Wafer carrying speed control method and control system

Abstract

The invention discloses a wafer carrying speed control method and system, and the method comprises the steps: obtaining the three-dimensional data of a carrying path, decomposing the data into a straight line and a bent segment, and obtaining geometric features; the curvature of the bent section is calculated and smoothly connected, and continuous curvature distribution is determined; dynamically constraining the speed upper limit of each point according to the curvature, and generating a preliminary speed curve; the speed of each axis is calculated through inverse kinematics, iterative optimization is carried out until the speed of each axis is coordinated, and a multi-axis cooperative speed sequence is generated; calculating attitude deviation based on real-time sensor data fusion; and dynamically adjusting the acceleration parameter, optimizing and reducing the deviation, and obtaining a final control instruction. According to the control method, through curvature dynamic planning, multi-axis collaborative optimization and real-time posture self-adaptive adjustment, synchronous improvement of the carrying speed and stability is systematically achieved.

Description

Technical Field

[0001] This invention relates to the field of wafer handling control technology, and in particular to a wafer handling speed control method and control system. Background Technology

[0002] In precision manufacturing, especially in semiconductor wafer handling, high-speed, high-precision operation of automated equipment is crucial for improving production line efficiency. With increasing wafer size and shorter process cycles, handling systems not only require faster movement speeds to reduce non-productive time but also must ensure extremely high stability during movement to prevent damage or contamination of brittle wafers due to vibration, slippage, or instability. However, increased movement speed is often accompanied by increased vibration and amplified dynamic errors, creating an inherent and irreconcilable contradiction between "high speed" and "high stability," which has become the most prominent challenge facing the industry today.

[0003] Existing wafer handling technologies have significant limitations in multi-axis motion control. Many traditional methods focus on motion optimization in a single direction, neglecting the coordination of multi-axis motion in three-dimensional space. When dealing with complex paths, this approach often fails to effectively balance speed changes in each direction. For example, when wafers are turning rapidly or moving vertically, the lack of coordinated control over multi-axis speeds can lead to sudden speed changes at certain critical points. These sudden changes not only increase the unevenness of the forces acting on the wafers but may also cause additional wear on the mechanical components of the equipment, reducing the long-term stability of the system.

[0004] How to achieve coordinated control of multi-axis speed in complex three-dimensional paths, and optimize speed adjustment according to trajectory curvature in curved sections, has become a key problem that wafer handling technology urgently needs to solve. Existing solutions have failed to coordinate and optimize motion paths, multi-axis dynamics and real-time status feedback at the system level, resulting in the equipment performance always being unable to break through the technical bottleneck between speed and stability. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the defect in the prior art that the wafer handling efficiency and reliability are difficult to balance due to the inherent contradiction between high speed and high stability. The present invention provides a wafer handling speed control method that can systematically improve the handling speed and operational stability at the same time through dynamic speed planning based on path curvature, multi-axis speed collaborative iterative optimization and adaptive adjustment based on real-time attitude feedback.

[0006] To address the aforementioned technical problems, this invention provides a wafer handling speed control method, comprising the following steps: The three-dimensional coordinate data of the object's transport path is obtained, and the path is divided into straight segments and curved segments using a preset path decomposition algorithm to obtain the geometric feature description of each segment. Based on the obtained geometric feature description, the trajectory curvature value is calculated for the curved segment, the straight segment and the curved segment are smoothly connected, and the continuous path curvature distribution is determined. Based on the path curvature distribution, a preliminary speed planning curve is generated by dynamically constraining the upper limit of the speed of the path points according to the curvature value. Based on the obtained preliminary velocity planning curve, the joint parameters of multi-axis motion are obtained, the velocity vector of each axis is calculated by the inverse kinematics algorithm, and iterative optimization is performed to ensure that the difference between the velocity vectors of each axis is within the coordination threshold, thereby generating a coordinated multi-axis velocity sequence. From the determined coordinated multi-axis velocity sequence, to meet the attitude stabilization requirements, the object tilt angle fed back by the real-time sensor is obtained, and the position and angle data are fused by Kalman filtering to obtain the attitude deviation estimate. Based on the attitude deviation estimate, the acceleration parameters in the velocity sequence are dynamically adjusted, and an optimization algorithm is used to reduce the attitude deviation, thereby determining the optimized velocity control command.

[0007] In one embodiment of the present invention, three-dimensional coordinate data of the object transport path is obtained, and the path is divided into straight segments and curved segments using a preset path decomposition algorithm to obtain a geometric feature description of each segment, including: Obtain the three-dimensional coordinate data of the object's transport path, collect spatial trajectory points from sensors or positioning systems, and obtain the original three-dimensional coordinate sequence; The three-dimensional coordinate sequence is processed by a preset path segmentation algorithm to obtain the coordinate sets of straight line segments and curve segments. For a set of coordinates of line segments, calculate the length and direction vector of each segment to obtain a geometric feature description of the line segments; Based on the set of curve segment coordinates, a spline interpolation algorithm is used to fit the curve trajectory, calculate the curvature and tangent vector, and obtain the geometric feature description of the curve segment. The changes in motion direction are extracted from the geometric features of straight and curved segments, and the angles and distances between path points are analyzed to obtain the dynamic feature sequence of the path. By using dynamic feature sequences, the K-means clustering algorithm is used to classify path segments, determine the spatial trajectory characteristics of each segment, and obtain a set of classified path segments. Based on the classified set of path segments, the geometric features of straight line segments and curved segments are integrated to generate a complete description of the path's geometric features.

[0008] In one embodiment of the present invention, based on the obtained geometric feature description, the trajectory curvature value is calculated for the curved segment, the straight segment and the curved segment are smoothly connected, and a continuous path curvature distribution is determined, including: Based on the geometric feature description of the curve segment, the curvature continuity constraint algorithm is used to optimize the arrangement of control points and generate a fitting curve with smooth curvature transition. Based on the geometric feature description of the straight line segment, the rate of curvature change at the connection point between the straight line and the curve is calculated. Based on the rate of curvature change, a Bézier curve splicing algorithm is used to generate a smooth transition segment at the connection point between the straight line segment and the curve segment. Based on the fitted curve and smooth transition segment of curvature smooth transition, calculate the curvature values ​​of continuous sampling points on the path to generate a discrete curvature distribution sequence; The discrete curvature distribution sequence is smoothed using a Gaussian filtering algorithm to eliminate local curvature fluctuations and obtain a continuous path curvature distribution.

[0009] In one embodiment of the present invention, the smoothing process of the discrete curvature distribution sequence using a Gaussian filtering algorithm includes: Set the Gaussian filter window size and standard deviation parameters, and adaptively adjust the filter strength based on the total path length and curvature change frequency; Convolution operations are used to filter discrete curvature sequences, eliminating local curvature fluctuations while preserving global curvature variation characteristics. The continuity of the filtered curvature distribution is checked to ensure that the first and second derivatives of the curvature distribution curve are continuous.

[0010] In one embodiment of the present invention, based on the path curvature distribution, a preliminary velocity planning curve is generated by dynamically constraining the upper limit of the velocity of path points according to the curvature value, including: The curvature values ​​of each path point are obtained from the curvature distribution of the path. A preset threshold is used to determine the curvature magnitude. If the curvature value is greater than the preset threshold, the set of speed limits is determined according to the mapping relationship between the curvature value and the speed limit. By combining the set of upper speed limits with the coordinate sequence of the path point set, the upper speed limit curve is fitted using spline interpolation to obtain the preliminary speed planning curve.

[0011] In one embodiment of the present invention, for the obtained preliminary velocity planning curve, joint parameters of multi-axis motion are acquired, velocity vectors of each axis are calculated using an inverse kinematics algorithm, and iterative optimization is performed to ensure that the differences between the velocity vectors of each axis are within a coordination threshold, thereby generating a coordinated multi-axis velocity sequence, including: Obtain the preliminary velocity planning curve, extract the joint parameters of multi-axis motion from it, and determine the initial motion state of each axis; Using the inverse kinematics algorithm, the joint velocity vectors of each axis are calculated for the extracted joint parameters to obtain the initial velocity vector set; Based on the initial set of velocity vectors, the joint velocity vectors of each axis are iteratively adjusted using a particle swarm optimization algorithm, with the goal of minimizing the differences between the velocity vectors of each axis, thereby obtaining an optimized set of velocity vectors. By using the optimized set of velocity vectors, a coordinated multi-axis velocity sequence is generated, and the motion trajectory of each axis is determined.

[0012] In one embodiment of the present invention, from a determined coordinated multi-axis velocity sequence, to meet attitude stabilization requirements, the object tilt angle fed back by real-time sensors is obtained, and the position and angle data are fused using Kalman filtering to obtain an attitude deviation estimate, including: Coordinated control commands are obtained from the multi-axis velocity sequence, and the velocities of each axis are synchronized using a preset motion planning algorithm to obtain a unified set of motion commands. Based on the unified motion command set and combined with sensor feedback data, the object tilt angle is collected in real time using accelerometers and gyroscopes to obtain the original attitude data. By fusing the position and angle data from the original attitude data using Kalman filtering, the state estimation is iteratively updated to obtain the optimized attitude deviation estimate.

[0013] In one embodiment of the present invention, based on the attitude deviation estimate, the acceleration parameters in the velocity sequence are dynamically adjusted, and an optimization algorithm is used to reduce the attitude deviation, thereby determining an optimized velocity control command, including: Attitude deviation estimates and velocity sequences are obtained from sensor data. The attitude deviation estimates are classified using a preset threshold. If the attitude deviation estimates are greater than the threshold, they are smoothed using a Kalman filter algorithm to obtain smoothed deviation data. Based on the deviation data, the acceleration parameters in the velocity sequence are extracted, and logical judgment is used to determine whether the acceleration parameters exceed the preset range. If they do, the acceleration parameters are adjusted by the gradient descent algorithm to obtain optimized acceleration parameters. Using the acceleration parameters and real-time feedback sensor data, a dynamic adjustment mechanism is employed to update the velocity sequence, generating an adjusted velocity sequence. Based on the speed sequence, logical judgments are used to generate control commands, resulting in optimized speed control commands.

[0014] In one embodiment of the present invention, the method further includes: performing global speed optimization on the speed control command based on the target transport time, and adjusting the speed parameters of multiple path segments in parallel to meet the time constraints, thereby determining the final speed control command to be executed.

[0015] The technical solution of the present invention has the following advantages compared with the prior art: The wafer handling speed control method described in this invention constructs a closed-loop optimization system from global path planning to local parameter adaptive adjustment. It can systematically improve the handling speed and operational stability simultaneously through dynamic speed planning based on path curvature, multi-axis speed collaborative iterative optimization, and adaptive adjustment based on real-time attitude feedback.

[0016] This method first plans the motion at its source. By decomposing the path and calculating the continuous curvature distribution, a speed baseline that strictly matches the geometry is generated, ensuring automatic speed reduction on high-curvature sections and mitigating the risk of instability during cornering from the outset. Furthermore, the method of this invention deeply understands the cooperative nature of multi-axis mechanical systems. Through inverse kinematics calculation and iterative optimization, the smooth trajectory at the end point is transformed into highly matched speed commands between each drive shaft, eliminating the inherent torsional vibration caused by incoordination between shaft movements and providing a solid dynamic foundation for high-speed operation.

[0017] This method introduces an adaptive control layer based on real-time sensor feedback. By fusing multi-source data through a Kalman filter, the system can accurately estimate dynamic attitude deviations that cannot be predicted by a model. Based on this real-time estimate, instead of relying on fixed control parameters, the system dynamically adjusts acceleration commands to actively counteract attitude errors. This means that the control process has the ability to intelligently respond to uncertain disturbances, thus maintaining extreme stability even under complex operating conditions. The entire technical solution is progressively advanced, from offline path velocity planning to online multi-axis coordination and adaptive adjustment, forming an organic whole that balances efficiency and stability.

[0018] Therefore, the beneficial effects of this method are as follows: curvature-based dynamic speed planning lays the foundation for smooth motion, avoiding speed shocks at sharp bends; multi-axis speed collaborative optimization eliminates vibration-inducing factors from a mechanical perspective; and real-time attitude feedback and adaptive adjustment endow the system with strong robustness against disturbances, ensuring precise control at high speeds. These beneficial effects and technical features complement each other, significantly improving handling speed and shortening the work cycle without sacrificing stability. On the contrary, they achieve higher dynamic accuracy and reliability, ultimately achieving the dual goals of improving production efficiency and ensuring product yield. Attached Figure Description

[0019] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of the wafer handling speed control method of the present invention; Figure 2 This is a flowchart of the steps for path acquisition and decomposition in this invention; Figure 3 This is a flowchart of the curvature calculation and smoothing process of the present invention; Figure 4 This is a flowchart of the preliminary speed planning steps of the present invention; Figure 5 This is a flowchart of the steps involved in the multi-axis coordination optimization of this invention; Figure 6 This is a flowchart of the attitude monitoring and feedback steps of the present invention; Figure 7 This is a flowchart of the steps for generating control commands in this invention; Figure 8 This is a structural framework diagram of the wafer handling speed control system of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0021] Reference Figure 1 As shown, this application proposes a wafer handling speed control method, including the following steps: The process involves acquiring 3D coordinate data of the object's transport path, dividing it into straight and curved segments using a pre-defined path decomposition algorithm, and obtaining geometric feature descriptions for each segment. Based on these descriptions, the trajectory curvature value is calculated for the curved segments, and the straight and curved segments are smoothly connected to determine a continuous path curvature distribution. Based on this curvature distribution, the upper limit of the path points' speed is dynamically constrained according to the curvature values, generating a preliminary speed planning curve. For this preliminary speed planning curve, the joint parameters of the multi-axis motion are acquired, and the velocity vectors of each axis are calculated using an inverse kinematics algorithm. These vectors are then iteratively optimized to ensure that the differences between the velocity vectors are within a coordination threshold, thus generating a coordinated multi-axis velocity sequence. From this coordinated multi-axis velocity sequence, and considering attitude stability requirements, the object's tilt angle is obtained from real-time sensor feedback. The position and angle data are fused using Kalman filtering to obtain an attitude deviation estimate. Based on this attitude deviation estimate, the acceleration parameters in the velocity sequence are dynamically adjusted, and an optimization algorithm is used to reduce the attitude deviation, thereby determining the optimized speed control command.

[0022] The path decomposition algorithm is a calculation method that divides a 3D trajectory into straight and curved segments. Specifically, it can be implemented using curvature-based clustering analysis to extract the geometric features of the path. The path curvature distribution is a continuous data sequence describing the degree of trajectory curvature, which can be generated using spline interpolation and Gaussian filtering to constrain the upper limit of velocity. The inverse kinematics algorithm is a mathematical method that maps the end effector velocity to joint space, which can be implemented using the inversion of the Jacobian matrix to calculate the velocity vectors of each axis. The coordination threshold refers to the allowable range of multi-axis velocity differences, which can be set according to the dynamic characteristics of the mechanical system to ensure motion coordination. The attitude deviation estimate is the optimized calculation result of the wafer tilt angle, which can be obtained by fusing multi-sensor data using Kalman filtering for feedback control adjustment.

[0023] Specifically, this method first performs geometric analysis on the transport path, decomposing the complex trajectory into straight and curved segments to establish a structured path feature model. A curvature continuity constraint algorithm is then used to process path connections, generating smooth curvature distribution curves to provide dynamic constraints for velocity planning. When generating the initial velocity curve, the upper limit of the velocity is dynamically adjusted based on the curvature values ​​at each point to prevent wafer slippage caused by excessive centrifugal force. Subsequently, the path velocity is converted into multi-axis joint velocity, and an iterative optimization algorithm is used to eliminate velocity vector differences, ensuring the synchronization of motion across axes. During the motion execution phase, attitude data is acquired in real time, and a filtering algorithm is used to eliminate measurement noise, accurately estimating the wafer tilt state. Finally, the acceleration parameters are dynamically adjusted based on the attitude deviation, forming a closed-loop adjustment mechanism for velocity control and attitude stability.

[0024] Compared to existing technologies, traditional methods use fixed velocity curves when handling complex paths, which cannot adapt to the differences in dynamic constraints caused by curvature variations. Existing multi-axis coordinated control is mostly based on open-loop calculations, lacking proactive optimization for velocity vector differences. This scheme achieves dynamic matching of velocity planning and path characteristics by combining path decomposition and curvature analysis; it employs a multi-axis optimization algorithm with a coordination threshold to effectively suppress abrupt changes in joint velocity; and it introduces a real-time attitude feedback adjustment mechanism to overcome the limitations of traditional methods that rely on preset parameters.

[0025] Through the above technical solution, this application achieves coordinated control of multi-axis velocities in a three-dimensional path, automatically adjusting motion parameters according to curvature changes in curved sections, and dynamically correcting acceleration commands through real-time feedback. This method significantly reduces the risk of sudden changes in wafer stress and attitude instability while ensuring handling efficiency, resolving the contradiction between high-speed motion and stability.

[0026] Reference Figure 2As shown, this application further proposes a specific implementation method for obtaining three-dimensional coordinate data of an object transport path, and dividing it into straight segments and curved segments using a preset path decomposition algorithm to obtain geometric feature descriptions of each segment. The method includes: obtaining three-dimensional coordinate data of the object transport path; collecting spatial trajectory points from sensors or positioning systems to obtain an original three-dimensional coordinate sequence; processing the three-dimensional coordinate sequence using a preset path segmentation algorithm to obtain a set of coordinates for straight segments and a set of coordinates for curved segments; calculating the length and direction vector of each segment for the set of coordinates for straight segments to obtain a geometric feature description of the straight segments; fitting the curve trajectory using a spline interpolation algorithm based on the set of coordinates for curved segments, calculating the curvature and tangent vector to obtain a geometric feature description of the curved segments; extracting changes in motion direction from the geometric feature descriptions of the straight and curved segments, analyzing the angles and distances between path points to obtain a dynamic feature sequence of the path; classifying the path segments using a K-means clustering algorithm based on the dynamic feature sequence, determining the spatial trajectory characteristics of each segment, and obtaining a set of classified path segments; and integrating the geometric features of the straight and curved segments based on the set of classified path segments to generate a complete geometric feature description of the path.

[0027] Among them, path segmentation algorithms refer to calculation methods that automatically divide a three-dimensional coordinate sequence into straight line segments and curved segment segments. Specifically, this can be implemented using judgment rules based on curvature thresholds or angle changes to eliminate subjective errors from manual segmentation. Spline interpolation algorithms are mathematical methods that generate continuous curves by fitting discrete trajectory points with polynomial functions. Specifically, this can be implemented using cubic spline interpolation or B-spline interpolation to accurately describe the geometric characteristics of curved paths. K-means clustering algorithms are unsupervised machine learning methods that automatically classify data based on features. Specifically, this can be implemented using Euclidean distance to measure the dynamic feature differences of trajectory segments to identify path segments with different spatial trajectory characteristics.

[0028] Specifically, the original 3D coordinate sequence is acquired through high-precision sensors, providing the basic data for path decomposition. The path segmentation algorithm automatically processes the coordinate sequence, classifying points that meet the straight-line determination criteria into straight-line segments and the rest into curve segments. For the set of straight-line segments, vector operations are used to calculate the length and direction of each segment, forming quantized parameters; for the set of curve segments, spline interpolation is used to generate continuous trajectories, and curvature and tangent vectors are calculated. The dynamic feature sequence is constructed by analyzing the directional changes of adjacent path points, reflecting the turning characteristics of the motion trajectory. The K-means clustering algorithm divides the path segments into different categories based on the dynamic features, automatically identifying the distribution patterns of straight and curve segments. Finally, the geometric feature parameters of each segment are integrated to form a complete path description model, providing structured input for subsequent speed control.

[0029] Compared to existing technologies, traditional methods rely on manual experience to segment path segments, which can easily introduce subjective errors leading to sudden speed changes. This solution achieves accurate path segment identification through automated path segmentation and feature extraction, combined with machine learning classification algorithms. Existing technologies often use simple circular arc approximations to describe the geometry of curve segments, while this solution employs spline interpolation for precise fitting, significantly improving the accuracy of curvature calculation for curved segments. Furthermore, the construction and cluster analysis of dynamic feature sequences overcome the limitation of traditional methods in ignoring changes in motion direction.

[0030] Through the above technical solutions, this application achieves refined decomposition and geometric feature extraction of 3D paths, effectively avoiding the problem of sudden changes in motion speed caused by inaccurate path segmentation. An automated straight-segment and curved-segment recognition mechanism improves the objectivity and accuracy of path classification. Precise curvature and tangent vector calculations provide a reliable geometric parameter basis for subsequent speed optimization, and dynamic feature clustering analysis ensures the rationality of path segment classification, thereby supporting the coordinated optimization of multi-axis motion control systems.

[0031] Reference Figure 3 As shown, this application further proposes a method for achieving a smooth, abrupt connection between straight and curved segments after path decomposition and eliminating local fluctuations in curvature distribution. The method includes the following steps: based on the geometric feature description of the curved segments, a curvature continuity constraint algorithm is used to optimize the arrangement of control points and generate a fitted curve for smooth curvature transition; based on the geometric feature description of the straight segments, the rate of curvature change at the connection point between the straight and curved segments is calculated; based on the rate of curvature change, a Bezier curve splicing algorithm is used to generate a smooth transition segment at the connection point between the straight and curved segments; based on the fitted curve and smooth transition segment, the curvature values ​​of continuous sampling points on the path are calculated to generate a discrete curvature distribution sequence; the discrete curvature distribution sequence is smoothed using a Gaussian filtering algorithm to eliminate local curvature fluctuations and obtain a continuous path curvature distribution.

[0032] The curvature continuity constraint algorithm ensures the continuity of curvature values ​​and their derivatives at connection points between adjacent curve segments through mathematical constraints. Specifically, it can be achieved by combining cubic spline interpolation with curvature differential equations to solve for control point coordinates, thus eliminating abrupt curvature changes at segmental connections in the fitted curve. The Bézier curve splicing algorithm utilizes the endpoint tangent vector control characteristics of Bézier curves to insert transition curves between straight and curved segments. This is achieved by adjusting the control point weights to match the rate of curvature change of the transition segment with the path segments on both sides, resulting in a smooth transition in geometry and motion parameters. The discrete curvature distribution sequence refers to the set of curvature values ​​obtained by sampling along the path at fixed intervals. This is achieved by calculating the curvature of each sampling point using numerical differentiation methods, reflecting the local geometric characteristics of the path. The Gaussian filtering algorithm uses a Gaussian function as weights to perform a weighted average of the discrete curvature sequence. This is achieved by adaptively adjusting the filter window width and standard deviation parameters, suppressing high-frequency noise while preserving the global trend characteristics of the curvature distribution.

[0033] Specifically, firstly, the control points of the curve segments are optimized using a curvature continuity constraint algorithm. This ensures that adjacent curve segments not only have positional continuity at connection points but also that the first derivative of curvature is continuous, thus eliminating curvature jumps common in traditional interpolation methods. Furthermore, in the boundary region between straight and curved segments, the position and weight of the Bézier curve control points are dynamically adjusted by calculating the rate of curvature change at the connection points, generating transition segments with continuously changing curvature derivatives to avoid abrupt acceleration changes caused by path geometric discontinuities. Subsequently, discrete sampling is performed along the optimized complete path with a fixed step size. The curvature value of each sampling point is calculated using numerical differentiation, forming a discrete sequence containing all local features of the path. Finally, Gaussian filtering parameters are adaptively selected based on the total path length and curvature change frequency. Convolution operations are performed on the discrete curvature sequence to eliminate local fluctuations while maintaining the macroscopic continuity of the curvature distribution. A derivative continuity check ensures that the filtered curvature curve meets the dynamic constraints.

[0034] Compared to existing technologies, traditional path smoothing methods typically focus only on positional continuity while neglecting curvature continuity. For example, when using linear interpolation to connect straight and curved segments, although the path geometry is continuous, abrupt curvature changes at the connection points can cause vibrations in the transport equipment. This solution, however, achieves dual continuity of curvature and its derivative at geometrical connections through a curvature continuity constraint algorithm and a Bézier curve splicing algorithm, fundamentally eliminating abrupt changes in motion parameters. Furthermore, existing curvature filtering methods often employ fixed-parameter mean filtering, which can easily lead to overly smoothed curvature distributions. This solution, however, uses adaptive Gaussian filtering, dynamically adjusting the filtering intensity based on path characteristics, effectively suppressing local noise while fully preserving the global characteristics of curvature changes.

[0035] Through the above technical solution, this application effectively solves the problem of abrupt curvature changes at the connection between straight and curved segments after path decomposition. By generating smooth transition segments with continuously changing curvature, sudden speed changes of the handling equipment at path turning points are avoided. Simultaneously, adaptive Gaussian filtering is used to process the discrete curvature sequence, eliminating local fluctuation interference and forming a continuous curvature distribution that conforms to dynamic constraints. This provides a stable and reliable curvature constraint basis for the subsequent speed planning module, thereby reducing vibration risks during wafer handling and improving motion stability.

[0036] In this embodiment, the discrete curvature distribution sequence is smoothed using a Gaussian filtering algorithm, including setting the Gaussian filter window size and standard deviation parameters, and adaptively adjusting the filtering intensity based on the total path length and curvature change frequency; convolution operation is used to filter the discrete curvature sequence to eliminate local curvature fluctuations while retaining global curvature change characteristics; the continuity of the filtered curvature distribution is checked to ensure that the first and second derivatives of the curvature distribution curve are continuous.

[0037] The Gaussian filter window size refers to the neighborhood range used to calculate the weighted average of each curvature point. This can be implemented using a sliding window approach, with the window size dynamically adjusted proportionally to the total path length. For example, the window size expands as the total path length increases, adapting to curvature variations at different scales. The standard deviation parameter controls the weight distribution of the Gaussian kernel function. It can be adaptively matched based on the frequency of curvature changes; increasing the standard deviation enhances smoothing when curvature fluctuates at high frequencies, and decreasing it when curvature fluctuates at low frequencies to preserve detail features. The convolution operation involves mathematically convolving the Gaussian kernel function with the discrete curvature sequence. This can be accelerated using discrete Fourier transform, achieving a balance between suppressing local noise and preserving the overall trend. The continuity check involves differentiating the filtered curvature distribution. Numerical differentiation algorithms can be used to verify the continuity of the first derivative and the absence of abrupt changes in the second derivative, ensuring the curvature curve satisfies kinematic constraints.

[0038] Specifically, in regions with long paths and gentle curvature changes, increasing the filter window size and decreasing the standard deviation parameter can eliminate minor fluctuations while maintaining the overall curvature trend. In regions with dense path curvature, decreasing the window size and increasing the standard deviation parameter can effectively suppress high-frequency noise interference. During convolution, the symmetric weight distribution of the Gaussian kernel function ensures that abrupt changes in the curvature sequence are weighted and averaged by the surrounding data, while preserving the global curvature variation amplitude. After filtering, the slope continuity is verified by calculating the first derivative of the curvature distribution, and the existence of step points in the second derivative is checked. If discontinuities are detected, the filter parameters are readjusted for iterative optimization until the curvature smoothness requirements of motion planning are met.

[0039] Compared to existing technologies, traditional methods typically employ fixed-parameter mean or median filtering to process curvature data. This fails to dynamically adjust the filtering intensity based on path geometry, leading to over-smoothing that results in the loss of key curvature features or residual local noise. This proposed solution utilizes a dual-factor driven parameter adaptive mechanism based on both total path length and curvature change frequency. This ensures precise matching of filtering intensity to path morphology. Combined with the mathematical properties of convolution operations, it eliminates local fluctuations while maintaining the accuracy of the global curvature distribution. Furthermore, by verifying differential continuity, the physical realizability of the curvature curve is validated from a kinematic perspective, resolving the acceleration abruptness problem caused by neglecting derivative continuity in traditional methods.

[0040] Through the above technical solution, this application can effectively eliminate local fluctuations in the path curvature distribution, ensure a smooth transition of curvature changes in the spatial and temporal domains, avoid drastic fluctuations in speed commands during wafer handling caused by curvature abrupt changes, thereby improving the stability of multi-axis coordinated motion and reducing the risk of inertial shock to the wafer caused by acceleration abrupt changes.

[0041] Reference Figure 4 As shown, this application further proposes a method based on path curvature distribution, dynamically constraining the speed upper limit of path points according to curvature values, and generating a preliminary speed planning curve. This includes obtaining the curvature values ​​of each path point from the path curvature distribution, judging the curvature magnitude using a preset threshold, and determining the speed upper limit set based on the mapping relationship between curvature values ​​and speed upper limits if the curvature value is greater than the preset threshold. Then, by combining the speed upper limit set with the coordinate sequence of the path point set, the speed upper limit curve is fitted using a spline interpolation method to obtain the preliminary speed planning curve.

[0042] The path curvature distribution refers to the sequence of curvature values ​​at various points along the path, obtained by smoothing the discrete curvature distribution sequence using a Gaussian filtering algorithm. It reflects the changing trend of the path's curvature. The mapping relationship between curvature values ​​and the upper speed limit refers to setting a corresponding upper speed limit based on the magnitude of the curvature value; the greater the curvature, the lower the corresponding upper speed limit. This relationship can be established through experimental data or a dynamic model. The spline interpolation method uses a cubic spline function to fit a curve to the discrete upper speed limit values, generating a continuously differentiable speed planning curve.

[0043] Specifically, the path curvature distribution after Gaussian filtering accurately reflects the degree of curvature at each point on the path. When the curvature value exceeds a preset threshold, it indicates a significant risk of centrifugal force in that area. In this case, the upper limit of speed is reduced according to the curvature-velocity mapping relationship to prevent the wafer from shifting due to excessive lateral acceleration. For path segments that do not exceed the threshold, a higher handling speed is allowed. This threshold is an engineering parameter determined through theoretical calculations, simulations, and experiments, based on the system's dynamic performance and ensuring wafer safety constraints and motion stability requirements. Spline interpolation converts the discrete upper limit of speed into a continuous curve, ensuring a smooth transition of speed changes between adjacent path points and eliminating mechanical vibrations caused by abrupt changes at discrete points in traditional segmented speed control.

[0044] Compared to existing technologies, traditional methods, which employ fixed speed thresholds or piecewise linear adjustments, cannot dynamically constrain speed based on continuous changes in path curvature. This results in insufficient speed reduction in high-curvature regions or wasted speed in low-curvature regions. This proposed solution overcomes the problem of abrupt speed changes by achieving second-order continuous differentiability of the speed curve on the global path through a dynamic mapping between curvature distribution and speed upper limits, combined with the mathematical properties of spline interpolation.

[0045] Through the above technical solution, this application achieves dynamic matching of speed and path curvature during wafer handling. It automatically reduces speed in curved areas to suppress the influence of centrifugal force, maintains efficient transportation in straight areas, and eliminates acceleration abrupt changes through continuous speed curve planning to ensure the stability of wafer orientation.

[0046] Reference Figure 5 As shown, this application further proposes a method for obtaining joint parameters for multi-axis motion based on the obtained preliminary velocity planning curve, calculating the velocity vector of each axis using an inverse kinematics algorithm, and iteratively optimizing the velocity vectors to ensure that the differences between the velocity vectors of each axis are within a coordination threshold, thereby generating a coordinated multi-axis velocity sequence. This includes: obtaining the preliminary velocity planning curve, extracting the joint parameters for multi-axis motion from it, and determining the initial motion state of each axis; calculating the joint velocity vector of each axis using the extracted joint parameters through an inverse kinematics algorithm to obtain an initial velocity vector set; iteratively adjusting the joint velocity vector of each axis using a particle swarm optimization algorithm based on the initial velocity vector set, with the goal of minimizing the differences between the velocity vectors of each axis, thereby obtaining an optimized velocity vector set; and generating a coordinated multi-axis velocity sequence using the optimized velocity vector set to determine the motion trajectory of each axis.

[0047] Among them, the inverse kinematics algorithm refers to a mathematical method that converts the motion parameters of the end effector into the motion parameters of each joint axis. Specifically, it can be implemented using Jacobian matrix inversion or numerical iteration methods, and is used to establish the kinematic relationship between path planning and multi-axis motion. The particle swarm optimization algorithm is a global optimization method based on swarm intelligence. Specifically, it can be implemented using inertia weight adjustment strategies or neighborhood topology structures, and is used to find the optimal solution with the minimum difference in the multi-axis velocity vector space. The coordination threshold refers to the maximum allowable difference range between multi-axis velocity vectors, which can be determined through experimental calibration or dynamic adjustment strategies, and is used to constrain the termination condition of the optimization process.

[0048] Specifically, the process begins by extracting multi-axis joint parameters, such as the displacement range and acceleration limits of each axis, from the initial velocity planning curve to determine the initial motion state of each axis. Then, an inverse kinematics algorithm is used to convert the velocity planning of the path points into velocity vectors for each joint axis. For example, the pseudo-inverse of the Jacobian matrix is ​​used to calculate the velocity components of each axis, forming an initial set of velocity vectors. Further, a particle swarm optimization algorithm is employed to iteratively adjust the initial velocity vectors. For instance, the fitness function is set as the standard deviation of the velocity vectors for each axis, and the velocity differences are gradually reduced through a swarm search mechanism. When the velocity vector differences of all axes reach a coordination threshold, the optimization process terminates. The final multi-axis velocity sequence ensures the synchronization of the motion trajectories of each axis in time and space.

[0049] Compared to existing technologies, traditional methods typically allocate multi-axis velocities in a fixed ratio, failing to dynamically adapt to motion coupling relationships under complex paths. This proposed solution, however, precisely establishes the mapping relationship between path and joint motion through inverse kinematics, and combines this with a swarm optimization algorithm to globally adjust velocity parameters, effectively solving the problem of uneven wafer stress caused by abrupt changes in multi-axis velocity. For example, existing technologies often experience velocity mismatch between axes when switching between straight paths and curved sections, while this solution uses real-time optimization to ensure that the velocity differences between axes remain within a coordinated threshold.

[0050] Through the above technical solution, this application achieves dynamic coordinated control of multi-axis motion speed, eliminating uneven wafer stress caused by sudden speed changes. During wafer handling, the motion trajectories of each joint axis are kept synchronized through an optimized speed sequence, avoiding mechanical shock caused by excessive single-axis speed. Simultaneously, the application of particle swarm optimization algorithm gives the multi-axis speed adjustment process globally optimal characteristics, effectively reducing dynamic errors of the equipment under complex paths.

[0051] Reference Figure 6As shown, this application further proposes a method for obtaining an estimated attitude deviation value by acquiring the object tilt angle fed back by real-time sensors from a determined coordinated multi-axis velocity sequence, in order to meet the attitude stabilization requirements, and fusing the position and angle data through Kalman filtering. The method includes: acquiring coordinated control commands from the multi-axis velocity sequence; synchronizing the velocities of each axis through a preset motion planning algorithm to obtain a unified motion command set; acquiring the object tilt angle in real time using an accelerometer and gyroscope based on the unified motion command set and sensor feedback data to obtain the original attitude data; and fusing the position and angle data in the original attitude data through Kalman filtering, iteratively updating the state estimation to obtain an optimized attitude deviation estimate value.

[0052] In a multi-axis motion system, coordinated control commands refer to the set of velocity parameters of each actuator. This can be achieved using a timestamp-based command synchronization protocol to ensure alignment of multi-axis motion commands in the time dimension. Motion planning algorithms are optimization methods used to eliminate multi-axis velocity conflicts. Specifically, dynamic time warping algorithms can be used to time-match the velocity sequences of each axis, resolving motion asynchrony issues caused by differences in inter-axis velocity. Accelerometers and gyroscopes are sensor combinations used to detect the spatial attitude of an object. This can be implemented using a MEMS inertial measurement unit, capturing the dynamic tilt state of the wafer through joint measurement of three-axis acceleration and angular velocity. Kalman filtering is a state estimation algorithm used for multi-source data fusion. Specifically, a linear Kalman filter can be used to weight and fuse position coordinates and angle data, eliminating sensor noise interference through an iterative process of prediction and correction.

[0053] Specifically, after the multi-axis velocity sequence is generated, the motion parameters of each axis are extracted through a command synchronization protocol, and the time phase of each axis velocity is adjusted based on a dynamic time warping algorithm to form a spatiotemporally consistent motion command set. During command execution, the acceleration and angular velocity data of the wafer are acquired in real time using a MEMS inertial measurement unit. Combined with the target position information in the motion command, a raw attitude dataset containing position error and angular deviation is constructed. Subsequently, the position coordinates and angle data are input into a Kalman filter, and the attitude parameters at the next moment are predicted through the state transition equation. Then, the predicted values ​​are compared with the sensor's measured values ​​through the observation equation, and the state estimate is updated with the goal of minimizing the covariance matrix. Finally, the optimized attitude deviation is output. This process continuously corrects dynamic errors through a closed-loop feedback mechanism to ensure the real-time performance and accuracy of attitude estimation.

[0054] Compared to existing technologies, traditional methods typically rely on a single sensor for attitude detection, such as indirectly calculating the wafer tilt angle solely from encoder position data, leading to error accumulation and response lag. Existing multi-axis control strategies often neglect the timing alignment of motion commands, causing inter-axis speed fluctuations to overlap and exacerbating attitude instability. In contrast, this solution, through synchronous processing of motion commands and fusion estimation of multi-source data, eliminates inter-axis motion conflicts while utilizing the dynamic correction capability of Kalman filtering to effectively suppress the impact of sensor noise on attitude detection, achieving high-precision real-time deviation estimation.

[0055] Through the above technical solution, this application can solve the attitude deviation problem caused by multi-axis asynchronous speed during wafer handling. By synchronizing motion commands and fusing multi-source data, the vibration amplitude of the wafer during high-speed movement is significantly reduced, improving the stability of the handling process. Simultaneously, through the iterative optimization mechanism of Kalman filtering, the interference of sensor noise on attitude estimation is reduced, providing reliable deviation data for subsequent dynamic adjustment and avoiding the risk of misadjustment caused by detection errors.

[0056] Reference Figure 7 As shown, this application further proposes a method for dynamically adjusting the acceleration parameters in the velocity sequence based on the attitude deviation estimate and using an optimization algorithm to reduce the attitude deviation, thereby determining the optimized speed control command. Specifically, this method includes: acquiring attitude deviation estimates and a velocity sequence from sensor data; classifying the attitude deviation estimates using a preset threshold; if the attitude deviation estimate is greater than the threshold, smoothing the data using a Kalman filter algorithm to obtain smoothed deviation data; extracting acceleration parameters from the velocity sequence based on the deviation data; using logical judgment to determine whether the acceleration parameters exceed a preset range; if so, adjusting the acceleration parameters using a gradient descent algorithm to obtain optimized acceleration parameters; updating the velocity sequence using a dynamic adjustment mechanism based on the acceleration parameters and real-time feedback sensor data to generate an adjusted velocity sequence; and after generating the control command based on the velocity sequence, performing global speed optimization on the speed control command based on the target transport time, adjusting the speed parameters of multiple path segments in parallel to meet time constraints.

[0057] The attitude deviation estimate refers to the wafer tilt deviation calculated by fusing position and angle data using Kalman filtering. This can be implemented using a six-axis sensor data fusion algorithm to quantify the degree of attitude instability during wafer handling. The gradient descent algorithm is a mathematical method that iteratively optimizes the attitude deviation by calculating the partial derivatives of the acceleration parameters. This can be implemented using an adaptive learning rate optimizer to find the optimal acceleration parameters while ensuring convergence speed. The dynamic adjustment mechanism refers to a control architecture based on real-time sensor feedback and updated motion commands using optimized parameters. This can be implemented using a feedforward-feedback composite control structure to achieve rapid response and stable execution of control commands. Global velocity optimization refers to a technique that balances the velocity parameters of each path segment using a multi-objective optimization algorithm. This can be implemented using the Lagrange multiplier method within a parallel computing framework to maximize motion efficiency while satisfying the total handling time constraint.

[0058] Specifically, when the sensor detects that the wafer tilt angle exceeds a safety threshold, which is a critical parameter determined based on control performance requirements, sensor accuracy, and system stability, the raw attitude data is first processed by Kalman filtering to eliminate measurement noise interference and obtain a reliable deviation estimate. Based on the comparison between the deviation data and the preset acceleration range, a gradient descent algorithm is triggered to iteratively optimize the current acceleration parameters. The adjustment amount in each iteration is determined by the partial derivative of the deviation with respect to acceleration. The optimized acceleration parameters are input into the dynamic adjustment module, which, combined with real-time wafer position information, generates an updated velocity sequence. After completing the local parameter adjustment, the system initiates a global optimization module to perform parallel calculations on the velocity parameters of multiple paths. Through constraint solving, it ensures that the total handling time does not exceed process requirements, ultimately generating a velocity control command that balances stability and efficiency.

[0059] Compared to existing technologies, traditional methods typically use fixed acceleration thresholds for velocity adjustment, failing to dynamically adjust parameters based on real-time attitude deviations, and lacking collaborative computation across multiple paths for global time optimization. This solution introduces a dynamic adjustment mechanism based on mathematical optimization, achieving closed-loop control of acceleration parameters and attitude deviations. Simultaneously, it employs a parallel computing architecture to handle the global optimization problem across multiple paths, effectively solving the technical challenge of balancing local adjustment and overall efficiency under high-speed motion.

[0060] Through the above technical solution, this application can correct acceleration parameters in real time during high-speed wafer handling to suppress attitude deviations and avoid wafer slippage or breakage caused by sudden vibrations. Simultaneously, global optimization ensures overall handling efficiency. This technical solution guarantees the motion stability of critical path segments and achieves coordinated adjustment of speed parameters across multiple path segments, ultimately reaching a balance between high-speed motion and smooth handling.

[0061] In light of practical needs, this application further proposes to perform global speed optimization of speed control commands based on target transport time. By adjusting the speed parameters of multiple path segments in parallel to meet time constraints, the final speed control command to be executed is determined.

[0062] The target transport time refers to the pre-set maximum allowable time threshold for the wafer transport process, which can be achieved using process cycle parameters or equipment operating cycle parameters, serving as a constraint for global optimization. Global speed optimization involves simultaneously considering the correlation of speed parameters between multiple path segments and making overall adjustments. This can be achieved using multi-objective optimization algorithms or constraint programming models to eliminate accumulated time errors caused by segmented optimization. Parallel adjustment refers to synchronously correcting the speed parameters of multiple path segments. This can be achieved using distributed computing frameworks or parallel processing threads to maintain the continuity of speed between different path segments. Time constraints transform the target transport time into mathematical constraints, which can be achieved using the Lagrange multiplier method or penalty function method to force the optimized speed parameters to meet the overall time requirements.

[0063] Specifically, after completing multi-axis speed coordination control and attitude deviation adjustment, the target transport time is input into the optimization model as a rigid constraint. A multivariate equation set containing acceleration parameters, speed limits, and time allocation relationships for each path segment is constructed, and a nonlinear programming algorithm is used to globally solve the speed control commands. During this process, speed parameter adjustments for each path segment are executed synchronously, and iterative calculations ensure that the sum of the acceleration and deceleration times for each segment converges to the target transport time. Simultaneously, by introducing speed smoothness constraints between path segments, the adjusted speed curve maintains motion stability while meeting time constraints.

[0064] Compared to existing technologies, traditional methods typically employ a sequential optimization strategy, adjusting speed parameters segment by segment before accumulating and verifying the results over time. This approach can easily lead to drastic speed adjustments at the end of the path to compensate for earlier time errors. In contrast, this proposed solution establishes a global optimization model, directly integrating time constraints into the joint optimization process of multiple path segments. This not only avoids the progressive propagation of time errors but also enables coordinated correction of speed parameters across different path segments through parallel computation mechanisms.

[0065] Through the above technical solution, this application can accurately control the overall handling time while ensuring the stability of the wafer handling posture, effectively solving the problem of time exceeding the limit caused by local speed optimization. By coordinating the speed parameters of multiple path segments from a global perspective, it avoids the speed abruptness phenomenon caused by segment-by-segment optimization in traditional methods, and ensures a balance between the operating efficiency and time controllability of the equipment in complex three-dimensional paths.

[0066] Reference Figure 8 As shown, in order to implement the above control method, this application also discloses a wafer handling speed control system, including: The path processing module is used to acquire the three-dimensional coordinate data of the wafer handling path and divide it into straight segments and curved segments through a path decomposition algorithm to obtain the geometric feature description of each path segment. The curvature analysis module, connected to the path processing module, is used to calculate the trajectory curvature value of the curved segment based on the geometric feature description, smoothly connect the straight segment and the curved segment, and determine the continuous path curvature distribution. The speed planning module, connected to the curvature analysis module, is used to dynamically constrain the upper limit of the speed of the path points based on the curvature distribution and the curvature value, thereby generating a preliminary speed planning curve. A multi-axis coordination optimization module, connected to the velocity planning module, is used to obtain joint parameters of multi-axis motion based on the initial velocity planning curve, calculate the velocity vector of each axis through inverse kinematics algorithm, and iteratively optimize it so that the difference between the velocity vectors of each axis is within the coordination threshold, thereby generating a coordinated multi-axis velocity sequence. The attitude monitoring and estimation module is connected to the multi-axis coordination and optimization module. It is used to obtain the wafer tilt angle fed back by the real-time sensor from the coordinated multi-axis velocity sequence, and to fuse the position and angle data through the Kalman filter algorithm to obtain the attitude deviation estimate. An adaptive control module, connected to the attitude monitoring and estimation module, is used to dynamically adjust the acceleration parameters in the velocity sequence based on the attitude deviation estimate, and to reduce the attitude deviation using an optimization algorithm, thereby outputting the final optimized speed control command to the transport controller.

[0067] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for controlling wafer handling speed, characterized in that: Includes the following steps: The three-dimensional coordinate data of the object's transport path is obtained, and the path is divided into straight segments and curved segments using a preset path decomposition algorithm to obtain the geometric feature description of each segment. Based on the obtained geometric feature description, the trajectory curvature value is calculated for the curved segment, the straight segment and the curved segment are smoothly connected, and the continuous path curvature distribution is determined. Based on the path curvature distribution, a preliminary speed planning curve is generated by dynamically constraining the upper limit of the speed of the path points according to the curvature value. Based on the obtained preliminary velocity planning curve, the joint parameters of multi-axis motion are obtained, the velocity vector of each axis is calculated by the inverse kinematics algorithm, and iterative optimization is performed to ensure that the difference between the velocity vectors of each axis is within the coordination threshold, thereby generating a coordinated multi-axis velocity sequence. From the determined coordinated multi-axis velocity sequence, to meet the attitude stabilization requirements, the object tilt angle fed back by the real-time sensor is obtained, and the position and angle data are fused by Kalman filtering to obtain the attitude deviation estimate. Based on the attitude deviation estimate, the acceleration parameters in the velocity sequence are dynamically adjusted, and an optimization algorithm is used to reduce the attitude deviation, thereby determining the optimized velocity control command.

2. The wafer handling speed control method according to claim 1, characterized in that: The three-dimensional coordinate data of the object's transport path is obtained, and then divided into straight and curved segments using a preset path decomposition algorithm to obtain the geometric feature descriptions of each segment, including: Obtain the three-dimensional coordinate data of the object's transport path, collect spatial trajectory points from sensors or positioning systems, and obtain the original three-dimensional coordinate sequence; The three-dimensional coordinate sequence is processed by a preset path segmentation algorithm to obtain the coordinate sets of straight line segments and curve segments. For a set of coordinates of line segments, calculate the length and direction vector of each segment to obtain a geometric feature description of the line segments; Based on the set of curve segment coordinates, a spline interpolation algorithm is used to fit the curve trajectory, calculate the curvature and tangent vector, and obtain the geometric feature description of the curve segment. The changes in motion direction are extracted from the geometric features of straight and curved segments, and the angles and distances between path points are analyzed to obtain the dynamic feature sequence of the path. By using dynamic feature sequences, the K-means clustering algorithm is used to classify path segments, determine the spatial trajectory characteristics of each segment, and obtain a set of classified path segments. Based on the classified set of path segments, the geometric features of straight line segments and curved segments are integrated to generate a complete description of the path's geometric features.

3. The wafer handling speed control method according to claim 2, characterized in that: Based on the obtained geometric feature description, the trajectory curvature value is calculated for the curved segment, the straight segment and the curved segment are smoothly connected, and the continuous path curvature distribution is determined, including: Based on the geometric feature description of the curve segment, the curvature continuity constraint algorithm is used to optimize the arrangement of control points and generate a fitting curve with smooth curvature transition. Based on the geometric feature description of the straight line segment, the rate of curvature change at the connection point between the straight line and the curve is calculated. Based on the rate of curvature change, a Bézier curve splicing algorithm is used to generate a smooth transition segment at the connection point between the straight line segment and the curve segment. Based on the fitted curve and smooth transition segment of curvature smooth transition, calculate the curvature values ​​of continuous sampling points on the path to generate a discrete curvature distribution sequence; The discrete curvature distribution sequence is smoothed using a Gaussian filtering algorithm to eliminate local curvature fluctuations and obtain a continuous path curvature distribution.

4. The wafer handling speed control method according to claim 3, characterized in that: The smoothing process of the discrete curvature distribution sequence using a Gaussian filtering algorithm includes: Set the Gaussian filter window size and standard deviation parameters, and adaptively adjust the filter strength based on the total path length and curvature change frequency. Convolution operations are used to filter discrete curvature sequences, eliminating local curvature fluctuations while preserving global curvature variation characteristics. The continuity of the filtered curvature distribution is checked to ensure that the first and second derivatives of the curvature distribution curve are continuous.

5. The wafer handling speed control method according to claim 1, characterized in that: Based on the path curvature distribution, a preliminary velocity planning curve is generated by dynamically constraining the upper limit of the path points according to the curvature values, including: The curvature values ​​of each path point are obtained from the curvature distribution of the path. A preset threshold is used to determine the curvature magnitude. If the curvature value is greater than the preset threshold, the set of speed limits is determined according to the mapping relationship between the curvature value and the speed limit. By combining the set of upper speed limits with the coordinate sequence of the path point set, the upper speed limit curve is fitted using spline interpolation to obtain the preliminary speed planning curve.

6. The wafer handling speed control method according to claim 1, characterized in that: Based on the obtained preliminary velocity planning curve, the joint parameters of the multi-axis motion are acquired. The velocity vector of each axis is calculated using an inverse kinematics algorithm, and iterative optimization is performed to ensure that the difference between the velocity vectors of each axis is within a coordination threshold, thereby generating a coordinated multi-axis velocity sequence, including: Obtain the preliminary velocity planning curve, extract the joint parameters of multi-axis motion from it, and determine the initial motion state of each axis; Using the inverse kinematics algorithm, the joint velocity vectors of each axis are calculated for the extracted joint parameters to obtain the initial velocity vector set; Based on the initial set of velocity vectors, the joint velocity vectors of each axis are iteratively adjusted using a particle swarm optimization algorithm, with the goal of minimizing the differences between the velocity vectors of each axis, thereby obtaining an optimized set of velocity vectors. By using the optimized set of velocity vectors, a coordinated multi-axis velocity sequence is generated, and the motion trajectory of each axis is determined.

7. The wafer handling speed control method according to claim 1, characterized in that: From the determined coordinated multi-axis velocity sequence, to meet the attitude stabilization requirements, the object tilt angle fed back by real-time sensors is obtained. The position and angle data are fused using Kalman filtering to obtain an estimated attitude deviation, including: Coordinated control commands are obtained from the multi-axis velocity sequence, and the velocities of each axis are synchronized using a preset motion planning algorithm to obtain a unified set of motion commands. Based on the unified motion command set and combined with sensor feedback data, the object tilt angle is collected in real time using accelerometers and gyroscopes to obtain the original attitude data. By fusing the position and angle data from the original attitude data using Kalman filtering, the state estimation is iteratively updated to obtain the optimized attitude deviation estimate.

8. The wafer handling speed control method according to claim 1, characterized in that: Based on the attitude deviation estimate, the acceleration parameters in the velocity sequence are dynamically adjusted, and an optimization algorithm is used to reduce the attitude deviation, thereby determining an optimized velocity control command, including: Attitude deviation estimates and velocity sequences are obtained from sensor data. The attitude deviation estimates are classified using a preset threshold. If the attitude deviation estimates are greater than the threshold, they are smoothed using a Kalman filter algorithm to obtain smoothed deviation data. Based on the deviation data, the acceleration parameters in the velocity sequence are extracted, and logical judgment is used to determine whether the acceleration parameters exceed the preset range. If they do, the acceleration parameters are adjusted by the gradient descent algorithm to obtain optimized acceleration parameters. Using the acceleration parameters and real-time feedback sensor data, a dynamic adjustment mechanism is employed to update the velocity sequence, generating an adjusted velocity sequence. Based on the speed sequence, logical judgments are used to generate control commands, resulting in optimized speed control commands.

9. The wafer handling speed control method according to claim 8, characterized in that: Also includes: Based on the target transport time, the speed control command is globally optimized. By adjusting the speed parameters of multiple path segments in parallel to meet the time constraints, the final speed control command to be executed is determined.

10. A wafer handling speed control system, characterized in that: The method for implementing any one of claims 1 to 9 includes: The path processing module is used to acquire the three-dimensional coordinate data of the wafer handling path and divide it into straight segments and curved segments through a path decomposition algorithm to obtain the geometric feature description of each path segment. The curvature analysis module, connected to the path processing module, is used to calculate the trajectory curvature value of the curved segment based on the geometric feature description, smoothly connect the straight segment and the curved segment, and determine the continuous path curvature distribution. The speed planning module, connected to the curvature analysis module, is used to dynamically constrain the upper limit of the speed of the path points based on the curvature distribution and the curvature value, thereby generating a preliminary speed planning curve. A multi-axis coordination optimization module, connected to the velocity planning module, is used to obtain joint parameters of multi-axis motion based on the initial velocity planning curve, calculate the velocity vector of each axis through inverse kinematics algorithm, and iteratively optimize it so that the difference between the velocity vectors of each axis is within the coordination threshold, thereby generating a coordinated multi-axis velocity sequence. The attitude monitoring and estimation module is connected to the multi-axis coordination and optimization module. It is used to obtain the wafer tilt angle fed back by the real-time sensor from the coordinated multi-axis velocity sequence, and to fuse the position and angle data through the Kalman filter algorithm to obtain the attitude deviation estimate. An adaptive control module, connected to the attitude monitoring and estimation module, is used to dynamically adjust the acceleration parameters in the velocity sequence based on the attitude deviation estimate, and to reduce the attitude deviation using an optimization algorithm, thereby outputting the final optimized speed control command to the transport controller.

Images