Collision detection method and system for wafer transfer robot

Abstract

The invention belongs to the technical field of semiconductor manufacturing safety control, and particularly relates to a collision detection method and system for a wafer transfer robot, and the method comprises the steps: receiving a joint angle sequence and height coordinates outputted by a robot trajectory planning module; cartesian coordinates of all nodes of the mechanical arm are calculated through a forward kinematics solver; preferentially checking whether the node is located in a predefined safety area or not by utilizing a point area judgment and line segment area detection algorithm; performing regionalized wafer collision detection on the wafer outside the safe region; if the wafer collision does not occur, constructing an oriented bounding box (OBB) for the risk connecting rod; detecting whether the vertex of the OBB crosses a boundary or not, and performing line segment intersection detection on each edge of the OBB and a chamber boundary line segment of a corresponding area; and finally outputting a collision judgment result, and deciding whether to execute motion or not by a robot control system. According to the invention, efficient and accurate collision detection is realized through a multi-stage detection mechanism and highly adaptive section selection, and the method has the advantages of high real-time performance, low cost, easy integration and the like.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of safety control of semiconductor manufacturing, in particular to a collision detection method and system of wafer transfer robot. BACKGROUND

[0002] In the field of semiconductor manufacturing, wafer transfer robots need to run at high speed in a complex multi-chamber environment, and the collision risk directly leads to the damage of high-value wafers and equipment downtime. There are three major technical routes in the traditional collision detection scheme:

[0003] 1. Real-time monitoring based on physical sensors: through laser radar, force sensor and other hardware to collect data in real time, but the vacuum chamber environment has requirements for vacuum degree and cleanliness, which limits the deployment of sensors, and high-precision sensors are high in cost and difficult to modify the chamber.

[0004] 2. General collision detection algorithm: using bounding volume hierarchy (BVH) or Gilbert-Johnson-Keerthi (GJK) algorithm, although it can be implemented in software, but the calculation complexity is O(n log n), which is difficult to meet the millisecond-level real-time response requirement.

[0005] 3. Cylinder coordinate robot joint value constraint algorithm: according to the characteristics of the cylindrical coordinate robot commonly used in wafer transmission field, a plurality of workspaces with limited azimuth angle, extension distance, lifting shaft and other attributes are set. By constraining the joint values of the end effector in the workspace, the motion path is constrained and the collision is warned. It has good real-time performance and simple logic, but it is not sensitive to the collision volume of other structures except the end effector, so it cannot be extended to the lifting multi-link planar joint robot of complex chamber model.

[0006] Therefore, there is an urgent need to design a collision detection method of wafer transfer robot which can be extended to complex chamber model. SUMMARY

[0007] The application aims to provide a collision detection method and system for lifting multi-link planar joint wafer transfer robot applied to complex structure chamber, which receives the joint angle values as input in the trajectory planning process, performs forward kinematics solving based on the preset mechanical structure parameters of the robot, and obtains the spatial pose of each link of the mechanical arm and the end effector. Through multi-level collision detection mechanism, real-time collision analysis is carried out between the geometric configuration obtained by solving and the predefined chamber model, and finally the collision risk judgment result is output. The robot control system decides whether to stop the current motion command according to the judgment result, so as to avoid the collision accident between the robot and the chamber structure, and ensure the safe transmission of wafers and stable operation of the equipment.

[0008] The application is a wafer handling robot collision detection method, comprising the following steps:

[0009] S1: receiving the joint angle sequence and height coordinates output by the robot trajectory planning module in real time;

[0010] S2: solving by a forward kinematics solver, based on the input joint angles and link lengths, calculating the Cartesian coordinates of each node of the robot arm; by point area judgment algorithm and line segment area detection algorithm, preferentially checking whether all nodes are located within the predefined safety area, if so, directly determining no collision risk; otherwise, according to the current height, adaptively selecting the corresponding chamber cross-section model;

[0011] S3: performing regional wafer collision detection on wafers located outside the predefined safety area, and if a collision is detected, terminating the process and alarming; if no wafer collision is detected, constructing an oriented bounding box (OBB) for the risk link whose end point is not in the safety area;

[0012] S4: detecting whether the vertices of the oriented bounding box (OBB) are out of bounds, if so, determining a collision; if the vertices are not out of bounds, performing regional detection on each edge of the oriented bounding box (OBB), and performing accurate line segment intersection detection with the corresponding region's chamber boundary line segment;

[0013] S5: outputting the final collision determination according to the detection result, and deciding by the robot control system whether to execute the movement.

[0014] Before performing step S1, a height-cross-section-region-line segment mapping database is also constructed, specifically:

[0015] S1-1: horizontally slicing the three-dimensional model of the semiconductor process machine chamber at different heights to obtain multiple horizontal cross-sections of height layers;

[0016] S1-2: converting the chamber cross-section boundaries of each height layer into a boundary representation composed of line segments;

[0017] S1-3: according to the chamber structure characteristics, dividing the horizontal cross-sections into three categories according to height intervals, including Process chamber cross-section, LoadLock chamber cross-section and Home cross-section;

[0018] S1-4: boundary line segment partitioning step: based on the chamber structure characteristics, by the preset longitudinal region boundary lines X1 and X2 and the transverse region boundary line Y, dividing the chamber cross-section boundaries of each height layer into six regions;

[0019] S1-5: for each chamber boundary line segment, establishing a mapping relationship between the line segment and the region, and forming a predefined safety area;

[0020] S1-6: Database construction step: associate and map the height interval, cross-section type, region number and boundary line segment to form a multi-level mapping database of height-cross-section-region-line segment, which is used for multi-level real-time collision detection. The coordinates of the 4 vertices of the OBB are received and added to the detection list. After traversal, the OBB list is output for subsequent collision detection.

[0021] In step S2, the point region judgment algorithm is as follows:

[0022] S2-1: Receive the Cartesian coordinates (x, y) of each node of the mechanical arm to be judged.

[0023] S2-2: For each node of the mechanical arm: compare the x coordinate of the point with the preset longitudinal region boundary line X1 and X2; compare the y coordinate of the point with the preset transverse region boundary line Y.

[0024] S2-3: Determine the region number of the point according to the comparison result and output the region number 1-6.

[0025] Wherein, the region division rule is: when xX1 and yY, it is judged as region 4; when X1≤xX2 and yY, it is judged as region 5; when xX2 and yY, it is judged as region 6.

[0026] S2-4: Output the region number, which is used to retrieve the corresponding region boundary line segment from the height-cross-section-region-line segment mapping database, providing a regionalized data basis for subsequent collision detection.

[0027] In step S2, the line segment region detection algorithm is as follows:

[0028] S3-1: For a given line segment starting point P1(x1, y1) and ending point P v (x2, y2), calculate the intersection point of the line segment and the region boundary line using the line segment parameter equation.

[0029] The line segment parameter equation is represented as:

[0030]

[0031] S3-2: For the vertical region boundary line x bound , calculate the intersection point parameter t:

[0032]

[0033] Wherein, ε is the floating point precision tolerance, ε=10 -7 ;

[0034] Calculate the y coordinate of the intersection point:

[0035] y intersect = y1 + t · (y2 - y1)

[0036] S3-3: For the horizontal region boundary line y bound , calculate the intersection parameter t:

[0037]

[0038] Calculate the x coordinate of the intersection point:

[0039] x intersect = x1 + t · (x2 - x1)

[0040] S3-4: Collect the line segment endpoints and each valid intersection point, and the region information of t ∈ [0, 1], output the set of all region numbers that the line segment passes through.

[0041] In step S2, the regionized wafer collision detection includes the following steps:

[0042] S4-1: Input wafer center coordinates P (x, y); provide point region determination algorithm to determine the region where the point is located;

[0043] S4-2: Obtain all chamber boundary line segments of the region and adjacent regions from the height-section-region-line segment mapping database;

[0044] S4-3: Based on vector projection, the shortest distance from the point to the line segment is:

[0045]

[0046] S4-4: Record the minimum value d min in all distances, compare d min and wafer radius R; if d min <R, it is determined that a collision occurs; otherwise, it is determined that no collision occurs.

[0047] In step S3, if no wafer collision is detected, an OBB construction algorithm is used to construct a directional bounding box OBB for the risk link whose endpoints are not in the safe region, including the following steps:

[0048] S5-1: Input the coordinates of the two endpoints of the risk link and the link width w;

[0049] S5-2: Calculate the direction vector of the link: v = endpoint2 coordinates - endpoint1;

[0050] S5-3: If the modulus of the direction vector is zero, generate a square oriented bounding box (OBB) centered at the endpoint 1 with a side length of the link width; otherwise, calculate the unit tangent vector t of the direction vector; calculate the unit normal vector n as the perpendicular vector of the tangent vector;

[0051] S5-4: Extend the endpoint of the link along ±t tangent by 1 / 2w to obtain two extension points; extend from each extension point along ±n normal by 1 / 2w to obtain four vertices, constituting an oriented bounding box (OBB).

[0052] The line segment intersection detection algorithm comprises the following steps:

[0053] S6-1: For two line segments A and B, the starting point of line segment A is a1, and the ending point is a2, the starting point of line segment B is b1, and the ending point is b2;

[0054] S6-2: Calculate the direction vector of line segment A as: Calculate the direction vector of line segment B as:

[0055] S6-3: Perform a fast rejection test to detect the overlap of the AABB bounding box, and output non-intersection if no overlap of the AABB bounding box is detected; otherwise, perform a cross-standing test;

[0056] S6-4: Perform the cross-standing test using the parametric equation method, and determine the intersection of the two line segments by calculating the intersection position.

[0057] The step 6-4 is specifically:

[0058] a) The cross-standing test is performed using the parametric equation method, and the expression is:

[0059]

[0060] wherein, represents the starting point of line segment A; represents the ending point of line segment A; represents the starting point of line segment B; represents the ending point of line segment B; is the cross product denominator, representing the direction relationship of the two line segments, and the calculation formula is t is the proportional position of the intersection point on line segment A; u is the proportional position of the intersection point on line segment B;

[0061] b) When t = 0, the intersection point is on the starting point of line segment A; when t = 1, the intersection point is on the ending point of line segment A;

[0062] c) When u = 0, the intersection point is on the starting point of line segment B; when u = 1, the intersection point is on the ending point of line segment B;

[0063] d) If t ∈ [0, 1] and u ∈ [0, 1], the line segments intersect; otherwise, they do not intersect.

[0064] A collision detection system of a wafer handling robot collision detection method, comprising:

[0065] A robot control system for outputting motion parameters including joint angle sequences and height coordinates in a trajectory planning process to an input module;

[0066] An input module for receiving joint angle sequences and Z-axis height parameters output by the robot control system in the trajectory planning process;

[0067] A forward kinematics solver for performing forward kinematics calculation on the joint angles based on preset mechanical structure parameters to obtain a set of Cartesian coordinates of each node of the robot arm;

[0068] A multi-stage real-time collision detection unit for performing real-time collision detection between the robot links, end effector, and wafer carrier and a predefined chamber model based on the set of node coordinates, determining whether there is a collision risk, and outputting a collision determination result to an output module;

[0069] An output module for feeding back the determination result to the robot control system to control whether the robot continues to execute the trajectory or aborts and alarms.

[0070] The multi-stage real-time collision detection unit comprises a node coordinate determination module, a cross-section height adaptive module, a regional wafer collision detector, an oriented bounding box (OBB) generator, a boundary out-of-bounds detector, a regional line segment retriever, and a line segment intersection detector.

[0071] The node coordinate determination module is configured to use a ray method to check whether all nodes in the set of spatial coordinates are located within a preset safety zone, and output a no collision determination when all nodes are within the safety zone.

[0072] The cross-section height adaptive module is configured to select a corresponding chamber cross-section model according to the Z-axis height parameter when there is a node that is not within the safety zone, and call the boundary line segments from a height-cross-section-region-line segment mapping database.

[0073] The regional wafer collision detector is configured to receive wafer center coordinates, determine the region where the wafer center is located through a point region determination algorithm, obtain all chamber boundary line segments of the region and adjacent regions from a height-cross-section-region-line segment mapping database, calculate the shortest distance from the wafer center to each boundary line segment based on a vector projection principle, and output a collision determination when the calculated shortest distance is less than the wafer radius.

[0074] An oriented bounding box (OBB) generator is configured to construct an OBB for a risk link whose end points are not in a predefined safe area: receiving two end point coordinates and a link width w of the risk link; calculating a direction vector of the link; if the direction vector module is zero, generating a square OBB with the end points as the center and the side length as w; otherwise, calculating a unit tangent vector and a unit normal vector of the link; expanding the end points of the link by ±1 / 2w along the tangent vector to generate two expanded points; expanding each expanded point by ±1 / 2w along the normal vector to obtain four vertex coordinates, which constitute an OBB bounding box;

[0075] A boundary out-of-bounds detector is configured to determine whether the vertex of the OBB bounding box is out of the boundary of the chamber by using a ray method;

[0076] A regionalized line segment retriever is configured to apply a line segment regional detection algorithm based on the region where the OBB bounding box is located, determine the region to which each side of the OBB bounding box belongs, and retrieve the chamber boundary line segment of the corresponding region from a height-section-region-line segment mapping database;

[0077] A line segment intersection detector is configured to determine the intersection relationship between the OBB side and the chamber boundary line segment based on a parametric equation method, and output a collision determination when intersection is detected.

[0078] The present application has the following advantages and benefits:

[0079] 1. The present application discards the physical sensor scheme and realizes collision detection by using software algorithm, completely avoids the restriction of sensor deployment in the vacuum chamber environment, does not need to modify the chamber structure, greatly reduces the system cost and maintenance difficulty;

[0080] 2. The present application has high calculation efficiency and good real-time performance: the present application optimizes the calculation complexity to the level close to O(1) through a multi-level collision detection mechanism (safe area filtering, regionalized line segment retrieval, height self-adaptive chamber selection), and can realize millisecond-level response (≤1ms) on a standard industrial computer, meeting the real-time requirements of high-speed wafer handling;

[0081] 3. The present application not only detects the end effector, but also covers all the link structures of the robot, accurately describes the robot envelope space through the oriented bounding box (Oriented Bounding Box, hereinafter referred to as OBB) generation technology, and is suitable for multi-link planar joint robots in complex chamber environment;

[0082] 4. The present application has a spatial dimension decoupling function: the present application realizes a height layering mapping mechanism according to the motion characteristics of the lifting planar joint robot and the structure characteristics of the semiconductor manufacturing equipment vacuum chamber, converts the three-dimensional collision detection problem into a height-related two-dimensional plane detection sequence, and realizes a stepwise improvement in calculation efficiency while maintaining detection accuracy.

[0083] 5.The application is seamlessly integrated with the robot control system through a standard data interface, intercepts dangerous paths in real time during trajectory planning, effectively avoids damage to high-value wafers and equipment downtime accidents, and ensures continuous and stable operation of the semiconductor production line. BRIEF DESCRIPTION OF DRAWINGS

[0084] Figure 1 Work timing flow chart of the application;

[0085] Figure 2 Real-time collision detection system working principle flow chart of the application;

[0086] Figure 3 Wafer robot and chamber structure schematic diagram of the application;

[0087] Figure 4 Chamber model preprocessing flow chart of the application;

[0088] Figure 5 Process cavity section and partition schematic diagram in the chamber model of the application;

[0089] Figure 6 LoadLock cavity section and partition schematic diagram in the chamber model of the application;

[0090] Figure 7 Home section and partition schematic diagram in the chamber model of the application;

[0091] Figure 8 Safe area defined on the chamber boundary at each height in the application;

[0092] Figure 9 Algorithm flow chart of the point area determination algorithm of the application;

[0093] Figure 10 Algorithm flow chart of the line segment area detection algorithm of the application;

[0094] Figure 11 Regional wafer collision detection method flow chart of the application;

[0095] Figure 12 Risk link determination and OBB list output method flow chart of the application;

[0096] Figure 13 Algorithm flow chart of the OBB construction algorithm of the application;

[0097] Figure 14 Chamber boundary, safe area, link, wafer, risk link OBB schematic diagram in the application;

[0098] Figure 15 Method flow chart of the line segment intersection detection algorithm of the application. DETAILED DESCRIPTION

[0099] The application will be further described in detail below with reference to the accompanying drawings and examples.

[0100] As Figure 1 shown, it is a working timing flow chart of the application, wherein, Figure 1 Label 1 in the figure is the trajectory planning module of the robot control system, and the application does not involve the improvement of the trajectory planning algorithm itself, which belongs to the basic functional module of the robot control system, and the function is to receive the target pose of the robot (such as the wafer transmission path point), generate the joint angle sequence through the kinematics algorithm, output the time-angle parameterized trajectory, generate the discrete path point, and provide the input data for the collision detection system. Label 2 is the main content of the application, which receives each joint angle and Z-axis height output by the trajectory planning module, and outputs the collision judgment to the robot control system, which decides to continue the motion or aborts the motion planning and alarms according to the judgment result.

[0101] As Figure 2 shown, it is a working principle flow chart of the real-time collision detection system of the application, and the wafer handling robot collision detection method of the application is realized based on the collision detection system, which includes:

[0102] The robot control system is used for outputting the motion parameters including the joint angle sequence and the height coordinate in the trajectory planning process to the input module.

[0103] The input module is used for receiving the joint angle sequence and the Z-axis height parameter output by the robot control system in the trajectory planning process; as Figure 2 shown by label 1 in the figure, it is the input module of the joint angle and the height, which receives the joint angle sequence (θ1-θ n ) and the height coordinate (Z) of the robot control system.

[0104] The forward kinematics solver is used for performing forward kinematics calculation on the joint angle based on the preset mechanical structure parameters to obtain a set of Cartesian coordinates of each node of the mechanical arm; as Figure 2 shown by label 2 in the figure, the forward kinematics solver calculates the Cartesian coordinates of each node based on the geometric parameters of the mechanical arm, and outputs a set of spatial positions of the base / joint / end effector {A0, A1,..., P n}.

[0105] The multi-stage real-time collision detection unit is used for performing real-time collision detection between the robot connecting rod, the end effector and the wafer carried by the robot and the pre-defined cavity model based on the set of node coordinates, judging whether there is a collision risk, and outputting the collision judgment result to the output module.

[0106] The output module is used to feed back the determination result to the robot control system to control whether the robot continues to execute the trajectory or stops and issues an alarm.

[0107] The multi-level real-time collision detection unit includes: a node coordinate determination module, a cross-section height adaptive module, a regionalized wafer collision detector, an oriented bounding box (OBB) generator, a boundary crossing detector, a regionalized line segment retrieval device, and a line segment intersection detector; corresponding to the following respectively: Figure 2 Labels 3 to 9 are shown;

[0108] Node coordinate determination module (e.g.) Figure 2 (As shown in label 3), it is used to verify whether all nodes in the spatial coordinate set are located within the preset safe zone using the ray method. When all nodes are within the safe zone, a no-collision judgment is output.

[0109] The cross-section height adaptive module (such as) Figure 2 As shown in label 4), it is used to select the corresponding chamber cross-section model based on the Z-axis height parameter when there are nodes outside the safe zone, and to call the boundary line segment from the height-cross-section-region-line segment mapping database.

[0110] Regionalized wafer collision detectors (e.g.) Figure 2 (As shown in label 5) is used to receive the wafer center coordinates, determine the region where the wafer center is located through a point region determination algorithm; obtain all chamber boundary line segments of the region and adjacent regions from the height-section-region-line segment mapping database; calculate the shortest distance from the wafer center to each boundary line segment based on the vector projection principle; when the calculated shortest distance is less than the wafer radius, output a collision determination.

[0111] Oriented bounding box OBB generator (e.g.) Figure 2 As shown in label 6), this is used to construct an oriented bounding box for risky links whose endpoints are not within a predefined safety area: It receives the coordinates of the two endpoints of the risky link and the link width w; calculates the link's direction vector; if the magnitude of the direction vector is zero, it generates a square OBB centered at the endpoint with a side length w; otherwise, it calculates the link's unit tangent vector and unit normal vector; it extends the link endpoints along the tangent vector by ±1 / 2w to generate two extension points; it extends each extension point along the normal vector by ±1 / 2w to obtain four vertex coordinates, forming the OBB bounding box.

[0112] Boundary crossing detectors (such as...) Figure 2 (As shown in label 7), used to determine whether the vertices of the OBB bounding box exceed the chamber boundary using the ray method;

[0113] Regionalized line segment searcher (e.g.) Figure 2(As shown in label 8), it is used to apply the line segment region detection algorithm based on the region where the OBB bounding box is located, determine the region to which each side of the OBB bounding box belongs, and retrieve the corresponding chamber boundary line segment from the height-section-region-line segment mapping database;

[0114] Line segment intersection detector (e.g.) Figure 2 As shown in label 9, it is used to determine the intersection relationship between the OBB edge and the chamber boundary line segment based on the parametric equation method, and outputs a collision determination when an intersection is detected.

[0115] like Figure 2 As shown, based on the above system, the present invention provides a collision detection method for a wafer handling robot, comprising the following steps:

[0116] S1: Receives the joint angle sequence and height coordinates output in real time by the robot trajectory planning module;

[0117] S2: Solve using a forward kinematics solver. Based on the input joint angles and link lengths, calculate the Cartesian coordinates of each node of the robotic arm. Using a point region judgment algorithm and a line segment region detection algorithm, first check whether all nodes are located within a predefined safe area. If so, directly determine that there is no collision risk. Otherwise, adaptively select the corresponding chamber cross-section model based on the current height.

[0118] S3: Perform regionalized wafer collision detection on wafers located outside the predefined safe area. If a collision is detected, terminate the process and issue an alarm; if no wafer collision is detected, construct an oriented bounding box (OBB) for risk links whose endpoints are not within the safe area.

[0119] S4: Detect whether the vertices of the oriented bounding box (OBB) have crossed the boundary. If they do, a collision is determined. If the vertices have not crossed the boundary, perform regional detection on each side of the OBB and perform precise line segment intersection detection with the cavity boundary line segments of the corresponding region.

[0120] S5: Output the final collision determination based on the detection results, and let the robot control system decide whether to execute the movement.

[0121] like Figure 3 As shown, the environment used in this embodiment is a cavity wall structure, which comes from a semiconductor process machine with 10 modular chambers, mainly rectangular chambers. The wafer robot uses lifting and telescopic movements to transfer wafers between the modular chambers to perform various process steps on the wafers, thereby ensuring an efficient and precise manufacturing process, reducing manual intervention and improving yield.

[0122] After obtaining the chamber structure model, it is necessary to preprocess the chamber model before executing step S1 and construct a height-section-region-line segment mapping database.

[0123] As Figure 4 shown, the present embodiment is first manually sliced;

[0124] S1-1: Draw the cross-sectional boundary, divide it into three groups according to different chamber boundary characteristics, and obtain multiple horizontal cross-sections of different height intervals, namely Process chamber cross-sections ( Figure 5 ) with height interval [51, 17), LoadLock chamber cross-sections ( Figure 6 ) with height interval [13, 50), and Home chamber cross-sections ( Figure 7 ) with other height intervals,

[0125] S1-2: Abstract the boundary line segments on the above cross-sections into point sets, and write them into the configuration file together with the height intervals, thus realizing the mapping from the real chamber to the virtual space.

[0126] According to the characteristics of the chamber structure, when the robot is in some positions, there is no possibility of collision with the chamber boundary, which is referred to as a safe zone in the present invention. By predefining the safe zone, the computational load can be greatly reduced in the area where collision detection is not necessary. The safe zone in this example is shown in Figure 8 .

[0127] S1-3: According to the characteristics of the chamber structure, it is also necessary to partition each chamber boundary sliced out, so that the collision detection algorithm only detects the chamber boundary in the area where the robot is located, reducing the computational load. In this example, according to the characteristics of the chamber model, by predefining two longitudinal region dividing lines (X1, X2) and one transverse region dividing line (Y), all the chamber boundaries at different heights are divided into six regions, as shown by the dashed lines in Figure 5 , Figure 6 and Figure 7 .

[0128] S1-4: For each chamber boundary line segment, establish a mapping relationship between the line segment and the region to form a predefined safe region;

[0129] S1-5: Database construction step: associate and map the height interval, cross-section type, region number, and boundary line segment to form a multi-level mapping database of height-cross-section-region-line segment, which is used to receive the coordinates of the 4 vertices of the OBB in multi-level real-time collision detection and add them to the detection list. After traversal, output the detected OBB list for subsequent collision detection.

[0130] The core of the boundary partitioning algorithm is to determine the region of each chamber boundary line segment composed of two points, which involves the region detection algorithm. The region detection algorithm consists of two parts: 1. Point region judgment algorithm; 2. Line segment region detection algorithm.

[0131] As Figure 9is a flow chart of the point region determination algorithm applied in the present embodiment. The algorithm is input with point Cartesian coordinates (x, y) and will output region number [1-6]. In step S2 of the present embodiment, the point region determination algorithm, in particular,

[0132] S2-1: Receive the Cartesian coordinates (x, y) of each node of the robot arm to be determined;

[0133] S2-2: For each node of the robot arm: compare the x coordinate of the point with the preset longitudinal region boundary line X1, X2; compare the y coordinate of the point with the preset transverse region boundary line Y;

[0134] S2-3: Determine the region number of the point according to the comparison result, and output the region number 1-6;

[0135] Wherein, the region division rule is: when x < X1 and y < Y, it is determined as region 1; when X1≤x < X2 and y < Y, it is determined as region 2; when x ≥ X2 and y < Y, it is determined as region 3; when x < X1 and y ≥ Y, it is determined as region 4; when X1≤x < X2 and y ≥ Y, it is determined as region 5; when x ≥ X2 and y ≥ Y, it is determined as region 6;

[0136] S2-4: Output the region number, which is used to retrieve the boundary line segment of the corresponding region from the height-section-region-line segment mapping database, to provide the regionalized data basis for subsequent collision detection.

[0137] As Figure 10 is a line segment region detection algorithm applied in the present embodiment, which is used in the calculation process. The line segment region detection algorithm in step S2 of the present embodiment has the following specific process:

[0138] S3-1: The algorithm is input with the Cartesian coordinates of the starting point (x1, y1) and the ending point (x2, y2) of the line segment, and will output the set of region numbers passed by the line segment. Wherein, the parameter equation method for calculating the intersection point coordinates of the line segment and the region boundary line is:

[0139] Line segment parameter equation

[0140] Let the starting point P1 (x1, y1) and the ending point P2 (x2, y2) of the line segment be represented by the parameter equation:

[0141]

[0142] S3-2: Intersection point perpendicular to region boundary line

[0143] Input: Region boundary line coordinate x bound , line segment end point (x1, y1), (x2, y2)

[0144] Computational formula:

[0145]

[0146] y intersect = y1 + t · (y2 - y1)

[0147] Output: intersection coordinates (x bound ,y intersect ), valid condition: t∈[0,1], where ε = 10 -7 is the floating-point precision tolerance;

[0148] S3-3: intersection of horizontal region boundary line

[0149] Input: region boundary line coordinates y bound , line segment endpoints (x1,y1),(x2,y2)

[0150] Computational formula:

[0151]

[0152] x intersect = x1 + t · (x2 - x1)

[0153] Output: intersection coordinates (x intersect ,y bound )

[0154] Valid condition: t∈[0,1]

[0155] Through the above line segment region detection algorithm, all regions to which the height chamber boundary line segments belong can be obtained. As shown in Figure 4 , finally, the height-section-region-line segment mapping database constructed is stored in a structured manner for collision detection calls.

[0156] Collision detection system description:

[0157] As shown in Figure 1 , the collision detection system works in the robot control system. The working sequence is that the trajectory planning module outputs joint angles θ and Z-axis height z loc and other parameters during the planning process, and the collision detection system receives the parameters to start collision detection.

[0158] As shown in Figure 2 , in step S2 of the embodiment, the collision detection system performs forward kinematics solving after receiving the parameters. The forward kinematics solver is input with the joint angles, and the module outputs the Cartesian coordinates of nodes A2, A3, P1 and P2, etc. through the preset link length calculation Figure 4 Computational formula:

[0159] A2 = A1 + L1 · [cosθ1, sinθ1]

[0160] A3 = A2 + L2 · [cos(θ1 + θ2), sin(θ1 + θ2)]

[0161] P1 = A3 + L3 · [cos(θ1 + θ2 + θ3), sin(θ1 + θ2 + θ3)]

[0162] P2 = A3 + L4 · [cos(θ1 + θ2 + θ4), sin(θ1 + θ2 + θ4)]

[0163] Wherein, the division of the nodes is as shown in Figure 3 Figure 3 Label 1 is the structure of the inner wall of the semiconductor process machine chamber, and the wafer transfer robot cannot generate friction and collision with any inner wall when performing movement. Figure 3 Label 2 is the base rotating shaft A1 of the lifting three-link plane joint wafer transfer robot, the horizontal position of the base rotating shaft A1 is fixed and unchanged, and it can perform vertical lifting movement within a certain range, Figure 3 Label 3 is the large arm L1, which performs horizontal circular motion around the base rotating shaft A1, and when the rotation angle is within a certain limited range, the large arm L1 will not have any collision risk with the chamber boundary. Figure 3 Label 4 is the elbow rotating shaft A2, Figure 3 Label 5 is the small arm L2, which performs horizontal circular motion around the elbow rotating shaft A2, and the L2 motion space is limited within the rectangular main chamber. Figure 3 Label 6 is the nested concentric end effector rotating shaft A3 and A4 of the upper and lower execution arms. Figure 3 Label 7 is the upper execution arm L3, which performs circular motion around the end effector rotating shaft A3, Figure 3 Label 8 is the lower execution arm L4, which performs circular motion around the end effector rotating shaft A4. The upper execution arm L3 is higher than the lower execution arm L4 in spatial position relationship by D (mm). The ends of the upper execution arm L3 and the lower execution arm L4 each carry a wafer with a size of 300 mm, i.e. label 9 and label 10, and the centers of the wafers are respectively at P1 and P2 in the horizontal direction. Figure 11 Label 11 is a module chamber on a part of the semiconductor process machine, and the wafers P1 and P2 and the upper and lower execution arms L3 and L4 carrying them can move into the module chamber at a suitable height. The robot transmits the wafers between various module chambers through lifting and stretching movements to perform various process processing steps on the wafers, thereby ensuring efficient and accurate manufacturing process, reducing manual intervention and improving yield.

[0164] ​According to the obtained node horizontal coordinates, input into the node coordinate judgment module, and use the ray method to check whether the node is contained in the preset safety zone. When all nodes are contained in the safety zone, it is determined that there is no collision, otherwise the cross-section height adaptive module is executed.

[0165] By setting the robot Z-axis height z loc Input into the cross-section height adaptive module, the motion boundary of each of the upper and lower execution arms L3 and L4 can be obtained from the height-cross-section-region-segment mapping database.

[0166] In this embodiment, P1 is the center coordinate of wafer 1 carried by the upper execution arm L3, and P2 is the center coordinate of wafer 2 carried by the lower execution arm L4. When there is a wafer center coordinate not in the preset safety zone, the wafer is subjected to regional wafer collision detection.

[0167] As shown in the step S3 of the embodiment, the specific steps of the regional wafer collision detection algorithm are as follows: Figure 12

[0168] S4-1: input wafer center coordinate P(x, y); provide point region judgment algorithm to determine the region where the point is located;

[0169] S4-2: obtain all chamber boundary line segments of the region and adjacent regions from the height-cross-section-region-segment mapping database;

[0170] S4-3: based on vector projection, the shortest distance of the point to the line segment is calculated as:

[0171]

[0172] S4-4: record the minimum value d min in all distances, compare d min with the wafer radius R; if d min <R, it is determined that a collision occurs; otherwise, it is determined that there is no collision.

[0173] If no collision of the wafer is detected, an OBB (Oriented Bounding Box) is constructed for the risky link, as shown in Figure 13 The specific process is as follows:

[0174] Receive all node coordinates of the robot as input, traverse all link node coordinates, if both endpoints of the link are in the preset safety zone, do not perform OBB construction for the link. If any endpoint coordinate is not in the preset safety zone, use the OBB construction algorithm to construct the OBB (Oriented Bounding Box) of the link, and receive the coordinates of the 4 vertices of the OBB output by the algorithm to add to the detection list. After the traversal is completed, output the OBB list to be detected.​

[0175] The OBB construction algorithm receives the coordinates of the two end points of the link and the link width as input. The OBB construction algorithm flow is as shown in Figure 14 , and specifically includes the following steps:

[0176] S5-1: input the two end point coordinates and the link width w of the risk link;

[0177] S5-2: calculate the direction vector of the link as: v = end point 2 coordinates - end point 1;

[0178] S5-3: if the modulus of the direction vector is zero, generate a square directional bounding box OBB centered at end point 1 with a link width; otherwise, calculate the unit tangent vector t of the direction vector; calculate the unit normal vector n as the perpendicular vector of the tangent vector;

[0179] S5-4: extend the end points of the link along ±t tangent by 1 / 2w to obtain two extended points; extend from each extended point along ±n normal by 1 / 2w to obtain four vertices, constituting a directional bounding box OBB.

[0180] As shown in Figure 14 , the spatial relationship between the chamber boundary, the safety zone, the link, the wafer, and the risk link OBB under a working condition is shown.

[0181] In Figure 3 , the poses of the links are completely consistent with Figure 2 . Among them, label 1 is the boundary of the chamber where the upper execution arm is located. Label 2 is the preset safety area. Label 3 is the preset area boundary of the upper execution arm. Label 4 is the position of the base rotation axis A1. Label 5 is the large arm L1. Label 6 is the first joint axis A2. Label 7 is the small arm L2. Label 8 is the horizontal coordinate coincident with the end effector rotation axis A3 and the end effector rotation axis A4. Label 9 is the lower execution arm L4. Label 10 is the center P2 of the wafer 2 carried by the lower execution arm L4. Label 11 is the upper execution arm L3. Label 12 is the center P1 of the wafer 1 carried by the upper execution arm L3. Label 13 is a 300mm wafer with P1 as the center. Label 14 is the OBB of the upper execution arm L3 constructed.

[0182] Since the end points of the links such as the large arm L1, the small arm L2, and the lower execution arm L4 are all located within the preset safety area, no OBB is constructed. Since P1 is located outside the preset safety area, it is necessary to construct the OBB of the link containing this end point, i.e., the upper execution arm L4.

[0183] As shown in Figure 15As shown, after obtaining the list of OBBs to be detected, a ray method is used to determine the out-of-bound, i.e., to determine whether each vertex position of all the OBBs to be detected is located outside the chamber boundary. If any vertex is located outside the chamber boundary, collision is directly determined. If no out-of-bound OBB vertex is detected, all the line segments in the list of OBBs to be detected are polled. First, the region where each line segment of the OBB is located is determined (using the line segment region detection algorithm). According to the region to which the line segment belongs, the chamber boundary line segment of the region is obtained from the height-section-region-line segment mapping database, and the line segment intersection detection algorithm is used to determine the intersection of the chamber boundary line segment and the OBB line segment one by one.

[0184] The line segment intersection detection algorithm is a general algorithm. As shown in Figure 15 The line segment intersection detection algorithm receives the start point coordinates and end point coordinates of two line segments, first performs a fast rejection experiment, and determines the overlap of the AABB bounding box. If no overlap of the AABB bounding box is detected, non-intersection is output. Otherwise, a cross-standing experiment is performed, and the intersection of the two line segments is determined by calculating the intersection position using the parametric equation method.

[0185] As shown in Figure 1 The specific steps of the line segment intersection detection algorithm are as follows:

[0186] S6-1: For two line segments A and B, the start point of line segment A is a1, and the end point is a2. The start point of line segment B is b1, and the end point is b2.

[0187] S6-2: The direction vector of line segment A is calculated as: The direction vector of line segment B is calculated as:

[0188] S6-3: A fast rejection experiment is performed to determine the overlap of the AABB bounding box. If no overlap of the AABB bounding box is detected, non-intersection is output. Otherwise, a cross-standing experiment is performed.

[0189] S6-4: The cross-standing experiment is performed using the parametric equation method, and the intersection of the two line segments is determined by calculating the intersection position.

[0190] a) The cross-standing experiment is performed using the parametric equation method, and the expression is:

[0191]

[0192] wherein, represents the start point of line segment A; represents the end point of line segment A; represents the start point of line segment B; represents the end point of line segment B; is the cross product denominator, representing the direction relationship of the two line segments, and the calculation formula is t is the proportional position of the intersection point on the segment A; u is the proportional position of the intersection point on the segment B;

[0193] b) When t = 0, the intersection point is at the start of the segment A; when t = 1, the intersection point is at the end of the segment A;

[0194] c) When y = 0, the intersection point is at the start of the segment B; when y = 1, the intersection point is at the end of the segment B;

[0195] d) If t ∈ [0, 1] and u ∈ [0, 1], the segments intersect; otherwise, they do not intersect.

[0196] After the parametric equation method is used to calculate the intersection of the segments, if the segments intersect, the segment intersection algorithm outputs intersection. Otherwise, it outputs no intersection.

[0197] When the segment intersection algorithm outputs intersection, the collision detection algorithm outputs collision. Otherwise, if the polling of all OBB segments is completed and the segment intersection algorithm does not output intersection, the collision detection algorithm outputs no collision.

[0198] After the calculation of the collision detection algorithm is completed, as ​ The robot control system receives the determination result of the collision detection system, and if the detected path point is determined to have no collision risk, the robot performs movement. Otherwise, the current round of motion planning is aborted and an alarm is given.

[0199] In summary, in combination with the provided embodiments, the present application provides a wafer handling robot collision detection method and system suitable for a complex structure cavity, which realizes comprehensive collision risk judgment of the mechanical arm link, end effector and wafer carrier in the trajectory planning stage by constructing a height-section-region-segment mapping database and combining a multi-level real-time collision detection mechanism. This method not only effectively avoids the deployment restrictions of traditional sensor solutions in a vacuum environment, but also significantly improves the detection efficiency and accuracy through regionalized detection and OBB precise modeling, meeting the real-time response requirements in high-speed wafer handling scenarios. The present application has good engineering applicability and system integration, providing reliable technical support for the safe operation of semiconductor manufacturing equipment and non-destructive wafer transmission.

[0200] Those skilled in the art can understand that the above description is only the preferred embodiment of the present application, and the features described in various embodiments of the present disclosure and / or claims can be combined or combined, even if such combination or combination is not explicitly described in the present disclosure. It is not intended to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacement for part of the technical features, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0201] Although preferred embodiments of the present application have been described, those skilled in the art can make further changes and modifications to these embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and changes of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and changes.

Claims

1. A collision detection method of a wafer handling robot, characterized by, Comprising the following steps: S1: receiving the joint angle sequence and height coordinates output in real time by the robot trajectory planning module; S2: solving by a forward kinematics solver, based on the input joint angles and link lengths, calculating the Cartesian coordinates of each node of the robot arm; by a point region judgment algorithm and a line segment region detection algorithm, preferentially checking whether all nodes are located within the predefined safety region, if so, directly determining no collision risk; Otherwise, according to the current height, adaptively selecting the corresponding chamber cross-section model; S3: for wafers located outside the predefined safety region, performing regional wafer collision detection, if collision is detected, terminating the process and alarming; if no wafer collision is detected, constructing an oriented bounding box (OBB) for the risk link whose end point is not in the safety region; S4: detecting whether the vertices of the OBB exceed the boundary, if so, determining collision; if the vertices do not exceed the boundary, performing regional detection on each edge of the OBB, and performing accurate line segment intersection detection with the chamber boundary line segment of the corresponding region; S5: outputting the final collision determination according to the detection result, and deciding by the robot control system whether to execute the motion.

2. The collision detection method of a wafer handling robot according to claim 1, wherein Before executing step S1, a height-cross-section-region-line segment mapping database is also constructed, specifically: S1-1: horizontally slicing the three-dimensional model of the semiconductor process machine chamber at different heights to obtain multiple horizontal cross-sections at different heights; S1-2: converting the chamber cross-section boundaries at each height into a boundary representation composed of line segments; S1-3: according to the chamber structure characteristics, dividing the horizontal cross-sections into three categories according to height intervals, including Process chamber cross-section, LoadLock chamber cross-section and Home cross-section; S1-4: boundary line segment partitioning step: based on the chamber structure characteristics, by presetting longitudinal region dividing lines X1 and X2 and transverse region dividing line Y, dividing the chamber cross-section boundaries of each height layer into six regions; S1-5: for each chamber boundary line segment, establishing a mapping relationship between the line segment and the region to form a predefined safety region; S1-6: database construction step: associating and mapping the height interval, cross-section type, region number and boundary line segment to form a multi-level mapping database of height-cross-section-region-line segment, which is used for multi-level real-time collision detection, receiving the coordinates of the 4 vertices of the OBB and adding them to the detection list, after traversal, outputting the OBB list to be detected for subsequent collision detection.

3. The collision detection method of a wafer handling robot according to claim 1, wherein In step S2, the point region judgment algorithm specifically comprises: S2-1: receiving the Cartesian coordinates (x, y) of each node of the robot arm to be determined; S2-2: for each node of the robot arm: comparing the x coordinate of the point with the preset longitudinal region dividing lines X1 and X2; comparing the y coordinate of the point with the preset transverse region dividing line Y; S2-3: determining the region number according to the comparison result, and outputting the region number 1-6; Wherein, the region division rule is: when x < X1 and y < Y, it is determined as region 1; when X1 <= x < X2 and y < Y, it is determined as region 2; when x >= X2 and y < Y, it is determined as region 3; when x < X1 and y >= Y, it is determined as region 4; when X1 <= x < X2 and y >= Y, it is determined as region 5; when x >= X2 and y >= Y, it is determined as region 6; S2-4: output the region number, the region number is used to retrieve the chamber boundary line segment of the corresponding region from the height-section-region-line segment mapping database, and provide the regionalization data basis for subsequent collision detection.

4. The collision detection method of a wafer handling robot according to claim 1, wherein, In step S2, the line segment region detection algorithm specifically comprises the following steps: S3-1: for a given line segment starting point P1(x1, y1) and ending point P1(x1, y1), the intersection point of the line segment and the region boundary line is calculated using the line segment parameter equation; The line segment parameter equation is expressed as: S3-2: For the vertical region boundary line x bound , calculate intersection parameter t: wherein ε is a floating point precision tolerance, ε = 10 -7 ; The y coordinate of the intersection point is calculated as: y intersect = y1 + t · (y2 - y1) S3-3: For horizontal region boundary line y bound , calculate intersection parameter t: The x coordinate of the intersection point is calculated as: x intersect = x1 + t · (x2 - x1) S3-4: collect the line segment end point and each valid intersection point, and the region information of t e [0, 1], output the set of all region numbers passed by the line segment.

5. The collision detection method of a wafer handling robot according to claim 1, wherein, In step S3, the wafer collision detection includes the following steps: S4-1: input the wafer center coordinate P(x, y); provide a point region determination algorithm to determine the region where the point is located; S4-2: obtain all chamber boundary line segments of the region and adjacent regions from the height-section-region-line segment mapping database; S4-3: the shortest distance from the point to the line segment is calculated based on vector projection as: S4-4: record the minimum value d of all distances min , compare d min with the wafer radius R; if d min <R, then determine a collision; otherwise determine no collision.

6. The collision detection method of a wafer handling robot according to claim 1, wherein, In step S3, if no wafer collision is detected, an OBB construction algorithm is used to construct a directional bounding box OBB for the risk link with the end points not in the safe region, comprising the following steps: S5-1: input the coordinates of the two end points of the risk link and the link width w; S5-2: the direction vector of the link is calculated as: v = endpoint 2 coordinate - endpoint 1; S5-3: if the modulus of the direction vector is zero, a square directional bounding box OBB is generated with the endpoint 1 as the center, and the side length is the link width; otherwise, the unit tangent vector t of the direction vector is calculated; the unit normal vector n is calculated as the perpendicular vector of the tangent vector; S5-4: expand the end points of the link along the ±t tangent by 1 / 2w to obtain two expanded points; expand each expanded point along ±n normal by 1 / 2w to obtain four vertices, which constitute a directional bounding box OBB.

7. The collision detection method of a wafer handling robot according to claim 1, wherein, The line segment intersection detection algorithm comprises the following steps: S6-1: for two line segments A and B, the starting point of line segment A is a1, and the ending point is a2, the starting point of line segment B is b1, and the ending point is b2; S6-2: Calculate the directional vector of line segment A as: Calculate the directional vector of line segment B as: S6-3: perform a quick rejection experiment to detect the overlap of the AABB bounding box, and output no intersection if no overlap of the AABB bounding box is detected; otherwise, perform a cross-standing experiment; S6-4: perform the cross-standing experiment using the parametric equation method, and determine the intersection of the two line segments by calculating the intersection position.

8. The collision detection method of a wafer handling robot according to claim 7, wherein, The step 6-4 specifically comprises: a) the cross-standing experiment is performed using the parametric equation method, and the expression is: wherein represents the start point of line segment A; represents the end point of line segment A; represents the start point of line segment B; represents the end point of line segment B; is the cross product of the direction vectors of the two line segments, calculated as t is the proportional position of the intersection point on line segment A; u is the proportional position of the intersection point on line segment B; b) when t = 0, the intersection point is on the starting point of line segment A; when t = 1, the intersection point is on the ending point of line segment A; c) When u = 0, the intersection point is at the start point of the line segment B; when u = 1, the intersection point is at the end point of the line segment B; d) If t is in [0, 1] and u is in [0, 1], the line segments intersect; otherwise, they do not intersect.

9. A collision detection system for implementing the wafer handling robot collision detection method according to any one of claims 1 to 8, characterized by Comprise: a robot control system for outputting motion parameters including joint angle sequences and height coordinates in a trajectory planning process to an input module; an input module for receiving joint angle sequences and Z-axis height parameters output by the robot control system in the trajectory planning process; a forward kinematics solver for performing forward kinematics calculation on the joint angles based on preset mechanical structure parameters to obtain a set of Cartesian coordinates of each node of the robot arm; a multi-level real-time collision detection unit for performing real-time collision detection between the robot links, the end effector, and the wafer carrier and a predefined chamber model based on the set of node coordinates, determining whether there is a collision risk, and outputting a collision determination result to an output module; an output module for feeding back the determination result to the robot control system to control whether the robot continues to execute the trajectory or aborts and alarms.

10. The collision detection system of claim 9, wherein The multi-level real-time collision detection unit comprises a node coordinate determination module, a cross-section height adaptive module, a regional wafer collision detector, an oriented bounding box (OBB) generator, a boundary out-of-bound detector, a regional line segment retriever, and a line segment intersection detector. The node coordinate determination module is configured to check whether all nodes in the set of spatial coordinates are located within a preset safety zone using a ray method, and output a no-collision determination when all nodes are within the safety zone. The cross-section height adaptive module is configured to select a corresponding chamber cross-section model according to the Z-axis height parameter when there is a node that is not within the safety zone, and call the boundary line segments from a height-cross-section-region-line segment mapping database. The regional wafer collision detector is configured to receive wafer center coordinates, determine the region where the wafer center is located through a point-region determination algorithm, obtain all chamber boundary line segments of the region and adjacent regions from the height-cross-section-region-line segment mapping database, calculate the shortest distance from the wafer center to each boundary line segment based on the vector projection principle, and output a collision determination when the calculated shortest distance is less than the wafer radius. The OBB generator is configured to construct an OBB for a risk link whose end points are not within a predefined safety region: receive the coordinates of the two end points of the risk link and the link width w; calculate the direction vector of the link; if the direction vector module is zero, generate a square OBB with the end points as the center and the side length as w; otherwise, calculate the unit tangent vector and the unit normal vector of the link; expand the end points of the link by ±1 / 2w along the tangent vector to generate two expanded points; expand each expanded point by ±1 / 2w along the normal vector to obtain four vertex coordinates, which constitute the OBB bounding box. The boundary out-of-bound detector is configured to determine whether the vertex of the OBB bounding box exceeds the chamber boundary using a ray method. The regional line segment retriever is configured to determine the regions to which the edges of the OBB bounding box belong based on the region where the OBB bounding box is located, and retrieve the chamber boundary line segments of the corresponding regions from the height-cross-section-region-line segment mapping database. The line segment intersection detector is configured to determine whether the edges of the OBB bounding box intersect with the chamber boundary line segments based on the regions to which the edges belong. Line segment intersection detector for determining intersection of OBB edge and chamber boundary line segment based on parametric equation method, outputting collision determination when intersection is detected.

Images