A digital twin modeling method for coupling of material bin position channel and transport body

Abstract

The application discloses a kind of material warehouse position passageway and transport body space-time coupling digital twin modeling method, it is related to intelligent logistics control technical field, including: based on the environment boundary of position and passageway geometry data, in combination with the motion trajectory of transport body and energy consumption discerns material quality, constructs digital twin model;According to material quality, calculate material deformation and deformation sensitivity, correct the initial geometry of material with super quadratic surface parameter to obtain envelope boundary with acceleration as variable;Along the gap normal between envelope boundary and environment boundary, use deformation sensitivity to map gap as algebraic constraint, construct continuous collision cost function about acceleration by non-negative mapping and curvilinear integral, solve control law in combination with model;Extract structure damping dissipation energy when executing control law to correct current energy consumption and re-identify material quality, update digital twin model.The application is used to solve the problem of insufficient material boundary representation accuracy, space interference is difficult to convert into constraint and parameter updating distortion.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics control technology, and more specifically, to a digital twin modeling method for spatiotemporal coupling of material storage channels and transport bodies. Background Technology

[0002] Existing AGV / AMR digital twin technology typically collects the motion state, load information, and structural response of the transporter, establishes a dynamic model, and updates the digital twin synchronously to reflect the dynamic changes of the transporter under different loads and operating conditions.

[0003] For example, the invention application with publication number CN122284594A discloses an AGV / AMR digital twin operation and maintenance system that integrates dynamic modeling. It obtains physical attributes such as cargo mass, center of mass, and structural deformation through visual semantic parsing, and solves the dynamic response by combining multibody dynamics to drive the digital twin to update its posture and geometric deformation. At the same time, it identifies and corrects model parameters such as mass and damping online.

[0004] However, the aforementioned existing technologies have the following technical problems when transporting easily deformable materials in confined spaces and narrow passages: 1. When the transport vehicle accelerates, decelerates, or changes direction, the material will undergo dynamic deformation and positional shift due to inertia. Existing digital twin technology usually relies on the delayed update of sensor data, which makes it difficult to accurately reflect the physical envelope boundary of the material changing in real time with the transport state, resulting in a deviation between the geometry in the model and the actual space occupancy.

[0005] 2. In narrow passages, the deformation of materials causes the actual usable space gap to change continuously. Existing models can usually only provide passive status monitoring or collision warnings, and it is difficult to transform this dynamic spatial interference risk into spatial constraints that the execution system can directly use. This makes it difficult for the transport body to accurately judge the safety margin when passing through narrow spaces, which can easily lead to frequent shutdowns or collision interference, affecting the continuity and safety of transport operations. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a digital twin modeling method for the spatiotemporal coupling of material storage channels and transport bodies. This method addresses the problems of insufficient accuracy in representing dynamic material boundaries, difficulty in converting spatial interference into constraints, and distortion in parameter updates by establishing a deformation sensitivity-generated envelope boundary, converting gaps into collision costs, and stripping damping dissipation energy from the closed-loop update model.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A digital twin modeling method for the spatiotemporal coupling of material storage channels and transport systems includes the following steps: Environmental boundaries are established based on the geometric data of storage locations and channels. Material quality is identified by combining the movement trajectory of the transport vehicle and energy consumption. A digital twin model including environmental boundaries and material quality is constructed. Based on the material mass, the material deformation and its deformation sensitivity to the acceleration of the transport body are calculated. The hyperquadratic surface parameters of the initial geometry of the material in the model are corrected to obtain the envelope boundary with acceleration as the variable. Along the normal of the gap between the envelope boundary and the environmental boundary, the gap is mapped to an algebraic constraint using deformation sensitivity, and a continuous collision cost function with respect to acceleration is constructed through nonnegative mapping and surface integral. The transport body control law is then solved in combination with the model. Extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material quality, and then update the digital twin model.

[0008] In a preferred embodiment, obtaining the envelope boundary with acceleration as the variable includes: Using the initial spatial pose matrix and scale parameters in the hyperquadratic surface parameters as reference values, and performing a first-order Taylor expansion with respect to acceleration based on deformation sensitivity, the homogeneous transformation matrix correction and scale correction are obtained. The initial spatial pose matrix and scale parameters are corrected using the homogeneous transformation matrix correction and scale correction, respectively, and then substituted into the preset hyperquadratic surface equation to generate a hyperquadratic surface expression with acceleration as the variable, which serves as the envelope boundary.

[0009] In a preferred embodiment, mapping the gap to an algebraic constraint using deformation sensitivity includes: calculating the shortest gap between the envelope boundary and the environment boundary, and extracting a unit normal vector pointing towards the environment boundary along the shortest gap direction; multiplying the deformation sensitivity by the transport body acceleration to obtain the deformation displacement increment, projecting the deformation displacement increment onto the unit normal vector to obtain the normal intrusion amount; and establishing the algebraic constraint based on the normal intrusion amount, the shortest gap, and a preset safety margin.

[0010] In a preferred embodiment, solving the transport control law using the model includes: concatenating the current pose, velocity vector, and material deformation of the transport body into an augmented state vector; predicting the state changes of the augmented state vector within a preset prediction time domain based on a digital twin model; introducing the continuous collision cost function into the objective function of the predictive control, and performing rolling optimization in conjunction with the state changes to obtain the transport control law.

[0011] In a preferred embodiment, the extraction of structural damping dissipation energy during the execution of the control law includes: obtaining the instantaneous relative velocity of the material relative to the transport body when the transport body executes the control law; and performing damping power integration by combining a preset damping coefficient with the instantaneous relative velocity to obtain the structural damping dissipation energy.

[0012] In a preferred embodiment, the step of identifying material mass by combining the motion trajectory of the transporter and energy consumption includes: establishing a nominal power equation characterizing the mapping relationship between driving power and motion state based on the nominal unloaded mass and nominal resistance parameters of the transporter, and obtaining theoretical energy consumption by integrating the velocity and acceleration in the motion trajectory; calculating the gradient of the residual between the theoretical energy consumption and the energy consumption with respect to the material mass to be identified, and iteratively updating the material mass estimate along the negative direction of the gradient to obtain the material mass.

[0013] In a preferred embodiment, the material deformation includes the relative displacement and attitude deflection of the material relative to the transport body. The calculation of material deformation based on the material mass includes: calculating the inertial load on the material based on the material mass and the acceleration of the transport body, and using the material gravity as an external excitation input to a preset structural dynamics equation to solve for the relative displacement and attitude deflection.

[0014] In a preferred embodiment, establishing the environmental boundary based on the geometric data of the warehouse and the passage includes: extracting spatial geometric features, including the depth dimension of the warehouse, the net clearance boundary of the passage, and the sidewall contour, from the point cloud data of the warehouse and the passage; performing surface fitting and Boolean operations on the spatial geometric features to generate an implicit environmental constraint function, which serves as the environmental boundary.

[0015] In a preferred embodiment, the construction of a digital twin model including environmental boundaries and material mass includes: under a preset unified spatial coordinate system, using the environmental boundaries as environmental boundary constraints, the material mass as entity parameters, and the initial motion state of the transport body and the initial geometry of the material as entity initialization parameters, constructing a joint state space and its state evolution equation as the digital twin model.

[0016] A digital twin modeling system for spatiotemporal coupling of material storage channels and transport vehicles includes the following modules: a twin modeling module, used to establish environmental boundaries based on storage location and channel geometric data, identify material quality by combining transport vehicle motion trajectory and energy consumption, and construct a digital twin model including environmental boundaries and material quality; an envelope update module, used to calculate material deformation and its deformation sensitivity to transport vehicle acceleration based on material quality, correct the hyperquadratic surface parameters of the initial material geometry in the model, and obtain an envelope boundary with acceleration as the variable; a control solution module, used to map the gap between the envelope boundary and the environmental boundary into algebraic constraints using deformation sensitivity, and construct a continuous collision cost function with respect to acceleration through nonnegative mapping and surface integration, and solve the transport vehicle control law in combination with the model; and a closed-loop update module, used to extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material quality, and update the digital twin model.

[0017] The technical effects and advantages of the digital twin modeling method for spatiotemporal coupling of material storage channels and transport bodies in this invention are as follows: This invention calculates the deformation sensitivity of material deformation state to the acceleration of the transport body and updates the initial geometry of the material to an envelope boundary with acceleration as the variable. This allows the digital twin model to dynamically represent the spatial boundary of the material based on the acceleration of the transport body, improving the accuracy of the material deformation state representation in the model. By combining deformation sensitivity, the gap distance between the envelope boundary and the environmental boundary is transformed into an algebraic constraint with respect to acceleration. Furthermore, a continuous collision cost function is constructed, allowing the spatial relationship between the dynamic boundary of the material and the environmental boundary to be incorporated into the control solution in the form of mathematical constraints. This enables the digital twin model to reflect the collision risk corresponding to acceleration changes, providing dynamic spatial constraints for transport control in restricted compartments and channels. Simultaneously, by extracting the structural damping dissipation energy generated when the transport body executes the control law, and using this to correct the current energy consumption and re-identify the material mass, the influence of material deformation and relative motion on the energy consumption identification results is reduced, improving the accuracy of updating the material mass parameters in the digital twin model and allowing the model parameters to be adjusted according to the actual transport state. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a digital twin modeling method for spatiotemporal coupling of material storage channels and transport bodies, provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the expansion of the material envelope boundary within a confined channel, provided in an embodiment of the present invention.

[0020] Figure 3 The graph shows the evolution of the transport body acceleration and the instantaneous relative velocity of the material, as provided in the embodiments of the present invention.

[0021] Figure 4 The total power consumption versus structural damping dissipation power decoupling curve is provided for embodiments of the present invention.

[0022] Figure 5 This is a schematic diagram of a digital twin modeling system module for spatiotemporal coupling of material storage channels and transport bodies, provided in an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1, Figure 1 This invention presents a digital twin modeling method for the spatiotemporal coupling of material storage channels and transport bodies, comprising the following steps: Step 1: Establish environmental boundaries based on the geometric data of the storage location and channels, identify material quality by combining the movement trajectory of the transport vehicle and energy consumption, and construct a digital twin model that includes both environmental boundaries and material quality.

[0025] 1.1 Constructing environmental boundaries based on position and channel geometry data, including: 1.1.1 From the point cloud data of the positions and channels, extract spatial geometric features including the depth of the positions, the net clearance boundary of the channels, and the contour of the side walls, as follows: Using the initial starting position of the transport vehicle during the environmental exploration phase as the coordinate origin, the point cloud data collected by the 3D LiDAR deployed on top of the transport vehicle and the data from the chassis inertial measurement unit (IMU) are used as inputs. The Cartographer simultaneous localization and mapping algorithm is used to construct a 3D occupancy grid map. During the mapping process, the cumulative error is eliminated by configuring the backend pose graph optimization and loop closure detection functions of Cartographer to ensure the global consistency of the constructed map. After the mapping is completed, the coordinate system of the map is fixed as a globally fixed static reference system as a preset unified spatial coordinate system.

[0026] Based on this, relying on the 3D LiDAR and RGB-D depth camera deployed on the top of the conveyor, the surrounding working space is scanned in real time during operation to collect 3D point cloud data of the silos and passages. The spatial resolution of the point cloud data reaches the millimeter level, for example, 2mm. To suppress interference from suspended dust and high-frequency physical noise from the sensors, the collected point cloud data is preprocessed by voxel downsampling filtering and statistical outlier removal. Specifically, the voxel grid size for voxel downsampling filtering is set to 0.05m, and the number of query neighbors for statistical outlier removal is set to 50, with a standard deviation threshold of 1.0, in order to simplify the point cloud and remove outliers.

[0027] Using preprocessed point cloud data as input, a plane fitting algorithm based on Random Sampling Consensus (RANSAC) is employed for plane segmentation. A RANSAC distance threshold of 0.05m is set. The algorithm iteratively extracts the plane model with the largest interior point support from the point cloud set, dividing the point cloud into multiple planes. Based on the normal vectors and relative positions of each plane, spatial geometric features are extracted from the segmented planes: planes with horizontal and relatively parallel normal vectors on both sides (i.e., the angle between the normal vectors is less than 5 degrees) are extracted as sidewall contours; the upper horizontal ceiling plane and the lower horizontal ground plane are jointly extracted as the channel clearance boundary; and the rear wall surface inside the storage area, perpendicular to the channel direction, is extracted as the storage area depth boundary plane. The vertical projection distance from this rear wall surface to the storage area entrance plane is calculated as the storage area depth dimension. Since the RANSAC algorithm directly fits the analytical parameters of each plane, this step ultimately outputs the spatial analytical algebraic equations for each plane corresponding to the aforementioned sidewall contours, channel clearance boundaries, and storage area depth boundaries.

[0028] 1.1.2 Perform surface fitting and Boolean operations on the spatial geometric features to generate implicit constraint functions for the environment, which serve as the environment boundary, as follows: Using the directed distance formula from a point to a plane in analytic geometry, the spatial analytic algebraic equations of each plane are transformed into corresponding directed distance functions, i.e. ,in It is a directed distance function. To pre-define coordinate points in a unified spatial coordinate system, , , , The plane analytic parameters are given by the spatial analytic algebraic equation. Since the normal vector is constrained to point inside the safe space when extracting the plane, the directed distance function is positive inside the safe space and negative when it crosses the physical boundary, thereby realizing implicit surface fitting of the discrete plane.

[0029] By minimizing the aforementioned directed distance functions (i.e., Boolean intersection), a collision-free 3D safe and feasible region enclosed by all faces is constructed. The boundary mathematical expression of this 3D safe and feasible region is then used as an implicit constraint function for the environment, serving as the environment boundary. The formula for this implicit constraint function is as follows: (1) in, For any point in a pre-defined unified spatial coordinate system The implicit constraint function value of the environment at that location. The directed distance function is fitted to the surface of the plane corresponding to the sidewall profile. The directed distance function is fitted to the surface of the plane corresponding to the channel clearance boundary. Fit a directed distance function to the surface of the plane corresponding to the depth dimension of the storage space; if and only if a point in space satisfies At this time, it indicates that the location is completely within a safe physical passage space without interference.

[0030] 1.2 Identify material quality by combining the movement trajectory and energy consumption of the transport vehicle, including: 1.2.1 Based on the nominal mass and nominal resistance parameters of the transport vehicle under no-load conditions, a nominal power equation characterizing the mapping relationship between driving power and motion state is established. The theoretical energy consumption is then obtained by integrating the velocity and acceleration along the motion trajectory, as detailed below: By reading the vehicle's factory nameplate and the system's underlying firmware, the nominal unloaded mass of the transport vehicle is obtained. A constant-speed coasting attenuation test is pre-performed on a standard flat physical surface; that is, the transport vehicle is accelerated to a specified speed and then power is cut off. The inertial measurement unit records the natural deceleration, calibrating the rolling resistance coefficient between the tires and the ground as the nominal resistance parameter. Simultaneously, by analyzing the pulse counts of the high-precision dual-side wheel encoders with a chassis resolution of 4096 lines / revolution, the instantaneous linear velocity and instantaneous angular velocity in the motion trajectory are calculated. The first-order time-difference derivative of this instantaneous linear velocity sequence is then performed, and combined with the longitudinal acceleration data measured by the inertial measurement unit, a Kalman filter is applied to extract the instantaneous acceleration in the motion trajectory.

[0031] Combining the kinetic energy theorem and the law of conservation of energy, a nominal power equation characterizing the mapping relationship between driving power and motion state is established. The formula for the nominal power equation is as follows: (2) in, for The nominal drive power at any given moment. The nominal mass of the transport vehicle under no-load conditions. For the quality of the material to be identified, The instantaneous acceleration extracted from the motion trajectory. The instantaneous linear velocity extracted from the trajectory of motion. These are the nominal resistance parameters. It is the constant of gravitational acceleration; To comprehensively measure the transmission efficiency parameter, based on the conventional electromechanical energy conversion characteristics of industrial AGV drive systems, its value ranges from 0.80 to 0.90, for example, 0.85.

[0032] The extracted instantaneous linear velocity and instantaneous acceleration are substituted into the nominal power equation; within the observation time window, time segments in which the transporter is in driving condition (i.e., instantaneous nominal driving power is greater than 0) are selected, and the nominal power equation is definitely integrated over time within the time segments to obtain the theoretical energy consumption.

[0033] 1.2.2 Calculate the gradient between theoretical energy consumption and the residual between energy consumption with respect to the mass of the material to be identified, and iteratively update the estimated material mass along the negative direction of the gradient to obtain the material mass, as follows: Using the CAN bus communication protocol in the underlying driver interface, the instantaneous drive current and bus voltage of the traction motor are collected in real time with a high-frequency sampling period of 100Hz and multiplied to obtain the instantaneous electric power. Within the time segment of the observation time window, the instantaneous electric power is integrated over time to calculate the actual energy consumption of the transport vehicle.

[0034] Based on the residual between the theoretical energy consumption and the energy consumption, an objective function for the energy consumption residual is constructed, and the formula for the objective function for the energy consumption residual is as follows: (3) in, Let the energy consumption residual objective function be... Theoretical energy consumption, This represents the actual energy consumption of the transport system.

[0035] The gradient of the energy consumption residual objective function with respect to the mass of the material to be identified is calculated, and the estimated material mass is iteratively updated along the negative direction of the gradient to obtain the material mass; the formula for calculating the material mass is as follows: (4) (5) in, Let be the gradient of the energy consumption residual objective function with respect to the quality of the material to be identified. Observation time window For the first Material quality estimate in the next iteration loop For the first Material quality estimate in the next iteration loop; The learning rate hyperparameter for non-negative gradient descent is a preset value, with units of kg. 2 / J 2 Its value is, for example, 0.01.

[0036] Based on the aforementioned energy consumption residual objective function, a gradient descent algorithm is used for optimization. Starting with the initial estimated mass, in each iteration, the current estimated value is substituted into the aforementioned gradient calculation formula to solve for the current gradient, and the material mass estimate is updated according to the formula. Through multiple iterations, the gradient is updated until the absolute value is less than a minimum threshold (e.g., 10). -4 If the maximum number of iterations is reached (e.g., 1000 times), the best estimate after convergence is the final identified material quality.

[0037] 1.3 Construct a digital twin model including environmental boundaries and material quality, including: The three-dimensional geometric contour model of the target material is retrieved in real time from the upper-level WMS warehouse management system or the enterprise CAD design database through the API interface; the three-dimensional coordinate set of all vertices in the three-dimensional geometric contour model is extracted, and the maximum and minimum values ​​of the vertex coordinates are found along the X-axis, Y-axis and Z-axis directions of the preset unified spatial coordinate system established in step 1.1.1, and the difference is calculated to obtain the length, width and height dimensions based on the axially aligned bounding box (AABB), and the three-dimensional bounding box dimensions are used as the initial geometry of the material.

[0038] In the unified spatial coordinate system, the transport body pose extracted based on the system's underlying SLAM localization module (i.e., TF coordinate transformation tree) is concatenated with the linear velocity and acceleration extracted from the motion trajectory in step 1.2.1 to form the kinematic variables of the joint state vector, which serve as the initial motion state of the transport body. The initial geometry of the material obtained above is assigned to the joint state vector. The initial motion state of the transport body and the initial geometry of the material together constitute the entity initialization parameters, thereby constructing and initializing the joint state space, which is represented as follows: (6) in, Let them be the joint state space vectors. The two-dimensional position in the pose of the transport body. The yaw angle is the position and orientation of the transporter. , These are the linear velocity and angular velocity extracted from the motion trajectory, respectively. The three-dimensional envelope scale assigned to the initial geometry of the material.

[0039] The material mass is assigned to the inertial mass parameter in the state evolution equation as an entity parameter; the environmental boundary (i.e., the implicit environmental constraint function) is used as the feasible region constraint for the position components of the state vector in spatial evolution, i.e., the environmental boundary constraint condition; then, a state evolution equation is configured for this joint state space, and the formula of the state evolution equation is as follows: (7) in, The time derivative of the joint state-space vector. For external control input vectors, such as chassis driving torque; The state transition function of the digital twin model is implemented in this embodiment using the kinematic and dynamic coupling equations of the differential-driven mobile robot. At the kinematic level, the coupling equations characterize the integral mapping relationship between the linear velocity and angular velocity of the transport body to the global position and yaw angle, and at the dynamic level, they characterize the mapping relationship between the drive control input and the inertial load to the motion acceleration.

[0040] The aforementioned configuration of entity parameters, initialization parameters, and the joint state space and its state evolution equation constrained by environmental boundary conditions together constitute the digital twin model.

[0041] This step identifies material quality by establishing a gradient iteration of energy consumption residuals with respect to quality. Combined with implicit environmental boundary constraints generated based on point cloud Boolean operations, the model initialization parameters can accurately reflect the current physical and geometric interference relationships, laying a reliable data and constraint foundation for constructing a high-fidelity digital twin joint state space with forward state evolution capabilities.

[0042] Step 2: Calculate the material deformation and its deformation sensitivity to the acceleration of the transport body based on the material mass, correct the hyperquadratic surface parameters of the initial geometry of the material in the digital twin model, and obtain the envelope boundary with acceleration as the variable.

[0043] 2.1 Calculate material deformation based on material mass, and the deformation sensitivity of material deformation to the acceleration of the transport body, including: 2.1.1 Calculate the material deformation based on the material mass. The material deformation includes the relative displacement and attitude deflection of the material relative to the transport body, as detailed below: Multiplying the material mass by the gravitational acceleration vector yields the material's gravity; calculating the negative scalar product of the material mass and the transport body's acceleration yields the inertial load on the material; superimposing the inertial load with the material's gravity vector yields a three-dimensional force vector; calculating the cross product of the material's center-of-mass eccentricity vector and the three-dimensional force vector yields a three-dimensional deflection moment vector; concatenating the three-dimensional force vector and the three-dimensional deflection moment vector into a 6×1-dimensional generalized total external excitation, which is placed on the right-hand side of the preset structural dynamics equation as the excitation source; the calculation formula for the preset structural dynamics equation is as follows: (8) in, For including material quality The generalized mass matrix (constructed from the mass of materials and the distribution of rigid body kinematics theory). This is the second-order acceleration term of the material's deformation vector. For the generalized damping matrix, This is the first-order velocity term of the material's deformation vector. It is the generalized stiffness matrix; Let be the material deformation vector, and ,in It is a relative displacement vector that includes the translation of the material along the X, Y, and Z axes of the spatial coordinate system. It is an attitude deflection vector that includes the roll angle, pitch angle and yaw angle of the material around the X-axis, Y-axis and Z-axis; For generalized total external incentives.

[0044] It should be noted that for specific flexible or bulk materials, their generalized stiffness matrix and generalized damping matrix can be obtained through offline system identification tests (such as impact sweep tests) or modal analysis tests.

[0045] The Newmark-β stepwise integration algorithm is used to numerically solve the above structural dynamics equations; specifically, the Newmark-β algorithm is assumed to be unconditionally stable, and the parameters are set to... , ,in, To control the parameters of the numerical damping algorithm, To control acceleration changes, parameters of the integral law are assumed. The algorithm starts with a given initial displacement and initial velocity, and sets a discrete time step. Within each discrete time step, based on the known mass matrix, stiffness matrix, and damping matrix, a linear algebraic combination is performed using the standard Newmark-β algorithm format to obtain the equivalent stiffness matrix. Simultaneously, combining the deformation displacement, deformation velocity, deformation acceleration at the current moment, and the total external excitation at the next moment, the equivalent load for that integral step is calculated. The system of linear algebraic equations composed of the equivalent stiffness matrix and the equivalent load is solved to obtain the deformation displacement at the next moment. Based on the deformation state variables at the current moment and the Newmark-β algorithm update formula, the deformation acceleration and deformation velocity at the next moment are solved sequentially. Through the iterative process of the above time steps, a 6×1 dimensional material deformation vector relative to the transport body under the current excitation is calculated.

[0046] 2.1.2 Calculate the deformation sensitivity of the material to the acceleration of the transport body, as follows: For each relative displacement and attitude deflection component in the material deformation vector, calculate the first-order partial derivative with respect to each axial component of the transport acceleration vector; substitute the condition that the transport acceleration is zero (i.e., the system is in equilibrium) into the above derivative results to obtain the Jacobian matrix constant that characterizes the rate of change of material deformation caused by unit acceleration at the equilibrium operating point, as the deformation sensitivity.

[0047] 2.2 Based on deformation sensitivity, the hyperquadratic surface parameters of the initial material geometry in the digital twin model are corrected to obtain the envelope boundary with acceleration as the variable, including: 2.2.1 Using the initial spatial pose matrix and scale parameters in the hyperquadratic surface parameters as reference values, and performing a first-order Taylor expansion with respect to acceleration based on deformation sensitivity, the homogeneous transformation matrix correction and scale correction are obtained, as follows: In terms of geometric mapping, the length, width, and height of the three-dimensional bounding box of the target material obtained in step 1 are directly corresponding to and assigned to the three semi-axis scale parameters of the hyperquadratic surface. At the same time, the geometric center coordinates of the three-dimensional bounding box are converted into a 3×1-dimensional translation vector, and the stationary azimuth angle (Euler angle) is converted into a 3×3-dimensional rotation matrix through trigonometric functions. The two are combined and spliced ​​into a 4×4-dimensional homogeneous transformation matrix, which serves as the initial spatial pose matrix of the hyperquadratic surface.

[0048] Taking the system in equilibrium (i.e., the state where the transporter acceleration is zero) as the reference point for the first-order Taylor expansion, the deformation sensitivity is introduced by using the chain rule, and the first-order Taylor expansions of pose and scale are performed with respect to the transporter acceleration.

[0049] For the scale parameter of the hyperquadratic surface, the first-order Taylor expansion formula of the scale parameter is as follows: (9) in, For dynamic scale parameters that depend on the acceleration variable of the transport body, For the acceleration of the transport medium; The initial scale parameters of the hyperquadratic surface are and , These are the half-axis length components of the initial scale parameter along the X, Y, and Z axes of the material-following local coordinate system; This is the deformation sensitivity constant matrix extracted at the equilibrium point where acceleration is zero in step 2.1.2; This is the increment of acceleration relative to the equilibrium state. Since the acceleration in the equilibrium state is zero, ; This is a scale correction amount; The matrix represents the spatial analytical partial derivative of the scale with respect to the material deformation vector. This matrix characterizes the coupling relationship between the envelope parameter and the material deformation. It can be obtained through offline calibration. For example, different motion stimuli are applied to the material, and the six-dimensional material deformation and envelope scale changes are recorded simultaneously using a vision or motion capture system. This constant matrix is ​​then obtained by fitting multiple linear regression.

[0050] For the spatial pose matrix of the hyperquadratic surface, the first-order Taylor expansion formula of the spatial pose matrix is ​​as follows: (10) in, For dynamic pose matrices that depend on acceleration variables, This is the initial spatial pose matrix; The spatial analytical partial derivative matrix of the pose with respect to the material deformation vector can be obtained by taking the analytical partial derivatives of the translation vector component and the rotation matrix component in the initial spatial pose matrix with respect to the six variables of the material deformation vector and combining them. This is the correction factor for the homogeneous transformation matrix.

[0051] 2.2.2 Using the homogeneous transformation matrix correction and scale correction, the initial spatial pose matrix and scale parameters are corrected respectively, and substituted into the preset hyperquadratic surface equation to generate a hyperquadratic surface expression with acceleration as the variable, which serves as the envelope boundary, as follows: Using the obtained homogeneous transformation matrix correction and scale correction, matrix addition is performed with the corresponding initial spatial pose matrix and initial scale parameters, respectively, as an approximate linearization calculation, to obtain the corrected dynamic pose matrix and dynamic scale parameters under real-time acceleration control. The inverse of the dynamic pose matrix is ​​multiplied by the vector of any coordinate point in a preset unified spatial coordinate system, transforming its coordinates to the local coordinate system of the hyperquadratic surface. Specifically... ,in These are the coordinates of a point in the transformed local coordinate system. It is the inverse of the dynamic pose matrix. To unify the coordinate points in the spatial coordinate system; the local coordinate system is a follower coordinate system with the geometric center of the aforementioned three-dimensional bounding box as the origin and the length, width, and height of the three-dimensional bounding box as the coordinate axes.

[0052] The predefined hyperquadratic surface equations, specifically implicit functions of the hyperquadratic surface, are constructed as follows: (11) in, The equation for the hyperquadratic surface is predefined; Let be the coordinates of a point in the transformed local coordinate system, and , These are the coordinate components of the point in the local coordinate system along each axis. , These are preset superquadratic parameters, for example, all of which are 0.1.

[0053] Substituting the transformed local coordinates and dynamic scale parameters into the preset hyperquadratic surface equation yields a hyperquadratic surface expression with acceleration as the variable; this hyperquadratic surface expression... Specifically as follows: (12) Since both the dynamic pose matrix and the dynamic scale parameter in this expression are functions of acceleration, the surface dynamically changes with instantaneous acceleration; satisfying... The set of all global coordinate points is the parameterized envelope boundary that varies with acceleration.

[0054] like Figure 2 As shown in the figure, a top-down two-dimensional projection is used to compare the static envelope boundary corresponding to the initial geometry of the transport body when it performs transient acceleration in a confined channel with the dynamic parameterized envelope boundary after correction using deformation sensitivity. It can be seen from the figure that after introducing deformation sensitivity, the digital twin model predicts the nonlinear boundary expansion and normal intrusion of the material in the channel normal direction due to the physical inertial force, providing a reliable geometric boundary for the subsequent calculation of collision cost.

[0055] This step calculates the sensitivity of material deformation to the acceleration of the transport body, and based on this, transforms the hyperquadratic surface parameters into a parameterized envelope boundary with acceleration as the variable. This establishes a causal relationship between material deformation and transport body acceleration, and the transient acceleration is mapped to the dynamic evolution of the material's three-dimensional boundary, realizing the coupling between the time-dimensional motion state and the spatial-dimensional geometric boundary. This solves the problem that existing digital twin models cannot reflect the real-time changes in spatial boundaries caused by dynamic inertia, enabling the model to parse the material-occupied boundary based on the underlying control input.

[0056] Step 3: Along the normal of the gap between the envelope boundary and the environment boundary, the gap is mapped to an algebraic constraint using deformation sensitivity, and a continuous collision cost function with respect to acceleration is constructed through non-negative mapping and surface integral. The transport control law is then solved using a digital twin model.

[0057] 3.1 Along the normal of the gap between the envelope boundary and the environmental boundary, the gap is mapped to an algebraic constraint using deformation sensitivity, including: 3.1.1 Calculate the shortest gap between the envelope boundary and the environment boundary, and extract the unit normal vector pointing towards the environment boundary along the direction of the shortest gap, as follows: For the hyperquadratic surface expression corresponding to the envelope boundary, the MarchingCubes algorithm is used to extract triangular meshes and obtain vertex sets on its zero isosurface. Specifically, the hyperquadratic surface expression is used as the algorithm input, and the resolution of spatial voxel partitioning is set (e.g., the mesh step size is set to 0.05m). The algorithm calculates the function values ​​of the eight vertices of each voxel by traversing and uses a lookup table to generate triangular patches at the edges where the positive and negative signs of the function values ​​intersect, thereby discretizing the continuous environmental mathematical function into an envelope vertex set, which serves as the convex set point cloud of the envelope boundary.

[0058] For the implicit functional equations corresponding to the environment boundary, the aforementioned MarchingCubes algorithm is used to extract triangular meshes from its zero isosurface to obtain a discrete initial set of environment vertices. Since the environment boundary usually has non-convex geometric features, the V-HACD (Approximate Convex Decomposition) algorithm is further introduced for geometric preprocessing. The extracted environment boundary triangular mesh model is used as the algorithm input, and the maximum allowable concavity error threshold (e.g., 0.01) and the maximum number of convex hulls generated by decomposition are set (e.g., 64). Based on the principle of hierarchical clustering, the algorithm automatically segments and decomposes the initial set of environment vertices into several independent clusters of environment convex polyhedral point clouds under the condition of satisfying the concavity error threshold.

[0059] The acquired envelope boundary convex set point cloud clusters and the point cloud clusters of each environmental convex polyhedron are used as two sets of spatial convex set vertices, respectively, and input into the Gilbert-Johnson-Keerthi (GJK) distance search algorithm; the distance convergence tolerance of the GJK algorithm is set (e.g., 10). -4 Given the maximum number of iterations (e.g., 100), the closest pair of coordinate points on the surfaces of the two convex sets is calculated by iteratively constructing and searching the nearest simplex that infinitely approximates the origin in the Minkowski difference set of the vertices of the two convex sets. After traversing all convex polyhedral point cloud clusters of the environment boundary and performing the above calculation, the pair of environment boundary points and envelope boundary points with the smallest global distance are selected, and the Euclidean distance between these two points is the shortest gap.

[0060] Subsequently, a direction vector is constructed with the envelope boundary point as the starting point and the environment point as the ending point. This direction vector is then normalized by dividing it by its own Euclidean norm (i.e., the vector length) to obtain a unit normal vector pointing towards the environment boundary along the shortest gap direction.

[0061] 3.1.2 Multiply the deformation sensitivity by the transport volume acceleration to obtain the deformation displacement increment. Project the deformation displacement increment onto the unit normal vector to obtain the normal intrusion, as follows: Since the deformation sensitivity is a 6×3 Jacobian matrix (containing 3D translational and rotational partial derivatives), the first three rows of the 3×3 submatrix corresponding to the translational displacement are extracted as the translational deformation sensitivity. This is then directly multiplied by the 3×1 transport acceleration vector to obtain a 3×1 deformation displacement increment vector. Next, this deformation displacement increment vector is multiplied by the previously extracted 3×1 unit normal vector to project the deformation displacement increment onto the unit normal vector, thus obtaining the scalarized normal intrusion. The formula for calculating the normal intrusion is as follows: (13) in, Normal intrusion quantity, For unit normal vector, This is the 3×3 translational deformation sensitivity submatrix extracted from the deformation sensitivity.

[0062] It should be noted that the above calculation of normal intrusion only extracts the translation component as a first-order linear approximation; since the effects of material attitude deflection and scale expansion have been expressed in step 2.2 through the dynamic parameter expansion of the hyperquadratic surface equation itself, removing attitude and scale derivatives when calculating normal intrusion here can effectively avoid the repeated calculation of deformation geometric increment.

[0063] 3.1.3 Based on the normal intrusion amount, the shortest gap, and the preset safety margin, algebraic constraints are established as follows: During the movement of the transport body, in order to ensure transport safety, the actual remaining gap between the material envelope boundary and the environmental boundary should not be less than the preset safety margin to compensate for the influence of positioning error, measurement error and dynamic uncertainty factors; based on this, the algebraic constraint is established, that is, the difference between the minimum gap and the normal intrusion is always greater than or equal to the preset safety margin; the preset safety margin is set according to the system positioning accuracy, for example, 0.05m.

[0064] 3.2 Based on algebraic constraints, a continuous collision cost function with respect to acceleration is constructed through nonnegative mapping and surface integration, including: To avoid local extremum problems that may arise from optimizing solely based on the single shortest gap in the algebraic constraints, the interference penalty is extended to the entire surface of the envelope boundary. For any global surface element on the envelope surface, a quadratic penalty function is introduced to calculate its local nonnegativity cost, as shown in the following formula: (14) in, For the envelope surface (i.e., satisfying) The global surface infinitesimal point, Let be the local non-negative cost scalar corresponding to the infinitesimal point of the surface. To pre-set a safety margin, This represents the local shortest gap from a point on the surface element to the boundary of the environment. Let be the local unit normal vector pointing from the infinitesimal point of the surface to the boundary of the environment; The quadratic penalty function has a cost of zero when the actual remaining gap is greater than or equal to the preset safety margin. When the actual remaining gap is less than the preset safety margin, the quadratic penalty function causes the cost to increase rapidly with the increase of the intrusion amount, thereby guiding the optimization process away from dangerous solutions.

[0065] It should be noted that the local shortest gap and local unit normal vector in the formula are obtained as follows: The three-dimensional coordinates of the global surface micro-element point are directly substituted into the implicit constraint function of the environment established in step 1.1.2 for evaluation, and the function output value is used as the local shortest gap; the spatial gradient vector of the implicit constraint function of the environment at the micro-element point is calculated, and normalized by dividing it by its own Euclidean norm (i.e., vector length), which is then used as the local unit normal vector; by substituting the above parameters, the formula terms are obtained. This represents the local normal intrusion of the global surface element under acceleration.

[0066] To prevent interference risks on the overall surface of the material, the obtained local non-negative cost is integrally applied to the three-dimensional surface of the envelope boundary to obtain the continuous collision cost function, as shown in the following formula: (15) in, Let be the continuous collision cost function with respect to acceleration; It is an envelope boundary surface, that is, it satisfies The surface set consisting of all global coordinate points; It is a surface area infinitesimal element.

[0067] 3.3 Combining the continuous collision cost function and the digital twin model, the transport control law is solved, including: 3.3.1 The current pose, velocity vector, and material deformation of the transport body are concatenated into an augmented state vector. Based on the digital twin model, the state changes of the augmented state vector within a preset prediction time domain are predicted, as follows: The 8×1 dimensional joint state space vector (containing the transport body pose, velocity vector, and initial three-dimensional envelope size of the material) from step 1 is concatenated with the 6×1 dimensional material deformation vector (composed of relative displacement and attitude deflection) from step 2 in the column vector dimension to obtain a 14×1 dimensional generalized augmented state vector. The state evolution equation of the digital twin model configured in step 1 is simultaneously concatenated with the preset structural dynamics equation from step 2 in the derivative solution dimension to construct a generalized augmented state transition function that drives the evolution of the generalized augmented state vector. Specifically, the first 8 dimensions of the function's derivative output are solved by the state evolution equation, and the last 6 dimensions are solved by the preset structural dynamics equation.

[0068] Based on the response delay time of the underlying actuators of the transporter and the braking physical distance at the maximum operating speed, the preset prediction time domain is set to 20 discrete step lengths, i.e., predicting the state evolution within the next 2 seconds. The optimal control sequence of the previous control cycle is extracted and shifted backward in the time dimension to form the reference control input sequence for the current cycle. Using the generalized augmented state vector as the initial state condition and the reference control input sequence as the driving force, the fourth-order Runge-Kutta numerical integration method is adopted. By calculating the four slope values ​​in each integration step and performing a weighted average, the generalized augmented state transition function is iteratively numerically integrated in the prediction time domain to obtain the future trajectory change sequence of the generalized augmented state vector.

[0069] 3.3.2 The continuous collision cost function is introduced into the objective function of predictive control, and rolling optimization is performed in conjunction with state changes to obtain the transport control law, as follows: The continuous collision cost function is incorporated into the objective function of model predictive control, and the calculation formula of the objective function of model predictive control is as follows: (16) in, The objective function of model predictive control is... To preset the prediction time domain length, To predict the future time interval index number of the walk within the time domain; , The first The L2 norm of the difference between the predicted augmented state vector and the expected reference state vector is the state tracking error accumulation term, which is used to constrain the transport body to closely follow the predetermined trajectory. , The pre-defined diagonal positive definite weight matrix. Desirable To emphasize the weight of location tracking, Desirable To allow for a moderate relaxation of control instructions; To control the input increment, its L2 norm is used as the control increment energy term to ensure the smoothness of the control command; This is the collision avoidance weight coefficient. A higher penalty weight, such as 10, can be assigned to suppress collisions. 4 ; For the first With the prediction step size, the collision cost scalar is calculated by substituting the corresponding instantaneous acceleration into the continuous collision cost function.

[0070] The aforementioned generalized augmented state transition function is used as a dynamic equality constraint in the optimization process. Specifically, at each prediction step, the new state vector generated by the optimizer must follow this joint evolution law, i.e., satisfy the equality constraint: (17) in, For the first Predict the new state vector of the step size. For generalized augmented state transition functions; For the first The control input command vector for predicting the step length includes the traction motor drive torque command at this moment. The system uses this drive torque command to multiply by the chassis mechanical transmission ratio and deduct the nominal driving resistance, then divides it by the total mass of the system identified in step 1 (i.e., the sum of the empty mass of the transport body and the mass of the material), and calculates the corresponding acceleration command forward according to Newton's second law.

[0071] Using the control input command sequence (including control input command vectors with various discrete step lengths) in the future prediction time domain as the optimization decision variable, the Sequence Quadratic Programming (SQP) algorithm is employed for optimization. The control input sequence corresponding to the future trajectory change sequence is extracted as the initial iteration point of the SQP algorithm, i.e., the hot-start point. Under the conditions of satisfying the aforementioned dynamic equality constraints and actuator saturation constraints (i.e., the maximum torque limit of the motor output and the maximum acceleration / deceleration capability limit of the controller), the objective function of the model predictive control is minimized in a rolling manner until the gradient descent norm of the SQP algorithm is less than a preset convergence tolerance (e.g., 10). -5 The optimal control input sequence in the prediction time domain is obtained by reaching the maximum number of iterations (e.g., 50 times). The first element of the optimal control input sequence (e.g., the torque of the chassis traction motor) is extracted as the transport body control law.

[0072] This step maps the spatial gap to an algebraic constraint on acceleration by utilizing deformation sensitivity, and constructs a continuous collision cost function that is incorporated into the predictive control objective function, thereby realizing the algebraic representation of the spatial geometric interference relationship. This enables the system to dynamically characterize the material deformation boundary in narrow channels and actively suppress deformation expansion and avoid collisions in advance.

[0073] Step 4: Extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material mass, and then update the digital twin model.

[0074] 4.1 Structural damping dissipation energy extracted during the execution of the control law, including: 4.1.1 Obtain the instantaneous relative velocity of the material with respect to the transport body when the control law is executed, as follows: The vehicle controller receives the control law of the transport vehicle through the industrial control bus and extracts the driving torque in the control law as the torque reference input. Combined with the real-time rotor position and stator phase current fed back by the motor encoder, the field-oriented control (FOC) algorithm is used to calculate and generate the stator voltage command through the decoupling transformation from the phase coordinate system to the orthogonal coordinate system. The command is then converted into a pulse width modulation (PWM) duty cycle signal to drive the underlying actuator to perform physical acceleration or braking.

[0075] During the execution cycle of this control law, the same Kalman filtering fusion method as in step 1.2.1 is used. The acceleration of the transport vehicle chassis in the global coordinate system is extracted in real time using the onboard inertial measurement unit (IMU) combined with chassis wheel encoder data. This acceleration is substituted into the calculation logic of step 2.1.1, and combined with the material mass and gravitational acceleration vector, the total external excitation is recalculated and placed on the right-hand side of the preset structural dynamics equation. Using the same Newmark-β stepwise integration algorithm, iterative calculations are performed step-by-step within the current control execution time window. Based on the algorithm update formula, the material deformation vector and its first-order velocity term at the current moment are solved sequentially. Furthermore, the first three translational components (i.e., the time derivatives corresponding to the relative displacement vector) in the first-order velocity term obtained by the algorithm are extracted to obtain the instantaneous relative velocity of the material's geometric center relative to the transport vehicle chassis within the current control execution time window.

[0076] 4.1.2 By integrating the damping power using the preset damping coefficient and the instantaneous relative velocity, the structural damping dissipation energy is obtained, as follows: A material vibration attenuation test is conducted in advance, that is, the material is pulsed and vibrated using a force hammer, its free vibration response is recorded and the logarithmic attenuation rate is calculated, and then the preset damping coefficient is derived; or, the value can be directly assigned by consulting the standard engineering material dynamics damping manual. For example, for common corrugated paper packaging materials in warehousing and logistics scenarios, the preset damping coefficient is recommended to be set to 150 N·s / m.

[0077] The instantaneous power is obtained by multiplying the preset damping coefficient by the square of the instantaneous relative velocity modulus, and then by performing a definite integral within the control execution time window to obtain the structural damping dissipation energy.

[0078] 4.2 Based on structural damping dissipation energy, correct the current energy consumption and re-identify material mass to update the digital twin model, including: Using the same method as obtaining the energy consumption of the transport vehicle in step 1.2.2, the instantaneous operating voltage and current of the traction motor are synchronously acquired via high-frequency acquisition through the underlying CAN bus, and the current energy consumption of the transport vehicle is calculated by integration. The current energy consumption is corrected using structural damping dissipation energy: the structural damping dissipation energy is divided by the comprehensive transmission efficiency parameter in step 1.2.1 to perform an equivalent electrical energy conversion, obtaining the equivalent electrical energy; the converted equivalent electrical energy is subtracted from the current energy consumption to obtain the effective traction energy consumption used to drive the rigid chassis. This effective traction energy consumption is fed back into the energy consumption residual objective function established in step 1.2.2, replacing the original actual energy consumption of the transport vehicle, and the residual iteration is performed again, and the material quality is identified, thereby driving the continuous updating of the entity parameters of the digital twin model.

[0079] like Figure 3 and Figure 4 As shown in the figure, the evolution curves of the kinematic state variables of the transporter are given within a typical control execution window. Figure 3 The timing curve of the underlying power consumption () Figure 4 ); Figure 3 The data evolution confirms that during the acceleration and deceleration of the transporter, the actual acceleration of the chassis will induce a significant instantaneous relative velocity (i.e., flexible sway) in the material. Figure 4 This visually presents the real-time distribution of structural damping dissipation power and effective traction power induced by the aforementioned swaying. Figure 4 The areas of each colored region below the curve (i.e., the definite integral of power within the time window) correspond to the effective traction energy consumption and structural damping dissipation energy, respectively; through the analysis of... Figure 4 By integrating and stripping the damping dissipation region in the chassis, the corresponding energy consumption is corrected in real time. This effectively avoids miscalculating it as the effective traction power of the chassis, providing a key data decoupling basis for high-fidelity material quality identification based on effective traction energy consumption.

[0080] It should be noted that the continuous updating of material mass is not only used to correct a single parameter, but also to gradually improve the consistency between the digital twin model and the actual physical state through closed-loop feedback. Specifically, the material mass re-identified after correction by structural damping dissipation energy is written back into the digital twin model. When calculating inertial loads, solving for material deformation states, and updating the envelope boundary based on this mass, the results are closer to the actual physical process. As the prediction error of the envelope boundary decreases, the judgment of collision risk by the prediction control layer becomes more reliable, thereby reducing the excessive expansion of the envelope boundary caused by the overestimation of material mass and reducing the resulting false shutdowns or excessive braking.

[0081] This step deducts the structural damping dissipation energy caused by material deformation from the total energy consumption of the transporter, thus preventing this part of the energy from being mistakenly included in the traction load. This reduces the risk of overestimating the material quality, thereby reducing the phenomenon of excessive expansion of the envelope boundary and false triggering of safety shutdown in the digital twin model, and improving the accuracy of entity parameter estimation in the digital twin model.

[0082] This embodiment combines environmental boundary construction, material quality identification, deformation envelope update, collision cost construction, predictive control solution, and energy consumption-based decoupling to form a closed-loop process of "modeling-control-update". This enables the digital twin model to be dynamically updated with the actual transportation process, improves the consistency between the model and the physical entity, and helps to reduce the collision risk caused by dynamic deformation of materials.

[0083] Example 2, Figure 5 A digital twin modeling system for the spatiotemporal coupling of material storage channels and transport systems is presented, including the following modules: The twin modeling module is used to establish environmental boundaries based on the geometric data of the warehouse and the channel, and to identify the material quality by combining the movement trajectory of the transport vehicle and energy consumption, thereby constructing a digital twin model that includes both environmental boundaries and material quality. The envelope update module is used to calculate the material deformation and its deformation sensitivity to the acceleration of the transport body based on the material mass, correct the hyperquadratic surface parameters of the initial geometry of the material in the model, and obtain the envelope boundary with acceleration as the variable. The control solution module is used to map the gap between the envelope boundary and the environmental boundary into an algebraic constraint using deformation sensitivity, and construct a continuous collision cost function with respect to acceleration through nonnegative mapping and surface integral, and solve the transport body control law in combination with the model. The closed-loop update module is used to extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material quality, and then update the digital twin model.

[0084] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0085] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0086] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0087] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0088] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0089] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital twin modeling method for spatiotemporal coupling of material storage channels and transport bodies, characterized in that, Includes the following steps: Environmental boundaries are established based on the geometric data of storage locations and channels. Material quality is identified by combining the movement trajectory of the transport vehicle and energy consumption. A digital twin model including environmental boundaries and material quality is constructed. Based on the material mass, the material deformation and its deformation sensitivity to the acceleration of the transport body are calculated. The hyperquadratic surface parameters of the initial geometry of the material in the model are corrected to obtain the envelope boundary with acceleration as the variable. Along the normal of the gap between the envelope boundary and the environmental boundary, the gap is mapped to an algebraic constraint using deformation sensitivity, and a continuous collision cost function with respect to acceleration is constructed through nonnegative mapping and surface integral. The transport body control law is then solved in combination with the model. Extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material quality, and then update the digital twin model.

2. The method according to claim 1, characterized in that, The process of obtaining the envelope boundary with acceleration as the variable includes: Using the initial spatial pose matrix and scale parameter in the hyperquadratic surface parameters as reference values, and performing a first-order Taylor expansion with respect to acceleration based on deformation sensitivity, the homogeneous transformation matrix correction and scale correction are obtained. The initial spatial pose matrix and scale parameters are corrected by using the homogeneous transformation matrix correction and scale correction, respectively, and then substituted into the preset hyperquadratic surface equation to generate a hyperquadratic surface expression with acceleration as the variable, which serves as the envelope boundary.

3. The method according to claim 1, characterized in that, The method of mapping the gap to an algebraic constraint using deformation sensitivity includes: Calculate the shortest gap between the envelope boundary and the environment boundary, and extract the unit normal vector pointing to the environment boundary along the direction of the shortest gap; The deformation sensitivity is multiplied by the transport body acceleration to obtain the deformation displacement increment, and the deformation displacement increment is projected onto the unit normal vector to obtain the normal intrusion amount; The algebraic constraints are established based on the normal intrusion amount, the shortest gap, and the preset safety margin.

4. The method according to claim 1, characterized in that, The process of solving the transport control law using the aforementioned model includes: The current pose of the transport body, the velocity vector, and the material deformation are concatenated into an augmented state vector. Based on the digital twin model, predict the state changes of the augmented state vector within a preset prediction time domain; The continuous collision cost function is introduced into the objective function of predictive control, and rolling optimization is performed in conjunction with the state changes to obtain the transport control law.

5. The method according to claim 1, characterized in that, The structural damping dissipation energy during the extraction of the control law includes: Obtain the instantaneous relative velocity of the material with respect to the transport body when the control law is executed; The damping power is integrated by combining the preset damping coefficient and the instantaneous relative velocity to obtain the structural damping dissipation energy.

6. The method according to claim 1, characterized in that, The method of identifying material quality by combining the movement trajectory of the transport vehicle and energy consumption includes: Based on the nominal mass and nominal resistance parameters of the transport body under no-load conditions, a nominal power equation characterizing the mapping relationship between driving power and motion state is established, and the theoretical energy consumption is obtained by integrating the velocity and acceleration in the motion trajectory. The gradient of the residual between the theoretical energy consumption and the energy consumption with respect to the mass of the material to be identified is calculated, and the estimated mass of the material is iteratively updated along the negative direction of the gradient to obtain the mass of the material.

7. The method according to claim 1, characterized in that, The material deformation includes the relative displacement and attitude deflection of the material relative to the transport body, and the calculation of material deformation based on material mass includes: The inertial load on the material is calculated based on the material mass and the acceleration of the transport body, and together with the material gravity, it is used as an external excitation input to the preset structural dynamics equation to solve for the relative displacement and attitude deflection.

8. The method according to claim 1, characterized in that, The establishment of environmental boundaries based on position and channel geometric data includes: From the point cloud data of the warehouse and the passage, extract spatial geometric features including the depth of the warehouse, the net clearance boundary of the passage, and the contour of the side wall; The spatial geometric features are fitted with surfaces and Boolean operations to generate implicit environmental constraint functions, which serve as the environmental boundaries.

9. The method according to claim 1, characterized in that, The construction of the digital twin model, which includes environmental boundaries and material quality, includes: Under a pre-defined unified spatial coordinate system, the environmental boundary is used as the environmental boundary constraint, the material mass is used as the entity parameter, and the initial motion state of the transport body and the initial geometry of the material are used as entity initialization parameters to construct a joint state space and its state evolution equation, which serves as the digital twin model.

10. A digital twin modeling system for spatiotemporal coupling of material storage channels and transport bodies, characterized in that, Includes the following modules: The twin modeling module is used to establish environmental boundaries based on the geometric data of the warehouse and the channel, and to identify the material quality by combining the movement trajectory of the transport vehicle and energy consumption, thereby constructing a digital twin model that includes both environmental boundaries and material quality. The envelope update module is used to calculate the material deformation and its deformation sensitivity to the acceleration of the transport body based on the material mass, correct the hyperquadratic surface parameters of the initial geometry of the material in the model, and obtain the envelope boundary with acceleration as the variable. The control solution module is used to map the gap between the envelope boundary and the environmental boundary into an algebraic constraint using deformation sensitivity, and construct a continuous collision cost function with respect to acceleration through nonnegative mapping and surface integral, and solve the transport body control law in combination with the model. The closed-loop update module is used to extract the structural damping dissipation energy when executing the control law, correct the current energy consumption, re-identify the material quality, and then update the digital twin model.

Images