A novel continuum robot and a shape prediction method thereof

Abstract

The application discloses a novel continuum robot and a shape prediction method thereof, wherein the continuum robot comprises a handle main body, a snake bone, a driving device and two driving cables; the snake bone is fixed to the front end of the handle main body, the driving device is arranged at the tail end inside the handle main body, the two driving cables are symmetrically arranged in the cable constraint grooves on the two side walls of the snake bone along the axial direction, and the driving device controls the deformation of the snake bone through the driving cables; the snake bone comprises a plurality of guide joints which are sequentially adjacent; each guide joint comprises a flexible hinge in the middle and one nested ring buckle structure on the left and right sides which rotates around the flexible hinge. The shape prediction method uses the real-time collected length change of the driving cable, obtains the predicted coordinates of a plurality of control points of the snake bone through a trained prediction model, and reconstructs the shape curve of the snake bone through a spline curve fitting algorithm. The continuum robot has good bending consistency, torsion resistance and local interference resistance, and has high shape prediction precision.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, specifically relating to a novel continuum robot and its shape prediction method. Background Technology

[0002] Surgical diagnosis and treatment of laryngeal diseases, such as biopsies and tumor resections, typically rely on rigid surgical instruments. However, the laryngeal anatomy is complex, the space is narrow, and the tissues are fragile. Traditional rigid instruments have significant limitations in terms of operational flexibility and adaptability, easily causing tissue damage. The pre-set curvature of existing rigid laryngoscopes is difficult to adapt to individual differences among patients.

[0003] With the development of transoral surgical robots, flexible continuum robots have become a research hotspot. Existing technologies, such as notched, serpentine, and concentric tube continuum robots, mostly suffer from limitations such as the trade-off between the central channel and overall size, as well as motion coupling, insufficient stiffness, or complex control. While some designs employing superelastic nickel-titanium alloys and contact-assisted structures can achieve high strength and self-support, their force-displacement hysteresis is significant, directly affecting control accuracy.

[0004] Furthermore, shape estimation and precise control of continuum robots present another major challenge. Traditional kinematic models, such as piecewise constant curvature models, are susceptible to wire friction and external loads, leading to significant errors in actual pose prediction. While data-driven methods can overcome some nonlinear issues, most existing methods have not been validated under large bending angles and are highly dependent on sensors, making it difficult to achieve high-precision shape prediction across the entire workspace in miniature, multi-jointed structures.

[0005] In terms of force sensing, existing drive force sensors mostly use commercial sensors, which are large in size and usually installed at the end of the system. Due to the accumulation of friction along the transmission path, the measurement error is large, making it difficult to meet the requirements of high-precision control. The development of fiber Bragg grating (FBG) sensing technology has provided a new solution for micro force sensing, but the integration of existing commercial FBG sensors with the robot body still has structural conflicts, which restricts their direct application in micro continuum robots.

[0006] In the prior art, patent application CN108078535A discloses a double-ringed rotating structure, which includes a mother rotating part and a daughter rotating part. The mother rotating part has a circular protrusion, an inner ringed annular groove, and an outer ringed annular protrusion, while the daughter rotating part has an inner ringed annular protrusion and an outer ringed annular groove. Relative rotation and axial positioning of adjacent joints are achieved through multiple nesting. Although this structure has good connection stability, the rigid protrusion structure cannot provide bending elasticity. The bending of the joint depends entirely on the gap between the ringed structures, resulting in poor bending consistency and cumulative motion errors. Summary of the Invention

[0007] This invention provides a novel continuum robot and its shape prediction method, aiming to solve one of the technical problems of existing micro continuum robots, such as complex structure, low control accuracy, lack of high-precision force sensing, and difficulty in achieving high-precision shape prediction of large bending angles in narrow cavities.

[0008] To achieve the above technical objectives, the present invention adopts the following technical solution: A novel continuum robot includes: a handle body, a snake skeleton, a drive device, and two drive cables; the snake skeleton is fixed to the front end of the handle body; the drive device is located at the tail end inside the handle body; the two drive cables are symmetrically arranged along the axial direction in the cable constraint grooves on both sides of the snake skeleton, with the first end fixedly connected to the tail end of the snake skeleton and the second end fixedly connected to the drive device, and the drive device controls the deformation of the snake skeleton through the drive cables. The snake bone adopts a tubular structure, with several sequentially adjacent guide joints on both the front and rear walls; each guide joint includes a flexible hinge in the middle and a nested ring structure on each of the left and right sides. Each nested ring structure includes a first fastening part at the lower end and a second fastening part at the upper end, which are connected as one unit by a flexible hinge. The first fastening part includes an inner ring-shaped protrusion and an outer ring-shaped groove, and the second fastening part includes an inner ring-shaped groove and an outer ring-shaped protrusion. The inner ring-shaped protrusion rotates within the inner ring-shaped groove, and the outer ring-shaped protrusion rotates within the outer ring-shaped groove.

[0009] Furthermore, the flexible hinge adopts an arc-shaped flexible hinge structure with a C-shaped cross section, which is composed of two concentric arcs, inner and outer.

[0010] Furthermore, the drive unit has a displacement feedback function to obtain changes in the length of the drive cable.

[0011] Furthermore, several guide wheels are installed inside the handle body to guide the direction of a section of the drive cable between the drive device and the snake bone.

[0012] Furthermore, force sensing units are respectively installed on the left and right side walls inside the handle body, including a U-shaped cantilever platform structure, pulleys and FBG sensors; The pulley is located at one end of the U-shaped cantilever platform structure near the axis, and the drive cable passes around the pulley with the resultant force perpendicular to the axis direction. The U-shaped cantilever platform structure has a groove in the direction perpendicular to the axis, and the FBG sensor is installed in the groove. When the drive cable is under stress, the pulley is driven to deform the U-shaped cantilever platform structure. The FBG sensor obtains the tension of the drive cable based on the deformation of the U-shaped cantilever platform structure.

[0013] Furthermore, the U-shaped cantilever platform structure includes The components include a fixed platform, an arc-shaped elastic body, a vertical connecting arm, and a pulley-fixed platform; The fixed platform is fixedly connected to the side wall of the gripping part; The two ends of the arc-shaped elastic body are connected by a vertical connecting arm. The fixed platform and the pulley fixed platform are integrated into one unit; A recess for mounting the FBG sensor is provided in The surface of the fixed platform and the pulley fixed platform.

[0014] Furthermore, the novel continuum robot also includes a sensor module and a shape prediction module; the sensor module is used to collect the length change and / or tension of the drive cable; the shape prediction module uses an artificial intelligence network and predicts the position of several control points of the snake skeleton based on the data collected by the sensor module, and then fits the shape curve of the snake skeleton.

[0015] Furthermore, the shape prediction module is also used to: adjust the predicted position of the control point by a global scaling factor based on the total length constraint of the snake bone and the base point distance constraint, and then fit the morphological curve of the snake bone based on the adjusted position.

[0016] A method for predicting the shape of the novel continuum robot described above includes: S1, Data Acquisition: The drive cable is controlled by the drive device to produce different length changes, and the bending images of the snake bone under different drive states are acquired. The coordinates of each control point on the center line of the snake bone are extracted as the true value through image processing. S2, Sample Construction: Construct a training sample set using the change in the length of the driving cable as the input feature vector and the coordinates of each control point as the target output vector; S3, Model Training: Using artificial intelligence algorithms, with the goal of minimizing the mean square error between the predicted and true coordinates of the control points, a nonlinear mapping model from the change in the length of the drive cable to the coordinates of each control point is established. S4, Preliminary prediction: Input the real-time collected changes in the length of the drive cable into the trained prediction model to obtain the predicted coordinates of each control point of the snake bone. S5, Shape Reconstruction: Based on the predicted coordinates of each control point, the morphological curve of the snake bone is reconstructed using a spline curve fitting algorithm.

[0017] Furthermore, after extracting the coordinates of each control point on the centerline of the snake bone through image processing, the obtained coordinates of each control point are corrected by geometric constraints to satisfy the preset conditions of constant distance from the base point to the first control point and constant total length of the snake bone, and the corrected coordinates are used as the true values; wherein, the base point is located at the connection point between the snake bone and the handle body, and the first control point is located at the center point of the first guide joint near the handle body. After the preliminary prediction of the predicted coordinates of each control point in step S4, the predicted coordinates of each control point are also corrected by geometric constraints to satisfy the preset conditions of constant distance from the base point to the first control point and constant total length of the snake bone. In step S5, the corrected coordinates are used to perform spline curve fitting to reconstruct the morphological curve of the snake bone.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. Excellent mechanical properties: The C-shaped cross-section arc hinge is based on the principle of compliant joint with notch, which simplifies the complex nonlinear model into a linear spring chain model, and has good bending consistency, torsional performance and resistance to local disturbances.

[0020] 2. Large bending range: The overall bending range reaches ±180°, which significantly improves the flexibility and operability of the instrument in confined spaces.

[0021] 3. High-precision shape prediction: The random forest and geometric constraint fusion method (RF-Geo) is adopted to achieve millimeter-level shape prediction of the micro multi-joint snake bone in the entire working space (the outer diameter of the snake bone is only 4 mm and the effective working length is 60 mm). The average absolute error of the end shape prediction is as low as 0.28 mm.

[0022] 4. Compact structure and small size: While setting cable restraint grooves on the side wall, it retains a central operating channel with a sufficiently large diameter, which can safely pass through narrow throats and accommodate surgical instruments, making it suitable for minimally invasive surgical scenarios.

[0023] 5. Integrated micro force sensing: The force sensing unit of the U-shaped cantilever platform based on FBG enables high-precision real-time monitoring of driving tension, providing a foundation for closed-loop control and force feedback.

[0024] 6. Passive stability of end-effector posture: The C-shaped hinge structure can absorb and compensate for interference through the compliant deformation of the middle rod, passively maintaining the end-effector orientation in complex interactions, thus improving operational safety and reliability. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall structure of the micro continuum robot in an embodiment of the present invention.

[0026] Figure 2This is a schematic diagram of the internal structure of the micro continuum robot in an embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram of the overall structure of the snake skeleton.

[0028] Figure 4 This is a schematic diagram of the cable constraint groove layout.

[0029] Figure 5 This is a partially enlarged schematic diagram of the snake bone assembly, showing the guide joint and flexible hinge structure.

[0030] Figure 6 This is a schematic diagram of the structural parameters of a C-shaped cross-section flexible hinge. Sub-figures (a) and (b) are a front view and a three-dimensional view of the flexible hinge, respectively.

[0031] Figure 7 This is a schematic diagram of the force sensing unit for a U-shaped cantilever platform.

[0032] Figure 8 This is a force-sensing diagram of the force sensing unit of the U-shaped cantilever platform.

[0033] Figure 9 This is a flowchart of the shape prediction method of the present invention.

[0034] Figure 10 This is a comparison diagram between the shape prediction result and the actual shape in an embodiment of the present invention. Sub-images (a) and (b) correspond to the snake bone bending to the left to its maximum angle and not bending (straight state), respectively. Sub-image (c) has the largest shape prediction error at the bending angle, which is 1.496 mm.

[0035] Reference numerals: 1-Continuous robot; 2-Snake skeleton; 3-Drive cable; 4-Guide wheel assembly; 5-Force sensing unit; 6-Drive device; 11-Snake skeleton fixed end; 12-Handle body; 13-Fiber optic cable routing groove; 21-Guide joint; 211-First fastening part; 212-Second fastening part; 22-Flexible hinge; 23-Cable constraint groove; 24-Tool channel; 25-External reserved interface end; 26-Snake skeleton fixed section; 51- 52-Fixed platform; 53-Circular arc elastomer; 54-Vertical connecting arm; 55-Pulley fixed platform; 56-Groove; 67-FBG sensor; 68-Miniature servo electric cylinder; 69-Connector. Detailed Implementation

[0036] The embodiments of the present invention will be described in detail below. These embodiments are based on the technical solutions of the present invention and provide detailed implementation methods and specific operation processes to further explain the technical solutions of the present invention.

[0037] Example 1

[0038] like Figure 1-4 As shown, this embodiment provides a novel continuum robot 1, including: a handle body 12, a snake skeleton 2, a drive device, and two drive cables 3; the snake skeleton 2 is fixed to the front end of the handle body 12; the drive device is disposed at the tail end inside the handle body 12; the two drive cables 3 are symmetrically arranged along the axial direction in the cable constraint grooves 23 on both sides of the snake skeleton 2, the first end is fixedly connected to the tail end of the snake skeleton 2 (i.e., the end away from the handle body), and the second end is fixedly connected to the drive device, and the drive device controls the deformation of the snake skeleton 2 through the drive cables 3.

[0039] 1. Snake bones.

[0040] The snake bone 2 has a tubular structure with a tool channel 24 formed in the center, such as... Figure 4 As shown, it allows passage for biopsy instruments, laser transmission fibers, flushing tubing, and other equipment. The tail end of the snake bone 2 is equipped with an external reserved interface end 25, such as... Figure 3 As shown, the snake bone 2 is provided with a snake bone fixing section 26 at its proximal end for connecting to the end effector. This section is used to connect with the snake bone fixing end 11 of the handle body 12.

[0041] like Figure 3 , Figure 5 As shown, the front and rear walls of the snake bone 2 each include several sequentially adjacent guide joints 21; each guide joint 21 includes one flexible hinge 22 and two nested ring structures.

[0042] 1.1 Nested interlocking structure. For example... Figure 5 As shown.

[0043] The nested ring-lock structure consists of a first locking part 211 at the lower end and a second locking part 212 at the upper end. The two locking parts are connected as one unit by a flexible hinge 22 at their respective middle positions. The first locking part 211 includes an inner ring groove and an outer ring protrusion that encircle the center point of the flexible hinge 22. The second locking part 212 includes an inner ring protrusion and an outer ring groove that encircle the center point of the flexible hinge 22.

[0044] The length of the inner annular groove is adapted to the length of the inner annular protrusion (the top of the inner annular groove and the inner annular protrusion form a certain angle, which is the bending stroke space, the same below), and the length of the outer annular groove is adapted to the length of the outer annular protrusion. The maximum rotation angle of adjacent guide joints is determined by the geometry of the sidewall contact surface. To prevent the protrusion from disengaging from the annular groove during bending, the fitting angle between the outer annular groove and the protrusion is increased so that: the inner annular protrusion rotates within the inner annular groove and its end remains inside the inner annular groove at the limit position of the rotation stroke, while the outer annular protrusion rotates within the outer annular groove and its end remains inside the outer annular groove at the limit position of the rotation stroke. This forms a double-ring nested structure, realizing the rotation, axial limiting, and radial anti-disengagement of the guide joint 21, and enhancing the stability of the rotation process.

[0045] Specifically, the nested ring structure is integrally formed using a three-dimensional laser cutting process based on the snake-bone tubular structure. The contact surfaces of the inner ring-shaped protrusion and the outer ring-shaped groove, as well as the contact surfaces of the inner ring-shaped groove and the outer ring-shaped protrusion, all employ a conical limiting surface structure. During the cutting process, the laser beam is always pointed towards the axis of the tube, causing a natural conical mating structure to form on both sides of the cut. The conical limiting surfaces in the two double-sided limiting structures on both sides are tilted in opposite directions, forming a self-locking mechanism that restricts the radial disengagement of adjacent joint units during bending.

[0046] In addition, a rotation gap is cut between the outer side of the outer annular groove of the first fastening part 211 and the outer side of the outer annular protrusion of the second fastening part 212 to prevent interference or collision when the two fastening parts rotate relative to each other.

[0047] 1.2 Flexible hinges. For example... Figure 6 As shown.

[0048] The flexible hinge 22 adopts a C-shaped cross-section arc flexible hinge structure. The C-shaped flexible hinge is composed of two concentric arcs, and bending deformation is achieved based on a single-cut micro-spring mechanism.

[0049] Among them, the flexible hinge 22 is also integrally formed by laser precision cutting process on the basis of the tubular structure of snake bone 2. Its cross-sectional thickness is distributed non-linearly along the axial direction, so that the multi-joint structure exhibits the equivalent mechanical properties of linear spring chain during bending.

[0050] In this embodiment, the flexible hinge 22 includes rigid segments at the upper and lower ends and an arc segment in the middle. The main structural parameters include the distance between the upper and lower rigid segments. outer diameter of the arc inner diameter of the arc central angle Effective deformation length Thickness at the thinnest point Cross-sectional width and radial height Among them, dimensionless parameters This device is used to characterize the flexibility of hinges. Based on the principle of single-cut micro-springs, the hinge is integrally formed using laser precision cutting technology. Its thickness has a nonlinear gradient distribution along the axial direction, which makes the multi-joint structure exhibit the equivalent mechanical properties of a linear spring chain during bending. This transforms the complex nonlinear bending behavior into a linear superposition relationship, significantly enhancing the system's resistance to local disturbances and motion consistency.

[0051] To quantitatively analyze the mechanical properties of this hinge, the following mathematical model is established in this embodiment. The hinge acts as a variable cross-section bending mechanism, and its thickness... The variation along the axial direction can be expressed as: ; According to the pseudo-rigid body model (PRBM), the bending stiffness coefficient K can be calculated using the following formula: ; in, The elastic modulus of the material. Let be the moment of inertia of the cross section. For a C-shaped cross section, the moment of inertia is... The expression is: ; Effective deformation length of the hinge Determined by geometric relations: ; Central angle satisfy: ; The above formula establishes a quantitative relationship between hinge geometry parameters and bending stiffness, providing a theoretical basis for kinematic modeling.

[0052] Compared with existing technologies, this invention sets the central area as a curved flexible hinge 22, while retaining the double-ring structure on both sides to achieve the function of limiting and preventing disengagement. This composite design of "central elasticity + rigid limiting on both sides" ensures both bending flexibility and prevents radial disengagement of the joint.

[0053] 2. Drive cable.

[0054] In this embodiment, the drive cable 3, such as Figure 2 , 4As shown in Figure 5, two symmetrically arranged metal wires or polymer fiber filaments are used. Each drive cable 3 passes through all guide joints 21 in sequence and is embedded in the cable constraint groove 23 at the side wall position of each guide joint 21. The ends of the two drive cables 3 are fixed to the farthest guide joint 21 by welding or mechanical clamping, and their proximal ends are guided by the guide wheel inside the handle and then connected to the drive device 6 through the force sensing unit 5.

[0055] In this embodiment, the cable constraint groove 23 is disposed on the left and right side walls of the snake bone 2 to accommodate and guide the movement trajectory of the drive cable 3. In this embodiment, due to the side wall layout, the cable constraint groove 23 only occupies the limited space around the pipe wall and will not encroach on the central area of ​​the snake bone, thereby ensuring unobstructed operation. In addition, the cable constraint groove 23 adopts an arc-shaped or wedge-shaped cross-section design, and its groove depth is slightly larger than the diameter of the drive cable 3, so that the drive cable 3 is confined in the groove while being able to slide freely along the axial direction, and keeps the drive cable 3 parallel to the snake bone axis, effectively reducing nonlinear friction caused by lateral displacement.

[0056] 3. Guide wheel assembly.

[0057] The handle body 12 has three guide wheels inside, such as Figure 2 As shown, the guide wheel assembly 4 is used to guide the direction of a section of the drive cable between the drive device 6 and the snake skeleton 2, preventing the drive cable 3 from being affected by obstacles and thus affecting the control accuracy of the snake skeleton 2. Specifically, this embodiment uses a miniature rolling bearing type guide wheel. The outer edge of the guide wheel is machined with circular or wedge-shaped grooves to accommodate and restrict the position of the drive cable 3, preventing the cable from coming off. The guide wheel is mounted on a fixed rotating shaft via a miniature rolling bearing, allowing for flexible rotation, thereby changing the transmission direction of the drive cable 3 and significantly reducing frictional resistance.

[0058] 4. Drive unit.

[0059] In this embodiment, the drive device 6 employs a linear drive mechanism with position feedback function, such as a servo motor combined with a ball screw module or a miniature linear driver. The drive device 6 receives instructions from the control system, precisely controls the extension and retraction of the drive cable 3, and provides real-time feedback of the actual displacement through a built-in photoelectric encoder or magnetic encoder, thereby acquiring the length change data of the drive cable 3 for predicting and controlling the deformation of the snake bone 2.

[0060] The drive device 6 in this embodiment uses a Time Robot LA series micro servo electric cylinder 61 (16mm stroke, D-type LVTTL serial interface). This cylinder features an integrated drive and control design, internally integrating a coreless motor, planetary reducer, lead screw mechanism, absolute position sensor, and servo controller. It has a feedforward compensated position closed-loop control function, achieving a repeatability accuracy of ±0.02 mm. The connector 62 connects the drive cable 3 to the actuator of the micro servo electric cylinder 61, accurately transmitting the linear motion of the cylinder to the drive cable 3. The micro servo electric cylinder 61 receives commands from the control system and precisely controls the extension and retraction of the push rod, thereby driving the cable 3 to generate displacement. Those skilled in the art should understand that the aforementioned micro servo electric cylinder 61 is merely a preferred example of this embodiment. Any micro linear drive device with position feedback function can be equivalently replaced, including but not limited to servo motor modules with ball screws, linear motors, piezoelectric ceramic actuators, shape memory alloy actuators, etc., as long as they can achieve precise displacement control and have position feedback function, they can be applied to this invention.

[0061] 5. Force sensing unit. For example... Figure 7 As shown.

[0062] Force sensing units 5 are respectively installed on the left and right side walls inside the handle body 12, including a U-shaped cantilever platform structure, a pulley, and an FBG sensor 56; wherein the pulley reuses the middle guide wheel in the guide wheel assembly 4. The pulley is located at one end of the U-shaped cantilever platform structure near the axis, and the drive cable 3 passes around the pulley with the resultant force perpendicular to the axis direction; the U-shaped cantilever platform structure has a groove in the direction perpendicular to the axis, and the FBG sensor 56 is installed in the groove 55; when the drive cable 3 is subjected to force, the pulley is driven by force to deform the U-shaped cantilever platform structure, and the FBG sensor 56 obtains the tension of the drive cable 3 according to the deformation of the U-shaped cantilever platform structure.

[0063] Specifically, the U-shaped cantilever platform structure includes The components include a fixed platform 51, an arc-shaped elastic body 52, a vertical connecting arm 53, and a pulley fixed platform 54; The fixed platform 51 is fixedly connected to the left / right side walls inside the gripping part; each end of the arc-shaped elastic body 52 is connected to... The fixed platform 51 and the pulley fixed platform 54 are integrated to form a U-shaped frame. Additionally, a recess 55 for mounting the FBG sensor 56 is provided. The surfaces of the type fixed platform 51 and the pulley fixed platform 54.

[0064] Each FBG sensor 56 employs a dual FBG differential structure: the working FBG is bonded to the force-bearing side, and the temperature-compensating FBG eliminates temperature drift through differential calculation (temperature sensitivity is reduced from 9.6 pm / °C to 0.27 pm / °C).

[0065] like Figure 8 As shown, when the tension F of the drive cable 3 is applied to the pulley fixed platform 54, the arc elastic body 52 bends and deforms, and the platform is displaced. The FBG sensor 56 generates strain. Displacement It has a linear relationship with the load F: Where C is the compliance coefficient. Strain With displacement The geometric relationship is satisfied, and the final wavelength shift and tension satisfy the following: Therefore, a linear mapping relationship between wavelength change and tension value can be established through calibration experiments. The FBG sensor 56 can be used to acquire the tension of the drive cable 3 in real time. Its pigtail is led out to the external demodulation equipment through the fiber optic cable routing slot 13. The acquired tension data provides basic information for force feedback control and shape prediction.

[0066] 6. Shape prediction module.

[0067] The novel continuum robot 1 also includes a shape prediction module; the shape prediction module uses an artificial intelligence network and predicts the position of several control points of the snake bone 2 based on the length change and / or tension data of the drive cable 3 collected by the sensor module, and then fits the shape curve of the snake bone 2.

[0068] Furthermore, the shape prediction module is also used to: adjust the predicted position of the control point by a global scaling factor based on the total length constraint of the snake bone and the base point distance constraint, and then fit the morphological curve of the snake bone 2 based on the adjusted position.

[0069] The detailed workflow of the shape prediction module in this embodiment is used to implement the preliminary coordinate prediction, coordinate correction and shape reconstruction of the shape prediction method described in Embodiment 2. It is further explained in Embodiment 2.

[0070] Preferably, the snake bone 2 is made of stainless steel, with an outer diameter of 4 mm, a wall thickness of 0.2 mm, a total length of 75 mm, and an effective working length of 60 mm; the number of guide joints 21 is 18, the bending angle of a single guide joint 21 is ±10°, and the overall bending range is ±180°.

[0071] Example 2

[0072] This embodiment provides a shape prediction method for the novel continuum robot described in Embodiment 1, referencing... Figure 9As shown, it includes:

[0073] S1, Data Acquisition: The drive cable is controlled by the drive device to produce different length changes, and the bending images of the snake bone under different drive states are acquired. The coordinates of each control point on the center line of the snake bone are extracted as the true value through image processing.

[0074] In this embodiment, the center of the snake's tail end, the center of the snake's tail end incision, and the center point of the flexible hinge of the 18 guide joints are taken as control points. The positions of the front and rear walls along the axial centerline and the connection points with the handle body are taken as base points, and are numbered sequentially from near the handle body to the snake's tail end. .

[0075] S2, Sample Construction: Construct a training sample set using the change in the length of the driving cable as the input feature vector and the coordinates of each control point as the target output vector.

[0076] Suppose there are m sets of sampled data (m=195 in this embodiment), for any i... For each sample, the input feature vector is composed of the length variations of the two driving cables. ,in These represent the length changes of the two drive cables relative to their initial positions, with the extension of the electric cylinder defined as positive and the retraction of the electric cylinder as negative. The initial position... The output target vector is a combination of the physical coordinates of 20 control points. .

[0077] Because perspective distortion or extraction errors may occur during image acquisition, in a more preferred embodiment, after extracting the coordinates of each control point on the snake's midline through image processing, the coordinates of each sample are processed before training. Perform geometric corrections. First, force the base points... To the first control point The distance is constant at a preset value. (In this embodiment) Secondly, forced from the first control point; To the last control point The cumulative line distance is kept constant at a preset value. (In this embodiment) The correction method is as follows: keeping the relative orientation angles between control points unchanged, the coordinates of all control points are adjusted through a global scaling factor to satisfy the two constraints mentioned above. This part modifies the coordinates in the sample in the same way as the principle of correcting the predicted coordinates in subsequent step S4; the detailed correction process is described in the predicted coordinate correction section.

[0078] Revised As the output target vector used for final training.

[0079] S3, Model Training: Using artificial intelligence algorithms, with the goal of minimizing the mean square error between the predicted and true coordinates of the control points, a nonlinear mapping model from the change in the length of the drive cable to the coordinates of each control point is established.

[0080] In this embodiment, the random forest algorithm is used to construct the prediction network. The random forest consists of B decision trees (B=100 in this embodiment), and its prediction function based on ensemble is: ; in Let represent the prediction function of the b-th decision tree.

[0081] Model training is performed by minimizing the mean squared error loss function: ; Where N is the number of training samples. For each decision tree, the Bootstrap method is used to randomly draw the same number of samples with replacement from the original training set as the training set for that tree. At the same time, when splitting each node, a subset of features is randomly selected (the size of the feature subset is usually sqrt(2)≈1) for optimal splitting. Each tree is recursively grown until the preset minimum number of samples in the leaf nodes is reached (5 in this embodiment).

[0082] S4, Preliminary prediction: Input the real-time collected changes in the length of the drive cable into the trained prediction model to obtain the predicted coordinates of each control point of the snake bone.

[0083] Based on the initial prediction results of random forest The predictions may not fully meet the robot's physical constraints (e.g., distance deviation from the base point to the first control point, total length deviation), therefore post-processing geometric constraint correction is introduced. This embodiment employs a post-processing constraint ensemble strategy to correct the random forest's predictions through geometric constraints. ; Where C represents the geometric constraint correction function. The correction steps are as follows:

[0084] (1) Calculate the predicted point sequence Predicted coordinates from the base point O to the first control point Predicted distance and the predicted coordinates of the first control point Predicted coordinates to the last control point Cumulative broken line distance .

[0085] (2) Calculate the distance scaling factor Multiply the coordinates of the first control point by This corrects the distance from base point O to the first control point to the true value. .

[0086] (3) Calculate the length scaling factor ,in This is the cumulative polyline distance from the first control point to the last control point after correction in step (2). Keeping the coordinates of the first control point unchanged, multiply the vectors of the other control points relative to the first control point by a length scaling factor. This corrects the cumulative distance from the first control point to the last control point to be... .

[0087] (4) Control point coordinate sequence after two scaling operations These are the final predicted coordinates of the snake-bone control points, which strictly meet the preset geometric constraints.

[0088] S5, Shape Reconstruction: Based on the predicted coordinates of each control point after correction, the morphological curve of the snake bone is reconstructed using a spline curve fitting algorithm.

[0089] Based on the corrected sequence of coordinates of all control points A smooth curve is fitted using a cubic spline interpolation algorithm, which represents the reconstructed complete snake skeleton shape. Cubic spline interpolation ensures that the curve passes through all control points and has continuous first and second derivatives, thus obtaining a smooth and physically conforming continuum shape.

[0090] In this embodiment, the Euclidean distance between the predicted shape endpoint and the actual endpoint is defined as the absolute error: ; in, and These are the predicted and actual coordinates of the 20th control point (endpoint control point), respectively. The relative error is expressed as the ratio of the absolute error to the total length of the robot.

[0091] Simultaneously, Dynamic Time Warping (DTW) distance is introduced to assess overall shape similarity: ; in and These represent the predicted and corrected control point coordinate sequences, respectively.

[0092] In this embodiment, a total of 195 training samples were collected. Figure 10 As shown, this is a comparison diagram between the snake-like morphology (dashed line) predicted by the method of this invention and the actual morphology (solid line). It can be seen that the two are highly consistent throughout the entire workspace. Furthermore, experimental results show that within the full bending range, as... Figure 10As shown, the mean error of the end-shape prediction is as low as 0.28 mm, the standard deviation is 0.29 mm, and the maximum error is 1.496 mm; the mean relative error is 0.46%, and the standard deviation is 0.48%. The overall performance reaches the millimeter-level accuracy threshold.

[0093] In summary, the micro-continuum robot and its shape prediction method provided by this invention achieve small size, large bending range, high-precision shape prediction, and excellent anti-interference performance through innovative arc-shaped flexible hinges, nested ring structures, U-shaped cantilever platform force sensing units, and a shape prediction algorithm combining random forest and geometric constraints (RF-Geo). Furthermore, by sensing changes in drive cable tension, tip contact detection and safety control can be achieved, meeting the clinical needs of minimally invasive surgeries such as laryngeal surgery.

[0094] The above embodiments are preferred embodiments of this application. Those skilled in the art can make various changes or improvements based on them. Without departing from the overall concept of this application, such changes or improvements should fall within the scope of protection claimed in this application.

Claims

1. A novel continuum robot, characterized in that, include: The handle body, the snake bone, the drive device, and two drive cables are provided. The snake bone is fixed to the front end of the handle body. The drive device is located at the tail end inside the handle body. The two drive cables are symmetrically arranged along the axial direction in the cable constraint grooves on both sides of the snake bone. The first end is fixedly connected to the tail end of the snake bone, and the second end is fixedly connected to the drive device. The drive device controls the deformation of the snake bone through the drive cables. The snake bone adopts a tubular structure, with several sequentially adjacent guide joints on both the front and rear walls; each guide joint includes a flexible hinge in the middle and a nested ring structure on each of the left and right sides. Each nested ring structure includes a first fastening part at the lower end and a second fastening part at the upper end, which are connected as one unit by a flexible hinge; the second fastening part includes an inner ring-shaped protrusion and an outer ring-shaped groove, and the first fastening part includes an inner ring-shaped groove and an outer ring-shaped protrusion. The inner ring-shaped protrusion rotates in the inner ring-shaped groove, and the outer ring-shaped protrusion rotates in the outer ring-shaped groove. The flexible hinge adopts an arc-shaped flexible hinge structure with a C-shaped cross section. The C-shaped flexible hinge is composed of two concentric arcs, and the flexible hinge is integrally formed by laser precision cutting process based on the tubular structure of snake bone. The novel continuum robot also includes a sensor module and a shape prediction module; The sensor module is used to collect data on changes in the length and / or tension of the drive cable; The shape prediction module uses an artificial intelligence network and predicts the positions of several control points of the snake bone based on the data collected by the sensor module, and then fits the shape curve of the snake bone. The shape prediction module is also used to: adjust the predicted position of the control points based on the total length constraint of the snake bone and the distance constraint of the base point through a global scaling factor, and then fit the morphological curve of the snake bone based on the adjusted position.

2. The novel continuum robot according to claim 1, characterized in that, The drive unit has a displacement feedback function to obtain the length change of the drive cable.

3. The novel continuum robot according to claim 1, characterized in that, Several guide wheels are installed inside the handle body to guide the direction of the drive cable between the drive device and the snake bone.

4. The novel continuum robot according to claim 1, characterized in that, Force sensing units, including a U-shaped cantilever platform structure, pulleys, and FBG sensors, are respectively installed on the left and right side walls inside the handle body. The pulley is located at one end of the U-shaped cantilever platform structure near the axis, and the drive cable passes around the pulley with the resultant force perpendicular to the axis direction. The U-shaped cantilever platform structure has a groove in the direction perpendicular to the axis, and the FBG sensor is installed in the groove. When the drive cable is under stress, the pulley is driven to deform the U-shaped cantilever platform structure. The FBG sensor obtains the tension of the drive cable based on the deformation of the U-shaped cantilever platform structure.

5. The novel continuum robot according to claim 1, characterized in that, The U-shaped cantilever platform structure includes The components include a fixed platform, an arc-shaped elastic body, a vertical connecting arm, and a pulley-fixed platform; The fixed platform is fixedly connected to the side wall of the gripping part; The two ends of the arc-shaped elastic body are connected by a vertical connecting arm. The fixed platform and the pulley fixed platform are integrated into one unit; A recess for mounting the FBG sensor is provided in The surface of the fixed platform and the pulley fixed platform.

6. A novel shape prediction method for a continuum robot, used for shape prediction of the novel continuum robot according to any one of claims 1-5, characterized in that, include: S1, Data Acquisition: The drive cable is controlled by the drive device to produce different length changes, and the bending images of the snake bone under different drive states are acquired. The coordinates of each control point on the center line of the snake bone are extracted as the true value through image processing. S2, Sample Construction: Construct a training sample set using the change in the length of the driving cable as the input feature vector and the coordinates of each control point as the target output vector; S3, Model Training: Using artificial intelligence algorithms, with the goal of minimizing the mean square error between the predicted and true coordinates of the control points, a nonlinear mapping model from the change in the length of the drive cable to the coordinates of each control point is established. S4, Preliminary prediction: Input the real-time collected changes in the length of the drive cable into the trained prediction model to obtain the predicted coordinates of each control point of the snake bone. S5, Shape Reconstruction: Based on the predicted coordinates of each control point, the morphological curve of the snake bone is reconstructed using a spline curve fitting algorithm.

7. The shape prediction method for the novel continuum robot according to claim 6, characterized in that, After extracting the coordinates of each control point on the centerline of the snake bone through image processing, the obtained coordinates of each control point are corrected by geometric constraints to satisfy the preset conditions of constant distance from the base point to the first control point and constant total length of the snake bone, and the corrected coordinates are used as the true values; wherein, the base point is located at the connection point between the snake bone and the handle body, and the first control point is located at the center point of the first guide joint near the handle body. After the preliminary prediction of the predicted coordinates of each control point in step S4, the predicted coordinates of each control point are also corrected by geometric constraints to satisfy the preset conditions of constant distance from the base point to the first control point and constant total length of the snake bone. In step S5, the corrected coordinates are used to perform spline curve fitting to reconstruct the morphological curve of the snake bone.

Images