frequency of the tremor had undergone substantial drift.

[0109] 2.4 Bidirectional Adaptive Bandwidth Adjustment and Hysteresis Logic

[0110] The bidirectional adaptive bandwidth adjustment mechanism in this embodiment is designed as follows:

[0111] Fatigue decline phase: When the spectral centroid is detected Due to fatigue from prolonged operation, the frequency continues to drift towards lower frequencies (e.g., attenuation from 10Hz to 6Hz), and the drift amplitude exceeds a preset threshold. (e.g., 1.5Hz) The system automatically lowers the cutoff frequency of the high-pass filter from the default 8Hz to 5.5Hz, dynamically widening the jitter capture bandwidth to ensure that low-frequency large-amplitude jitter can still be fully included in the compensation range.

[0112] Recovery phase: After a short rest by the operator, the following was detected. It rebounded towards higher frequencies and exceeded the recovery threshold. (e.g., 7.5Hz) The system automatically adjusts the cutoff frequency from 5.5Hz back to 8Hz to restore the standard bandwidth, avoiding unnecessary low-frequency noise interference introduced by an excessively wide filter bandwidth.

[0113] Hysteresis range design: To prevent the cutoff frequency from frequently switching near the threshold boundary (such as oscillating back and forth around 8Hz), the trigger threshold is lowered. ) and rebound trigger threshold ( A hysteresis interval of 1.0 Hz is provided between ). When When the frequency is between 6.5 and 7.5 Hz, the cutoff frequency remains unchanged.

[0114] 2.5 Discussion of Boundary Conditions for the Zero Mean Assumption

[0115] In this embodiment, the underlying assumption for decoupling is that the flutter residual signal is a zero-mean random process. This assumption holds under the following conditions:

[0116] (1) The operator’s active control intention movement can be approximated as a smooth low-frequency signal (main frequency <3Hz) within a 100ms time window, which is fully separated from the 8-12Hz vibration signal in the frequency domain.

[0117] (2) Minimum Jerk optimization can accurately extract the low-frequency components of the intended motion without “overfitting” the high-frequency tremor components.

[0118] (3) When a doctor performs rapid, wide-ranging maneuvers (such as moving instruments significantly to switch surgical areas), transient motion may deviate from the assumption that the intended motion contains only low-frequency components. In this case, the system monitors the acceleration amplitude and briefly suspends tremor compensation when it exceeds a preset acceleration threshold. The preset acceleration threshold is determined by statistically analyzing the distribution of acceleration amplitude under normal fine manipulation and selecting the upper boundary value of a specific confidence interval (such as a 95% confidence interval) as the preset acceleration threshold. The duration of suspending tremor compensation is an adaptive duration, during which the system continuously monitors the real-time acceleration amplitude. The condition for resuming compensation after the transient motion ends is as follows: when the system detects that the real-time acceleration amplitude has fallen below a specific proportion (such as 80%) of the preset acceleration threshold and has been maintained within this range for more than a set time window (such as 50ms), the transient motion is determined to have ended, and the system automatically resumes tremor compensation. The approximately 200ms mentioned in the previous scheme is only a typical empirical reference duration for suspending large-scale movements. The actual suspension and recovery movements are completely adaptively determined by the above threshold conditions, thereby avoiding misjudging rapid intentional movements as tremors.

[0119] Example 3: Anisotropic Space Constraint Protection

[0120] This embodiment describes in detail the space constraint protection mechanism based on anisotropic damping strategy.

[0121] 3.1 Shortest safe distance Real-time monitoring

[0122] The minimum safe distance from the instrument tip to the cavity wall or surrounding tissue is continuously calculated in real time using a binocular depth camera. In this embodiment, the shortest safe distance Specifically, this is achieved by calculating the shortest Euclidean distance in three-dimensional space. The specific implementation is as follows:

[0123] (1) Extract the three-dimensional spatial coordinates of the instrument end effector from the depth image. .

[0124] (2) Segment the surface point cloud of the cavity wall and surrounding tissue in the depth image to obtain the three-dimensional spatial coordinates of each point on the tissue surface.

[0125] (3) Calculation The shortest Euclidean distance to each point cloud on the tissue surface: .

[0126] 3.2 Three-level security threshold system

[0127] Based on surgical safety requirements, thresholds are set according to Table 2;

[0128] Table 2. Three-level security threshold settings for different organization types.

[0129] Level 1 (Highest Alert) Major blood vessels (aorta, vena cava, etc.) 10mm If the injury could lead to massive bleeding, the vibration displacement must be limited to the greatest extent possible. Level 2 (Medium Alert) The walls of hollow organs such as the intestinal wall and gallbladder wall 5mm Damage may result in perforation or leakage of contents. Level 3 (Regular Alert) Nerve bundles, ureters, etc. 3mm Damage may lead to long-term functional impairment

[0130] When the value is less than the corresponding safety threshold, the system automatically identifies it as a high-risk boundary condition and triggers anisotropic damping strategy.

[0131] 3.3 Vector Decomposition of Anisotropic Damping

[0132] The core idea of ​​the anisotropic damping strategy is to distinguish between the direction of tremor disturbance and the direction of movement intended by the physician, apply high damping only to the direction of tremor disturbance, and maintain normal operating feel in the direction of intended movement.

[0133] The specific mathematical implementation is as follows:

[0134] (1) Define the unit vector of the direction of movement of the doctor's intention to actively control the movement. At time t, based on the smoothed intention trajectory obtained from minimum Jerk intention decoupling, the intention motion velocity vector at the current time is extracted. (t), normalized to obtain .

[0135] (2) Define the direction of the jitter disturbance. The jitter residual signal... The instantaneous direction vector e of (t) is normalized to Generally, tremor signals are mostly bidirectional oscillations, with the direction of their principal component being orthogonal or nearly orthogonal to the intended direction of motion.

[0136] (3) Projective decomposition of the flutter disturbance vector. Two quantities are known:

[0137] Projection calculation involves two steps:

[0138] Parallel component (along the direction of the doctor's intention):

[0139] ;

[0140] The dot product of the instantaneous direction vector e results in a scalar representing the length of the flutter residual projected onto the intended direction, which is then multiplied by the unit vector. , obtained It is a three-dimensional vector parallel to the direction of the intention.

[0141] Orthogonal components (perpendicular to the intended direction, i.e., the direction of the actual dangerous tremor disturbance):

[0142] ;

[0143] Subtracting the parallel component from the total thrill residuals leaves the component perpendicular to the intended direction. This component points in the dangerous direction that could potentially scratch surrounding tissue.

[0144] (4) Adjust the differential gain in each direction

[0145] Core logic: Amplify damping only in the dangerous direction (orthogonal direction), and maintain normal feel in the safe direction (intended direction).

[0146] Intended direction (parallel component):

[0147] ;

[0148] The original differential gain remains unchanged. When the surgeon moves the instrument along the surgical direction, the system does not add additional damping, and the handling is unaffected.

[0149] Direction of flutter disturbance (orthogonal component):

[0150] ;

[0151] in:

[0152] Original differential gain coefficients (differential term coefficients in the PD controller)

[0153] : Preset safe distance threshold (e.g., 10mm near large blood vessels)

[0154] The shortest Euclidean distance (in mm) from the instrument tip to the nearest tissue surface, as currently measured.

[0155] α: An adjustable sensitivity coefficient (e.g., 3) that controls the amplification rate of damping as distance decreases.

[0156] exp(): An exponential function with base e.

[0157] The actual effect is:

[0158] when Much larger When the instrument is far from the tissue, it is safe: If ) is negative, exp(negative number) approaches 0. near There is almost no additional damping.

[0159] when near When (instrument is close to tissue): ( As the expression approaches 0, exp(0) = 1. near ×e, the damping begins to increase.

[0160] when Less than When (instrument approaches tissue, danger): ( When the exponent is positive, the exponent term rapidly increases, and the damping increases sharply.

[0161] when When the value approaches 0, it indicates that the instrument is almost touching the tissue. ≈1, exp(α) is greatly amplified, the damping tends to be at its maximum, and the instrument is physically prevented from continuing to approach the tissue.

[0162] The final form of the entire compensation torque in the orthogonal direction:

[0163]

[0164] In the formula It is the orthogonal tremor component obtained from step (3). This is the amplified differential gain obtained from step (4). Damping is precisely applied in the dangerous direction, while the original damping is still used in the safe direction. .

[0165] (5) The damping term in the final compensation torque is decomposed as follows:

[0166] ;

[0167] The effect of this damping strategy is to maintain a light and easy feel of control in movements that are in the same direction as the doctor's control, while applying exponentially high damping in the direction of tremor disturbance that is orthogonal to the direction of control, thus physically limiting the instantaneous displacement amplitude in the direction of tremor.

[0168] 3.4 Clinical Scenario Examples

[0169] Taking the anatomy of the Calot's Triangle during cholecystectomy as an example: When the instrument tip is within 10mm of the common hepatic artery (Level 1 threshold trigger), the system applies damping of approximately 5-10 times the normal value in the direction of thrill disturbance. This compresses the displacement amplitude of the instrument tip in the thrill direction to the sub-millimeter level, while maintaining normal damping during the surgeon's active manipulation along the bile duct. The surgeon experiences almost no additional resistance. Once the instrument has safely moved out of the 10mm range, the damping automatically returns to normal.

[0170] Example 4: Patient physiological motor feedforward compensation and multi-device synergy

[0171] 4.1 Depth Image Extraction of Patient Physiological Motion

[0172] During prolonged thoracoscopic and laparoscopic surgeries, the patient's spontaneous breathing and heartbeat cause periodic movements of the diaphragm, chest wall, and intra-abdominal organs. These movements, transmitted through tissues, affect the relative position of the instrument tip to the target tissue. This embodiment extracts information about these periodic movements from depth images:

[0173] (1) In consecutive frames of the depth image, set the region of interest (ROI, such as a 20×20 pixel window) in the diaphragm or chest wall region.

[0174] (2) Track the average change of depth values ​​within the ROI frame by frame to obtain a one-dimensional time series. This represents the periodic displacement of the cavity wall caused by respiratory movements.

[0175] (3) To Perform a Fast Fourier Transform (FFT) to identify the dominant respiratory frequency (typically 0.15-0.4 Hz, corresponding to an adult respiratory rate of 9-24 breaths / minute) and the dominant cardiac frequency (typically 1.0-1.5 Hz, corresponding to a heart rate of 60-90 beats / minute).

[0176] (4) Establish a sinusoidal fitting model:

[0177] ;

[0178] in and The amplitudes of breathing and heartbeat (estimated via online least squares fitting). and For frequency, and Let δ be the phase, and δ be the prediction lead (usually taken as...). To ensure that the predicted displacement and the compensation torque are aligned in time.

[0179] 4.2 Feedforward superposition to sliding mode compensation torque

[0180] Predicted physiological movement displacement As a feedforward quantity, it is added to the sliding mode compensation torque of Example 1 to obtain an enhanced compensation torque:

[0181] ;

[0182] in The feedforward gain matrix maps the displacement prediction values ​​to the equivalent joint torque. As in Example 1, based on tremor and The reference compensation torque.

[0183] The physical effect of this scheme is that the compensating torque not only counteracts the instrument shaking caused by the operator's hand tremor, but also counteracts the relative movement of tissue and instrument caused by the patient's breathing / heartbeat, thus achieving dual stability of tremor suppression and physiological motion compensation.

[0184] 4.3 Intraoperative stiffness self-calibration

[0185] This embodiment also includes an intraoperative stiffness self-calibration step:

[0186] (1) When the end of the instrument contacts a tissue surface with known stiffness (such as bony structures - rib surface, vertebral body surface - whose stiffness is much greater than that of soft tissue, and can be regarded as rigid contact), the system determines that it enters the calibration mode.

[0187] (2) In calibration mode, the system records the contact force. (Calculated from IMU acceleration signals and instrument mass) and deformation displacement measured from depth images .

[0188] (3) Based on linear elasticity relationship: The stiffness coefficients in the corresponding directions of the K(t) matrix are updated online.

[0189] (4) After calibration, use It replaces the initial stiffness value calculated based on the theoretical cantilever beam formula, thereby improving the accuracy of subsequent vibration compensation.

[0190] 4.4 Expansion of Multi-device Collaboration Scenarios

[0191] When two or more surgical instruments (such as a main operating instrument + an auxiliary traction instrument) are present in the surgical field, this embodiment further introduces a multi-instrument collaboration mechanism:

[0192] (1) Each device independently performs steps S1-S4 of claim 1 and calculates its own vibration compensation torque.

[0193] (2) Each instrument shares the following data via a network (such as the real-time Ethernet bus inside the surgical robot): the current spatial coordinates of the instrument's end effector. Current predicted seismic displacement (t+ Current safe distance .

[0194] (3) Add a collision prediction module between instruments: Based on the spatial coordinates of the ends of each instrument and the predicted vibration displacement, calculate the shortest distance between any two instruments in real time. ,when When the collision safety threshold is less than 1, a mutual repulsion term is superimposed on each of the compensation torques.

[0195] (4) The mutual repulsion term is calculated using the artificial potential field method:

[0196] #imgpt178#;

[0197] Where η is the repulsive force gain coefficient. This repulsive force term ensures that vibration compensation will not cause accidental collisions between two instruments, thus guaranteeing the safety of multi-instrument collaborative operation.

[0198] (5) The data audit module of each instrument records each other’s spatial position and tremor data simultaneously. After the operation, the multi-instrument coordinated motion trajectory of the entire surgical field can be reconstructed to analyze the tremor coupling effect in multi-instrument operation.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent compensation and control of micro-vibrations in thoracoscopic and laparoscopic surgical instruments, characterized in that, Includes the following steps: S1: Synchronously acquire inertial sensing data and endoscopic visual depth image data from the end effector of the robotic arm; S2: Based on a preset visual time delay constant The cantilever length L(t) of the device entering the card sleeve is calculated in real time by the positive kinematics of the robotic arm. A nonlinear kinematic model containing the time-varying stiffness matrix K(t) is constructed to describe the actual spatial position of the device end under the combined action of time delay and flexible deformation. S3: The nonlinear kinematic model is used to decouple the end-effector motion data from the motion intent, separating the active control intention motion data from the physiological high-frequency tremor data; S4: Based on the separated high-frequency tremor data and the visual time delay constant The vibration displacement is predicted in advance by feedforward, and the dynamic reverse compensation torque is calculated in real time by sliding mode control algorithm and output to the servo motor of the robotic arm joint to counteract the vibration displacement. S5: The shortest safe distance from the instrument tip to the cavity wall or surrounding tissue, as monitored in real time based on intraoperative depth images. The compensation control parameters are dynamically adjusted.

2. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 1, characterized in that, In step S3, a third-order linear state-space model is constructed with the device end position, velocity, and acceleration as state variables. Jerk is used as process noise, and its variance weighting matrix is ​​set to penalize large Jerk values. Using a Kalman filter framework, a smoothed operational intention baseline is estimated in real time based solely on historical observation data as the active control intention motion data. The reconstructed actual motion trajectory is subtracted from the active control intention motion data to obtain a zero-mean residual signal as the physiological high-frequency tremor data.

3. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 2, characterized in that, In step S3, the motion intention decoupling further includes a fatigue time-frequency adaptive adjustment step: performing wavelet packet transform on the physiological high-frequency tremor data to dynamically identify the center frequency of the frequency band with the highest tremor energy proportion; when it is detected that the center frequency drifts to a lower frequency direction due to long-term operational fatigue and exceeds a preset variance threshold, the filter cutoff frequency is automatically lowered to broaden the tremor capture bandwidth; when it is detected that the center frequency rises back to a higher frequency direction and exceeds a preset recovery threshold, the filter cutoff frequency is automatically raised to restore the standard bandwidth.

4. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 1, characterized in that, In step S2, the nonlinear kinematic model adopts an incremental form, and its expression is: ； in, The actual spatial position of the end point with deformation and time delay compensation. Let J(q) be the total time delay constant for visual transmission and image processing, J(q) be the Jacobian matrix of the robot arm joint space, q be the joint angle, and q̇ be the joint angular velocity. The equivalent external force at the end measured by the inertial sensor, K(t) is the time-varying stiffness matrix; the incremental integration interval is from To t, s < hour An undefined problem.

5. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 4, characterized in that, The time-varying stiffness matrix K(t) is solved as follows: the tamper point corresponds to the remote motion center of the robotic arm, which is a fixed three-dimensional coordinate point known through preoperative calibration in the robot base coordinate system; the system obtains the angles of each joint in real time through the joint encoder of the robotic arm, and calculates the three-dimensional spatial distance of the instrument end relative to the remote motion center in real time by combining the forward kinematics (FK), which is the cantilever length L(t); at the same time, the cantilever length L(t) is recursively estimated based on the speed of the instrument entering and exiting the tamper sleeve to compensate for the visual time delay of the depth image itself.

6. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 3, characterized in that, In step S4, the dynamic reverse compensation torque The calculation is based on the sliding mode control algorithm, and the formula is as follows: ； in, The dominant angular frequency of the flutter extracted through time-frequency analysis The phase of the flutter displacement signal is rotated in the future direction by a phase lead compensator. · The time-delay compensation advance prediction value obtained from radians The time derivative of the flutter displacement signal is used to predict the flutter velocity signal after time delay compensation. and These are the proportional gain coefficient and the derivative gain coefficient of the controller, respectively. Let λ be the saturation function of the boundary layer thickness Φ, and λ be the sliding mode switching gain, whose values ​​satisfy... d(t) is the upper bound of the system modeling error and external disturbance.

7. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 1, characterized in that, In step S5, the spatial constraint protection adopts an anisotropic damping strategy: when the shortest safe distance When the resistance is below the safety threshold, normal damping is maintained in the direction parallel to the direction of the doctor's intended movement, while the differential gain coefficient is exponentially amplified in the direction of the tremor disturbance orthogonal to that direction. To selectively apply high damping.

8. The intelligent compensation and control method for micro-vibrations of thoracoscopic and laparoscopic surgical instruments according to claim 1, characterized in that, The control method further includes a patient physiological motion feedforward compensation step: extracting the periodic displacement of the cavity wall caused by the patient's breathing and / or heartbeat from the depth image data, establishing a periodic motion prediction model, and superimposing the predicted displacement as a feedforward quantity into the reverse compensation torque in step S4 to counteract the disturbance of the patient's physiological motion to the device end.

9. A smart compensation and control system for minute vibrations of thoracoscopic and laparoscopic surgical instruments, characterized in that, include: The multi-source sensing module is used to simultaneously acquire inertial sensing data and endoscopic visual depth image data from the end effector of the robotic arm. The nonlinear kinematics modeling module is used for modeling based on a preset visual time delay constant. The cantilever length L(t) of the device entering the card sleeve is calculated in real time by the positive kinematics of the robotic arm. A nonlinear kinematic model containing the time-varying stiffness matrix K(t) is constructed to describe the actual spatial position of the device end under the combined action of time delay and flexible deformation. The motion intention decoupling module is used to decouple the end motion data from the nonlinear kinematic model, separating the active control intention motion data from the physiological high-frequency tremor data. An adaptive torque compensation module is used to compensate for high-frequency vibration data and the visual time delay constant. The vibration displacement is predicted in advance by feedforward, and the dynamic reverse compensation torque is calculated in real time by sliding mode control algorithm and output to the servo motor of the robotic arm joint to counteract the vibration displacement. The dynamic closed-loop fine-tuning module is used to monitor the shortest safe distance between the instrument tip and the cavity wall or surrounding tissue in real time based on the depth images described during the procedure. The compensation control parameters are dynamically adjusted.

10. The intelligent compensation and control system for micro-vibration of thoracoscopic and laparoscopic surgical instruments according to claim 9, characterized in that, The motion intent decoupling module integrates: a minimum jerk optimizer, which constructs a minimum jerk target functional within a sliding time window to optimize and fit the actual motion trajectory in order to extract the operation intent baseline; and a wavelet packet transformer, which performs time-frequency decomposition on high-frequency tremor data and dynamically identifies the center frequency of the frequency band with the highest tremor energy proportion. It also includes a bidirectional adaptive frequency band adjuster, which automatically lowers or raises the filter cutoff frequency based on the detected center frequency drift direction.

Images