Full-automatic multi-size wafer loading platform for docking interaction with AGV trolley

Abstract

The present application relates to the technical field of semiconductor, and especially relates to a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley. The full-automatic multi-size wafer loading platform comprises a loading platform, a lifting cover, a visual imaging module and a control system. In the present application, a reference mark array is arranged at the bottom edge of the inner side of the lifting cover, the wafer box image collected by the visual module at the center of the top is used, the actual refraction parameter in the current environment is identified online by using a light incidence-refraction inversion model, the wafer box image collected is corrected reversely at the pixel level, the dynamic imaging distortion caused by the quartz glass / air multi-layer interface is eliminated, the position deviation of the wafer box is determined through regional brightness analysis, the wafer box is automatically translated and corrected in X-Y linkage by the correction mechanism, and thus the visual positioning precision and the system robustness in the sealed state are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of semiconductor technology, and particularly relates to a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley. BACKGROUND

[0002] In the existing semiconductor wafer transmission system, after the AGV trolley transports the wafer box to the loading platform, the wafer box is usually roughly positioned by mechanical limiting or manual intervention. Such a mode is difficult to meet the demand of the high-precision robot for taking and placing the box, especially in the mixed use scene of multi-size wafer boxes (such as 8-inch and 12-inch wafer boxes), because of the lack of automatic identification and deviation correction capability, it is easy to cause the deviation of taking the box, jamming and even collision. Although some wafer loading platforms try to introduce a visual positioning system to improve this problem, due to the factors such as window refraction, shielding or uneven illumination, the existing visual positioning system will cause imaging distortion and contrast reduction after the lifting cover is closed, which seriously affects the positioning accuracy. Especially when transparent quartz glass is used as the sealing interface, the refraction effect of the multi-layer interface will produce dynamic drift with the change of environmental temperature and humidity, and the traditional offline calibration compensation method cannot effectively cope with this real-time distortion problem. Therefore, the present application provides a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley. SUMMARY

[0003] The present application aims to solve the problems of the existing wafer box alignment system, such as the nonlinear and time-varying deviation of the visual imaging coordinates caused by the dynamic drift of the refractive index of the quartz glass / air multi-layer interface with the change of temperature and humidity after the lifting sealing cover is closed, the large positioning error caused by the fact that the traditional offline calibration cannot be compensated, and the inability to meet the demand of the high-precision robot for taking and placing the wafer box, and proposes a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley.

[0004] The present application provides a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley, comprising: a loading platform for carrying a wafer box; a lifting cover that can be lowered to a closed position to seal the wafer box; a visual imaging module arranged at the center of the inner top surface of the lifting cover for acquiring an original image of the wafer box; and a control system for acquiring the original image of the wafer box by the visual imaging module under the closed state of the lifting cover, combining the reference mark array and the environmental temperature and humidity parameters, calculating the current refraction parameters in real time through the light incidence-refraction inversion model, correcting the original image pixel by pixel based on the constructed inverse mapping lookup table to generate a corrected image, performing regional gray scale analysis on the corrected image, identifying the outline coverage area of the wafer box, and calculating the positional deviation of the geometric center relative to the preset alignment center, and controlling the alignment mechanism to perform X-Y bidirectional translation alignment of the wafer box according to the positional deviation, and triggering the locking mechanism to fix the loading platform after the alignment is completed.

[0005] Optionally, the bottom edge of the lifting cover is further provided with a ring-shaped fluororubber sealing ring for forming an airtight contact with the top surface of the platform when the cover is lowered to cover the wafer box on the platform.

[0006] Optionally, the ring-shaped fluororubber sealing ring forms a uniform linear contact pressure with the mirror-polished top surface of the platform through a set Shore hardness and pre-compression amount.

[0007] Optionally, the outer side of the ring-shaped fluororubber sealing ring is arranged with a plurality of vacuum detection ports, each of which is connected with a differential pressure sensor for real-time monitoring of the airtightness of the ring-shaped fluororubber sealing ring, and if the pressure drop exceeds a threshold value within a preset pressure maintaining period, an alarm is triggered and the homing operation is paused.

[0008] Optionally, the reference mark array includes at least four targets embedded in the inner side of the ring-shaped fluororubber sealing ring and spatially non-overlapping with the outer contour of the wafer box.

[0009] Optionally, the light incidence-refraction inversion model includes: constructing a light tracking model based on Snell's law; using an improved Levenberg-Marquardt algorithm to iteratively solve the refractive index and medium thickness parameters in the current environment, taking the difference between the measured coordinates and the calibrated coordinates of the reference marks as the error function.

[0010] Optionally, the construction of the reverse mapping lookup table includes: for each pixel point, calculating its corresponding actual object coordinate according to the camera intrinsic parameter matrix, the environmental temperature and humidity, and the actual refraction parameters; and generating the corrected image from the original image by a bicubic interpolation algorithm.

[0011] Optionally, the regional gray scale analysis includes: dividing the corrected image into a plurality of rectangular sub-regions; calculating the average gray scale value of each rectangular sub-region and comparing it with a preset proportion threshold based on the dynamic background mean value calculated for the image edge sub-region; marking the rectangular sub-regions below the preset proportion threshold as dark regions, calculating the center of the wafer box based on the dark region distribution matrix using a centroid fitting algorithm, and comparing it with a preset alignment center to generate the position deviation amount of the X-axis and Y-axis.

[0012] Optionally, the control system further includes continuously collecting multiple frames of original images, performing pixel-level mean filtering, and establishing a dynamic background gray scale model to suppress local reflection and dust interference.

[0013] Optionally, the control system identifies the size type of the wafer box according to the position deviation amount, calls the corresponding preset dynamic parameter model for feedforward compensation, and automatically switches the PID control parameters according to the load mass.

[0014] The application has the advantages and beneficial effects that the application provides a full-automatic multi-size wafer loading platform for docking interaction with an AGV trolley, reference mark arrays are arranged at the bottom edge of the inner side of the lifting cover, wafer box images collected by the top center vision module are used to identify actual refraction parameters in the current environment online by using a light incidence-refraction inversion model, pixel-level coordinate reverse correction is performed on the collected wafer box images, dynamic imaging distortion caused by the quartz glass / air multi-layer interface is eliminated, and the wafer box position deviation is determined through regional brightness analysis, and the wafer box is automatically executed X-Y linkage translation correction by the correction mechanism, so that the vision positioning accuracy and system robustness in the sealed state are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 It is a front view of the full-automatic multi-size wafer loading platform for docking interaction with the AGV trolley.

[0016] Figure 2 It is a structural schematic view of the full-automatic multi-size wafer loading platform for docking interaction with the AGV trolley.

[0017] Figure 3 It is a partial structural schematic view of the full-automatic multi-size wafer loading platform for docking interaction with the AGV trolley.

[0018] Figure 4 It is a bottom view of the lifting cover.

[0019] Figure 5 It is a top view of the correction mechanism.

[0020] In the figure, the carrier 1, the lifting cover 2, the wafer box 3, the Y-direction correction assembly 4, the correction block 5, the operation area 6, the annular fluorine rubber sealing ring 7, the differential pressure sensor 8, the vision imaging module 9, the servo sliding table 10, the X-direction sliding groove 11, the X-direction sliding block 12, the X-direction servo motor 13, the Y-direction sliding seat 14, the Y-direction sliding block 15, and the Y-direction servo motor 16. DETAILED DESCRIPTION

[0021] The embodiments of the application will be described in detail below, examples of the embodiments are shown in the drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the application, and cannot be understood as a limitation on the application, and the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0022] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood in a broad sense, for example, can be fixedly connected, or can be detachably connected, or integrally connected; can be mechanically connected, or can be electrically connected; can be directly connected, or can be indirectly connected through an intermediate medium, or can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0023] With reference to the accompanying Figures 1-4 The present application provides a full-automatic multi-size wafer loading platform for docking and interacting with an AGV, mainly comprising a loading platform 1, a lifting cover 2, a visual imaging module 9 and a control system.

[0024] In some optional implementation manners of the present application, the loading platform 1 is used for carrying a wafer box 3.

[0025] In some optional implementation manners of the present application, the lifting cover 2 can be lowered to a closed position to seal the wafer box 3.

[0026] In some optional implementation manners of the present application, the bottom edge of the lifting cover 2 is further provided with an annular fluororubber sealing ring 7, which is used to form an airtight contact with the top surface of the loading platform 1 when the wafer box 3 on the loading platform 1 is covered.

[0027] In some optional implementation manners of the present application, the annular fluororubber sealing ring 7 forms a uniform linear contact pressure with the mirror-polished top surface of the loading platform 1 through a set Shore hardness and pre-compression amount.

[0028] In some optional implementation manners of the present application, the outer side of the annular fluororubber sealing ring 7 is arranged with a plurality of vacuum detection ports, each of which is connected with a differential pressure sensor 8, which is used to monitor the airtightness inside the annular fluororubber sealing ring 7 in real time. If the pressure drop exceeds the threshold value within the preset pressure maintaining period, an alarm is triggered and the homing operation is paused.

[0029] In some optional implementations of the present application, a protective sealing safety mechanism is provided, i.e. an annular fluororubber sealing ring 7 is arranged between the lifting cover 2 and the carrier 1. Specifically, the annular fluororubber sealing ring 7 is integrated at the bottom edge of the lifting cover 2. The Shore hardness of the annular fluororubber sealing ring 7 is controlled within the range of 70A±5, and the pre-compression amount is set to 1.8mm. When the lifting cover 2 is lowered to the closed position, uniform linear contact pressure is formed between the lifting cover 2 and the top surface of the carrier 1, thereby achieving airtight isolation of the Class 1 clean environment. The contact area between the top surface of the carrier 1 and the annular fluororubber sealing ring 7 is mirror-polished, and the surface roughness Ra is ≤0.2μm. The lifting cover 2 is driven to vertically lift by the servo slide 10, and is equipped with a high-precision displacement sensor to real-time feedback the stroke position. When the annular fluororubber sealing ring 7 is completely compressed in place, the control system triggers the pressure maintaining mode, and maintains the output force of the servo slide 10 constant to compensate for material creep. A vacuum detection port is arranged outside the annular fluororubber sealing ring 7 in a ring shape, and is connected with a micro differential pressure sensor 8 for real-time monitoring of the airtightness of the annular fluororubber sealing ring 7. If the pressure drop exceeds 3Pa within the 5-second pressure maintaining period, it is determined that the sealing fails and an alarm is triggered, and the homing operation is temporarily suspended to prevent the risk of contamination.

[0030] In some optional implementations of the present application, for example, the lifting cover 2 is a rectangular cavity structure with open back and bottom, with a height of 400 mm and a width of 320 mm, and a ring-shaped fluororubber sealing ring 7 is embedded in the bottom edge, which is made of perfluoroether rubber (FFKM) and has excellent chemical resistance and a Shore hardness of 70A±3, ensuring that it remains stable in elasticity under long-term contact with clean nitrogen and a small amount of volatile organic substances. The ring-shaped fluororubber sealing ring 7 is installed in the trapezoidal groove precisely processed on the bottom of the lifting cover 2, and the pre-compression amount is set to 1.8 mm. When the lifting cover 2 is driven by the servo slide 10 to vertically descend to the closed position, the ring-shaped fluororubber sealing ring 7 forms a uniform linear contact pressure with the top surface of the stage 1, and the standard deviation of the measured contact pressure distribution is less than ±5%, meeting the airtight isolation requirements of Class 1 clean environment. The contact area between the top surface of the stage 1 and the ring-shaped fluororubber sealing ring 7 is treated by ultra-precision grinding and mirror polishing, with a surface roughness Ra=0.18μm, and is provided with a flow guide groove (not shown) for guiding the residual air between the interfaces to be quickly discharged during compression, avoiding the "air cushion effect" to cause sealing delay. The end of the flow guide groove is connected to a vacuum exhaust channel, which can start a miniature vacuum pump for auxiliary air exhaust after the lifting cover 2 is compressed, further improving the sealing response speed. The measured results show that it only takes 2.3 seconds from the start of descent to the completion of sealing. The lifting of the lifting cover 2 is driven by two sets of synchronous servo slides 10, with a pushing force of 500N and a stroke of 100mm, and is equipped with a high-precision grating displacement sensor with a resolution of 1μm, which provides real-time feedback on the current position of the lifting cover 2. The control system controls the servo slide 10 to drive the lifting cover 2 to descend at a constant speed according to the preset stroke curve, and switches to a low-speed buffer mode (such as reducing the speed to 5mm / s) when it is 5mm away from the end point to prevent impact. When the displacement sensor detects that the lifting cover 2 has completely compressed the ring-shaped fluororubber sealing ring 7 to the set position (such as within ±0.05mm of the end of the stroke), the control system triggers the pressure maintaining mode, maintaining the output force of the servo slide 10 at 300N to compensate for the creep relaxation of the fluororubber material under continuous pressure (i.e. the measured 24-hour creep amount is about 0.12mm). Four vacuum detection ports are arranged in a ring around the outside of the ring-shaped fluororubber sealing ring 7, which are distributed at equal intervals and connected to miniature differential pressure sensors 8 (range ±100Pa, accuracy ±0.5Pa) for real-time monitoring of the airtightness of the ring-shaped fluororubber sealing ring 7. During testing, 50Pa of slightly positive pressure nitrogen is injected into the ring-shaped fluororubber sealing ring 7, and a 5-second pressure maintaining period is entered. If the average pressure drop during this period exceeds 3Pa, the seal is determined to have failed and an alarm is triggered, and the subsequent visual acquisition and homing operation is suspended to prevent the risk of contamination. The measured results show that under normal sealing conditions, the pressure drop within 5 seconds is only 1.2Pa, and the system's reliable recognition rate reaches 99.98%.

[0031] In some optional implementations of the present application, reference is made to the accompanying Figure 4A visual imaging module 9 is arranged at the center of the inner top surface of the lifting cover 2 to collect the original image of the wafer box 3.

[0032] In some optional implementations of the present application, the visual imaging module 9 can integrate a 50 million pixel global shutter CMOS industrial camera and a coaxial ring-shaped LED light source, the optical axis of which is perpendicular to the plane of the stage 1, the working distance is 350 mm, and the imaging field of view can cover a diameter of up to 300 mm, meeting the full-width collection requirements of wafer boxes 3 of various sizes (8 inches, 12 inches, etc.).

[0033] In some optional implementations of the present application, the visual imaging module 9 is integrated at the center of the inner top surface of the lifting cover 2, which includes a 50 million pixel global shutter CMOS industrial camera (model: Basler ace2 Pro) equipped with a F-mount telecentric lens with a focal length of 35 mm, a working distance of 350 mm, and an imaging field of view with a diameter of up to 305 mm, which can completely cover the top surface area of a 12-inch wafer box 3 (outer diameter 298 mm). The light source uses a coaxial ring-shaped LED array with a wavelength of 630 nm and a brightness adjustable range of 2000-6000 cd / m 2 , and vertical illumination is achieved through a half-transmission half-reflection mirror to effectively reduce the shadows caused by oblique light. The optical axis is strictly perpendicular to the plane of the stage, with an inclination error of less than 0.1°, ensuring that the image has no perspective distortion.

[0034] In some optional implementations of the present application, the control system is configured to, in the closed state of the lifting cover 2, acquire the original image of the wafer box 3 through the visual imaging module 9; based on the imaging positions of the reference marker array in the original image and in combination with the environmental temperature and humidity parameters, calculate the actual refraction parameters of the current optical system in real time through a light incidence-refraction inversion model; according to the actual refraction parameters, construct an inverse mapping lookup table to perform pixel-by-pixel coordinate correction on the original image to generate a corrected image; perform regional gray scale analysis on the corrected image to identify the contour coverage area of the wafer box 3 and calculate the positional deviation of the geometric center of the wafer box 3 relative to the preset alignment center; according to the positional deviation, control the normalizing mechanism to perform X-Y bidirectional translation normalization on the wafer box 3; after the normalization is completed, trigger the locking mechanism to fix the stage 1.

[0035] In some optional implementations of the present application, the reference Figure 4 The reference marker array includes at least four targets embedded in the inner side of the ring-shaped fluororubber sealing ring 7 and does not overlap the outer contour of the wafer box 3 in space.

[0036] In some optional implementation modes of the present application, the light incidence-refraction inversion model comprises: constructing a light tracking model based on Snell's law; and using an improved Levenberg-Marquardt algorithm to iteratively solve the refractive index and medium thickness parameters in the current environment, with the difference between the measured coordinates of the reference marks and the calibration coordinates as an error function.

[0037] In some optional implementation modes of the present application, the construction of the reverse mapping lookup table comprises: for each pixel point, calculating its corresponding actual object coordinate according to the camera intrinsic parameter matrix, the environmental temperature and humidity, and the actual refraction parameters; and correcting the image to be generated from the original image by a bicubic interpolation algorithm.

[0038] In some optional implementation modes of the present application, the regional gray scale analysis comprises: dividing the corrected image into a plurality of rectangular sub-regions; calculating the average gray scale value of each rectangular sub-region and comparing it with a preset proportion threshold of the dynamic background mean value calculated based on the image edge sub-region; marking the rectangular sub-regions lower than the preset proportion threshold as dark regions, calculating the wafer box center based on the dark region distribution matrix using a centroid fitting algorithm, and comparing it with the preset alignment center to generate the position deviation amount of the X-axis and Y-axis.

[0039] In some optional implementation modes of the present application, the reference mark array and the refraction compensation method involved in the present application comprise the following steps.

[0040] 1. Reference mark array arrangement: at least four micro reference marks are symmetrically arranged on the inner side edge of the bottom of the lifting cover 2, which are high-contrast cross targets, made of weather-resistant ceramic, and embedded in the inner side non-contact area of the sealing ring.

[0041] 2. Real-time identification of refraction parameters: (1) based on Snell's law, a light tracking model of the air-transparent medium-air three-layer refraction path is constructed, and the formula is: , wherein, n1 is the refractive index of the air medium layer outside the lifting cover 2, n2 is the refractive index of the quartz glass medium layer, n3 is the refractive index of the air medium layer inside the lifting cover 2, θ1 is the light incidence angle of the air medium layer outside the lifting cover 2, θ2 is the light refraction angle of the quartz glass medium layer, θ3 is the light incidence angle of the air medium layer inside the lifting cover 2, di is the thickness of the i-th medium layer, is the pixel coordinate, representing a pixel point on the imaging plane, used to generate the origin or direction of the ray, The parameter coordinates of the ray are used to refer to the position parameters of the intersection point of the light ray and the object surface. (2) The improved Levenberg-Marquardt algorithm is used to take the difference between the measured coordinates of the reference marks and the calibration coordinates as an error function, and iteratively solve the refractive index and medium thickness parameters in the current environment, and the formula is: wherein, is the measured mark coordinate of the reference mark, is the calibration coordinate, f() is the light ray tracing mapping function, and k is the number of measured mark coordinates of the reference mark.

[0042] 3. Dynamic distortion correction: (1) For each pixel point (u, v) on the original image plane, the corresponding actual object coordinate (x, y) is calculated according to the camera intrinsic matrix, the environmental temperature and humidity, and the actual refraction parameters, and the calculation formula is: wherein, K is the camera intrinsic matrix, δ is the refraction compensation term, T is the current environmental temperature, and RH is the current environmental humidity. (2) The bicubic interpolation correction is implemented, and the formula is: wherein, is the target pixel point calculated by bicubic interpolation, is the pixel point on the original image plane, i is the horizontal coordinate position of the target pixel point in the original image, j is the vertical coordinate position of the target pixel point in the original image, and W(i) and W(j) are the weight coefficients of the Catmull-Rom interpolation kernel function in the parameter space.

[0043] 4. Wafer box image correction: The edge point detection is performed on the original image to obtain the pixel points of the wafer box edge, the corresponding actual object coordinates are searched as the correction coordinates through the reverse mapping lookup table, the target pixel points are obtained after the bicubic interpolation correction, and the corrected image is obtained based on the target pixel points.

[0044] In some optional implementation manners of the present application, the specific implementation process of the reference mark array and the refraction compensation method is as follows: (1) calibration stage: the reference mark image is collected under the standard environment (23℃, 45%RH), the center coordinates of each mark are measured, and the initial refraction parameters are recorded = (145.12, -2.34), = (1.89, 146.78), etc. = 1.458, d = 12.5mm. (2) Real-time compensation process: the current environmental monitoring value is T = 25.3℃, RH = 52%, the mark coordinates = (144.87, -2.41), = (1.76, 146.52) are collected, and the actual refraction parameters =1.453, d=12.3mm, the reverse mapping table of 4096x3072 is generated, and the time consumption is 18ms. (3) Wafer box image correction: the wafer edge point (2048, 1536) is detected in the original image, the correction coordinates (2045.2, 1532.7) are obtained by table lookup, the gray value is calculated by bicubic interpolation, and the final positioning error is reduced from 0.15mm to 0.02mm.

[0045] In some optional implementation modes of the present application, the specific implementation process of the wafer box positioning method based on regional brightness analysis is as follows, for example, the correction image is divided into 10x10 rectangular sub-regions according to the preset grid, each rectangular sub-region corresponds to a physical size of 30.5mmx30.5mm, and the average gray value of each rectangular sub-region is calculated, and the formula is as follows: , wherein, is the average gray value of the i-th row and j-th column rectangular sub-region, is the total number of pixels in the region; the global background mean value is calculated by selecting the gray value of the four rectangular sub-regions (i=1, 10 or j=1, 10) on the four edges, and the formula is as follows: , wherein, is the average gray value of the first row and first column rectangular sub-region (i.e. the upper left corner rectangular sub-region), is the average gray value of the first row and tenth column rectangular sub-region (i.e. the lower left corner rectangular sub-region), is the average gray value of the tenth row and first column rectangular sub-region (i.e. the upper right corner rectangular sub-region), is the average gray value of the tenth row and tenth column rectangular sub-region (i.e. the lower right corner rectangular sub-region), and the preset brightness threshold is 0.6x The average gray value of each rectangular sub-region is compared with the preset proportion threshold based on the dynamic background mean value calculated based on the image edge sub-region. If the average gray value of the rectangular sub-region is less than the preset proportion threshold, it is determined as a "dark area", and is marked as a wafer box covering area. For example, in an actual test, the preset proportion threshold is calculated to be 180, and it is found that the average gray value of 36 rectangular sub-regions in the center 6x6 region is lower than the preset proportion threshold, which constitutes a continuous covering area, and it is preliminarily judged as a 12-inch wafer box 3. Based on the dark area distribution matrix , the geometric center coordinates of the wafer box outer contour are calculated by using the centroid fitting algorithm, and the formula is as follows: , , wherein, , represents the center coordinates of each sub-region, and the preset alignment center is located at the geometric center of the stage 1, and the position deviation amount of the X-axis and Y-axis is calculated and generated , In an actual operation, it is measured that = 1.24mm, = -0.86mm, and it is calculated that = 1.24mm, = -0.86mm, and the deviation calculation accuracy can reach ±0.02mm through repeated measurement verification.

[0046] In some optional implementations of the present application, the control system further comprises continuously collecting multiple frames of original images, performing pixel-level mean filtering, and establishing a dynamic background gray model to suppress local reflection and dust interference.

[0047] In some optional implementations of the present application, to improve the anti-interference ability, the control system introduces a multi-frame image averaging and dynamic background modeling mechanism, specifically: 5 frames of images are continuously collected, pixel-level mean filtering is performed, and a background gray model is established at the same time, and the formula is: , wherein a is an update coefficient, and the value is 0.9, is the current background gray value, is the last frame background gray value, which is used to suppress the transient abnormality caused by local reflection or dust. The actual measurement shows that under the condition of slight reflection at the edge of the wafer box, this mechanism reduces the misjudgment rate from 12.7% to 1.3%.

[0048] In some optional implementations of the present application, the control system identifies the wafer box size type according to the position deviation amount, and calls the corresponding preset dynamic parameter model for feedforward compensation, and automatically switches the PID control parameters according to the load mass.

[0049] In some optional implementations of the present application, reference is made to the accompanying Figure 5 The present application relates to a correcting mechanism composed of an X-direction correcting assembly and a Y-direction correcting assembly 4. The X-direction correcting assembly includes an X-direction sliding groove 11 arranged on a loading platform, two X-direction sliding blocks 12 slidingly connected to the X-direction sliding groove 11, and an X-direction servo motor 13 driving the X-direction sliding blocks 12 to move in the X-direction. The Y-direction correcting assembly 4 includes a Y-direction sliding seat 14 arranged on the X-direction sliding block 12, two Y-direction sliding blocks 15 slidingly connected to the Y-direction sliding seat 14, a correcting block 5 arranged on the Y-direction sliding block 15, and a Y-direction servo motor 16 driving the Y-direction sliding blocks 15 to move in the Y-direction. It should be noted that after each X-Y bidirectional translation correction, the X-direction correcting assembly and the Y-direction correcting assembly 4 return to the initial position to wait for the next X-Y bidirectional translation correction. The control system generates a and The linkage trajectory is planned using a feedforward and PID composite control algorithm: First, the wafer box size type is determined based on the deviation (estimated by the area of ​​the coverage area), and the corresponding dynamic parameter model is called for feedforward compensation to eliminate the inertial coupling effect; then, XY bidirectional translation correction is performed, and the PID parameters can be automatically switched according to the current load mass to achieve fast and stable centering; after correction is completed, the mechanical lock pin is triggered to insert into the positioning hole at the bottom of the platform 1 to ensure that there is no micro-displacement during subsequent robot operation, and finally the "positioning completed" signal is output.

[0050] In some optional implementations of this application, the specific implementation process of the feedforward and PID composite control algorithm, for example, the control system based on the generated... =1.24mm and =-0.86mm, the wafer cassette size is estimated based on the dark area coverage area. For example, if the number of dark area coverage sub-regions is ≥30, it is determined to be 12 inches; otherwise, it is determined to be 8 inches. In this embodiment, the coverage is 36 rectangular sub-regions, which is confirmed as a 12-inch mode. Then, the corresponding dynamic parameter model is called for feedforward compensation. It is known that the mass m of the 12-inch FOUP is 8.5kg, and the X-axis acceleration is... 1.8 m / s 2 At that time, inertial force The calculation formula is: The calculated inertial force is 15.3 N. Based on this, the feedforward controller preloads the X-axis drive torque to eliminate hysteresis and coupled oscillations at startup. Subsequently, XY bidirectional translation correction is performed using a combined feedforward and PID control algorithm, with the following formula: ,in, For feedforward control gain, For proportional gain, For integral gain, For differential gain, This represents the current position error. This is the first derivative of the current position error. The second derivative of the current position error; the PID parameters adaptively switch according to the size, such as in the 8-inch mode. , , 12-inch mode , , In this embodiment, the X-axis needs to move 1.24mm in the positive direction, and the Y-axis needs to move 0.86mm in the negative direction. The entire process takes 1.4 seconds, and the final positioning error is... =0.01mm, = -0.02mm, which meets the accuracy requirement of the robot for taking the box. After the alignment is completed, the control system triggers the mechanical locking pin to be inserted into the two positioning holes (hole diameter Φ6H7) at the bottom of the platform 1, with a locking force of 300N, to avoid micro displacement when the lifting cover 2 is accidentally collided by human. After the locking state is confirmed by the Hall sensor, the control system outputs the "positioning completed" signal to the upper computer, allowing the robot to enter the operation area 6 at the back of the platform 1 to take the wafer box.

[0051] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can still be modified or replaced equivalently without departing from the spirit and scope of the present application. Any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered within the protection scope of the claims of the present application.

Claims

1. A fully automated multi-size wafer loading platform that interfaces and interacts with AGV (Automated Guided Vehicle) trolleys, characterized in that, include: A stage, used to hold a wafer cassette; The lifting cover can be lowered to a closed position to seal the wafer cassette; The bottom edge of the lifting cover is also provided with an annular fluororubber sealing ring, which is used to form an airtight contact with the top surface of the stage when it is lowered to cover the wafer box on the stage. The visual imaging module is located at the center of the top surface inside the lifting housing and is used to acquire raw images of the wafer cassette. The control system is used to acquire the original image of the wafer cassette through the vision imaging module when the lifting cover is closed. Combined with the reference marker array and ambient temperature and humidity parameters, the system calculates the current refraction parameters in real time through the light incident-refraction inversion model. Based on the constructed inverse mapping lookup table, the system performs pixel-by-pixel coordinate correction on the original image to generate a corrected image. The system performs regional grayscale analysis on the corrected image to identify the outline coverage area of ​​the wafer cassette and calculates the positional deviation of the geometric center relative to the preset alignment center. Based on the positional deviation, the system controls the correction mechanism to perform XY bidirectional translation correction on the wafer cassette. After the correction is completed, the system triggers the locking mechanism to fix the stage. The reference marker array includes at least four targets, which are embedded inside the annular fluororubber sealing ring and do not spatially overlap with the outer contour of the wafer cassette. The light incident-refraction inversion model includes: Based on Snell's law, a ray tracing model is constructed. An improved Levenberg-Marquardt algorithm is used, with the difference between the measured coordinates and the calibrated coordinates of the reference mark as the error function, to iteratively solve the refractive index and medium thickness parameters under the current environment; The construction of the reverse mapping lookup table includes: For each pixel, its corresponding actual object coordinates are calculated based on the camera intrinsic parameter matrix, ambient temperature and humidity, and actual refraction parameters. The corrected image is generated by mapping the original image using a bicubic interpolation algorithm.

2. The fully automated multi-size wafer loading platform that interacts with AGV vehicles as described in claim 1, characterized in that, The annular fluororubber sealing ring, through a set Shore hardness and pre-compression amount, forms a uniform line contact pressure with the mirror-polished top surface of the platform.

3. The fully automated multi-size wafer loading platform for docking and interacting with AGV vehicles as described in claim 2, characterized in that, The annular fluororubber sealing ring has multiple vacuum detection ports arranged circumferentially on its outer side. Each vacuum detection port is connected to a differential pressure sensor to monitor the internal airtightness of the annular fluororubber sealing ring in real time. If the pressure drop exceeds the threshold within the preset pressure holding period, an alarm is triggered and the correction operation is suspended.

4. The fully automated multi-size wafer loading platform for docking and interacting with AGV vehicles as described in claim 1, characterized in that, The grayscale analysis of the region includes: The corrected image is divided into several rectangular sub-regions; Calculate the average gray value of each rectangular sub-region and compare it with a preset proportional threshold of the average dynamic background value calculated based on the image edge sub-regions; Rectangular sub-regions below a preset ratio threshold are marked as dark areas. The center of the wafer box is calculated using a centroid fitting algorithm based on the dark area distribution matrix and compared with a preset alignment center to generate the positional deviation of the X-axis and Y-axis.

5. The fully automated multi-size wafer loading platform for docking and interacting with AGV vehicles as described in claim 1, characterized in that, The control system also includes continuously acquiring multiple frames of original images, performing pixel-level mean filtering, and establishing a dynamic background grayscale model to suppress local reflections and dust interference.

6. The fully automated multi-size wafer loading platform for docking and interacting with AGV vehicles as described in claim 1, characterized in that, The control system identifies the wafer cell size type based on the position deviation and calls the corresponding preset dynamic parameter model for feedforward compensation. At the same time, it automatically switches the PID control parameters according to the load mass.

Images