Self-adaptive sensing anti-shake regulation and control method and system for spine surgery auxiliary mechanical arm

Abstract

The invention discloses a self-adaptive sensing anti-shake regulation and control method and system for a spinal surgery auxiliary mechanical arm, and relates to the technical field of spinal surgery, and the method comprises the steps: collecting multi-dimensional data of a mechanical arm tail end actuator through a multi-dimensional sensor group, and transmitting the multi-dimensional data to an AeroStead space-level stability calculation platform; the platform carries out filtering processing on the data, and vibration of the mechanical arm and disturbance components caused by external interference are separated; constructing a spine mechanical arm rigid-flexible coupling dynamic disturbance observation and compensation model by using the disturbance component to observe disturbance and generate a compensation amount; calling a spinal operation area three-dimensional reconstruction and dynamic obstacle avoidance mixed density network model, and fusing different images to generate an operation area three-dimensional dynamic model and an obstacle avoidance adjustment signal; a mechanical arm motion control instruction is generated through a force sense position cooperation self-adaptive control algorithm; and the mechanical arm is driven to move for closed-loop regulation and control. According to the method, through fusion of multiple models and algorithms, the disturbance compensation precision, the operation area modeling accuracy and the control collaboration are improved, and the operation safety is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of spinal surgery technology, and in particular to an adaptive sensing anti-shake control method and system for spinal surgery assistive robotic arms. Background Technology

[0002] In spinal surgery, the confined surgical space and complex anatomical structures of the surgical area place extremely high demands on the precision and stability of the robotic arm. Traditional manual operation is susceptible to factors such as surgeon's hand tremors and operator fatigue, making it difficult to guarantee high-precision operation over extended periods. Surgical assistive robotic arms, however, can improve operational stability through automated control. With technological advancements, robotic arms need to integrate multi-dimensional data acquisition, dynamic disturbance processing, three-dimensional surgical area modeling, and adaptive control to address issues such as vibrations of the robotic arm's own structure, external environmental interference, and dynamic changes in the surgical area during surgery. The application of technologies such as the AeroStead spatial stability computing platform, rigid-flexible coupling dynamic model, hybrid density network model, and force-position collaborative algorithm has become key to achieving adaptive sensing and anti-shake control of the robotic arm, aiming to provide more precise and safer auxiliary support for spinal surgery and meet the stringent clinical requirements for surgical precision and safety.

[0003] Existing technologies for assistive robotic arms in spinal surgery have two significant drawbacks: First, they lack the ability to compensate for dynamic disturbances. Most robotic arms only consider the dynamic characteristics of rigid structures and do not fully integrate the flexible deformation of joints and the elastic deformation of links to construct a complete disturbance observation and compensation model. This makes it difficult to compensate for disturbances caused by structural flexibility during robotic arm movement in real time and accurately, leading to positional deviations in the end effector and affecting surgical precision. Second, they lack coordination between dynamic obstacle avoidance and control in the surgical area. Existing technologies rely heavily on single image data when reconstructing the three-dimensional surgical area. Their ability to integrate preoperative and intraoperative images to generate dynamic models is weak, and the coordination and adaptation between obstacle avoidance signals and force-sensory position control are insufficient. They cannot quickly adjust control parameters according to real-time changes in the surgical area, which can easily lead to delayed obstacle avoidance response or mismatch between control commands and obstacle avoidance requirements, increasing surgical risks. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides an adaptive sensing anti-shake control method and system for a spinal surgery assistive robotic arm.

[0005] The technical solution adopted in this invention is an adaptive sensing and anti-shake control method for a spinal surgery-assisted robotic arm, comprising: Step S1, collecting real-time position data, attitude data, and interaction force data generated by the contact between the robotic arm's end effector and spinal surgical area tissues during movement using a multi-dimensional sensor group mounted on the spinal surgery-assisted robotic arm, and transmitting the collected data to the AeroStead spatial stabilization computing platform; Step S2, the AeroStead spatial stabilization computing platform performs real-time filtering on the received real-time position data, attitude data, and interaction force data to separate the disturbance components caused by the vibration of the robotic arm's own structure and the disturbance components caused by external environmental interference; Step S3, constructing a rigid-flexible coupling dynamic disturbance observation and compensation model for the spinal robotic arm based on the separated disturbance components, and using this model to observe the dynamic disturbances caused by the flexible deformation of the robotic arm joints and the elastic deformation of the connecting rods in real time, and generating corresponding disturbance compensation amounts; Step S4. A hybrid density network model combining 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area is invoked. Preoperative CT images of the spinal surgical area are fused with real-time intraoperative local images and input into this model to generate a 3D dynamic model of the spinal surgical area. Simultaneously, based on the predicted trajectory of the robotic arm, potential collision areas are marked in the 3D dynamic model, and obstacle avoidance adjustment signals are output. In step S5, the disturbance compensation amount generated in step S3 and the obstacle avoidance adjustment signal output in step S4 are input into a force-position collaborative adaptive control algorithm. This algorithm dynamically adjusts control parameters based on the position deviation, posture deviation, and interaction force deviation of the robotic arm's end effector, generating motion control commands for different joints of the robotic arm. In step S6, the motion control commands generated in step S5 are transmitted to the drive system of the spinal surgery assistive robotic arm, driving different joints of the robotic arm to move according to the control commands. Simultaneously, the joint motion state data fed back from the drive system is received in real time and fed back to the AeroStead spatial stability computing platform for closed-loop control.

[0006] Furthermore, the expressions for the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal manipulator are as follows: ,in, The inertia matrix of the rigid-flexible coupled dynamic model. Let the joint angle vector of the robotic arm be... The joint angular acceleration vector. The matrix of Coriolis force and centrifugal force. The joint angular velocity vector. For flexible deformation stiffness matrix, Here is the damping matrix. This is the joint driving torque vector. For dynamic disturbance torque; ,in, For disturbance compensation torque, These are the observed values ​​for the inertia matrix, the Coriolis force and centrifugal force matrix, the flexible deformation stiffness matrix, and the damping matrix, respectively. Let the joint's expected angle vector be... Let be the desired angular velocity vector of the joint. Let be the desired angular acceleration vector of the joint. Let be the desired joint driving torque vector.

[0007] Furthermore, the expressions for the three-dimensional reconstruction and dynamic obstacle avoidance hybrid density network model of the spinal surgical area are as follows: ,in, This is the probability distribution of 3D vertex coordinates output by the hybrid density network. The coordinate vector of the vertex of the three-dimensional model of the spinal surgical area. The input is the fused image feature vector. For model parameters, For the first The weights of the Gaussian components, The function is a Gaussian distribution. For the first The mean vector of Gaussian components, For the first The covariance matrix of Gaussian components; ,in, For dynamic obstacle avoidance cost function, The set of coordinates of obstacles in the surgical area. For predicting the coordinate set of the robotic arm's motion trajectory, For the cost function parameters, These are the weighting coefficients. The function for calculating Euclidean distance is... For the first The coordinates of the obstacle. For the first on the trajectory One predicted coordinate, For indicator functions, This is a collection of dangerous areas within the surgical region.

[0008] Furthermore, the expressions for the force-position cooperative adaptive control algorithm are as follows: ,in, For position control output, , These are the proportional coefficient, integral coefficient, and derivative coefficient, which are dynamically adjusted according to the position deviation. This is the position deviation vector. Let the vector be the rate of change of position deviation. For time; ,in, To control the output amount by force sensing, , These are the proportional coefficient and differential coefficient, respectively, which are dynamically adjusted according to force deviation. The force deviation vector, Let the force deviation change rate vector be... This is an estimate of the Jacobian matrix for the robotic arm. The end-effector interaction force vector, , For the final control command, These are the position and force control weighting coefficients.

[0009] Furthermore, the disturbance separation algorithm expression of the AeroStead spatial level stability calculation platform is as follows: ,in, The raw data collected, For useful signal components, For the disturbance component, For noise components, These are estimates of the disturbance components. These are the filter weight coefficients. The sampling period is This is the length of the filtering window. This is a disturbance trend compensation term.

[0010] Furthermore, the parameters for the adaptive sensing and stabilization control of the spinal surgery assistive robotic arm include the joint stiffness coefficient of the robotic arm. Joint damping coefficient End effector force sensing threshold Position control accuracy Attitude control accuracy AeroStead spatial level stable sensing computing platform sampling frequency Number of iterations for hybrid density networks The parameters satisfy the following relationship: For the dynamic response time of the robotic arm, , This represents the minimum distance between obstacles in the surgical area.

[0011] Further, step S3 includes the following sub-steps: S31, extracting motion parameters of different joints of the robotic arm, including joint angular displacement, angular velocity, and angular acceleration, and simultaneously collecting strain data of the robotic arm links, using these data as input data for the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm; S32, based on the input joint motion parameters and link strain data, using the finite element analysis method to calculate the degree of elastic deformation of the robotic arm links, and determining the influence coefficient of flexible deformation on dynamic characteristics; S33, substituting the influence coefficient into the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm, and observing the magnitude and direction of the dynamic disturbance generated during the movement of the robotic arm in real time through iterative calculation of the model; S34, based on the observed magnitude and direction of the disturbance, combined with the robotic arm joint driving capability parameters, calculating and generating a compensation amount that can offset the disturbance, ensuring that the compensation amount is within the output range of the robotic arm driving system.

[0012] Further, step S4 includes the following sub-steps: S41, segmenting the preoperative CT image data of the spinal surgical area, extracting the contour features of the vertebral bodies, intervertebral discs and surrounding important tissues, and simultaneously enhancing the local image data acquired in real time during the operation to highlight the local details of the surgical area; S42, registering and fusing the segmented preoperative CT image features with the enhanced intraoperative real-time image features to establish a unified coordinate system, aligning the two types of image data under the same coordinate system to form fused image features; S43, inputting the fused image features into the three-dimensional reconstruction and dynamic obstacle avoidance hybrid density network model of the spinal surgical area, the model generates three-dimensional point cloud data of the spinal surgical area through multi-layer neural network calculation, and then constructs a three-dimensional dynamic model of the spinal surgical area; S44, simulating the movement process of the robotic arm in the three-dimensional dynamic model of the spinal surgical area according to the motion trajectory planning results of the robotic arm end effector, marking the areas on the motion trajectory that may collide with the tissues in the surgical area, and generating corresponding obstacle avoidance adjustment signals.

[0013] Further, step S5 includes the following sub-steps: S51, acquiring the desired position, desired posture, and desired interaction force of the robotic arm end effector, comparing it with the real-time position, posture, and interaction force data collected in step S1, and calculating the position deviation, posture deviation, and interaction force deviation; S52, inputting the calculated deviations into the parameter adjustment module of the force-position collaborative adaptive control algorithm, which dynamically adjusts the proportional, integral, and derivative coefficients of the algorithm according to the magnitude and trend of the deviations; S53, using the adjusted control parameters, generating control components for robotic arm position adjustment and control components for interaction force control through the operation of the force-position collaborative adaptive control algorithm; S54, setting the weight ratio of position control and force control according to the surgical operation requirements of the surgical area, fusing the position control component and the force control component according to the ratio, and generating motion control commands for different joints of the robotic arm.

[0014] An adaptive sensing and anti-shake control system for a spinal surgery assistive robotic arm is presented. This system, applied to the adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms, includes: a robotic arm motion and force data acquisition unit, equipped with a multi-dimensional sensor array, positioned relative to the robotic arm's end effector and the spinal surgical area, for acquiring robotic arm motion data and interaction force data, and transmitting the data to an AeroStead spatial stabilization calculation and processing unit; the AeroStead spatial stabilization calculation and processing unit, which is the AeroStead spatial stabilization calculation platform, connected to the robotic arm motion and force data acquisition unit via a data bus, receiving data transmitted from the robotic arm motion and force data acquisition unit, and bidirectionally connected to a dual-model storage and retrieval unit and an adaptive control algorithm storage unit, calling the stored models and algorithms for data processing; and a dual-model storage and retrieval service unit, connected to the AeroStead spatial stabilization calculation and processing unit, internally storing observations and compensations for rigid-flexible coupling dynamic disturbances of the spinal robotic arm. The system includes a model, a 3D reconstruction of the spinal surgical area, and a dynamic obstacle avoidance hybrid density network model, which provides model call services to the AeroStead spatial level stability computing unit. A force-position control algorithm storage unit, connected to the AeroStead spatial level stability computing unit, stores force-position collaborative adaptive control algorithms and provides algorithm call services to the AeroStead spatial level stability computing unit. A control command conversion and drive unit, connected to the AeroStead spatial level stability computing unit via a control bus, receives motion control commands output by the AeroStead spatial level stability computing unit, connects to the robotic arm drive system, and converts the control commands into drive signals to drive the robotic arm movement. A joint motion state feedback transmission unit, connected to both the robotic arm drive system and the AeroStead spatial level stability computing unit, collects joint motion state data of the robotic arm and feeds the data back to the AeroStead spatial level stability computing unit, forming a closed-loop control link.

[0015] Beneficial Effects: This invention proposes an adaptive sensing and anti-shake control method and system for spinal surgery assistive robotic arms. It collects data using a multi-dimensional sensor array and separates disturbance components using the AeroStead spatial stabilization calculation platform. Then, utilizing a rigid-flexible coupling dynamic disturbance observation and compensation model for the spinal robotic arm, it fully considers joint flexibility deformation and link elastic deformation, generating disturbance compensation in real time. This significantly improves the dynamic disturbance compensation capability, effectively offsetting disturbances caused by structural flexibility during robotic arm movement, avoiding end effector position deviations, and overcoming the shortcomings of insufficient disturbance compensation in existing technologies. Furthermore, it leverages three-dimensional reconstruction of the spinal surgical area and dynamic avoidance... The obstacle avoidance hybrid density network model integrates preoperative CT images and intraoperative real-time images to generate a three-dimensional dynamic model. Combined with a force-position collaborative adaptive control algorithm, it enables dynamic adaptation of obstacle avoidance signals and control parameters, significantly improving the coordination of dynamic obstacle avoidance and control in the surgical area. It can quickly respond to changes in the surgical area and adjust control commands, avoiding obstacle avoidance lag or command mismatch problems, and overcoming the shortcomings of existing technologies in terms of coordination. The whole system continuously feeds back joint motion status data through a closed-loop control link, further improving the stability and accuracy of the robotic arm, providing more reliable auxiliary support for spinal surgery, and meeting the stringent clinical requirements for surgical precision and safety. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0017] Figure 2 This is a flowchart of method step S3 of the present invention;

[0018] Figure 3 This is a flowchart of method step S4 of the present invention;

[0019] Figure 4 This is a flowchart of step S5 of the method of the present invention;

[0020] Figure 5 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 1 As shown, the adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms includes:

[0023] Step S1: Collect real-time position data, posture data, and interaction force data generated by the contact with the spinal surgical area tissues of the end effector of the robotic arm during the movement process through the multi-dimensional sensor group mounted on the spinal surgery assisting robotic arm, and transmit the collected multiple sets of data to the AeroStead spatial stability computing platform.

[0024] Specifically, step S1 is the data foundation of the entire control method, providing accurate and comprehensive raw data support for all subsequent disturbance processing, model calculation, and control command generation. If the data acquisition in this step is incomplete or lacks accuracy, it will directly affect the control effect of all subsequent steps, and may even lead to deviations in surgical assistance operations. This step requires the use of a multi-dimensional sensor group mounted on the robotic arm to synchronously acquire key data during the movement of the end effector, including three categories: position, posture, and interaction force. These three types of data reflect the spatial position state, angular posture state, and contact state with the surgical area tissue, respectively. The combination of these three can fully present the real-time operating status of the robotic arm, providing comprehensive data input for subsequent disturbance separation, model construction, and control command generation, and is a prerequisite for achieving adaptive perception anti-shake control. In the specific implementation process, the robotic arm end effector is first equipped with a multi-dimensional sensor group consisting of displacement sensors, attitude sensors, and force sensors. The displacement sensors are laser displacement sensors with an accuracy of 0.001 mm, the attitude sensors are six-axis gyroscopes with a sampling frequency of 1000 Hz, and the force sensors are piezoelectric force sensors with a range of 0 to 50 N and an accuracy of 0.01 N, ensuring that the accuracy of various data acquisitions meets the needs of spinal surgery assistance. After the robotic arm begins to move, the displacement sensors collect the X, Y, and Z axis coordinate data of the end effector in three-dimensional space in real time, with the sampling interval set to 0.001 seconds; the attitude sensors simultaneously collect the pitch angle, yaw angle, and roll angle data of the end effector, generating a set of attitude data every 0.001 seconds; the force sensors continuously collect the interaction force data generated by the contact between the end effector and the spinal surgical area tissues, including axial force and radial force, also recording data at 0.001-second intervals. The collected position, attitude, and interaction force data are transmitted to the AeroStead spatial stability computing platform at a rate of 1000 frames per second via high-speed data transmission lines to ensure real-time data transmission and avoid data delays affecting subsequent processing steps.

[0025] Step S2: The AeroStead spatial stability calculation platform performs real-time filtering on the received real-time position data, attitude data, and interaction force data to separate the disturbance components caused by the vibration of the robotic arm's own structure and the disturbance components caused by external environmental interference.

[0026] Specifically, step S2 is a crucial data preprocessing step. It filters the acquired raw data to separate useful signals from disturbances and noise components, providing a clean data foundation for subsequent disturbance observation model construction and control command generation. Without this step, the vibrations of the robotic arm itself and external interference mixed in the raw data will directly enter the subsequent model calculations, leading to deviations in the model output and affecting control accuracy. This step relies on the signal processing capabilities of the AeroStead spatial stability computing platform. A specific filtering algorithm distinguishes between useful signals generated by the normal movement of the robotic arm and disturbance components caused by structural vibrations of the robotic arm and external environmental vibrations. Simultaneously, it removes noise generated during data acquisition, ensuring that the data input to the model only includes effective information reflecting the actual operating state of the robotic arm, laying the foundation for accurate disturbance observation and the generation of reasonable control commands. In the specific implementation process, the AeroStead spatial stability computing platform first receives the raw position, attitude, and interaction force data transmitted in step S1, and then activates its built-in adaptive filtering algorithm. The filtering window length is set to 50 sampling points to balance the filtering effect and real-time performance. For position data, the algorithm analyzes the rate of position change of adjacent sampling points to identify abrupt changes exceeding the normal motion rate range. This type of data is determined to be a disturbance component caused by the vibration of the robotic arm's own structure. Disturbances with vibration frequencies concentrated in the range of 5 to 20 Hz are effectively separated. For attitude data, the algorithm compares the attitude angle fluctuation amplitude with a preset stable attitude threshold and determines that the portion exceeding 0.01 degrees is a disturbance component, mainly originating from vibrations caused by the transmission gap of the robotic arm joints. For interaction force data, the algorithm eliminates instantaneous pulse forces through smoothing processing, and determines signals with force fluctuations exceeding 0.1 N and durations less than 0.005 seconds as disturbance components caused by external environmental interference. After filtering, the platform stores the three types of disturbance components separately, while transmitting the useful signal components after removing disturbances and noise to subsequent steps. The entire filtering process takes less than 0.002 seconds, ensuring that real-time control requirements are met.

[0027] Step S3: Construct a dynamic disturbance observation and compensation model for the rigid-flexible coupling of the spinal manipulator based on the separated disturbance components. Through this model, the dynamic disturbance caused by the flexible deformation of the manipulator joints and the elastic deformation of the connecting rods is observed in real time, and the corresponding disturbance compensation amount is generated.

[0028] Specifically, step S3 involves constructing a specialized dynamic model to observe the dynamic disturbances of the robotic arm caused by its structural flexibility in real time, and generating targeted compensation amounts to offset the impact of disturbances on the robotic arm's motion accuracy. This overcomes the shortcomings of existing technologies that only consider rigid structures and ignore flexible deformation, resulting in insufficient disturbance compensation. Based on the disturbance components separated in step S2, and combined with the structural characteristics of the robotic arm itself, this step constructs a rigid-flexible coupling dynamic disturbance observation and compensation model for the spinal robotic arm. It fully considers the two main flexible factors: joint flexible deformation and link elastic deformation, accurately capturing the magnitude and direction of dynamic disturbances, and then generating compensation amounts that can be directly used to adjust the robotic arm's motion. This ensures that the robotic arm's end effector always maintains a stable motion state, improving the accuracy of surgical assistance operations. In the specific implementation process, the basic framework of the model is first constructed based on the structural parameters of the robotic arm. The robotic arm joints adopt a flexible hinge structure with a stiffness coefficient set at 5000 N / m, and the links are made of carbon fiber composite materials with an elastic modulus of 200 GPa. These parameters serve as the initial inputs to the model. After the model starts, it first receives the disturbance component separated in step S2. Combined with the real-time joint angular displacement (range 0 to 180 degrees), angular velocity (maximum 10 degrees / second), and angular acceleration (maximum 5 degrees / second²) data of the robotic arm, it calculates the disturbance torque caused by the joint's flexible deformation in real time through dynamic equations. The disturbance torque corresponding to a maximum deformation of 0.05 mm is approximately 2 N·m. Simultaneously, it observes the end-effector position deviation caused by the elastic deformation of the connecting rod; a maximum deformation of 0.03 mm corresponds to a position deviation of 0.04 mm. After observation, the model generates a disturbance compensation amount based on the magnitude of the disturbance torque and position deviation, combined with the maximum output torque (5 N·m) and maximum adjustment speed (15 degrees / second) of the robotic arm joint drive motor. The torque adjustment range of the compensation amount is 0 to 3 N·m, and the position compensation range is 0 to 0.06 mm, ensuring that the compensation amount is within the drive system's load-bearing capacity. The generated compensation amount is transmitted in real-time to step S5 for subsequent control command adjustments. The entire observation and compensation amount generation process takes less than 0.003 seconds, meeting the real-time control requirements.

[0029] Step S4: Call the hybrid density network model of three-dimensional reconstruction and dynamic obstacle avoidance of the spinal surgical area. The CT image data of the spinal surgical area acquired before the operation and the local image data acquired in real time during the operation are fused and input into the model to generate a three-dimensional dynamic model of the spinal surgical area. At the same time, based on the prediction results of the robotic arm motion trajectory, potential collision areas are marked in the three-dimensional dynamic model and obstacle avoidance adjustment signals are output.

[0030] Specifically, step S4 constructs a 3D dynamic model of the surgical area and marks potential collision zones to provide real-time obstacle avoidance guidance for the robotic arm's movement. Simultaneously, it outputs obstacle avoidance adjustment signals to ensure the robotic arm does not collide with surgical tissue during surgery, overcoming the shortcomings of existing technologies such as single 3D reconstruction data and delayed obstacle avoidance response. This step utilizes a specialized hybrid density network model, fusing preoperative and intraoperative image data to generate a 3D dynamic model reflecting the real-time state of the surgical area. Compared to single-image modeling, this significantly improves model accuracy and dynamic adaptability. Furthermore, combined with robotic arm trajectory prediction, collision risk areas are marked in advance, and the output obstacle avoidance adjustment signals directly guide subsequent control command generation, enhancing the safety and adaptability of the robotic arm's movement. In the specific implementation process, preoperative CT image data of the spinal surgical area is prepared first, with a resolution of 512×512 pixels and a slice thickness of 0.5 mm, covering the entire spinal surgical area. Intraoperative real-time local image data is acquired using ultrasound images at a frame rate of 30 frames / second and a resolution of 640×480 pixels. Two types of image data are input into a hybrid density network model for 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area. The model first segments the preoperative CT images, extracts the contours of tissues such as vertebral bodies and intervertebral discs, and then registers them with intraoperative ultrasound images, with the registration error controlled within 0.1 mm. The data are then fused to generate a 3D dynamic model, which is updated 10 times / second to ensure that it reflects real-time changes in the surgical area. Subsequently, the model receives predicted motion trajectory data from the robotic arm. The trajectory consists of a series of target coordinates of the end effector, with adjacent coordinates spaced 0.1 mm apart and the motion speed set at 5 mm / second. The model uses a collision detection algorithm to calculate the distance between each coordinate point on the trajectory and the tissue in the surgical area. When the distance is less than 2 mm, it is marked as a potential collision area. At the same time, an obstacle avoidance adjustment signal is generated, which includes a position offset (range 0 to 1 mm) and a motion direction correction angle (range 0 to 5 degrees). The adjustment signal is then transmitted to step S5. The entire modeling, collision detection, and signal generation process takes less than 0.005 seconds to ensure timely obstacle avoidance response.

[0031] Step S5: Input the disturbance compensation amount generated in step S3 and the obstacle avoidance adjustment signal output in step S4 into the force-feed position cooperative adaptive control algorithm. The algorithm dynamically adjusts the control parameters based on the position deviation, attitude deviation and interaction force deviation of the end effector of the robotic arm, and generates motion control commands for different joints of the robotic arm.

[0032] Specifically, step S5 integrates the disturbance compensation amount and obstacle avoidance adjustment signal. Through a specialized adaptive control algorithm, it dynamically adjusts control parameters to generate precise joint motion control commands for the robotic arm, achieving coordinated control of force and position. This overcomes the shortcomings of existing technologies, such as fixed control parameters and insufficient coordination between obstacle avoidance and control. This step takes the disturbance compensation amount from step S3 and the obstacle avoidance adjustment signal from step S4 as inputs. Combined with the real-time deviation data of the robotic arm's end effector, it uses a force-position coordinated adaptive control algorithm to optimize proportional, integral, and derivative control parameters in real time. This ensures that the generated control commands can both cancel disturbances and meet obstacle avoidance requirements, guaranteeing a dual improvement in the robotic arm's motion accuracy and safety. In the specific implementation process, the real-time position deviation, attitude deviation, and interaction force deviation of the robotic arm's end effector are first acquired. The position deviation ranges from 0 to 0.1 mm, the attitude deviation ranges from 0 to 0.05 degrees, and the interaction force deviation ranges from 0 to 1 N. These deviation data serve as the initial inputs to the algorithm. After the algorithm starts, it first adjusts the proportional coefficient based on the position deviation: the proportional coefficient is set to 10 when the deviation is greater than 0.05 mm and adjusted to 5 when the deviation is less than 0.05 mm. It then adjusts the integral coefficient based on the interaction force deviation: the integral coefficient is set to 2 when the deviation is greater than 0.5 N and adjusted to 1 when the deviation is less than 0.5 N. Finally, it adjusts the differential coefficient based on the attitude deviation: the differential coefficient is set to 3 when the deviation is greater than 0.02 degrees and adjusted to 1.5 when the deviation is less than 0.02 degrees. Subsequently, the algorithm integrates the disturbance compensation amount (torque 0 to 3 N·m, position 0 to 0.06 mm) from step S3 and the obstacle avoidance adjustment signal (offset 0 to 1 mm, angle 0 to 5 degrees) from step S4 into the control calculation, generating position control components and force control components respectively. The position control component corresponds to the joint angular displacement adjustment amount (0 to 2 degrees), and the force control component corresponds to the joint driving torque adjustment amount (0 to 1 N·m). Finally, the algorithm sets the weight ratio of position and force control to 7:3 according to the surgical operation requirements, merges the two types of control components, and generates motion control commands for the six joints of the robotic arm. The commands include the joint target angle (0 to 180 degrees) and the movement speed (0 to 10 degrees / second). The entire algorithm operation takes less than 0.004 seconds to ensure that the control commands are generated in real time.

[0033] Step S6: Transmit the motion control commands generated in step S5 to the drive system of the spinal surgery assistive robotic arm, drive different joints of the robotic arm to move according to the control commands, and at the same time receive the joint motion status data fed back by the drive system in real time, and feed it back to the AeroStead spatial stability computing platform for closed-loop control.

[0034] Specifically, step S6 is the closed-loop control implementation stage. It converts the generated control commands into the actual movements of the robotic arm and forms a closed-loop control through state feedback, ensuring the robotic arm moves with the expected precision. Simultaneously, it corrects any potential deviations in real time, providing the ultimate guarantee for the adaptive anti-shake mechanism of the entire control method. This step transmits the control commands generated in step S5 to the drive system, driving the robotic arm joints. Simultaneously, it receives joint motion state data from the drive system and transmits it back to the computing platform, forming a closed-loop link of "command output - motion execution - state feedback - command correction." Compared to open-loop control, this significantly improves the stability and precision of the robotic arm's movements, ensuring that surgical assistance operations continuously meet requirements. In the specific implementation process, the motion control commands generated in step S5 are first transmitted to the robotic arm drive system via the control bus. The drive system includes six servo motors, each corresponding to one joint. The motors have a rated power of 100 watts, a rated torque of 5 N·m, and a maximum speed of 300 rpm. After receiving the command, the drive system converts the target joint angle (0 to 180 degrees) and movement speed (0 to 10 degrees / second) into motor control signals, controlling the motor to rotate and drive the robotic arm joints. The motor position control accuracy is 0.001 degrees, and the speed control accuracy is 0.01 degrees / second, ensuring precise response of the joint movement to the command. Simultaneously, the drive system's built-in encoder collects real-time data on the actual joint movement status, including the actual angle (accuracy 0.001 degrees), actual speed (accuracy 0.01 degrees / second), and motor current (range 0 to 5 amps). This data is transmitted to the AeroStead spatial stability computing platform at a rate of 1000 frames per second via the feedback line. The computing platform compares the feedback data with the target data in the command. If the angle deviation exceeds 0.005 degrees or the speed deviation exceeds 0.05 degrees / second, the deviation correction mechanism is immediately activated to adjust subsequent control commands, achieving closed-loop control. The entire process of command transmission, drive movement, and status feedback is controlled within 0.002 seconds, ensuring the real-time performance of the closed-loop control and maintaining the stable movement of the robotic arm.

[0035] Preferably, the expressions for the rigid-flexible coupling dynamic perturbation observation and compensation model of the spinal robotic arm are as follows: ,in, The inertia matrix of the rigid-flexible coupled dynamic model. Let the joint angle vector of the robotic arm be... The joint angular acceleration vector. The matrix of Coriolis force and centrifugal force. The joint angular velocity vector. For flexible deformation stiffness matrix, Here is the damping matrix. This is the joint driving torque vector. For dynamic disturbance torque; ,in, For disturbance compensation torque, These are the observed values ​​for the inertia matrix, the Coriolis force and centrifugal force matrix, the flexible deformation stiffness matrix, and the damping matrix, respectively. Let the joint's expected angle vector be... Let be the desired angular velocity vector of the joint. Let be the desired angular acceleration vector of the joint. Let be the desired joint driving torque vector.

[0036] Specifically, a dynamic disturbance observation and compensation model for a rigid-flexible coupled spinal manipulator is developed. This model accurately quantifies the dynamic disturbances caused by the flexible deformation of joints and the elastic deformation of links, and generates appropriate compensation amounts. This addresses the problem of inaccurate disturbance compensation in existing technologies that only consider rigid structures, providing model support for improving the motion stability of the manipulator. In implementation, the basic parameters of the model are first set based on the structural parameters of the manipulator. The stiffness coefficient of the flexible joint hinge is set to 5000 N / m, and the elastic modulus of the carbon fiber composite material of the link is set to 200 GPa, ensuring that the initial parameters of the model match the actual structure of the manipulator. During model operation, the disturbance components separated in step S2 are received. Combined with real-time data such as joint angular displacement (0 to 180 degrees), angular velocity (maximum 10 degrees / second), and angular acceleration (maximum 5 degrees / second²), the disturbance torque caused by the flexible deformation of the joint is observed through dynamic calculations. When the deformation reaches a maximum of 0.05 mm, the corresponding disturbance... The torque is approximately 2 N·m, and the end position deviation (0.04 mm) caused by the elastic deformation of the connecting rod (maximum 0.03 mm) is observed. Then, based on the observation results, combined with the maximum output torque of the drive motor (5 N·m) and the maximum adjustment speed (15 degrees / second), a disturbance compensation amount with a torque adjustment range of 0 to 3 N·m and a position compensation range of 0 to 0.06 mm is generated. This ensures that the compensation amount is within the load range of the drive system, and the entire observation and compensation amount generation process is controlled within 0.003 seconds, meeting the real-time control requirements and providing a precise disturbance cancellation basis for the subsequent control command generation.

[0037] Preferably, the expressions for the hybrid density network model of the three-dimensional reconstruction and dynamic obstacle avoidance of the spinal surgical area are as follows: ,in, This is the probability distribution of 3D vertex coordinates output by the hybrid density network. The coordinate vector of the vertex of the three-dimensional model of the spinal surgical area. The input is the fused image feature vector. For model parameters, For the first The weights of the Gaussian components, The function is a Gaussian distribution. For the first The mean vector of Gaussian components, For the first The covariance matrix of Gaussian components; ,in, For dynamic obstacle avoidance cost function, The set of coordinates of obstacles in the surgical area. For predicting the coordinate set of the robotic arm's motion trajectory, For the cost function parameters, These are the weighting coefficients. The function for calculating Euclidean distance is... For the first The coordinates of the obstacle. For the first on the trajectory One predicted coordinate, For indicator functions, This is a collection of dangerous areas within the surgical region.

[0038] Specifically, a hybrid density network model combining 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area is used. This model integrates preoperative and intraoperative imaging data to construct a 3D dynamic model that reflects the state of the surgical area in real time. It also accurately marks potential collision areas and outputs obstacle avoidance signals, addressing the problems of single modeling data and delayed obstacle avoidance response in existing technologies. This provides environmental support for the surgical area to ensure safe movement of the robotic arm. In practice, preoperative CT images of the spinal surgical area (512×512 pixels, 0.5 mm slice thickness) and intraoperative ultrasound images (30 frames / second, 640×480 pixels) are prepared to ensure the image data covers the surgical area and meets accuracy standards. After the model is started, the preoperative CT images are segmented to extract the contours of vertebral bodies, intervertebral discs, and other tissues. These are then registered with the intraoperative ultrasound images, with the registration error controlled within 0.1 mm. The models are then fused to generate a 3D dynamic model, with an update frequency of 10 times / second to ensure real-time reflection of changes in the position of tissues in the surgical area. Next, the model receives data from the robotic arm. The trajectory prediction data (adjacent coordinate interval 0.1 mm, movement speed 5 mm / s) is used to calculate the distance between each coordinate point on the trajectory and the surgical area tissue through a collision detection algorithm. When the distance is less than 2 mm, it is marked as a potential collision area. At the same time, obstacle avoidance adjustment signals including position offset (0 to 1 mm) and movement direction correction angle (0 to 5 degrees) are generated. The entire modeling, collision detection and signal generation process takes less than 0.005 seconds to ensure that the obstacle avoidance signal can be transmitted to step S5 in a timely manner, providing a safety basis for the surgical area for control command adjustment and avoiding collision between the robotic arm and the surgical area tissue.

[0039] Preferably, the expressions for the force-position cooperative adaptive control algorithm are as follows: ,in, For position control output, , These are the proportional coefficient, integral coefficient, and derivative coefficient, which are dynamically adjusted according to the position deviation. This is the position deviation vector. Let the vector be the rate of change of position deviation. For time; ,in, To control the output amount by force sensing, , These are the proportional coefficient and differential coefficient, respectively, which are dynamically adjusted according to force deviation. The force deviation vector, Let the force deviation change rate vector be... This is an estimate of the Jacobian matrix for the robotic arm. The end-effector interaction force vector, , For the final control command, These are the position and force control weighting coefficients.

[0040] Specifically, the force-position coordinated adaptive control algorithm integrates disturbance compensation and obstacle avoidance adjustment signals, dynamically optimizes control parameters, and generates joint motion control commands that balance accuracy and safety. This addresses the problems of fixed control parameters and poor coordination between force and position control in existing technologies, providing key algorithmic support for achieving adaptive anti-shake in robotic arms. In practical implementation, the algorithm first acquires real-time deviation data from the robotic arm's end effector: position deviation 0 to 0.1 mm, attitude deviation 0 to 0.05 degrees, and interaction force deviation 0 to 1 N. This data serves as the initial basis for parameter adjustment. Subsequently, the control parameters are dynamically adjusted based on the deviation magnitude: the proportional coefficient is set to 10 when the position deviation is greater than 0.05 mm and to 5 when it is less than 0.05 mm; the integral coefficient is set to 2 when the interaction force deviation is greater than 0.5 N and to 1 when it is less than 0.5 N; and the derivative coefficient is set to 3 when the attitude deviation is greater than 0.02 degrees and to 1.5 when it is less than 0.02 degrees, ensuring that the parameters are optimized in real-time according to the working conditions. Then, the disturbance compensation amount (torque 0 to 3 N·m, position 0 to 0.06 mm) from step S3 is incorporated into the algorithm. The obstacle avoidance adjustment signal of S4 (offset 0 to 1 mm, angle 0 to 5 degrees) generates position control components (corresponding to joint angular displacement adjustment 0 to 2 degrees) and force control components (corresponding to joint drive torque adjustment 0 to 1 N·m). Then, according to the surgical operation requirements, the weight ratio of position and force control is set to 7:3. The two types of components are fused to generate motion control commands for the six joints of the robotic arm. The commands include the joint target angle (0 to 180 degrees) and the movement speed (0 to 10 degrees / second). The entire algorithm calculation process takes less than 0.004 seconds to ensure that the control commands can be transmitted to the drive system in real time. This not only cancels disturbances to ensure motion accuracy, but also meets obstacle avoidance requirements to ensure surgical safety, and achieves coordinated control of force and position.

[0041] Preferably, the disturbance separation algorithm expression of the AeroStead spatial level stability calculation platform is: ,in, The raw data collected, For useful signal components, For the disturbance component, For noise components, These are estimates of the disturbance components. These are the filter weight coefficients. The sampling period is This is the length of the filtering window. This is a disturbance trend compensation term.

[0042] Specifically, the AeroStead spatial stability computing platform's disturbance separation algorithm precisely filters the raw data collected in step S1, separating useful signals, disturbance components, and noise. This provides clean data for subsequent model calculations and control command generation, solving the problem of decreased control accuracy caused by mixed interference in the raw data. It is a key technology in data preprocessing. In specific implementation, the computing platform first receives the raw position, attitude, and interaction force data transmitted in step S1 (transmission rate 1000 frames / second) to ensure that the data enters the processing flow in real time. Then, it starts the built-in adaptive filtering algorithm, setting the filtering window length to 50 sampling points to ensure both filtering effect and real-time performance. For position data, by analyzing the rate of change of adjacent sampling points, it identifies abrupt changes exceeding the normal motion rate range, and determines the components with vibration frequencies of 5 to 20 Hz as vibration disturbances of the robotic arm structure. For attitude data, it compares with a preset stable attitude threshold (angle fluctuation amplitude 0.01 degrees), and removes data exceeding the threshold. The threshold is partially determined to be a disturbance caused by joint transmission clearance; for interactive force data, instantaneous pulse force is eliminated through smoothing, and signals with force fluctuations exceeding 0.1 N and durations less than 0.005 seconds are identified as external environmental interference disturbances; at the same time, random noise generated during data acquisition is removed, and the three types of disturbance components are stored separately, while the useful signal components are transmitted to subsequent steps; the entire filtering and separation process takes less than 0.002 seconds, ensuring that clean data can be input into the subsequent model in a timely manner, avoiding interference signals from affecting the accuracy of model calculations, and laying a data foundation for subsequent disturbance observation, 3D modeling, and control command generation.

[0043] Preferably, the parameters for the adaptive sensing and stabilization control of the spinal surgery assistive robotic arm include the joint stiffness coefficient of the robotic arm. Joint damping coefficient End effector force sensing threshold Position control accuracy Attitude control accuracy AeroStead spatial level stable sensing computing platform sampling frequency Number of iterations for hybrid density networks The parameters satisfy the following relationship: For the dynamic response time of the robotic arm, , This represents the minimum distance between obstacles in the surgical area.

[0044] Specifically, the adaptive sensing and anti-shake control parameters of the spinal surgery assistive robotic arm clearly define the value range and constraint relationship of each key parameter to ensure parameter matching and performance compliance of all components of the entire control system. This avoids control failure or insufficient accuracy due to parameter mismatch and is the parameter guarantee for stable system operation. In specific implementation, the baseline values ​​of each core parameter are first determined. The joint stiffness coefficient of the robotic arm is set to 5000 N / m based on the joint flexible hinge structure, and the joint damping coefficient is set to 10 N·s / m to ensure that the joint movement has both flexibility and stability. The force sensing threshold of the end effector is set to 0.01 N, matching the accuracy of the force sensor (0.01 N), ensuring accurate sensing of minute interactive forces. The position control accuracy is set to 0.001 mm, and the attitude control accuracy is set to 0.001 degrees to meet the stringent accuracy requirements of spinal surgery for the robotic arm. The sampling frequency of the AeroStead spatial stabilization computing platform is set to 1000 Hz, matching the sensor sampling interval (0.001 Hz). To ensure consistency between data acquisition and processing rates, the hybrid density network iteration count is set to 100 times, ensuring modeling accuracy while keeping modeling time within 0.005 seconds. Simultaneously, parameter constraints are clearly defined: the sampling frequency of the computing platform is no less than twice the reciprocal of the robotic arm's dynamic response time (dynamic response time 0.001 seconds, sampling frequency no less than 1000 Hz), and the position control accuracy does not exceed 1 / 5 of the minimum distance between obstacles in the surgical area (minimum obstacle distance 0.005 mm, accuracy no more than 0.001 mm). This ensures that all parameters are mutually compatible, jointly supporting the system to achieve adaptive perception and anti-shake control, meeting surgical assistance needs.

[0045] Preferred, such as Figure 2 As shown, step S3 includes the following sub-steps: S31, extracting motion parameters of different joints of the robotic arm, including joint angular displacement, angular velocity, and angular acceleration, and simultaneously collecting strain data of the robotic arm links, using these data as input data for the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm; S32, based on the input joint motion parameters and link strain data, using the finite element analysis method to calculate the degree of elastic deformation of the robotic arm links, and determining the influence coefficient of flexible deformation on dynamic characteristics; S33, substituting the influence coefficient into the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm, and observing the magnitude and direction of the dynamic disturbance generated during the movement of the robotic arm in real time through iterative calculation of the model; S34, based on the observed magnitude and direction of the disturbance, combined with the robotic arm joint driving capability parameters, calculating and generating a compensation amount that can offset the disturbance, ensuring that the compensation amount is within the output range of the robotic arm driving system.

[0046] Specifically, step S3 standardizes the construction and application process of the rigid-flexible coupling dynamics disturbance observation and compensation model for the spinal manipulator through step-by-step procedures. This ensures that the model can accurately observe disturbances and generate reasonable compensation amounts, solving the problem of low disturbance observation accuracy caused by unclear model application procedures in existing technologies, and further improving the disturbance compensation effect. In specific implementation, S31 first extracts the motion parameters of each joint of the manipulator and the strain data of the connecting rods. The joint angular displacement range is 0 to 180 degrees, the maximum angular velocity is 10 degrees / second, and the maximum angular acceleration is 5 degrees / second². The connecting rod strain data is collected by strain gauges attached to the surface of the connecting rods at a sampling frequency of 1000 Hz. This data is used as model input. S32 uses the finite element analysis method based on the input data, setting the element type to solid elements and the mesh size to 0.1 mm, to calculate the degree of elastic deformation of the connecting rods. When the joint torque reaches 2 N·m, the maximum elastic deformation of the connecting rod is 0.03 mm. Based on this, the effect of flexible deformation on… The influence coefficient of the dynamic characteristics is 0.8; S33 substitutes the influence coefficient into the model and observes the disturbance in real time through iterative calculation (iteration step size 0.001 seconds). When the robotic arm moves to a joint angle of 120 degrees, a dynamic disturbance torque of 1.5 N·m is observed, with the direction opposite to the joint rotation; S34 combines the maximum output torque of the drive motor of 5 N·m to calculate and generate a disturbance compensation amount of 1.5 N·m, ensuring that the compensation amount is within the motor's load range. The entire step-by-step implementation process takes less than 0.003 seconds, ensuring the real-time nature of disturbance observation and compensation, and providing a precise basis for subsequent control command adjustments.

[0047] Preferred, such as Figure 3 As shown, step S4 includes the following sub-steps: S41, segmenting the preoperative CT image data of the spinal surgical area, extracting the contour features of the vertebral bodies, intervertebral discs and surrounding important tissues, and simultaneously enhancing the local image data acquired in real time during the operation to highlight the local details of the surgical area; S42, registering and fusing the segmented preoperative CT image features with the enhanced intraoperative real-time image features to establish a unified coordinate system, aligning the two types of image data under the same coordinate system to form fused image features; S43, inputting the fused image features into the hybrid density network model of the three-dimensional reconstruction and dynamic obstacle avoidance of the spinal surgical area, the model generates three-dimensional point cloud data of the spinal surgical area through multi-layer neural network calculation, and then constructs a three-dimensional dynamic model of the spinal surgical area; S44, simulating the movement process of the robotic arm in the three-dimensional dynamic model of the spinal surgical area according to the motion trajectory planning results of the robotic arm end effector, marking the areas on the motion trajectory that may collide with the tissues in the surgical area, and generating corresponding obstacle avoidance adjustment signals.

[0048] Specifically, step S4 clarifies the application process of the hybrid density network model for 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area step by step, ensuring that the model can fuse multi-source images to generate an accurate 3D dynamic model and output effective obstacle avoidance signals. This solves the problem of untimely obstacle avoidance caused by the disconnect between modeling and obstacle avoidance processes in existing technologies, and improves the safety of robotic arm movement. In specific implementation, S41 segments the preoperative CT images (512×512 pixels resolution, 0.5 mm slice thickness) and uses a threshold segmentation method (threshold range 100-300 HU) to extract the contours of the vertebral body and intervertebral disc. Histogram equalization is used to enhance details in the intraoperative ultrasound images (30 frames / second frame rate, 640×480 pixels resolution). S42 uses a registration algorithm based on feature point matching, selecting 10 feature points on the vertebral body edge to register the preoperative and intraoperative images, controlling the registration error within 0.1 mm to form fused image features. S43 inputs the fused features into the model, and the model uses... A network structure with 3 hidden layers (128 neurons per layer) generates 3D point cloud data (point cloud density 100 points / square millimeter) through 100 iterations to construct a 3D dynamic model. S44 simulates the motion of the robotic arm (speed 5 mm / s, adjacent coordinate interval 0.1 mm) using a distance detection algorithm (detection step size 0.01 mm). When a point on the trajectory is less than 2 mm away from the surgical area tissue, it is marked as a collision area, generating an obstacle avoidance signal with a position offset of 0.5 mm and a direction correction of 3 degrees. The entire sub-step takes less than 0.005 seconds, ensuring timely output of the obstacle avoidance signal.

[0049] Preferred, such as Figure 4 As shown, step S5 includes the following sub-steps: S51, acquiring the desired position, desired posture, and desired interaction force of the robotic arm end effector, comparing it with the real-time position, posture, and interaction force data collected in step S1, and calculating the position deviation, posture deviation, and interaction force deviation; S52, inputting the calculated deviations into the parameter adjustment module of the force-position collaborative adaptive control algorithm, which dynamically adjusts the proportional, integral, and derivative coefficients of the algorithm according to the magnitude and trend of the deviations; S53, using the adjusted control parameters, generating control components for robotic arm position adjustment and control components for interaction force control through the operation of the force-position collaborative adaptive control algorithm; S54, setting the weight ratio of position control and force control according to the surgical operation requirements of the surgical area, fusing the position control component and the force control component according to the ratio, and generating motion control commands for different joints of the robotic arm.

[0050] Specifically, step S5 standardizes the application process of the force-position collaborative adaptive control algorithm step by step, ensuring that the algorithm can integrate multi-source information to dynamically adjust parameters and generate precise control commands. This solves the problem of poor control command adaptability caused by the ambiguity of the algorithm application process in existing technologies, and improves the motion accuracy and coordination of the robotic arm. In specific implementation, S51 acquires the desired position (error allowable range ±0.05 mm), desired posture (error allowable range ±0.02 degrees), and desired interaction force (0-3 N) of the robotic arm end effector, and compares it with the real-time data collected in step S1 to calculate the position deviation of 0.08 mm, posture deviation of 0.03 degrees, and interaction force deviation of 0.6 N. S52 inputs the deviation amount into the algorithm parameter adjustment module, and adjusts the parameters according to the deviation size. When the position deviation is >0.05 mm, the proportional coefficient is set to 10; when the interaction force deviation is >0.5 N, the integral coefficient is set to 2; and when the posture deviation is >0. At 0.2 degrees, the differential coefficient is set to 3; S53 uses the adjusted parameters to generate position control components (joint angular displacement adjusted by 1.5 degrees) and force control components (driving torque adjusted by 0.8 N·m); S54 sets the position and force control weights to 7:3 according to surgical needs, and merges the two types of components to generate control commands for 6 joints. The commands include target angle (e.g., 90 degrees for joint 1) and movement speed (8 degrees / second). The entire step-by-step calculation takes less than 0.004 seconds, ensuring that the control commands can adapt to disturbance compensation and obstacle avoidance requirements in real time, and ensuring the precise and stable movement of the robotic arm.

[0051] The rigid-flexible coupling dynamic disturbance observation and compensation model for spinal robotic arms is a dynamic analysis and control model specifically designed for the structural characteristics of robotic arms used in spinal surgery. It aims to accurately capture the dynamic disturbances caused by the flexible deformation of joints and the elastic deformation of links during the movement of the robotic arm and generate targeted compensation quantities, rather than the traditional dynamic model that only considers rigid structures. The implementation process requires first initializing the model based on the actual structural parameters of the robotic arm, setting the stiffness coefficient of the joint flexible hinge to 5000 N / m and the elastic modulus of the connecting rod carbon fiber composite material to 200 GPa. Then, it receives joint angular displacement (range 0 to 180 degrees), angular velocity (maximum 10 degrees / second), angular acceleration (maximum 5 degrees / second²), and connecting rod strain data (sampling frequency 1000 Hz) collected by multi-dimensional sensors. The influence coefficient of the connecting rod elastic deformation (maximum 0.03 mm) on the dynamic characteristics is calculated by finite element analysis. Then, it is substituted into the dynamic equation for iterative calculation (iteration step size 0.001 seconds). The disturbance torque (range 0 to 2 N·m) and position deviation (range 0 to 0.04 mm) are observed in real time. Finally, the disturbance compensation amount of 0 to 3 N·m is generated by combining the maximum output torque of the drive motor of 5 N·m. The model's function is to counteract disturbances caused by the flexibility of the robotic arm structure in real time, avoiding positional deviations of the end effector. This model breaks through the limitations of traditional rigid dynamic models, improving the disturbance compensation accuracy to the 0.001 mm level, and providing dynamic support for the high-precision operations required for spinal surgery.

[0052] The hybrid density network model for 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area is a model for surgical area environment modeling and obstacle avoidance analysis built on a hybrid density network architecture, integrating multi-source image data. It can generate a 3D dynamic model reflecting the real-time state of the surgical area and mark potential collision zones, unlike traditional modeling models that rely on single images. Its implementation requires acquiring preoperative CT images of the spinal surgical area (512×512 pixels, 0.5 mm slice thickness) and intraoperative real-time ultrasound images (30 frames / second, 640×480 pixels). Threshold segmentation (threshold 100-300 HU) is used to extract the contours of the vertebral body and intervertebral disc from the CT images. Histogram equalization is performed on the ultrasound images to enhance details. Then, feature point matching (selecting 10 vertebral body edge feature points) is used to register the two types of images (registration error ≤ 0.1 mm), forming a fused image. The model employs a feature input model with a 3-layer hidden layer (128 neurons per layer) network structure. After 100 iterations, it generates 3D point cloud data (density 100 points / mm²), constructs a 3D dynamic model (updated 10 times / second), and finally combines this with the robotic arm's motion trajectory (speed 5 mm / second, adjacent coordinate interval 0.1 mm). A distance detection algorithm (step size 0.01 mm) marks collision areas ≤2 mm away, outputting obstacle avoidance signals with position offsets of 0-1 mm and angle corrections of 0-5 degrees. This model provides the robotic arm with real-time surgical area environmental information and obstacle avoidance guidance, reducing surgical area modeling errors to within 0.1 mm and obstacle avoidance response time to 0.005 seconds, significantly improving surgical operation safety.

[0053] The force-position cooperative adaptive control algorithm is a cooperative control algorithm that balances the positional accuracy and force perception of the robotic arm, and can dynamically adjust control parameters. It integrates disturbance compensation and obstacle avoidance signals to generate adaptive control commands, unlike traditional control algorithms with fixed parameters. Its implementation requires first acquiring the real-time position deviation (0 to 0.1 mm), attitude deviation (0 to 0.05 degrees), and interaction force deviation (0 to 1 N) of the robotic arm's end effector. The control parameters are then dynamically adjusted according to the deviation magnitude: the proportional coefficient is set to 10 when the position deviation is >0.05 mm and 5 when it is ≤0.05 mm; the integral coefficient is set to 2 when the interaction force deviation is >0.5 N and 1 when it is ≤0.5 N; the derivative coefficient is set to 3 when the attitude deviation is >0.02 degrees and 1.5 when it is ≤0.02 degrees. Finally, the algorithm incorporates observation and compensation for the rigid-flexible coupling dynamics of the spinal robotic arm. The model generates torque compensation of 0 to 3 N·m and position compensation of 0 to 0.06 mm, as well as position offset of 0 to 1 mm and angle correction of 0 to 5 degrees from the hybrid density network model of 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area. These are used to generate position control components (joint angular displacement adjustment of 0 to 2 degrees) and force control components (driving torque adjustment of 0 to 1 N·m), respectively. Finally, the two types of components are fused with a weight ratio of 7:3 to generate motion control commands for 6 joints (target angle of 0 to 180 degrees and speed of 0 to 10 degrees / second), with a computation time of ≤0.004 seconds. The algorithm aims to achieve synergistic optimization of the robotic arm's position accuracy and force control, improving position control accuracy to 0.001 mm and force control accuracy to 0.01 N, meeting the stringent requirements of multi-dimensional control of the robotic arm in spinal surgery.

[0054] The AeroStead air-level stabilization computing platform possesses a core computing unit with high-speed data processing, signal filtering, and disturbance separation capabilities. Unlike ordinary data processing platforms, it provides data preprocessing and computational support for the entire adaptive sensing and anti-shake control system. Its implementation requires first receiving raw position, attitude, and interaction force data collected by multi-dimensional sensors via a high-speed data bus (transmission rate 1000 frames / second). It then activates a built-in adaptive filtering algorithm (filter window length 50 sampling points) to separate structural vibration disturbances of 5 to 20 Hz from the position data by analyzing motion rates. For the attitude data, it separates joint transmission clearance disturbances by comparing with a stability threshold of 0.01 degrees. For the interaction force data, it separates external interference with fluctuations >0.1 N and duration <0.005 seconds through smoothing processing, while simultaneously removing random noise and transmitting the useful signal. The data is then fed into subsequent models. The platform also needs to schedule the computation of the spinal robotic arm rigid-flexible coupling dynamic disturbance observation and compensation model, the spinal surgical area 3D reconstruction and dynamic obstacle avoidance hybrid density network model, and the force-position collaborative adaptive control algorithm. It controls the computation timing and data interaction of each model algorithm to ensure the entire control process takes ≤0.01 seconds. Furthermore, the platform receives joint motion state data (angle accuracy 0.001 degrees, velocity accuracy 0.01 degrees / second) from the drive system, compares it with the target data, and initiates a correction mechanism if the deviation is >0.005 degrees or >0.05 degrees / second. The platform's role is to achieve real-time data processing, collaborative model scheduling, and closed-loop control, controlling data processing latency to within 0.002 seconds, ensuring the real-time performance and stability of the entire system, and providing hardware and algorithmic support for the collaborative operation of various core technologies.

[0055] like Figure 5The aforementioned adaptive sensing and anti-shake control system for a spinal surgery-assisted robotic arm includes: a robotic arm motion and force data acquisition unit equipped with a multi-dimensional sensor array, positioned relative to the robotic arm's end effector and the spinal surgical area, for acquiring robotic arm motion data and interaction force data, and transmitting the data to an AeroStead spatial stabilization calculation and processing unit; an AeroStead spatial stabilization calculation and processing unit, which is an AeroStead spatial stabilization calculation platform connected to the robotic arm motion and force data acquisition unit via a data bus, receiving data transmitted from the robotic arm motion and force data acquisition unit, and bidirectionally connected to a dual-model storage and retrieval unit and an adaptive control algorithm storage unit, respectively, to call the stored models and algorithms for data processing; and a dual-model storage and retrieval service unit connected to the AeroStead spatial stabilization calculation and processing unit, internally storing observations of rigid-flexible coupling dynamic disturbances of the spinal robotic arm. The system includes a compensation model, a 3D reconstruction model of the spinal surgical area, and a dynamic obstacle avoidance hybrid density network model, which provide model call services to the AeroStead spatial stability computing unit. A force-position control algorithm storage unit, connected to the AeroStead spatial stability computing unit, stores the force-position collaborative adaptive control algorithm and provides algorithm call services to the AeroStead spatial stability computing unit. A control command conversion and drive unit, connected to the AeroStead spatial stability computing unit via a control bus, receives motion control commands output by the AeroStead spatial stability computing unit, connects to the robotic arm drive system, and converts the control commands into drive signals to drive the robotic arm movement. A joint motion state feedback transmission unit, connected to both the robotic arm drive system and the AeroStead spatial stability computing unit, collects joint motion state data of the robotic arm and feeds the data back to the AeroStead spatial stability computing unit, forming a closed-loop control link.

[0056] An adaptive sensing and anti-shake control method and system for spinal surgery assistive robotic arms comprehensively collects position, attitude, and interaction force data of the robotic arm's end effector through a multi-dimensional sensor array. The AeroStead spatial stability calculation platform accurately separates the disturbance components of the robotic arm's own structural vibration and external environmental interference. Then, relying on a rigid-flexible coupling dynamic disturbance observation and compensation model for the spinal robotic arm, it fully considers the impact of joint flexibility deformation and link elastic deformation on dynamics, generating targeted disturbance compensation in real time. This end-to-end disturbance control mechanism effectively counteracts disturbances caused by structural flexibility during robotic arm movement, preventing positional deviations of the end effector. Compared to existing technologies that only focus on rigid structures, it significantly improves the accuracy and real-time performance of dynamic disturbance compensation, providing a stable motion benchmark for surgical operations.

[0057] This method and system demonstrate significant advantages in surgical area adaptation and control coordination, addressing the shortcomings of existing technologies in terms of coordination and significantly improving surgical safety and operational adaptability. Utilizing a 3D reconstruction of the spinal surgical area and a dynamic obstacle avoidance hybrid density network model, it integrates preoperative CT images and intraoperative real-time image data to generate a 3D model reflecting dynamic changes in the surgical area. This model accurately marks potential collision zones and outputs obstacle avoidance adjustment signals. Simultaneously, the obstacle avoidance adjustment signals and disturbance compensation are input into a force-position collaborative adaptive control algorithm. The algorithm dynamically adjusts control parameters based on position, posture, and force deviation, achieving precise adaptation between obstacle avoidance requirements and control commands. This collaborative "modeling-obstacle avoidance-control" mechanism can rapidly respond to changes in the surgical area, avoiding delayed obstacle avoidance response or mismatches between commands and requirements. It significantly improves the dynamic obstacle avoidance capability and control coordination of the surgical area, reduces surgical risks, and meets the stringent clinical requirements for surgical safety.

[0058] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive sensing and anti-shake control method for a robotic arm assisting spinal surgery, characterized in that, include: Step S1: The multi-dimensional sensor group mounted on the spinal surgery assistive robotic arm collects real-time position data, posture data, and interaction force data generated by the contact with the spinal surgical area tissue during the movement of the robotic arm's end effector, and transmits the collected data to the AeroStead spatial stability computing platform; Step S2: The AeroStead spatial stability computing platform performs real-time filtering on the received real-time position data, posture data, and interaction force data to separate the disturbance components caused by the vibration of the robotic arm's own structure and the disturbance components caused by external environmental interference; Step S3: Based on the separated disturbance components, a rigid-flexible coupling dynamic disturbance observation and compensation model for the spinal robotic arm is constructed. This model is used to observe the dynamic disturbances caused by the flexible deformation of the robotic arm joints and the elastic deformation of the links in real time, and generate corresponding disturbance compensation amounts; Step S4: The three-dimensional reconstruction of the spinal surgical area and the dynamic obstacle avoidance hybrid density are called. The network model utilizes preoperative CT images of the spinal surgical area and real-time intraoperative local images to generate a three-dimensional dynamic model of the spinal surgical area. Simultaneously, based on the predicted trajectory of the robotic arm, potential collision areas are marked in the three-dimensional dynamic model, and obstacle avoidance adjustment signals are output. Step S5: The disturbance compensation amount generated in step S3 and the obstacle avoidance adjustment signal output in step S4 are input into the force-position collaborative adaptive control algorithm. This algorithm dynamically adjusts control parameters based on the position deviation, posture deviation, and interaction force deviation of the robotic arm's end effector, generating motion control commands for different joints of the robotic arm. Step S6: The motion control commands generated in step S5 are transmitted to the drive system of the spinal surgery assistive robotic arm, driving different joints of the robotic arm to move according to the control commands. Simultaneously, the joint motion state data fed back from the drive system is received in real time and fed back to the AeroStead spatial stability computing platform for closed-loop control.

2. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, The expressions for the rigid-flexible coupling dynamic perturbation observation and compensation model of the spinal manipulator are as follows: ,in, The inertia matrix of the rigid-flexible coupled dynamic model. Let the joint angle vector of the robotic arm be... The joint angular acceleration vector. The matrix of Coriolis force and centrifugal force. The joint angular velocity vector. For flexible deformation stiffness matrix, Here is the damping matrix. This is the joint driving torque vector. For dynamic disturbance torque; ,in, For disturbance compensation torque, These are the observed values ​​for the inertia matrix, the Coriolis force and centrifugal force matrix, the flexible deformation stiffness matrix, and the damping matrix, respectively. Let the joint's expected angle vector be... Let be the desired angular velocity vector of the joint. Let be the desired angular acceleration vector of the joint. Let be the desired joint driving torque vector.

3. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, The expressions for the hybrid density network model of the three-dimensional reconstruction and dynamic obstacle avoidance of the spinal surgical area are as follows: ,in, This is the probability distribution of 3D vertex coordinates output by the hybrid density network. The coordinate vector of the vertex of the three-dimensional model of the spinal surgical area. The input is the fused image feature vector. For model parameters, For the first The weights of the Gaussian components, The function is a Gaussian distribution. For the first The mean vector of Gaussian components, For the first The covariance matrix of Gaussian components; ,in, For dynamic obstacle avoidance cost function, The set of coordinates of obstacles in the surgical area. For predicting the coordinate set of the robotic arm's motion trajectory, For the cost function parameters, These are the weighting coefficients. The function for calculating Euclidean distance is... For the first The coordinates of the obstacle. For the first on the trajectory One predicted coordinate, For indicator functions, This is a collection of dangerous areas within the surgical region.

4. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, The expressions for the force-position cooperative adaptive control algorithm are as follows: ,in, For position control output, , These are the proportional coefficient, integral coefficient, and derivative coefficient, which are dynamically adjusted according to the position deviation. This is the position deviation vector. Let the vector be the rate of change of position deviation. For time; ,in, To control the output amount by force sensing, , These are the proportional coefficient and differential coefficient, respectively, which are dynamically adjusted according to force deviation. The force deviation vector, Let the force deviation change rate vector be... This is an estimate of the Jacobian matrix for the robotic arm. The end-effector interaction force vector, , For the final control command, These are the position and force control weighting coefficients.

5. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, The disturbance separation algorithm expression of the AeroStead spatial level stability computing platform is as follows: ,in, The raw data collected, For useful signal components, For the disturbance component, For noise components, These are estimates of the disturbance components. These are the filter weight coefficients. The sampling period is This is the length of the filtering window. This is a disturbance trend compensation term.

6. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, The parameters for the adaptive sensing and stabilization control of the spinal surgery assistive robotic arm include the joint stiffness coefficient of the robotic arm. Joint damping coefficient End effector force sensing threshold Position control accuracy Attitude control accuracy AeroStead spatial level stable sensing computing platform sampling frequency Number of iterations for hybrid density networks The parameters satisfy the following relationship: For the dynamic response time of the robotic arm, , This represents the minimum distance between obstacles in the surgical area.

7. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, Step S3 includes the following sub-steps: S31, extracting motion parameters of different joints of the robotic arm, including joint angular displacement, angular velocity, and angular acceleration, and simultaneously collecting strain data of the robotic arm links, using these data as input data for the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm; S32, based on the input joint motion parameters and link strain data, using the finite element analysis method to calculate the degree of elastic deformation of the robotic arm links, and determining the influence coefficient of flexible deformation on dynamic characteristics; S33, substituting the influence coefficient into the rigid-flexible coupling dynamic disturbance observation and compensation model of the spinal robotic arm, and observing the magnitude and direction of the dynamic disturbance generated during the movement of the robotic arm in real time through iterative calculation of the model; S34, based on the observed magnitude and direction of the disturbance, combined with the robotic arm joint driving capability parameters, calculating and generating a compensation amount that can offset the disturbance, ensuring that the compensation amount is within the output range of the robotic arm drive system.

8. The adaptive sensing and anti-shake control method for spinal surgery assistive robotic arms according to claim 1, characterized in that, Step S4 includes the following sub-steps: S41, segmenting the preoperative CT image data of the spinal surgical area, extracting the contour features of the vertebral bodies, intervertebral discs and surrounding important tissues, and simultaneously enhancing the local image data acquired in real time during the operation to highlight the local details of the surgical area; S42, registering and fusing the segmented preoperative CT image features with the enhanced intraoperative real-time image features to establish a unified coordinate system, aligning the two types of image data under the same coordinate system to form fused image features; S43, inputting the fused image features into a hybrid density network model for 3D reconstruction and dynamic obstacle avoidance of the spinal surgical area, and generating 3D point cloud data of the spinal surgical area through multi-layer neural network calculations, thereby constructing a 3D dynamic model of the spinal surgical area; S44, simulating the movement process of the robotic arm in the 3D dynamic model of the spinal surgical area based on the motion trajectory planning results of the robotic arm end effector, marking the areas on the motion trajectory that may collide with the tissues in the surgical area, and generating corresponding obstacle avoidance adjustment signals.

9. The adaptive sensing and anti-shake control method for a spinal surgery assistive robotic arm according to claim 1, characterized in that, Step S5 includes the following sub-steps: S51, acquiring the desired position, desired posture, and desired interaction force of the robotic arm end effector, comparing it with the real-time position, posture, and interaction force data collected in step S1, and calculating the position deviation, posture deviation, and interaction force deviation; S52, inputting the calculated deviations into the parameter adjustment module of the force-position collaborative adaptive control algorithm, which dynamically adjusts the proportional, integral, and derivative coefficients of the algorithm according to the magnitude and trend of the deviations; S53, using the adjusted control parameters, generating control components for robotic arm position adjustment and control components for interaction force control through the operation of the force-position collaborative adaptive control algorithm; S54, setting the weight ratio of position control and force control according to the surgical operation requirements of the surgical area, fusing the position control component and the force control component according to the ratio, and generating motion control commands for different joints of the robotic arm.

10. An adaptive sensing and anti-shake control system for a spinal surgery assistive robotic arm, characterized in that, This system is applied to the adaptive sensing and anti-shake control method for a spinal surgery assistive robotic arm as described in claim 1, comprising: a robotic arm motion and force data acquisition unit, which is equipped with a multi-dimensional sensor group and is positioned relative to the robotic arm end effector and the spinal surgical area, for acquiring robotic arm motion data and interactive force data, and transmitting the data to the AeroStead spatial stabilization calculation and processing unit; an AeroStead spatial stabilization calculation and processing unit, which is the AeroStead spatial stabilization calculation platform, connected to the robotic arm motion and force data acquisition unit via a data bus, receiving the data transmitted by the robotic arm motion and force data acquisition unit, and bidirectionally connected to a dual-model storage and retrieval unit and an adaptive control algorithm storage unit, respectively, to call the stored models and algorithms for data processing; and a dual-model storage and retrieval service unit, which is connected to the AeroStead spatial stabilization calculation and processing unit, internally storing the spinal robotic arm rigid-flexible coupling dynamic disturbance observation and compensation model and the spinal surgical area. A 3D reconstruction and dynamic obstacle avoidance hybrid density network model provides model call services to the AeroStead spatial level stability computing and processing unit; a force-position control algorithm storage unit, connected to the AeroStead spatial level stability computing and processing unit, internally stores the force-position collaborative adaptive control algorithm and provides algorithm call services to the AeroStead spatial level stability computing and processing unit; a control command conversion and drive unit, connected to the AeroStead spatial level stability computing and processing unit via a control bus, receives motion control commands output by the AeroStead spatial level stability computing and processing unit, connects to the robotic arm drive system, and converts the control commands into drive signals to drive the robotic arm movement; a joint motion state feedback transmission unit, connected to both the robotic arm drive system and the AeroStead spatial level stability computing and processing unit, collects the joint motion state data of the robotic arm and feeds the data back to the AeroStead spatial level stability computing and processing unit, forming a closed-loop control link.

Images