Automatic loading and unloading control system for semiconductor material plate

Abstract

The invention relates to the technical field of semiconductor manufacturing, in particular to an automatic loading and unloading control system for semiconductor material plates. A clamping force dynamic control module; the path planning and obstacle avoidance module is used for dividing safety levels of loading and unloading areas on the basis of material plate stacking parameters, and generating an obstacle avoidance track by adopting an improved path search algorithm; optimizing an acceleration curve of the path according to the joint load data of the mechanical arm; the closed-loop verification module is used for carrying out deviation comparison on the actual pose of the tail end of the mechanical arm and the target coordinates, and if the deviation exceeds the limit, the visual positioning and morphology detection module is triggered to cooperatively reset to the path planning and obstacle avoidance module; the visual positioning and morphology detection module and the closed loop verification module are executed in sequence, and output parameters of the previous module serve as input parameters of the next module. Through close cooperation of the modules, efficient and accurate automatic loading and unloading control is achieved, the automation level is improved, and operation safety and accuracy are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor manufacturing technology, and particularly to an automatic loading and unloading control system for a semiconductor carrier plate. Background Art

[0002] In the field of semiconductor manufacturing, the automatic loading and unloading of carrier plates (such as ceramic or metal substrates for holding wafers) is a key link in processes such as chip packaging and testing. With the development of semiconductor devices towards high integration and miniaturization, the thickness of the carrier plate has gradually decreased (<0.3 mm), and the requirement for surface flatness has become increasingly stringent (warpage tolerance <0.1 mm), which poses extremely high requirements for the accuracy, stability, and environmental adaptability of the loading and unloading system. However, the existing technology has the following defects, resulting in frequent problems such as wafer breakage, positioning deviation, and low efficiency during the loading and unloading process:

[0003] Most existing loading and unloading systems adopt a single-wavelength light source and a monocular vision positioning scheme, and segment the edge of the carrier plate through a fixed threshold. However, there are often metal coatings or polishing treatments on the surface of the semiconductor carrier plate, resulting in strong reflection interference. The single-wavelength light source is prone to losing edge information due to overexposure in the reflection area;

[0004] Existing loading and unloading systems usually plan the trajectory of the robotic arm based on a fixed path template or a static environment assumption, without integrating dynamic parameters such as the stacking inclination angle of the carrier plate and the load of the robotic arm joints. For example, an increase in the stacking inclination angle of the carrier plate will reduce the safe movement space, and the traditional A* algorithm does not incorporate the inclination angle parameter into the path cost function, resulting in a conflict between the planned path and the actual position of the carrier plate. Summary of the Invention

[0005] Based on the above objectives, the present invention provides an automatic loading and unloading control system for a semiconductor carrier plate, including:

[0006] Visual positioning and topography detection module:

[0007] Eliminate the reflection interference on the surface of the carrier plate through multispectral imaging, and calculate the center coordinates of the carrier plate in combination with the sub-pixel edge positioning algorithm;

[0008] Generate a three-dimensional point cloud model of the surface of the carrier plate based on binocular vision, and extract the warpage height and edge inclination angle parameters;

[0009] Clamping force dynamic control module:

[0010] Adjust the output torque of the jaw servo motor through a fuzzy control algorithm according to the surface topography parameters and real-time pressure distribution data;

[0011] When a local pressure abnormal change is detected, trigger the high-frequency micro-vibration mechanism to adjust the pose of the carrier plate;

[0012] Path planning and obstacle avoidance module:

[0013] Divide the safety levels of the loading and unloading areas based on the pallet stacking parameters, and use an improved path search algorithm to generate an obstacle avoidance trajectory;

[0014] Optimize the acceleration curve of the path according to the load data of the robotic arm joints;

[0015] Closed-loop verification module:

[0016] Compare the deviation between the actual pose of the robotic arm end and the target coordinates. If the limit is exceeded, trigger the collaborative reset of the visual positioning and morphology detection module to the path planning and obstacle avoidance module;

[0017] The visual positioning and morphology detection module to the closed-loop verification module are executed in sequence, and the output parameters of the previous module are used as the input parameters of the next module.

[0018] Preferably, the specific process of the multispectral imaging includes:

[0019] According to the light reflection characteristics of the pallet material, calibrate the reflectance difference curve in the visible and infrared bands through experiments;

[0020] Select the wavelength combination with the reflectance difference exceeding the preset threshold, and control the annular array light source to alternately light up according to the sequence of the wavelength combination;

[0021] Extract the edge contours of the multi-wavelength images collected by the binocular camera, dynamically allocate weight coefficients according to the signal-to-noise ratio of the edge contrast at each wavelength, and generate a synthetic image resistant to specular reflection interference through weighted fusion.

[0022] Preferably, the sub-pixel edge positioning algorithm includes:

[0023] Detect the gradient direction of the synthetic image, and screen the candidate edge points with the gradient amplitude exceeding the dynamic threshold;

[0024] Perform bilinear interpolation on the candidate edge points to calculate the sub-pixel offset, and generate a three-dimensional point cloud in combination with the binocular disparity model;

[0025] Fit the point cloud plane equation based on the least squares method, calculate the edge inclination angle through the angle between the plane normal vector and the reference plane, and calculate the warping height through the height range of the point cloud.

[0026] Preferably, the parameter adjustment logic of the fuzzy control algorithm includes:

[0027] Calibrate the clamping force-deformation relationship curve of different material pallets through experiments, and establish a fuzzy rule base with material hardness and friction coefficient as input variables;

[0028] Take the deviation amount and deviation change rate between the real-time pressure distribution data and the target pressure threshold as fuzzy input variables, and output the torque correction amount of the servo motor;

[0029] The target pressure threshold is dynamically calculated according to the distribution gradient of the surface warpage height to ensure that the clamping force compensates in the opposite direction to the deformation trend of the material plate.

[0030] Preferably, the triggering conditions of the high-frequency micro-vibration mechanism include:

[0031] Real-time monitor the pressure gradient change rate of each area in the pressure sensor array. When the pressure gradient change rate of a local area is detected to exceed the safety threshold, it is determined as a slip risk;

[0032] The safety threshold is dynamically adjusted according to the product of the surface warpage height of the material plate and the material friction coefficient. The greater the warpage height or the lower the friction coefficient, the smaller the safety threshold.

[0033] Preferably, the obstacle avoidance strategy of the improved path search algorithm includes:

[0034] Obtain the three-dimensional contour data of the obstacle through the ToF sensor, and expand the obstacle boundary outward to form an expansion area. The distance for forming the expansion area is determined according to the motion envelope radius of the end effector of the robotic arm;

[0035] During the path search process, mark the grids within the expansion area as non-passable nodes, and use the dynamic weight A* algorithm to bypass the non-passable nodes;

[0036] The dynamic weight includes the path length weight, the joint torque demand weight, and the acceleration smoothness weight, and each weight coefficient is dynamically allocated according to the current load inertia of the robotic arm.

[0037] Preferably, the specific process of the acceleration curve optimization includes:

[0038] Perform fifth-order polynomial interpolation on the angular displacement sequence of each joint of the robotic arm to generate a continuously differentiable angular acceleration curve;

[0039] According to the maximum instantaneous torque limit of the joint motor, calculate the derivative extreme value of the interpolation curve. If it exceeds the torque limit, reallocate the interpolation nodes;

[0040] Through iterative adjustment of the interpolation coefficients, the actual torque demand is always lower than the preset safety factor of the motor rated torque.

[0041] Preferably, the deviation comparison logic of the closed-loop verification module includes:

[0042] Set the dual judgment conditions of the position deviation threshold and the pressure deviation threshold. The position deviation threshold is dynamically calculated according to 1 / 2 of the material plate size, and the pressure deviation threshold is linearly scaled according to the material brittleness grade;

[0043] When the position deviation exceeds the threshold, the vision positioning module is triggered to recalibrate the coordinates of the material plate; when the pressure deviation exceeds the threshold, the output torque of the clamping force control module is preferentially adjusted.

[0044] Preferably, the system further includes a data collaboration interface:

[0045] The coordinate parameters of the material plate output by the vision positioning module and the input parameters of the path planning module are synchronized in real time through shared memory, and the synchronization period is adaptively adjusted according to the movement speed of the robotic arm;

[0046] The pressure data of the clamping force control module and the robotic arm load data of the path planning module are aligned by a unified timestamp to ensure the timing consistency of the control instructions.

[0047] Preferably, the reset logic of the system includes:

[0048] When the closed-loop verification module detects three consecutive reset instructions, it switches to the safe mode and starts the manual intervention protocol;

[0049] In the safe mode, the robotic arm retracts to the initial position along the preset safe path, and at the same time, the power of the gripper servo motor is turned off and the clamping force is released.

[0050] Advantages of the present invention:

[0051] The high-precision position information provided by the vision positioning module provides reliable basic data for the clamping force control and path planning modules, enabling the entire system to be adjusted according to the actual situation. The clamping force dynamic control module can automatically adjust the output according to the changes in the actual pressure and shape, greatly reducing the need for manual intervention and at the same time reducing the risk of damage to the material plate. The path planning and obstacle avoidance module makes the movement of the robotic arm more stable through intelligent path generation and acceleration optimization, thereby improving the overall loading and unloading efficiency and the service life of the mechanical equipment. The real-time deviation correction mechanism of the closed-loop verification module can ensure that each action conforms to the predetermined goal and responds quickly in case of abnormal situations, avoiding system out-of-control caused by errors. Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It is the structural block diagram of the system of the present invention;

[0054] Figure 2It is the flowchart of the sub-pixel edge localization algorithm in the system of the present invention;

[0055] Figure 3 It is the flowchart of the acceleration curve optimization in the system of the present invention. Specific embodiments

[0056] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0057] Please refer to Figures 1-3 , an automatic loading and unloading control system for semiconductor wafers is provided in an embodiment of the present invention. First, the reflection interference on the surface of the wafer is eliminated by multi-spectral imaging technology. The differences in the reflection characteristics of light with different wavelengths are used to effectively extract edge information, and the sub-pixel edge localization algorithm calculates the center coordinates of the wafer. The accuracy of this step is the basis for the accuracy of the entire loading and unloading system. Then, based on binocular vision technology, a three-dimensional point cloud model of the wafer surface is generated. Through the processing of the point cloud data, the system can extract key parameters such as warping height and edge inclination angle, and these parameters will be transmitted to the subsequent modules as inputs for clamping force control and path planning.

[0058] The clamping force dynamic control module receives the surface topography data and real-time pressure distribution data from the vision positioning and topography detection module, and uses the fuzzy control algorithm to adjust the output torque of the jaw servo motor in real time. In this way, the jaws can dynamically adjust the clamping force according to the deformation of the wafer surface, thereby avoiding damage to the wafer caused by improper clamping. If the system detects an abnormal change in the local pressure, the micro-vibration mechanism will be automatically triggered to adjust the position of the wafer to ensure the clamping stability and further avoid wafer slippage or damage.

[0059] The path planning and obstacle avoidance module divides the safety levels of the loading and unloading area by combining the wafer stacking parameters, and uses an improved path search algorithm to generate an obstacle avoidance trajectory. During the path planning process, this module optimizes the acceleration curve of the path according to the joint load data of the robotic arm to ensure that the robotic arm will not exceed the load capacity during the movement process, thereby avoiding mechanical failures. By dynamically adjusting the acceleration curve of the path, the impact and vibration during the movement are reduced, and the smoothness of the loading and unloading process is improved.

[0060] The closed-loop verification module performs real-time deviation comparison between the actual pose of the end of the robotic arm and the target coordinates. When the pose deviation exceeds the set threshold, the closed-loop verification module triggers the collaborative reset of the aforementioned visual positioning and morphology detection module and the path planning and obstacle avoidance module to ensure that the robotic arm is repositioned and the path is replanned, ensuring the continuous and stable operation of the system. In this way, the system can automatically correct its operation when a deviation occurs, preventing misoperation of the material plate caused by pose deviation.

[0061] In a possible implementation manner, first, the system calibrates the reflectivity difference of the material plate in the visible light and infrared bands through experiments according to the material of the material plate. The reflectivity difference of material plates with different materials in different bands will significantly affect the quality of the image. Therefore, by experimentally measuring the reflectivity curve, the reflection characteristics of the surface of the material plate can be accurately grasped. In this step, the experiment measures the reflectivity of the material plate at different bands and records its reflection characteristics. The finally formed reflectivity difference curve provides a basis for subsequent light source selection and image processing.

[0062] Through the analysis of the reflectivity difference curve, the system can select a wavelength combination in the visible light and infrared bands whose reflectivity difference exceeds the preset threshold. This wavelength combination can maximize the recognition of the surface details of the material plate while avoiding the influence of specular reflection interference. After selecting the appropriate wavelength combination, the system controls the ring array light source to alternately light up in these wavelength orders. The alternating lighting of the light source enables the imaging system to capture the detailed information of the material plate at different bands. Especially in the area with strong surface reflection, the reflection interference is reduced through wavelength selection.

[0063] After the binocular camera is alternately illuminated by light sources of different bands, multi-wavelength images are collected. In this step, the system extracts the edge contours of these images. Since the contrast and signal-to-noise ratio of the edge information may be different in different bands, in order to ensure the quality of the synthesized image, the system dynamically assigns weight coefficients according to the contrast and signal-to-noise ratio of the edges in each band. Specifically, the band with a high signal-to-noise ratio will be given a higher weight, and the band with a low signal-to-noise ratio will be given a lower weight. These weighted images are fused to finally generate a synthesized image resistant to specular reflection interference.

[0064] In a possible implementation manner, first, the system performs gradient direction detection on the synthesized image processed by weighted fusion. Gradient direction detection identifies the regions with significant changes in the image by calculating the local gradient values of the image, and these regions usually correspond to the edges of objects. To further improve the detection accuracy, the system screens out candidate edge points whose gradient amplitudes exceed the threshold according to the set dynamic threshold. The setting of the dynamic threshold takes into account the changes in ambient light and image quality, so it can adapt to the image processing requirements in different situations and ensure the screening accuracy of edge points.

[0065] Once the candidate edge points are screened out, the next task is to accurately determine the actual position of the edge. In this step, the system performs bilinear interpolation calculations on the candidate edge points to obtain sub-pixel-level offsets. Through the bilinear interpolation method, detailed calculations can be carried out between image pixels to determine the accurate position of the edge, achieving sub-pixel-level positioning accuracy. Subsequently, using the binocular disparity model, the system combines the image information obtained from two perspectives, calculates the three-dimensional coordinates of each edge point, and generates a three-dimensional point cloud. The three-dimensional point cloud provides a detailed description of the surface structure of the material plate and is the basis for subsequent plane fitting and warpage height detection.

[0066] Based on the generated three-dimensional point cloud, the system then uses the least squares method to perform plane fitting on the point cloud. The least squares method can effectively fit the point cloud data and find an optimal plane equation to represent the surface of the material plate. By fitting the plane equation, the system can calculate the plane normal vector and solve for the edge inclination angle through the angle between the normal vector and the reference plane, that is, the degree of inclination of the surface of the material plate relative to the reference plane. At the same time, the system can also calculate the warpage height of the material plate based on the height range difference of the point cloud, that is, the height change of the surface of the material plate due to manufacturing or other factors.

[0067] In a possible implementation, first, in order to accurately control the clamping force of material plates of different materials, the system experimentally calibrates the clamping force-deformation relationship curves of material plates of different materials through experiments. These experimental data help the system establish the deformation characteristics of each material under different clamping forces. These relationship curves are used as important bases for further establishing the fuzzy rule base of the fuzzy control algorithm. The input variables in the fuzzy rule base mainly include material hardness and friction coefficient, which directly affect the relationship between the clamping force and deformation of the material plate.

[0068] Materials with high hardness are often not easily deformed and require a smaller clamping force to prevent damage, while materials with a large friction coefficient may require a larger clamping force to ensure stable grasping. Through the rule base of the fuzzy control system, the system can flexibly adjust the clamping force according to these input variables (material hardness, friction coefficient) to ensure that the clamping process will not cause excessive deformation or damage to the material plate.

[0069] During the actual working process, the system will collect pressure distribution data in real time, compare it with the preset target pressure threshold, and calculate the deviation amount and deviation change rate. The target pressure threshold is usually adjusted according to the preset clamping force requirements on the surface of the material plate based on the actual situation. Through the pressure distribution detected in real time, the system can obtain feedback information on the current clamping state of the material plate. Taking these deviation amounts and change rates as input variables of the fuzzy control algorithm, the system will output the torque correction amount of the servo motor to adjust the clamping force.

[0070] This process ensures that the system can make dynamic adjustments based on real-time feedback without relying on static preset conditions. Through the fuzzy control algorithm, the clamping force can be intelligently adjusted according to the changes in the pressure distribution, avoiding damage to the material plate or unstable clamping caused by uneven pressure or inappropriate clamping force.

[0071] To better adapt to the changes in the surface of the material plate, the system also dynamically calculates the distribution gradient of the warpage height of the material plate surface and adjusts the target pressure threshold. The distribution gradient of the warpage height refers to the change rate or non-uniformity of the warpage of the material plate surface. Since warpage usually leads to uneven pressure distribution in different areas of the material plate, the system needs a compensation strategy to ensure the uniformity of the clamping force.

[0072] In actual operation, areas with larger warpage may require smaller clamping force to avoid further deformation, while areas with smaller warpage or flat areas can apply greater clamping force. By dynamically calculating the distribution gradient of the warpage height of the material plate surface, the system can perform reverse compensation, adjust the target pressure threshold, and automatically correct the distribution of the clamping force based on the current warpage state to ensure the reverse compensation between the deformation trend of the material plate and the clamping force.

[0073] In a possible implementation, first, through a pressure sensor array arranged in different areas, the change rate of the pressure gradient in each area is monitored in real time. The pressure gradient refers to the pressure difference between each point on the surface of the material plate, and the change rate of the pressure gradient reflects the change speed of the pressure distribution over time or position. By monitoring the change of the pressure gradient, the system can accurately detect the non-uniformity of the surface pressure of the material plate.

[0074] The pressure sensor array obtains pressure data, converts it into an analyzable value, and then calculates the pressure gradient of each area. If the pressure gradient in a certain area changes too fast, this usually means that the pressure distribution in some parts of the material plate surface has changed drastically, which may cause slippage or instability on the material plate surface. At this time, the system will identify potential slippage risks and respond.

[0075] When the system monitors that the change rate of the pressure gradient in a local area exceeds the set safety threshold, it determines that there is a slippage risk in this area. The slippage risk means that the pressure distribution of the material plate in this area may be uneven, resulting in the material plate being displaced or offset during the automatic loading and unloading process, thus affecting the accuracy or safety of the loading and unloading.

[0076] The safety threshold in the system is not a fixed value. It will be dynamically adjusted according to the actual surface condition and material properties of the material plate. The adjustment of the safety threshold is mainly based on two factors: the warpage height of the material plate surface and the material friction coefficient. Specifically, the greater the warpage height, the more uneven the surface of the material plate; the lower the friction coefficient, the more likely the material plate is to slip during the clamping process. Therefore, when the warpage height is greater or the friction coefficient is lower, the system will automatically lower the safety threshold to more sensitively detect the slip risk.

[0077] When the warpage height of the material plate surface increases, the pressure distribution in some areas is more likely to show violent fluctuations. The system needs to lower the safety threshold to capture the abnormality of the pressure gradient change earlier. On the contrary, when the warpage height is small and the friction coefficient is high, the slip risk is relatively low, and the safety threshold can be appropriately increased to avoid overreaction.

[0078] When the system detects that the pressure gradient change rate in a local area exceeds the safety threshold through the pressure sensor and determines it as a slip risk, the system will trigger a micro-vibration response through the high-frequency micro-vibration mechanism. This high-frequency micro-vibration can adjust the state of the material plate through local tiny vibrations, redistribute the pressure in the local area, and help eliminate the slip risk caused by uneven pressure distribution, thus avoiding problems such as positioning deviation or unstable clamping during the loading and unloading of the material plate.

[0079] In a possible implementation, the system obtains the three-dimensional contour data of the obstacles in the environment through the ToF sensor. The ToF sensor can accurately capture the distance between the obstacle and the sensor by emitting laser and measuring its reflection time, thereby generating a three-dimensional image of the obstacle. These data can describe the spatial position and shape of the obstacle in detail, providing basic information for the path planning algorithm.

[0080] After obtaining the three-dimensional contour data of the obstacle, the system expands the boundary of the obstacle to form an expanded area. This expansion process is to simulate the actual movement range of the end effector of the robotic arm by expanding the boundary of the obstacle outward by a certain distance. The specific expansion distance is determined by the movement envelope radius of the end effector of the robotic arm. The movement envelope radius refers to the spatial range occupied by the end effector of the robotic arm (such as a gripper or a suction cup) during movement. Therefore, the expanded area should cover the space where collisions may occur within this area.

[0081] During the path search process, all grid nodes located within the inflated area are marked as impassable nodes, meaning these areas cannot be used as traversable paths during path planning. In this way, the path search algorithm can avoid collisions between the robotic arm and obstacles. To achieve obstacle avoidance, the system adopts the dynamic weight A* algorithm, which is an improved version of the A* algorithm. It can consider more dynamic factors when calculating the path and find an optimal path that bypasses the impassable nodes.

[0082] The dynamic weight A* algorithm is an extension of the A algorithm and is mainly applied to path planning problems, especially in fields such as robot navigation and autonomous driving. Different from the standard A algorithm, the dynamic weight A* algorithm controls the balance in the search process by dynamically adjusting the weight of the heuristic function to meet different environmental or task requirements.

[0083] In the dynamic weight A* algorithm, the system uses dynamic weights to optimize path selection. The dynamic weights include the following aspects:

[0084] Path length weight: The shorter the total length of the path, the better. The system tends to select shorter paths to reduce movement time.

[0085] Joint torque requirement weight: Considering the dynamic characteristics of the robotic arm, paths with smaller joint torque requirements are preferred during path planning, which helps reduce energy consumption and improve the motion stability of the robotic arm.

[0086] Acceleration smoothness weight: The smoother the acceleration change of the path, the better. The system smooths the acceleration curve of the path to avoid sudden acceleration changes, which helps extend the service life of the robotic arm and reduce unnecessary mechanical wear.

[0087] These weight coefficients are allocated in real time according to the current load inertia of the robotic arm. When the load inertia is large, the system increases the joint torque requirement weight and the smoothness weight to avoid excessive torque burden and acceleration changes. When the load is small or the robotic arm is in an unloaded state, the path length weight is relatively increased, and the system tends to select a path with a shorter length to improve efficiency.

[0088] In a possible implementation, first, the system performs quintic polynomial interpolation on the angular displacement sequences of each joint of the robotic arm. Quintic polynomial interpolation is a commonly used curve fitting technique that can generate a smooth, continuous, and differentiable acceleration curve between given discrete angular displacement data points. Through this interpolation method, it can be ensured that the angular velocity and angular acceleration of each joint of the robotic arm change continuously during the transition process, avoiding acceleration discontinuities and mutations, which are crucial for the motion stability of the robotic arm.

[0089] Since each joint of the robotic arm is driven by a motor, and the maximum instantaneous torque of the motor is limited, there may be points in the generated angular acceleration curve where the derivative extreme value (i.e., the maximum rate of change of acceleration) exceeds the maximum instantaneous torque limit of the motor. To avoid the motor from overloading at these points, the system calculates the derivative extreme value of the interpolation curve and compares it with the maximum torque limit of the joint motor. If it is found that the torque demand at a certain point exceeds the motor's bearing capacity, the system will take measures to redistribute the interpolation nodes, adjust the shape of the curve, and reduce the severity of the acceleration change, thereby ensuring that the motor's torque does not exceed its maximum bearing limit during the entire movement process.

[0090] To further optimize the acceleration curve and ensure that each joint can operate smoothly throughout the movement process, the system iteratively adjusts the interpolation coefficient to gradually adjust the shape of the interpolation curve, so that the actual torque demand always remains within the preset safety factor of the motor's rated torque. This iterative process optimizes the acceleration curve by repeatedly calculating the load conditions of each joint and adjusting the interpolation coefficient in real time to ensure that the system's load does not exceed the motor's rated load.

[0091] In a possible implementation, to effectively monitor the deviation during the system operation, the closed-loop verification module needs to set two key deviation thresholds:

[0092] Position deviation threshold: This threshold is dynamically calculated based on the size of the wafer plate, specifically 1 / 2 of the wafer plate size. The reason for this is that the sizes of semiconductor wafer plates may vary, and by dynamically calculating the position deviation threshold, the system can always maintain an appropriate error tolerance range when processing wafer plates of different sizes. The specific calculation method can be: based on the length and width dimensions of the current wafer plate, dynamically obtain half of this size as the maximum allowable position deviation, so as to control and correct the position error during the movement of the wafer plate.

[0093] Pressure deviation threshold: This threshold is linearly scaled according to the brittleness grade of the wafer plate material. Wafer plates of different materials have different pressure-bearing capacities, and materials with higher brittleness require more precise pressure control. Therefore, by linearly scaling the pressure deviation threshold according to the brittleness grade of the material, the system can control the pressure deviation in real time according to the different requirements of the wafer plate material and avoid damage to the wafer plate caused by excessive pressure.

[0094] When the system detects that the position deviation exceeds the set threshold, the vision positioning module will be triggered to recalibrate the coordinates of the pallet. The vision positioning module monitors the position and attitude of the pallet in real time through a camera or other sensors. Once it finds that the position deviation exceeds the set value, the system will correct the coordinates of the pallet by recalibrating the position. This process can effectively correct the positioning deviation caused by the operation error of the robotic arm or external interference, ensuring that the pallet can be accurately placed and retrieved.

[0095] When the system detects that the pressure deviation exceeds the set threshold, the system will first adjust the output torque of the clamping force control module. The clamping force control module is responsible for adjusting the clamping force of the fixture on the pallet. When the pressure deviation is too large, it may mean that the clamping force is inappropriate, which poses a threat to the safety of the pallet. By adjusting the output torque, the clamping force of the fixture can be effectively controlled to avoid applying excessive pressure to the pallet, thus ensuring that the pallet will not deform or break during handling, especially in the case of brittle materials.

[0096] In a possible implementation, the vision positioning module is responsible for obtaining the real-time coordinate information of the pallet through image recognition technology, while the path planning module needs to plan the motion trajectory of the robotic arm based on this coordinate information. To ensure the high consistency of the data between the two, the system adopts the method of shared memory to synchronize the pallet coordinate parameters output by the vision positioning module and the input parameters of the path planning module in real time.

[0097] Since the motion speed of the robotic arm and the size and characteristics (such as weight, material, etc.) of the pallet will affect the motion cycle, the system designs an adaptive synchronization cycle mechanism. Specifically, the adjustment of the synchronization cycle is automatically optimized based on the current motion speed of the robotic arm. For example, when the motion speed of the robotic arm is relatively fast, the synchronization cycle may be shortened to ensure that the real-time coordinates output by the vision positioning module can be transmitted to the path planning module in a timely manner, avoiding trajectory planning errors caused by data delay. When the motion speed is slow, the synchronization cycle can be appropriately extended to improve the working efficiency of the system.

[0098] The beneficial effect of this design is that it can ensure the high synchronization of visual information and path planning information within the system, eliminate the problem of inaccurate control caused by information lag, and thus improve the accuracy and real-time performance of the automatic loading and unloading process.

[0099] The clamping force control module is responsible for adjusting the clamping force of the fixture, while the path planning module needs to optimize the trajectory planning and control strategy according to the load situation of the robotic arm. To ensure the coordination between the clamping force and the load of the robotic arm, the system accurately aligns the pressure data output by the clamping force control module with the load data of the path planning module by means of unified timestamp alignment.

[0100] All data (including pressure data and load data) is timestamped to ensure that at the same point in time, the path planning module can receive the correct pressure and load data. This alignment effectively avoids control timing errors caused by inconsistent data updates, such as excessive or insufficient clamping force, or incorrect judgment of the robotic arm's load, thereby preventing improper operations.

[0101] The beneficial effect of this step is that by aligning data with a unified timestamp, it can ensure that each control instruction of the system has consistent timing, thereby improving the overall coordination and response speed of the system to load, pressure, and clamping force, and optimizing the accuracy and stability of the entire automatic loading and unloading process.

[0102] Through the design of the data collaboration interface, the system can ensure real-time information sharing and precise synchronization between different modules, avoiding misoperations caused by data delay, inconsistent timing, or information lag. Specifically, the synchronous adjustment of the vision positioning module and the path planning module enables the robotic arm to adjust its movement path in real time to adapt to the changes in the position of the pallet, while the alignment mechanism of the clamping force and the load ensures that the robotic arm will not cause improper clamping or equipment damage due to pressure mismatch during operation.

[0103] This collaborative mechanism enables the system to maintain a high degree of consistency between multiple operation links, reduce errors, improve operation efficiency, enhance the accuracy and reliability in the automation process. At the same time, it can also improve the adaptability of the system. When facing semiconductor pallets of different sizes, weights, and materials, it can automatically adjust the operation strategy to ensure a safe and efficient loading and unloading process.

[0104] In a possible implementation, the system monitors the device status in real time through a closed-loop verification module, especially the issuance of reset instructions. The closed-loop verification module continuously receives instruction information from other parts of the system (such as the robotic arm control module, gripper control module, etc.) and verifies the reset instructions. When this module detects three consecutive reset instructions, it indicates that the system may have encountered an error or fault that cannot be automatically recovered.

[0105] When certain faults or abnormal conditions occur in the system (such as robotic arm jamming, sensor error, operation error, etc.), reset instructions will be issued and the system will attempt to resume operation. If three consecutive reset instructions are issued within a short period of time, the closed-loop verification module will consider that the current state cannot be resolved by a normal reset, and the system should enter the safe mode at this time.

[0106] Once the closed-loop verification module detects three consecutive reset instructions, the system will automatically switch to the safe mode and initiate the manual intervention protocol. The design goal of the safe mode is to isolate the device from the potential dangerous state, prevent the robotic arm from continuing to operate, and avoid causing greater equipment damage or personal injury.

[0107] After entering the safe mode, the system will prompt the operator for manual intervention to check and repair system failures. The operator can obtain specific fault information through the system control interface and take appropriate repair measures. At this time, the system will not continue to perform automatic operations until the manual intervention is completed.

[0108] In the safe mode, the robotic arm will retract along a preset safe path to its initial position. The preset safe path is a path that is pre-planned and does not involve dangerous areas, usually avoiding high-risk or unstable operating areas, such as equipment damage areas or personnel activity areas. This design ensures that the robotic arm will not continue to perform operations that may cause accidents in a fault state.

[0109] The preset safe path ensures that the robotic arm can return to the starting position safely and smoothly, avoiding collisions with other equipment or personnel during this process. This path is usually carefully designed and verified by system engineers to ensure that it has minimal risk and maximally avoids any unsafe factors during the movement of the robotic arm.

[0110] The safe mode also includes the function of turning off the power of the gripper servo motor and releasing the clamping force. The purpose of this is to ensure that the robotic arm does not continue to apply unnecessary clamping force during the return to the initial position, avoiding damage to the material plate or other objects by the gripper. At the same time, cutting off the power of the gripper servo motor can prevent the gripper from continuing to operate under power in a fault state, reducing electrical risks.

[0111] Releasing the clamping force can effectively prevent the gripper from failing to release in time during a system failure, thus avoiding damage to mechanical components or the material plate due to excessive clamping force or the gripper not releasing.

[0112] The design of the reset logic, through steps such as monitoring the reset instruction, switching to the safe mode, retracting the safe path, and turning off the gripper power, ensures that the system can quickly enter a safe state when encountering abnormalities and provides effective guarantee for manual intervention, thus greatly improving the safety and reliability of the automatic loading and unloading control system of semiconductor material plates.

[0113] This invention covers any substitutions, modifications, equivalent methods, and solutions made within the essence and scope of this invention. To enable the public to have a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments of this invention, and those skilled in the art can fully understand this invention even without the description of these details. Additionally, well-known methods, processes, procedures, components, and circuits are not described in detail to avoid unnecessary confusion to the essence of this invention.

[0114] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An automatic loading and unloading control system for semiconductor material plates, characterized in that: include: Visual positioning and shape detection module: Multi-spectral imaging is used to eliminate the interference of reflective light on the surface of the sheet, and the center coordinates of the sheet are calculated by combining the sub-pixel edge positioning algorithm; Generate a 3D point cloud model of the sheet surface based on binocular vision and extract the warpage height and edge inclination parameters; Gripping force dynamic control module: According to the surface topography parameters and real-time pressure distribution data, the output torque of the gripper servo motor is adjusted through the fuzzy control algorithm; When abnormal changes in local pressure are detected, the high-frequency micro-vibration mechanism is triggered to adjust the position of the sheet; Path planning and obstacle avoidance module: The safety level of the loading and unloading area is divided based on the stacking parameters of the sheet materials, and the obstacle avoidance trajectory is generated using an improved path search algorithm; Optimize the acceleration curve of the path based on the robot arm joint load data; Closed-loop verification module: Compare the deviation between the actual position of the end of the robot arm and the target coordinates. If the deviation exceeds the limit, the coordinated reset of the visual positioning and shape detection module to the path planning and obstacle avoidance module is triggered; The visual positioning and shape detection module to the closed-loop verification module are executed in sequence, and the output parameters of the previous module are used as the input parameters of the next module.

2. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The specific process of the multispectral imaging includes: According to the light reflection characteristics of the material of the plate, the reflectivity difference curve in the visible light and infrared bands is calibrated through experiments; Selecting a wavelength combination whose reflectivity difference exceeds a preset threshold, and controlling the annular array light source to light up alternately according to the sequence of the wavelength combination; The edge contours of the multi-wavelength images collected by the binocular camera are extracted, and the weight coefficients are dynamically allocated according to the signal-to-noise ratio of the edge contrast at each wavelength. The weighted fusion is used to generate a synthetic image that is resistant to reflection interference.

3. The automatic loading and unloading control system for semiconductor material plates according to claim 2, characterized in that: The sub-pixel edge positioning algorithm includes: Perform gradient direction detection on the synthetic image and select candidate edge points whose gradient amplitude exceeds the dynamic threshold; Perform bilinear interpolation on candidate edge points to calculate sub-pixel offsets, and generate a 3D point cloud using a binocular disparity model. The point cloud plane equation is fitted based on the least squares method, the edge inclination is calculated by the angle between the plane normal vector and the reference plane, and the warping height is calculated by the point cloud height range.

4. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The parameter adjustment logic of the fuzzy control algorithm includes: Through experimental calibration of the clamping force-deformation relationship curves of plates of different materials, a fuzzy rule library with material hardness and friction coefficient as input variables was established; The deviation between the real-time pressure distribution data and the target pressure threshold value and the deviation change rate are used as fuzzy input variables to output the servo motor torque correction value; The target pressure threshold is dynamically calculated based on the distribution gradient of the surface warpage height to ensure reverse compensation of the clamping force and the deformation trend of the sheet.

5. The automatic loading and unloading control system for semiconductor material plates according to claim 4, characterized in that: The triggering conditions of the high-frequency micro-vibration mechanism include: Monitor the pressure gradient change rate of each area in the pressure sensor array in real time. When it is detected that the pressure gradient change rate of a local area exceeds the safety threshold, it is determined to be a slip risk. The safety threshold is dynamically adjusted according to the product of the warping height of the sheet surface and the friction coefficient of the material. The greater the warping height or the lower the friction coefficient, the smaller the safety threshold.

6. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The obstacle avoidance strategy of the improved path search algorithm includes: The three-dimensional contour data of the obstacle is obtained by using a ToF sensor, and the boundary of the obstacle is expanded outward to form an expansion area, wherein the distance of the expansion area is determined according to the motion envelope radius of the end effector of the robot arm; During the path search process, the grids in the expansion area are marked as inaccessible nodes, and the dynamic weight A* algorithm is used to bypass the inaccessible nodes; The dynamic weights include path length weight, joint torque demand weight and acceleration smoothness weight, and each weight coefficient is allocated in real time according to the current load inertia of the robot arm.

7. The automatic loading and unloading control system for semiconductor material plates according to claim 6, characterized in that: The specific process of the acceleration curve optimization includes: Perform quintic polynomial interpolation on the angular displacement sequence of each joint of the robot arm to generate a continuous and differentiable angular acceleration curve; According to the maximum instantaneous torque limit of the joint motor, the derivative extreme value of the interpolation curve is calculated, and if the torque limit is exceeded, the interpolation node is reallocated; By iteratively adjusting the interpolation coefficient, the actual torque demand is always lower than the preset safety factor of the motor rated torque.

8. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The deviation comparison logic of the closed-loop verification module includes: Setting dual judgment conditions of position deviation threshold and pressure deviation threshold. The position deviation threshold is dynamically calculated according to 1 / 2 of the sheet size, and the pressure deviation threshold is linearly scaled according to the material brittleness grade. When the position deviation exceeds the threshold, the visual positioning module is triggered to recalibrate the material plate coordinates; when the pressure deviation exceeds the threshold, the output torque of the clamping force control module is adjusted first.

9. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The system also includes a data collaboration interface: The material plate coordinate parameters output by the visual positioning module and the input parameters of the path planning module are synchronized in real time through shared memory, and the synchronization cycle is adaptively adjusted according to the movement speed of the robot arm; The pressure data of the gripping force control module and the robot arm load data of the path planning module are aligned through a unified timestamp to ensure the timing consistency of the control instructions.

10. The automatic loading and unloading control system for semiconductor material plates according to claim 1, characterized in that: The reset logic of the system includes: When the closed-loop verification module detects three consecutive reset instructions, it switches to safe mode and starts the manual intervention protocol; In safety mode, the robot arm returns to its initial position along a preset safety path, while the power supply to the gripper servo motor is turned off and the gripping force is released.

Images