CN-122020895-A - Mechanism-data hybrid drive-based continuum robot winding capture contact modeling solving method

CN122020895ACN 122020895 ACN122020895 ACN 122020895ACN-122020895-A

Abstract

A mechanism-data hybrid driving-based continuous robot winding capture contact modeling solving method belongs to the technical field of continuous robot winding capture dynamics modeling. Firstly, designing a continuum robot which is convenient for winding a captured object, establishing a multi-body dynamics model containing contact force, then, establishing a data-driven contact force prediction model based on the physical characteristics of the multi-body dynamics model of the continuum robot and a contact force complementary equation, and finally, coupling the trained contact force prediction model with the multi-body dynamics model of the continuum robot based on the contact force prediction model and solving the coupling model. According to the invention, the data driving model is integrated into the multi-body dynamics simulation, so that an optimal solving flow of experiment-training-simulation-retraining is formed, the problem of insufficient precision of contact force prediction in winding scenes due to nonlinearity and physical constraint can be solved, and the method can be expanded to interaction scenes of flexible robots and unstructured environments.

Inventors

PENG HAIJUN
LIU ZONGCHENG
SONG NINGNING
HAN XIANHAO
HE CHENYU
YANG HAO
ZHANG JIE

Assignees

大连理工大学

Dates

Publication Date: 20260512
Application Date: 20260127

Claims (8)

1. The mechanism-data hybrid driving-based continuum robot winding capture contact modeling solving method is characterized by comprising the following steps of: The first step, designing a continuum robot which is convenient for winding and capturing objects, and the specific steps are as follows: The continuous robot takes nickel-titanium alloy rods as a central skeleton, adopts nickel-titanium alloy rods with different rigidities for winding and enveloping different parts, and has the diameter of the nickel-titanium alloy rods as d a , wherein a rigid joint disc is arranged in the middle of each arm joint, two adjacent arm joints are connected through a hollow universal joint, a motor is arranged at the rigid joint disc on the initial arm joint, and the whole continuous robot is driven to move through a driving rope inserted into a rope penetrating hole of the rigid joint disc; Secondly, based on the continuum robot structure constructed in the first step, establishing a multi-body dynamics model containing contact force; Thirdly, based on the physical characteristics of the multi-body dynamics model and the contact force complementary equation of the continuum robot constructed in the second step, a data-driven contact force prediction model is built, and the method comprises the following steps of: Providing a contact force modeling strategy, wherein the primary objective is to construct a structure of a contact force prediction model, and obtaining an accurate contact force prediction model by training experimental data in the structure of the contact force prediction model; the method comprises the steps of adopting an artificial neural network ANN as a basic modeling tool, wherein a general calculation graph of the artificial neural network comprises an input layer, a hidden layer and an output layer, wherein parameters of the input layer comprise all variables related to target output; And fourthly, coupling the trained contact force prediction model with the multi-body dynamics model of the continuum robot based on the contact force prediction model established in the third step, and solving the coupling model.
2. The mechanism-data hybrid drive-based continuum robot winding capture contact modeling solution method of claim 1, wherein in the continuum robot of the first step, the rigid joint disc radius on each arm joint And hole center distance The distance between two adjacent rigid joint discs changes with the distance from the base end to the whole in a decreasing rule The distance from the base end is changed as a whole to be increased with the rule of decreasing first and then increasing.
3. The mechanism-data hybrid driving-based continuum robot winding capture contact modeling solving method according to claim 2, wherein the second step is specifically: Step 2.1, making assumption for multi-body dynamics model, adopting generalized coordinates Describing the motion of the continuum robot, wherein the geometry of the continuum robot is a spatial motion, wherein And Respectively the current arm sections Corresponding local coordinate system Relative to a global coordinate system Is a rotation angle and coordinates of (a); is an arm section Arm joint Bending deformation angle of (C) Is a local coordinate system Specific: Representing arm joints Corresponding local coordinate system Relative to a global coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Relative to a global coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; Representing arm joints Corresponding local coordinate system Relative to arm joints Corresponding local coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Relative to arm joints Corresponding local coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Relative to a global coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Relative to a global coordinate system In (a) The rotation angle of the shaft; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; Representing arm joints Corresponding local coordinate system Origin relative to global coordinate system Origin at Coordinates on the axis; step 2.2 kinetic energy of the continuum robot The method comprises the following steps: (1) ; Wherein, the 、 The kinetic energy of the arm joint, the nickel-titanium alloy rod and the driving rope are respectively calculated; Is a mass matrix of the continuum robot arm segment, Is that Is the first derivative of (a); Generalized coordinates for each arm segment; Curvature vector caused by bending of nickel-titanium alloy rod The method comprises the following steps: (2) ; Wherein, the And Respectively correspond to Curvature in two principal planes; First, the Length variation of root driving rope And curvature of The relation of (2) is: (3) ; Wherein, the For driving the angle of the rope in the plane of the cross section; for driving the total number of ropes, here due to The driving ropes are uniformly distributed, ; As a continuous function; Is the line coordinate along the nickel-titanium alloy rod; for driving the length of the rope; potential energy of drive rope The method comprises the following steps: (4) ; Wherein, the Is the first The stiffness coefficient of the root drive rope; elastic potential energy of nickel-titanium alloy rod in continuum robot The method comprises the following steps: (5) ; Wherein, the A strain vector that is a beam element; Is a stress vector; is the first The volume of the section nickel-titanium alloy rod; Total weight potential energy of arm joints in continuum robot The method comprises the following steps: (6) ; Wherein, the Is the first The mass of the individual arm segments; is the first Generalized coordinates of the individual arm segments; gravitational acceleration; is the first The centroid height of the arm segments; The total potential energy of the continuum robot The method comprises the following steps: (7) ; The kinetic equation of the continuum robot can be obtained as follows: (8) ; Wherein, the , Is the derivative of the system mass array, Is that Is a second derivative of (2); Representing the partial derivative of the potential energy of the continuum robot to the generalized coordinates; A generalized force vector corresponding to the contact force between the object and the arm section of the continuum robot and between the arm sections of the continuum robot; First, the Geometric constraints of a continuum robot by a drive rope The method comprises the following steps: , (9) ; Wherein, the Is that The length of the driving rope from the base end to the tail end at any time; The length of the driving rope from the base end to the tail end at the initial moment; The center distance of rope penetrating holes on two adjacent sections of discs; Is that Time driving rope Is a driving length of (a); Representing time; when the drive rope is tensioned, equation (9) takes the equal sign for time And (3) derivative obtaining: + (10) ; Wherein the method comprises the steps of To constrain jacobian matrix A corresponding row vector; limiting the relative torsion movement between two adjacent arm joints through a geometric equation and the joint intersection point coincidence at the joint of the two arm joints, for the first Geometric constraints at individual gimbals Expressed as: (11) ; Wherein, the Is the first The universal joints are connected with the central point; And Respectively under global coordinate system Lower part of each arm section and the first arm section The upper universal joint connection point coordinates of the arm joints; And Respectively the first Arm segment and the first Attitude coordinates of the arm sections; step 2.3, the core of the contact problem is the processing constraint, described by the complementary problem in non-smooth mechanics: (12) ; Wherein the method comprises the steps of Is an extended gap function; a distance discriminant function related to the generalized coordinates; a distance discriminant function related to generalized speed; The constraint on the speed layer is: (13) ; three states of tangential contact force, including adhesion, slip, critical, are uniformly expressed as linear complementation problem: (14) ; Wherein, the Tangential relative velocity vector for the point of contact; Is a generalized velocity vector; is the static friction coefficient; a generalized force scalar that is a normal contact force; is a tangential contact force generalized force vector; A unit direction vector that is the tangential relative velocity; step 2.4, adding state quantity of the wound captured object to the continuous robot multi-body dynamics equation proposed by the formula (8), and introducing Lagrange multiplier And Representing normal contact force vectors and tangential contact force vectors, Is the restraining force vector at the universal joint, In order to drive the rope tension vector, And A jacobian matrix containing tangential force complementary constraints and containing normal force complementary constraints, respectively: (15) ; Wherein the method comprises the steps of Is in the formula (8) The term minus the term of the relevant force vector of the drive rope, Is that For generalized coordinates A transpose of the partial derivatives of (a); for generalized coordinates Is a transpose of the partial derivative of (a).
4. A mechanism-data hybrid driving based continuum robot winding capture contact modeling solution method according to claim 3, wherein in step 2.1, it is assumed that: (a) No external collision and self-contact condition exists at the hollow universal joint; (b) Each arm segment deformation obeys an equal curvature assumption; (c) In the winding and capturing process, all arm sections are not contacted with each other, the torsion angle between the arm sections is 0, and only bending deformation is considered; (d) The friction force between the driving rope of the continuum robot and the rope threading hole is small and negligible relative to the contact friction force between the continuum robot and the external environment.
5. The mechanism-data hybrid drive-based continuum robot winding capture contact modeling solving method according to claim 4, wherein the third step specifically comprises: Step 3.1, determining input variables and output variables of a contact force prediction model; Firstly, the contact and motion states of a continuum robot under different winding object configurations are used for obtaining the contact point speeds, the contact point accelerations, the driving rope pulling forces and the change rates of the arm joint contact point speeds, the contact point accelerations, the driving rope pulling forces and the change rates of the driving rope pulling forces at different moments as the input of a contact force prediction model, the contact resultant force between a rigid joint disc on a corresponding arm joint and an object is used as the output of the contact force prediction model, and finally the contact resultant force of each arm joint is decomposed into normal contact force and tangential contact force components under the condition of meeting physical constraint, so that the validity of a contact force prediction model frame is verified; step 3.2, designing a contact force prediction model training frame based on the input variable and the output variable of the contact force prediction model constructed in the step 3.1; The input variables are contact point speed, contact point acceleration, driving rope tension and change rate thereof, the input variables are closely related to contact resultant force, and the target output result is the contact resultant force which can be decomposed into normal contact force and tangential contact force components; For the framework of the contact force prediction model, the target result is the contact resultant force, labeled as The actual output result is marked as Therefore loss in the contact force prediction model training process The following are provided: (17) ; Wherein the method comprises the steps of Representing the number of data points, Representing data sequence number, loss The method comprises the steps of respectively corresponding to a first term and a second term in a formula (17) through mean square error loss and physical constraint loss, wherein in each iteration process, the weights and the offsets are updated through formulas (18) to (19), and the method is specifically as follows: (18) ; (19) ; Wherein, the Represents a learning rate, the value of which is between 0 and 1; in summary, a data-driven contact force prediction model and a training framework thereof are established based on the physical characteristics of the multi-body dynamics model and the contact force complementary equation of the continuum robot constructed in the second step.
6. The mechanism-data hybrid driving-based continuum robot winding capture contact modeling solution method according to claim 5, wherein in the step 3.1: The contact force for a certain arm segment is expressed as: (16) ; Wherein, the Is that The resultant force scalar of contact of a certain arm segment with an object at the moment, Is that The speed and the and of the contact point of the arm section and the object at the moment The velocity vector formed by the addition of the dimensional zero matrix, For the number of arm segments of the robot, Is that The moment is the scalar of the acceleration of the arm segment at the point of contact with the object, Is that The tension vector of the drive rope is constantly taken, Is that The rate of change vector of the tension of the drive rope at the moment.
7. The mechanism-data hybrid driving-based continuum robot winding capture contact modeling solving method according to claim 6, wherein the fourth step is: step 4.1, for the multi-body dynamics model of the continuum robot and the wound captured object, the state quantity and the constraint quantity of each arm section of the continuum robot and the wound captured object are updated once after a time step every time a driving rule is given, and collision contact detection is performed once through a geometric relationship, wherein a detection function is set as follows Because the continuum robot has K arm sections Is set as The method comprises the steps of measuring the contact condition of a node disc on each arm section and an object in a matrix, determining the position coordinates of the contact point after the contact condition is detected, then solving the speed and the acceleration of the contact point through a difference method, merging the driving rope tension and the change rate thereof, inputting the combined driving rope tension into a contact force prediction model together, outputting a contact resultant force by the contact force prediction model, decomposing the contact resultant force into a normal contact force and a tangential contact force through the normal speed and the tangential speed of the contact point, applying the normal contact force and the tangential contact force on a winding capture object, simultaneously solving the position coordinates and the action direction vector of the contact point under a local coordinate system established at the centroid of each arm section, returning the position coordinates and the action direction vector to a multi-body dynamics equation of the continuous body robot together with the contact force for iterative solution, returning to the step one for solving the next state if the contact condition is converged, otherwise returning to the step two, recalculating the state of the continuous body robot and repeating the subsequent steps to finally obtain the latest states of the continuous body robot and the winding capture object; Step 4.2, assume Time of day (time) The joint disc on each arm joint contacts with the object Let the mass center of the arm joint be The mass center position of the object is The kth segment has a radius of The radius of the object is The coordinates of the contact point positions are respectively expressed as follows based on the section disc and the object: (20) ; (21) ; Wherein the method comprises the steps of , A rotation matrix vector derived from the gesture quaternion, Is a common normal vector, and specifically comprises the following steps: (22) ; If the acquisition is successful, for each All should satisfy: (23) ; while the speed of the contact point is expressed by the difference method as: (24) ; the acceleration at the point of contact is expressed by the difference method as: (25) ; Step 4.3, setting the combined contact resultant force function as Will be Is set as Dimension matrix and the same ( ) And ( ) Are all set as The dimension matrix is used to determine the dimensions of the matrix, And Are all set as Dimension matrix of ( ) And ( ) Each contact point is regarded as a tangent point between the circumference of the joint disc and an object, the three-dimensional coordinates of the contact point at the current moment can be obtained by utilizing the geometric relationship under the condition that the radius of each joint disc, the radius of the object, the coordinates of the central point of each joint disc, the direction cosine of the central axis of the object and the coordinates of the mass center of the object are known, and then the speed of the contact point at the current moment and the acceleration of the contact point are obtained by utilizing a difference method; Can be made by And Differential results, namely: (26) ; since the coupling model composed of the contact force prediction model and the multi-body dynamics model of the continuum robot can directly output the contact force, the dynamics equation is expressed as: (27) ; Wherein the method comprises the steps of Comprises external forces which can be directly solved, such as gravity, elastic force and the like of the continuum robot and the wound captured object, For the assembled double-sided constraint condition, For bilateral constraint of the transpose of the corresponding gradient matrix, Lagrangian multiplier for bilateral constraint; after all the above steps are completed, by combining The speed and the acceleration of the contact point of the wound capturing object on the corresponding section plate are input into a trained contact force prediction model together with the tension of the driving rope and the change rate of the tension, so that the real-time prediction of the contact force can be realized, and the dynamic performance of the continuous robot and the wound capturing object can be predicted by means of a multi-body dynamics model of the continuous robot.
8. The mechanism-data hybrid drive based continuum robot winding capture contact modeling solution of claim 7, wherein the convergence condition in step 4.1 is that the equation (27) residual is less than 1e -4 .

Description

Mechanism-data hybrid drive-based continuum robot winding capture contact modeling solving method Technical Field The invention belongs to the technical field of continuous robot winding capture dynamics modeling, relates to a contact force prediction framework based on a physical constraint neural network, and particularly relates to a mechanism-data hybrid driving-based continuous robot winding capture contact modeling solving method, which integrates a multi-body dynamics and data driving method to realize high-precision modeling and solving of continuous robot winding capture. Background The field of dynamic interaction between the continuum robot and the outside is in a rapid development stage, the unique structure and the driving mode provide a new way for solving the operation task under the complex environment, the continuum robot can flexibly shuttle like a snake by virtue of high flexibility and flexibility, capture and operation on the target are realized, the operation range of the robot is effectively expanded, the task which the traditional rigid robot can not complete is solved, however, when the continuum robot interacts with the outside, particularly when capturing a winding object, a plurality of difficulties exist due to the structural characteristics and the complexity of the operation environment, and the method comprises the following steps: In terms of the structure and dynamic characteristics of the robot, the dynamic model of the continuum robot has strong nonlinearity, and the motions of all joint modules are mutually coupled, so that the motion rule of the traditional linear control method is difficult to describe accurately, meanwhile, the redundant degree of freedom can enable the continuum robot to have various motion paths when capturing a wound object, and if the continuous robot is improperly controlled, the contact position or winding mode of the continuum robot and the object can be uncontrollable, and the capturing stability is influenced. In terms of uncertainty of the wound object and the environment, the contact of the continuum robot with the object belongs to dynamic rigid interactions, and the contact point position and the contact force change with time. For example, when the continuum robot winds an object, the movement of the contact point changes the stress balance of the continuum robot and the object, so that the subsequent motion track is affected, and a cycle association process of 'contact-stress-deformation-contact' is formed, which is difficult to deal with by the traditional open loop control. At the same time the winding pattern of the object, e.g. spiral winding, cross winding, may affect the stability of the continuum robot. If winding is loose, the object may fall off, if too tight, the rope may be overloaded or the object may be damaged, and adjusting the winding force in real time requires an accurate force prediction and dynamic control strategy. In the aspect of prediction and modeling, the real-time contact force is difficult to predict, when an object is wound, the contact state of the continuum robot and the object is in multipoint contact, the contact force distribution is complex, the traditional force sensor can only measure concentrated force, the force information of the whole contact surface is difficult to obtain, and control feedback is incomplete. In terms of robustness and adaptability, when the winding state of an object suddenly changes, such as the object falls off, and external force is interfered, a control algorithm needs a rapid adjustment strategy. For example, the traditional PID control is difficult to deal with a system with strong nonlinearity and time-varying parameters, oscillation or out of control can occur, a large amount of trial and error is required in an unknown environment by a learning algorithm such as reinforcement learning, and in practical application, an object damage or system failure can be caused by the trial and error of a continuum robot, so that the online optimization capability of the algorithm is limited. The continuous robot has remarkable advantages in scenes such as minimally invasive surgery and complex environment operation by means of flexible structures and high-degree-of-freedom characteristics, accurate contact force sensing and environment interaction state identification are core preconditions of safe and efficient operation, the current related sensing technology mainly expands around four paths through sensor integration, model driving, data driving and multiparty fusion, remarkable progress is achieved, but challenges are still faced in aspects such as complex working condition adaptability, lightweight design and multiparameter decoupling, wherein the sensor integration technology directly captures contact signals through special elements such as implanted PSS flexible sensors, piezoresistive electronic skin, FBG sensor arrays and SMA stiffness adjustable sensing frameworks, high-pr