Collision prediction method, collision prediction device, and welding system

Abstract

The purpose of the present invention is to provide a collision prediction method with higher accuracy, including accessories of a robot. A collision prediction method for predicting a collision between a multi-joint robot and a surrounding object comprises: a first acquisition step for acquiring configuration information and attitude information of the robot; a second acquisition step of acquiring point group data including the robot and the peripheral object; a classification step for classifying the point group data acquired in the second acquisition step into point group data corresponding to the robot and point group data corresponding to the peripheral object on the basis of configuration information and attitude information of the robot; and a prediction step for predicting collision between the robot and the surrounding object on the basis of the classification results of the classification in the classification step.

Description

Technical Field

[0001] The invention relates to a collision prediction method, a collision prediction device and a welding system. Background Art

[0002] In the past, when a robot performs an action on an object to be operated, the robot controls the action in a manner that does not cause a collision with peripheral devices located in the vicinity. Collision prediction is not limited to the action of the robot, but can also be used when generating an action program for controlling its action. Examples of peripheral devices of the robot include a positioner for holding the object to be operated, a camera for photographing the object to be operated, and the like. In addition to the peripheral devices, cables and the like as accessories of the robot are also conceived as components that may cause collisions during the action of the robot. It is also desirable to perform collision prediction and control the action of such accessories.

[0003] For example, Patent Document 1 discloses, as a collision prediction method for a robot, a method of avoiding collision by measuring position data of objects around the robot using a three-dimensional sensor and storing interfering coordinates.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent No. 6816070

[0007] In the method of Patent Document 1, only the front end of the robot is used as the collision avoidance object, so the collision prediction of the robot parts other than the front end and the attachments cannot be realized. In addition, when the environment of the work site changes, such as the posture change of the positioner, a high-precision collision prediction is also sought, but in the previous method, there is a possibility of reduced processing efficiency and misjudgment of collision. Summary of the invention

[0008] Problems to be solved by the invention

[0009] In view of the above-mentioned problems, an object of the present invention is to provide a more accurate collision prediction method involving the attachments of a robot.

[0010] Solutions to Solve Problems

[0011] In order to solve the above-mentioned problems, the present invention has the following structure. That is, a collision prediction method is used to predict the collision between a multi-joint robot and surrounding objects, wherein:

[0012] The collision prediction method has the following features:

[0013] A first acquisition step is to acquire the structure information and posture information of the robot;

[0014] A second acquisition step of acquiring point group data including the robot and the surrounding objects;

[0015] a classification step of classifying the point group data acquired in the second acquisition step into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; and

[0016] The prediction step predicts a collision between the robot and the surrounding object based on the classification result obtained by the classification step.

[0017] In addition, another embodiment of the present invention has the following structure. That is, a collision prediction device performs collision prediction between a multi-joint robot and surrounding objects, wherein:

[0018] The collision prediction device comprises:

[0019] A first acquisition mechanism, which acquires structural information and posture information of the robot;

[0020] A second acquisition mechanism, which acquires point group data including the robot and the surrounding objects;

[0021] a classification unit that classifies the point group data acquired by the second acquisition unit into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; and

[0022] A prediction mechanism predicts a collision between the robot and the surrounding object based on the classification result of the classification mechanism.

[0023] In addition, another embodiment of the present invention has the following structure. That is, a welding system includes a collision prediction device for predicting a collision between a multi-jointed robot and a surrounding object and a welding robot as the robot, wherein:

[0024] The collision prediction device comprises:

[0025] A first acquisition mechanism, which acquires structural information and posture information of the robot;

[0026] A second acquisition mechanism, which acquires point group data including the robot and the surrounding objects;

[0027] a classification unit that classifies the point group data acquired by the second acquisition unit into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; and

[0028] A prediction mechanism predicts a collision between the robot and the surrounding object based on the classification result of the classification mechanism.

[0029] Effects of the Invention

[0030] According to the present invention, it is possible to achieve more accurate prediction of collision between a robot including its appendages and surrounding objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram showing an example of a system configuration according to one embodiment of the present invention.

[0032] Figure 2 This is a block diagram showing a configuration example of a robot control device according to an embodiment of the present invention.

[0033] Figure 3 This is a block diagram showing an example of the functional configuration of a robot control device according to one embodiment of the present invention.

[0034] Figure 4 This is a conceptual diagram for explaining the movable range of a robot according to one embodiment of the present invention.

[0035] Figure 5A This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0036] Figure 5B This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0037] Figure 5C This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0038] Figure 5D This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0039] Figure 5E This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0040] Fig. 6A This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0041] Figure 6B This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0042] Figure 6C This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0043] Fig. 7AThis is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0044] Figure 7B This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0045] Figure 7C This is an example diagram for explaining a method for classifying point cloud data according to one embodiment of the present invention.

[0046] Fig. 8A This is an example diagram for explaining a collision prediction method according to one embodiment of the present invention.

[0047] Figure 8B This is an example diagram for explaining a collision prediction method according to one embodiment of the present invention.

[0048] Figure 8C This is an example diagram for explaining a collision prediction method according to one embodiment of the present invention.

[0049] Fig. 9 This is a flowchart of a collision prediction process according to one embodiment of the present invention.

[0050] Description of Reference Numerals

[0051] 1. Welding system

[0052] 10. Welding Robot

[0053] 11 Welding torch

[0054] 12 Wire feeding device

[0055] 13 Welding wire

[0056] 16, 17 Cables

[0057] 20 Robot control device

[0058] 201 CPU

[0059] 202 Memory

[0060] 202A Control Procedure

[0061] 203 Operation panel

[0062] 204 Ministry of Communications

[0063] 205 Robot connection

[0064] 206 Positioner connection

[0065] 30 Power supply unit

[0066] 40 Camera

[0067] 50 Locator. DETAILED DESCRIPTION

[0068] Hereinafter, the present specific embodiment will be described with reference to the accompanying drawings, etc. It should be noted that the embodiment described below is used to illustrate an embodiment of the present invention and is not intended to limit the present invention. In addition, the structures described in each embodiment may not all be structures necessary to solve the problems of the present invention. In addition, in each of the drawings, the same constituent elements are indicated by marking the same reference numerals to show the corresponding relationship.

[0069] <First embodiment>

[0070] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. In this embodiment, a welding system using a multi-joint robot as a welding robot is described, but the present invention is not limited thereto. For example, any robot that performs operations such as holding an operation object can be applied to the present invention.

[0071] [Structure of welding system]

[0072] Figure 1 FIG. 1 is a diagram showing a configuration example of welding system 1 according to the present embodiment. Figure 1 The welding system 1 shown is configured to include a welding robot 10, a robot control device 20, a power supply device 30, a camera 40, and a positioner 50. The welding robot 10 is configured as a 6-axis multi-joint robot, and a welding torch 11 is installed at its front end. A welding wire 13 is supplied to the welding torch 11 from a welding wire feeding device 12. The welding wire 13 is sent out from the front end of the welding torch 11 toward the welding part. The power supply device 30 supplies power to the welding wire 13. The robot control device 20 is configured to be able to communicate with the welding robot 10 and its peripheral devices, and to control various devices related to welding. The power supply device 30 is provided with a current sensor (not shown) for detecting the welding current flowing from the welding wire 13 during welding to the workpiece (not shown) held by the positioner 50, and a voltage sensor (not shown) for detecting the arc voltage between the welding wire 13 and the workpiece. In addition, the welding robot 10 is provided with a cable 16 related to the control signal, a cable 17 for the welding wire 13, and the like as accessories.

[0073] The power supply device 30 has a processing unit and a storage unit (not shown). The processing unit is composed of, for example, a CPU (Central Processing Unit). In addition, the storage unit is composed of, for example, a volatile or non-volatile memory such as an HDD (Hard Disk Drive), a ROM (Read Only Memory), and a RAM (Random Access Memory). The processing unit controls the power applied to the welding wire 13 by executing a computer program for power supply control stored in the storage unit. The power supply device 30 is also connected to the welding wire feeding device 12, and the processing unit controls the feeding speed and feeding amount of the welding wire 13.

[0074] The camera 40 is, for example, disposed around the welding robot 10 and is used to photograph the surroundings of the welding robot 10. The camera 40 is configured to be able to obtain point group data as three-dimensional data. As the camera 40, for example, a ToF (Time of Flight) camera, a stereo camera, or a LiDAR (Light Detection And Ranging) sensor can be used. Each of the above sensors has different characteristics, so they can be used differently according to the measurement environment and the measurement object. In the present embodiment, the camera 40 is disposed above the welding robot 10 and is configured to be able to photograph the welding robot 10 located below and its surroundings. It should be noted that the installation position and the shooting range of the camera 40 are not limited thereto, and can be arbitrarily set in a manner including the movable range of the welding robot 10 described later.

[0075] The ToF camera that can be used as the camera 40 calculates the distance for each pixel by irradiating a laser to the measurement object and measuring the reflected laser by the imaging element. The measurable distance of the ToF camera is about tens of cm to several meters. The stereo camera uses multiple images captured by multiple (for example, two) cameras to calculate the distance based on their parallax. The measurable distance of the stereo camera is about tens of cm to several meters. LiDAR calculates the distance by irradiating the surroundings with laser and measuring the reflected laser. The measurable distance of LiDAR is about tens of cm to several meters.

[0076] It should be noted that a plurality of cameras 40 may be provided, and the point group data captured by each camera may be used separately for collision prediction described below. Alternatively, the point group data captured by the plurality of cameras 40 may be subjected to coordinate transformation processing to unify and integrate the coordinates, and then collision prediction may be performed using the integrated point group data. By using a plurality of cameras 40, blind spots can be suppressed, and collision prediction with higher accuracy can be achieved.

[0077] Each part constituting the welding system 1 is connected to be communicable by various wired / wireless communication methods. The communication method here is not limited to one, and a plurality of communication methods may be combined and connected.

[0078] [Structure of robot control device]

[0079] Figure 2 2 is a diagram showing a configuration example of a robot control device 20 for controlling the motion of a welding robot 10. The robot control device 20 is configured to include a CPU 201 for controlling the entire device, a memory 202 for storing data, an operation panel 203 including a plurality of switches, a communication unit 204, a robot connection unit 205, and a positioner connection unit 206. The memory 202 is configured by a volatile or non-volatile storage device such as a ROM, a RAM, or a HDD. A control program 202A for controlling the welding robot 10 is stored in the memory 202. The CPU 201 controls various motions performed by the welding robot 10 and the positioner 50 by executing the control program 202A. At this time, the robot control device 20 performs collision prediction using a collision prediction method described later, and also controls the posture, motion trajectory, etc. of the welding robot 10 based on the prediction result. When controlling the welding robot 10, the robot control device 20 can determine the posture information such as the position, coordinates, and angle of each part of the welding robot 10 at that time.

[0080] The communication unit 204 is configured to include a communication module for wired or wireless communication. The communication unit 204 is used for data and signal communication with the power supply device 30, the camera 40, etc. The communication method and specification used in the communication unit 204 are not particularly limited, and a plurality of methods may be combined or different for each connected device. For example, the current value of the welding current detected by the current sensor not shown and the voltage value of the arc voltage detected by the voltage sensor not shown are given to the CPU 201 from the power supply device 30 via the communication unit 204. In addition, point group data as three-dimensional data as described later is obtained from the camera 40.

[0081] A driving circuit of the welding robot 10 is connected to the robot connection part 205 . The CPU 201 outputs a control signal based on the control program 202A to a driving circuit (not shown) included in the welding robot 10 via the robot connection part 205 .

[0082] The positioner connection portion 206 is connected to a driving circuit of the positioner 50 . The CPU 201 outputs a control signal based on the control program 202A to a driving circuit (not shown) included in the positioner 50 via the positioner connection portion 206 .

[0083] [Functional structure]

[0084] Figure 3 2 is a block diagram showing an example of the functional configuration of the robot control device 20. Each function can be realized by, for example, the CPU 201 reading and executing the control program 202A. It should be noted that the functional configuration shown here is an example, and other functions may be included.

[0085] The data acquisition unit 301 acquires the point group data captured by the camera 40. The data acquisition unit 301 may also notify the camera 40 of the capturing timing, capturing settings, etc. of the point group data. Furthermore, the data acquisition unit 301 acquires the posture information such as the position, coordinates, angle, etc. of each part constituting the welding robot 10 at the time point corresponding to the timing of capturing the point group data. The setting acceptance unit 302 accepts various settings related to the collision prediction of the present embodiment. For example, it may be configured to be able to accept the capturing settings made by the camera 40. The data management unit 303 maintains and manages the point group data acquired by the data acquisition unit 301 and the various settings received by the setting acceptance unit 302. In addition, the data management unit 303 also manages the device information such as the three-dimensional model representing the shape of the welding robot 10 and the device structure. In the device information managed by the data management unit 303, in addition to the structural information on the predetermined specifications such as the shape and structure of the welding robot 10, the posture information such as the position, coordinates, angle, etc. of each part that can change with the action may also be included.

[0086] The pre-processing unit 304 performs pre-processing on the acquired point group data. The pre-processing here may vary depending on the point group data used, and may include, for example, filtering, outlier removal, clustering, coordinate transformation, etc. Therefore, the pre-processing may be omitted depending on the point group data. The data classification unit 305 performs classification processing on the point group data. The details of the classification processing will be described later.

[0087] The collision prediction unit 306 performs collision prediction using the data to which the classification process is applied by the data classification unit 305. Details of the collision prediction will be described later. The prediction result output unit 307 outputs the prediction result obtained by the collision prediction unit 306. The prediction result can be output in a manner used for control within the system or in a manner that can be recognized by the operator. The robot control unit 308 controls the action of the welding robot 10 based on the prediction result of the collision prediction unit 306.

[0088] [Robot's movable range]

[0089] Figure 4 1 is a conceptual diagram for explaining the movable range of the welding robot 10. The area occupied by the multi-axis welding robot 10 changes according to its posture, etc. Figure 4, an area 400 in the XY plane is set as the shooting range of the camera 40. In addition, within the area 400, an area 402 indicated by the hatching indicates the movable range of the welding robot 10 in any posture. Therefore, an area 401 indicating a range outside the area 402 indicates that the welding robot 10 is outside the movable range that cannot exist in any posture.

[0090] Note that the positional relationship between the camera 40 and the welding robot 10 changes depending on the installation position and orientation of the camera 40. Therefore, the movable range of the welding robot 10 viewed from the camera 40 is set to an appropriately specified movable range depending on the system configuration and the like.

[0091] In this embodiment, in the pre-processing, at least a part of the point group data located in the area 401 outside the movable range is removed from the point group data acquired by the camera 40. That is, in the area outside the movable range, it is possible to process so that the welding robot 10 does not collide with the surrounding objects. This reduces the processing load and prevents errors in collision prediction.

[0092] [Classification of point group data]

[0093] The classification of the point group data of the present embodiment is described below. In the present embodiment, in order to be able to identify the welding robot 10 and other objects located in its periphery, the point group data corresponding to the welding robot 10 and the point group data corresponding to other objects in the point group data captured by the camera 40 are classified and used. And collision prediction is performed based on these classifications. It should be noted that the "other objects" here are not limited to peripheral devices such as the locator 50, and can include all surrounding objects such as the floor and the ceiling that may cause a collision.

[0094] In the present embodiment, a classification method capable of executing multiple point group data is configured. Depending on the classification method, the accuracy and processing time are different, so it is possible to switch according to the purpose, etc. In the present embodiment, four classification examples are shown, namely, classification based on a clustering method (also referred to as a "first classification method"), classification based on the difference of the point group of the time series (also referred to as a "second classification method"), classification based on only the front end of the welding robot 10 as an object (also referred to as a "third classification method"), and classification based on the difference in distance from the point group at the previous time (also referred to as a "fourth classification method").

[0095] (First classification method)

[0096] use Figure 5A to Figure 5E To illustrate the process of the first classification method. Figure 5A 4 is an example of point cloud data 500 including point cloud data 501 captured by the camera 40 at a certain time point. Figure 5AThe point cloud data 501 shown is clustered using, for example, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering, which is a well-known method.

[0097] Figure 5B An example of point group data 500 after clustering is shown. The point group data 500 is classified into a plurality of points such as point group data 511, 512, and 513 and is shown so as to be identifiable. Figure 5B The processing result is used to set information corresponding to the structural position of the joints of the welding robot 10. The joint positions can be determined based on information such as the specifications and posture of the welding robot 10, information of the three-dimensional model, and the like. Figure 5C This is a conceptual diagram showing a state in which information of the positions of the joints of the overlapping welding robot 10 is shown. Here, an example in which four pieces of information 521a, 521b, 521c, and 521d are set is shown. And, a point group located within a predetermined range is determined based on the information 521a, 521b, 521c, and 521d. Figure 5D An example is shown in which ranges 531a, 531b, 531c, and 531d of predetermined radius are set for the information 521a, 521b, 521c, and 521d, respectively.

[0098] Furthermore, all the categories of point group data included in the ranges 531a, 531b, 531c, and 531d are labeled as point group data representing the welding robot 10. In addition, the other point group data are labeled as objects other than the welding robot 10 (other objects). As a result, Figure 5E As shown, the data is classified into point group data 541 corresponding to the welding robot 10 and point group data 542 corresponding to other objects.

[0099] It should be noted that in the above description, Figure 5C As shown in the figure, an example of determining point group data using information on the positions of the joints of the welding robot 10 is shown. In addition, an arbitrary point group (hereinafter referred to as a "simulated point group") may be set, and the point group data of the welding robot 10 may be included in the arbitrary point group. That is, the welding robot 10 is set as an object occupying an area larger than the actual area, and a collision prediction is performed in the later stage. For example, after setting the positions of a plurality of joints, they may be connected by straight lines, and an arbitrary number of simulated point groups may be set in a straight line at a predetermined interval around them. The straight line may correspond to the connecting rod of the welding robot 10, for example.

[0100] In addition, an arbitrary simulated point group may be set for the periphery of the point group data corresponding to the welding robot 10, and the simulated point group may be included to be determined as the point group data of the welding robot 10. Such a simulated point group may be set in a predetermined shape corresponding to a cable or the like located behind the arm, for example, according to the posture of the welding robot 10. In this case, the set simulated point group may be set in size and shape based on the pre-determined cable specifications, the posture of the arm, and the like.

[0101] In addition, it is also possible to set an arbitrary simulated point group at a predetermined distance based on the position of the point group data corresponding to a specific part of the welding robot 10, and include the simulated point group to be determined as the point group data of the welding robot 10. Such a simulated point group may correspond to, for example, a cable located under the arm of the welding robot 10. In this case, the size and shape of the set simulated point group may be set based on the predetermined cable specifications.

[0102] (Second classification method)

[0103] use Figure 6A to Figure 6C The second classification method will be described. First, as an initial state, point group data around the welding robot 10 is acquired, and the point group data corresponding to the welding robot 10 and the point group data corresponding to other objects are clustered using the first classification method. Fig. 6A The result of clustering the point group data 600 using the first classification method is shown, and point group data 601 corresponding to the welding robot 10 and point group data 602 corresponding to other objects are obtained. These are set as initial settings.

[0104] Next, when processing the point cloud data at the time point when the collision prediction is performed (recorded as "current time point" for convenience), it is determined whether the other object (here, the locator 50) that may move has moved since the time point when it was set to the initial state. For example, it is possible to determine whether the locator 50 has moved based on the control signal, etc. If the other object has moved, the first classification method is used to classify the point cloud data at the time point. On the other hand, if the locator 50 has not moved, the following processing is performed.

[0105] The shortest distance between each point is detected in the point group included in the initial point group data and the point group included in the point group data at the current time point, and the point group whose shortest distance is less than the threshold is deleted, thereby extracting only the changed point group data. Figure 6BIn the point group data 611, point group data corresponding to the welding robot 10 is determined to be point group data that has not changed from the initial setting. Point group data 613 is point group data corresponding to the welding robot 10, which is determined to be point group data that has changed from the initial setting. Point group data 612 is point group data corresponding to other objects, and is set to be point group data that has not changed from the initial setting. And, Figure 6C Shown from Figure 6B The point cloud data shown is a state where the point cloud data 611 determined to be unchanged is deleted. This point cloud data is used to perform collision prediction at a later stage.

[0106] It should be noted that the timing of the initial state is not particularly limited, and for example, the initial state may be set when the operation of welding system 1 starts, or the initial state serving as a reference may be updated at regular intervals.

[0107] (Third classification method)

[0108] use Figure 7A to Figure 7C , the third classification method is explained. First, obtain Figure 5A Point group data of the entire welding system 1 as shown. In such point group data, based on the structural information of the welding robot 10, the tip of the welding robot 10 and the joint closest to the tip are specified. Fig. 7A This is data 700 of the acquired point group data focusing on the portion around the tip of the welding robot 10. Here, a point 702a corresponding to the tip of the welding robot 10 and a point 702b corresponding to the joint are identified.

[0109] Furthermore, based on the positional relationship between the point 702a and the point 702b, a part of the three-dimensional model of the welding robot specified in advance is arranged. Figure 7B An example of configuring a three-dimensional model is shown, and area 711 shows the range in which the three-dimensional model is configured. On the other hand, point group data 712 corresponds to point group data not included in area 711. It should be noted that the three-dimensional model configured here may be a simple model instead of a detailed three-dimensional model.

[0110] Figure 7C Shows deleted Figure 7BThe state of the point group data contained in the area 711 shown. Here, the point group data 722 corresponding to the position of the front end of the welding robot 10 in the point group data contained in the area 711 is retained. In addition, the point group data 721a, 721b, and 721c outside the area 711 are all processed as point groups corresponding to other objects. In addition, the point group data 722 equivalent to the front end of the welding robot 10 and the other point group data are used to perform collision prediction in the later stage. It should be noted that the range (the amount of point group data) processed as the front end of the welding robot 10 (equivalent to the point group data 722) is not particularly limited, and can be specified based on the size of the welding robot 10, the processing load, etc.

[0111] (Fourth classification method)

[0112] The fourth classification method is described. As with the second classification method, first, as an initial state, the point group data around the welding robot 10 is obtained, and the point group data corresponding to the welding robot 10 and the point group data corresponding to other objects are clustered using the first classification method. This is set as the initial setting. Next, referring to the newly obtained point group data at the current time point, it is determined which of the point group data corresponding to the welding robot 10 and the point group data of other objects is close to the initial setting. In addition, the point group data at the current time point is classified into the party closer to the welding robot 10 and the party closer to the other objects. At this time, a threshold for the distance is pre-set. When the distance between the point group data at the current time point and the point group data of the welding robot 10 set initially exceeds the threshold, the point group data at the current time point is processed as the point group data of other objects. After the determination of the point group data is completed, the initial setting is updated using the point group data at the current time point, and it is set as the point group data of the previous moment. Afterwards, when new point group data is obtained, it is compared with the point group data of the previous moment and the classification is repeated.

[0113] It should be noted that in the fourth classification method, the above-mentioned simulated point group can also be included in the initial setting and compared with the point group data at the current time point. In this case, the simulated point group can also be included in the setting of the point group data at the previous time point.

[0114] In the first classification method, the parts that can be predicted to collide are the main body (including the front end, the joints) and the accessories (including the cables) of the welding robot 10. In addition, the distance between objects that can be classified is the longest among the three classification methods. In addition, the execution time is also the longest among the four classification methods. On the other hand, in the first classification method, there are obstacles, etc., so that classification can be performed even when the point group corresponding to the welding robot 10 is broken.

[0115] In the second classification method, the parts that can be predicted to collide are equivalent to the main body (including the front end, the joints) and the accessories (including the cables) of the welding robot 10. In addition, the distance between the objects that can be classified becomes the middle of the first classification method and the third classification method. In addition, the execution time also becomes the middle of the first classification method and the third classification method. In the second classification method, when the objects other than the welding robot 10 move, the execution time becomes longer, but when the objects other than the welding robot 10 do not move, clustering is not used, so the execution time is shorter than the first classification method.

[0116] In the third classification method, the part where collision can be predicted is the main body (front end) of the welding robot 10. In addition, the distance between objects that can be classified is the shortest among the four classification methods. In addition, the execution time is also the shortest among the four classification methods.

[0117] In the fourth classification method, the distance between the point group data of the welding robot 10 in the previous state and the point group data at the current time point is referred to, so it is difficult to be affected by the intrusion of people into the work area, the movement of the positioner, etc. In addition, in the fourth classification method, the execution time is shortened compared with the second classification method.

[0118] [Collision prediction method]

[0119] In the present embodiment, collision prediction is performed using point group data classified using the above-mentioned multiple types of classification methods. In the present embodiment, three prediction methods are used: a collision prediction method based on the shortest distance between point groups (hereinafter also referred to as a "first prediction method"), a collision prediction method using overlapping voxels (hereinafter also referred to as a "second prediction method"), and a collision prediction method based on a time series (hereinafter also referred to as a "third prediction method").

[0120] These collision prediction methods can be switched in accordance with the above-mentioned classification methods, or can be configured so that the operator can pre-specify which method to use. For example, the first prediction method can use the classification results of any of the first to fourth classification methods. The second prediction method can use the classification results of the first, second, and fourth classification methods. The third prediction method can use the classification results of the first classification method.

[0121] (First prediction method)

[0122] Fig. 8A1 is a diagram for explaining a method of collision prediction based on the shortest distance between point groups as a first prediction method. Here, as a result of classification processing, points 801a, 801b, and 801c included in the point group data corresponding to the welding robot 10 and points 802a, 802b, and 802c included in the point group data corresponding to other objects are shown.

[0123] In the first prediction method, first, a threshold value for the distance immediately before a collision is set. The threshold value here can be predetermined based on the motion speed of the welding robot 10, the processing interval of the collision prediction, etc. Next, the shortest distance between points in the point group data is derived. Fig. 8A In the example of , the distance between point 801c included in the point group data of the welding robot 10 and point 802c included in the point group data of the other object is derived as the shortest distance 803. The shortest distance here is the distance in three-dimensional space. The method for deriving the shortest distance is not particularly limited, and for example, the well-known kd tree method can be used. And, when the shortest distance 803 is below the threshold, it is determined to be immediately before the collision.

[0124] It should be noted that the distance between point groups may not be just one shortest distance, and for example, the average of multiple groups with shorter distances between point groups may be used. In addition, the shortest distance may be derived after removing the deviation value between point groups.

[0125] As described above, the shortest distance may be derived by including a simulated point group in the point group data corresponding to the welding robot 10. In this case, the point group of an arbitrary range in the point group data corresponding to the welding robot 10 including the simulated point group may be reduced to speed up the processing.

[0126] In addition, the point group for the shortest distance derivation may be selected based on the moving direction of the welding robot 10. The moving direction of the welding robot 10 can be determined according to the direction and posture of welding. For example, the derivation range of the shortest distance between the point group data of the welding robot 10 and the point group data of other objects may be determined by the following process.

[0127] (i) Calculate the movement vector of the front end position (e.g., TCP (Tool Center Point)), joint position, and link point group (including simulation point group) in the welding robot 10. For example, the movement direction of the welding robot 10 can be obtained by referring to the teaching data. Alternatively, the movement vector can be calculated by calculating the difference between the current position vector and the past position vector. It should be noted that the reference position of the welding robot 10 when calculating the movement vector is not limited to the above, and may also include other parts.

[0128] (ii) A three-dimensional shape of a predetermined size and shape is configured along the position vector calculated by (i). For example, a cylinder as a predetermined three-dimensional shape can be configured along the direction of the moving vector with the TCP at the current time point as the center of the bottom surface. Alternatively, a rectangular parallelepiped as a predetermined three-dimensional shape can be configured in a manner parallel to the connecting rod with a point of the simulated point group of the connecting rod at the current time point as the center of the bottom surface. The three-dimensional shape used here is not limited to a cylinder or a cube, and any shape can be used. In addition, the structure of the three-dimensional shape can also be changed according to the position of the configured welding robot 10.

[0129] (iii) Extracting point group data located within the three-dimensional shape arranged in (ii) from point group data corresponding to other objects, and setting this point group data as a target for derivation of the shortest distance from the point group data corresponding to the welding robot 10 .

[0130] By using the above-described method, point group data used for collision prediction can be reduced, and processing speed can be increased.

[0131] (Second prediction method)

[0132] Figure 8B 1 is a diagram for explaining a collision prediction method using overlapping voxels as a second prediction method. Points 811a, 811b, and 811c included in the point cloud data corresponding to the welding robot 10 are shown as a result of the classification process. In addition, voxels 812 specified based on point cloud data corresponding to other objects are shown.

[0133] In the second prediction method, first, a threshold value for the number of points for the welding robot 10 and a size of a voxel based on point group data of other objects are set for determining that the object is immediately before a collision. The threshold value here can be predetermined based on the motion speed of the welding robot 10, the processing interval of the collision prediction, etc. In addition, the size of the voxel can also be predetermined based on the safety of the collision prediction, etc. The larger the size of the voxel, the more likely it is that the object is immediately before a collision even when the welding robot 10 is separated from the other object.

[0134] Next, the point group data of other objects are voxelized based on the set size. Figure 8B In addition, the number of points corresponding to the welding robot 10 overlapping with the voxel 812 is counted. Figure 8B In the example of , point 811c overlaps, and the overlap number is "1". When the overlap number is equal to or greater than a set threshold, it is determined that a collision is just before.

[0135] It should be noted that voxelization is not limited to the above, and the point group data corresponding to the welding robot 10 may be voxelized, or both the point group data corresponding to the welding robot 10 and the point group data corresponding to other objects may be voxelized.

[0136] Figure 8C An example of voxelizing both the point group data corresponding to the welding robot 10 and the point group data corresponding to other objects is shown. Points 822a, 822b, and 822c are points included in the point group data corresponding to the welding robot 10, and thus voxel 821a is set. In addition, points 823a, 823b, and 823c are points included in the point group data corresponding to other objects, and thus voxel 821b is set.

[0137] like Figure 8C As shown in the example, when collision prediction is performed using voxels of point group data corresponding to the welding robot 10 and voxels of point group data corresponding to other objects, first, it is determined whether the voxels overlap. And, in the case of overlap, the number of points contained in each voxel is counted, and an evaluation value is derived based on their number. And, if the evaluation value is above a predetermined threshold, it can be determined that a collision is imminent. The method for deriving the evaluation value is not particularly limited, but for example, it may be the product of the number of points contained in each voxel. Figure 8C In the case of the example of , it can be derived as (the number of points included in the voxel 821a: 3)×(the number of points included in the voxel 821b: 3)=(evaluation value: 9).

[0138] (Third prediction method)

[0139] The third prediction method is a time series collision prediction method. First, a threshold value is set for determining a point group immediately before a collision for the welding robot 10. The threshold value here can be predetermined based on the motion speed of the welding robot 10, the processing interval of the collision prediction, and the like.

[0140] In addition, the number of points of the point group data corresponding to the welding robot 10 corresponding to the initial posture of the welding robot 10 is maintained in advance. Next, the number of points of the point group data corresponding to the welding robot 10 at the time point (recorded as "current time point" for convenience) when the collision prediction is performed is obtained. And the difference between the number of points included in the point group data corresponding to the welding robot 10 in the initial posture and the number of points included in the point group data corresponding to the welding robot 10 at the current time point is calculated. When the difference is greater than a preset threshold, it is determined that a collision is imminent. After the determination, the number of points of the point group data corresponding to the welding robot 10 at the current time point is used to update the maintained number of points of the point group data. In the subsequent collision prediction processing, the updated value is used to repeat.

[0141] The first prediction method is easy to confirm the collision prediction position, but has a tendency that the processing execution time increases as the number of point group data corresponding to the welding robot 10 increases. The second prediction method is less easy to confirm the collision prediction position than the first prediction method, but can reduce the processing execution time.

[0142] The third prediction method makes it easy to confirm the predicted collision position and can suppress the execution time of the processing. In addition, the third prediction method has the characteristic of being able to cope with the case of slight misclassification of the point group. Therefore, in the third prediction method, even if there is a misjudgment in the first classification method, since the increase or decrease in the number of points is small, the influence of the classification accuracy on the judgment result can be suppressed. It should be noted that the third prediction method is intended to be applied when other objects are larger than a specified size. In addition, it is assumed that the number of points in the point group data corresponding to the welding robot 10 increases or decreases significantly according to the change in the posture of the welding robot 10, so it is preferable to switch whether to apply the third prediction method according to the change in posture.

[0143] [Processing Flow]

[0144] Fig. 9 The flowchart of the collision prediction process of the welding robot 10 of this embodiment is shown. This process flow can be realized, for example, by the CPU 201 of the robot control device 20 reading and executing the control program 202A. In addition, this process flow can be started at the same time as the operation of the welding system 1 starts, and the welding robot 10 continues to operate based on an arbitrary operation program in parallel with this process flow.

[0145] In S901, the robot control device 20 performs initial settings for the time series data. The processing here performs the initial settings used in the second classification method described above. More specifically, the classification results of the point group data of the welding robot 10 and other objects in the basic posture are set. It should be noted that this step can also be omitted according to the method used in the classification processing in the later stage.

[0146] In S902 , the robot control device 20 obtains device information, such as a preset three-dimensional model of the welding robot 10 , device configuration information, real-time posture information and position information of the welding robot 10 , and position information of the positioner 50 .

[0147] In S903 , the robot control device 20 acquires the point cloud data captured by the camera 40 .

[0148] In S904, the robot control device 20 applies pre-processing to the point group data obtained in S903. The pre-processing here may include, for example, coordinate transformation processing, downsampling processing, out-of-range point group removal processing, and deviation value removal processing. In the coordinate transformation processing, for example, coordinate transformation between different coordinate systems such as a camera coordinate system and a system coordinate system is performed. In addition, when point group data captured by multiple cameras 40 are integrated and used, coordinate transformation may be performed in a manner that makes them correspond. In the downsampling processing, the acquired point group data is reduced to a desired resolution according to the processing load, etc. In the out-of-range point group removal processing, for example, Figure 4 As shown, there is no need to perform collision prediction for the area 401 outside the range where the welding robot 10 moves, so the point group data of these areas are removed from the acquired point group data. In the deviation value removal process, points with values ​​exceeding a predetermined threshold are removed from the point group data. It should be noted that the type of pre-processing is an example, and other types of processing may also be performed. In addition, all or part of the pre-processing may be omitted according to the acquired point group data.

[0149] In S905, the robot control device 20 performs classification processing on the point group data to be targeted. As described above, in this embodiment, three classification methods can be used, and which method to use can be a classification method arbitrarily set by the operator, or can be selectively switched on the robot control device 20 side according to the processing load, the motion speed of the welding robot 10, the motion content, etc.

[0150] In S906, the robot control device 20 performs collision prediction based on the classification result performed in S905. As described above, in this embodiment, three prediction methods can be used, and which method to use can be selectively switched according to the classification method used in S905. At this time, the classification method arbitrarily set by the operator can be used, and the robot control device 20 side can also switch according to the processing load, the motion speed of the welding robot 10, the motion content, etc.

[0151] In S907, the robot controller 20 determines whether it is a state immediately before a collision based on the prediction result of S906. If it is a state immediately before a collision (S907: Yes), the processing of the robot controller 20 proceeds to S910. On the other hand, if it is not a state immediately before a collision (S907: No), the processing of the robot controller 20 proceeds to S908.

[0152] In S908, the robot controller 20 updates the time series data based on the processing results of S905 and S906. It should be noted that, depending on the classification process used, this process may be omitted, similar to the process of S901. Thereafter, the process of the robot controller 20 proceeds to S909.

[0153] In S909, the robot control device 20 outputs the prediction result in S906. As the prediction result, for example, in addition to not being immediately before a collision, the distance between the welding robot 10 and the other object and information on the processed point group data may also be included. In addition, there is no particular limitation on the output method, and the user may be notified via the operation panel 203 or the like, or the structure may be saved as history information in the memory 202. After that, the processing of the robot control device 20 returns to S902, and the processing is repeated.

[0154] In S910 , the robot control device 20 sets the welding robot 10 to the state immediately before the collision and controls the operation thereof to stop. Thereafter, the processing of the robot control device 20 proceeds to S911 .

[0155] In S911, the robot control device 20 outputs the intention to stop the welding robot 10 based on the prediction result in S906 and its prediction result. As the prediction result, for example, in addition to the information immediately before the collision, the distance between the welding robot 10 and the other object and the information of the processed point group data may also be included. In addition, there is no particular limitation on the output method, and the user may be notified via the operation panel 203 or the like, or the structure may be saved as history information in the memory 202. And, the present processing flow is terminated.

[0156] It should be noted that, even in the above-described process flow, when the operation of the welding robot 10 is completed, the process flow may be terminated.

[0157] As described above, according to the present embodiment, it is possible to achieve more accurate prediction of collision between the robot including the robot's accessories and surrounding objects.

[0158] <Other embodiments>

[0159] In the present invention, it can also be implemented by the following processing: a program or application for implementing the functions of one or more of the above-mentioned embodiments is supplied to a system or device using a network or storage medium, and one or more processors in a computer of the system or device reads and executes the program.

[0160] Alternatively, the circuit may be implemented by a circuit that implements one or more functions. Examples of the circuit that implements one or more functions include ASIC (Application Specific Integrated Circuit) and FPGA (Field Programmable Gate Array).

[0161] As described above, the following matters are disclosed in this specification.

[0162] (1) A collision prediction method for predicting a collision between a multi-jointed robot (e.g., 10) and a surrounding object (e.g., 50), wherein:

[0163] The collision prediction method has the following features:

[0164] A first acquisition step is to acquire the structure information and posture information of the robot;

[0165] A second acquisition step of acquiring point group data including the robot and the surrounding objects;

[0166] a classification step of classifying the point group data acquired in the second acquisition step into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; and

[0167] The prediction step predicts a collision between the robot and the surrounding object based on the classification result obtained by the classification step.

[0168] According to this configuration, it is possible to achieve more accurate prediction of collision between the robot including the robot's accessories and surrounding objects.

[0169] (2) The collision prediction method according to (1), wherein:

[0170] In the classification process, at least one of the following classification methods can be performed:

[0171] A first classification method for clustering the point group data obtained in the second acquisition step, and classifying the point group data into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the result of the clustering and the position of the robot determined by the structural information and posture information obtained in the first acquisition step;

[0172] a second classification method, using the first classification method on the point group data obtained in the second acquisition step to obtain first point group data corresponding to the robot, comparing the first point group data with the pre-held second point group data and removing the point group whose shortest distance is less than a predetermined threshold from the first point group data to classify it as the point group data corresponding to the robot;

[0173] a third classification method, predicting the position of the front end portion of the robot in the point group data acquired in the second acquisition step based on the structural information and the posture information, and classifying only the point group corresponding to the front end portion of the robot among the point groups overlapping with the three-dimensional model when the three-dimensional model of the robot is configured at the predicted position as the point group data corresponding to the robot; and

[0174] A fourth classification method is to obtain first point group data corresponding to the robot by using the first classification method on the point group data obtained in the second acquisition process, and to compare the first point group data with past point group data corresponding to the robot and past point group data corresponding to the surrounding objects to set the first point group data to the same classification as the point group data of the closer party.

[0175] According to this configuration, point cloud data can be classified using a plurality of classification methods, thereby achieving more appropriate collision prediction in accordance with processing load and processing accuracy.

[0176] (3) The collision prediction method according to (2), wherein:

[0177] The collision prediction method further comprises a determination step of determining whether the surrounding object has moved.

[0178] In the classification step, when it is determined that the surrounding object is moving, the first classification method is used, and when it is determined that the surrounding object is not moving, the second classification method is used.

[0179] According to this configuration, it is possible to perform collision prediction by switching the classification method according to the behavior of peripheral devices located around the robot.

[0180] (4) The collision prediction method according to (2) or (3), wherein:

[0181] In the prediction process, at least one of the following prediction methods can be performed:

[0182] A first prediction method compares the shortest distance between the point group data corresponding to the robot and the point group data corresponding to the surrounding objects with a predetermined threshold value, thereby predicting a collision;

[0183] A second prediction method includes setting at least one of the point group data corresponding to the robot and the point group data corresponding to the surrounding objects as voxels, determining overlap based on the voxels, and thereby predicting a collision; and

[0184] A third prediction method predicts a collision based on a difference in the number of points between current point group data corresponding to the robot and previous point group data corresponding to the robot.

[0185] According to this configuration, collision prediction can be performed using a plurality of types of prediction methods for point cloud data classified by a plurality of classification methods, thereby achieving more appropriate collision prediction in accordance with processing load and processing accuracy.

[0186] (5) The collision prediction method according to any one of (1) to (4), wherein:

[0187] The collision prediction method further comprises a pre-processing step of performing a predetermined pre-processing on the point group data acquired in the second acquisition step.

[0188] The predetermined pre-processing includes a process of removing point group data located outside the movable range of the robot.

[0189] The classification step classifies the point cloud data pre-processed in the pre-processing step.

[0190] According to this configuration, it is possible to perform collision prediction with higher accuracy and with a reduced processing load, focusing on the range in which a collision of the robot may occur.

[0191] (6) The collision prediction method according to any one of (1) to (5), wherein:

[0192] The collision prediction method further comprises a setting step of setting a simulated point group for the point group data classified as point group data corresponding to the robot in the classification step.

[0193] In the prediction step, collision prediction between the robot and the surrounding object is performed using point group data including the simulation point group set in the setting step,

[0194] The pseudo point group is arranged in a straight line connecting the position of the front end of the robot and the position of the joint of the robot, or is arranged in a predetermined shape based on a position corresponding to a predetermined part of the robot.

[0195] According to this configuration, it is possible to perform collision prediction by taking into account a region corresponding to a desired member or part in addition to the point cloud data corresponding to the robot.

[0196] (7) The collision prediction method according to any one of (1) to (6), wherein:

[0197] The collision prediction method also comprises:

[0198] a step of acquiring information on a moving direction of a predetermined position of the robot; and

[0199] an extraction step, for the point group data corresponding to the robot, setting a predetermined range along the moving direction of the robot based on the predetermined position, and extracting the point group data corresponding to the surrounding objects contained in the range,

[0200] In the prediction step, collision prediction between the robot and the surrounding objects is performed using the point group data corresponding to the surrounding objects and the point group data corresponding to the robot extracted in the extraction step.

[0201] According to this configuration, the point group data used for prediction can be limited and processed in consideration of the movement direction of the robot, and the amount of point group data can be reduced, thereby achieving high-speed processing.

[0202] (8) The collision prediction method according to any one of (1) to (7), wherein:

[0203] The posture information includes at least one of a position and an angle of a front end portion of the robot and a position and an angle of a joint portion of the robot.

[0204] According to this configuration, collision prediction can be performed with high accuracy based on the specifications of the robot and the real-time state of the robot.

[0205] (9) The collision prediction method according to any one of (1) to (8), wherein:

[0206] The structural information includes a three-dimensional model representing the shape of the robot.

[0207] According to this configuration, collision prediction can be performed with high accuracy using a three-dimensional model corresponding to the specifications of the robot.

[0208] (10) A collision prediction device (e.g., 20) for predicting a collision between a multi-jointed robot (e.g., 10) and a surrounding object (e.g., 50), wherein:

[0209] The collision prediction device comprises:

[0210] A first acquisition mechanism (e.g., 301), which acquires the structure information and posture information of the robot;

[0211] A second acquisition mechanism (e.g., 301), which acquires point group data including the robot and the surrounding objects;

[0212] a classification unit (e.g., 305) that classifies the point group data acquired by the second acquisition unit into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; and

[0213] The prediction unit (eg, 306 ) predicts a collision between the robot and the surrounding objects based on the classification result obtained by the classification unit.

[0214] According to this configuration, it is possible to achieve more accurate prediction of collision between the robot including the robot's accessories and surrounding objects.

[0215] (11) A welding system (e.g., 1), wherein:

[0216] The welding system includes the collision prediction device (eg, 20) described in (10) and a welding robot (eg, 10) as the robot.

[0217] According to this configuration, with respect to the welding robot as a target, it is possible to achieve more accurate prediction of a collision between the welding robot including the accessories of the welding robot and surrounding objects.

Claims

1. A collision prediction method for predicting collisions between a multi-joint robot and surrounding objects, wherein: The collision prediction method has the following features: A first acquisition step is to acquire the structure information and posture information of the robot; A second acquisition step of acquiring point group data including the robot and the surrounding objects; a classification step of classifying the point group data acquired in the second acquisition step into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; as well as The prediction step predicts a collision between the robot and the surrounding object based on the classification result obtained by the classification step.

2. The collision prediction method according to claim 1, wherein: In the classification process, at least one of the following classification methods can be performed: A first classification method for clustering the point group data obtained in the second acquisition step, and classifying the point group data into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the result of the clustering and the position of the robot determined by the structural information and posture information obtained in the first acquisition step; a second classification method, using the first classification method on the point group data obtained in the second acquisition step to obtain first point group data corresponding to the robot, comparing the first point group data with the pre-held second point group data and removing the point group whose shortest distance is less than a predetermined threshold from the first point group data to classify it as the point group data corresponding to the robot; a third classification method, based on the structural information and the posture information, predicting the position of the front end portion of the robot in the point group data obtained in the second acquisition step, and classifying only the point group corresponding to the front end portion of the robot among the point groups overlapping with the three-dimensional model when the three-dimensional model of the robot is configured at the predicted position as the point group data corresponding to the robot; as well as A fourth classification method is to obtain first point group data corresponding to the robot by using the first classification method on the point group data obtained in the second acquisition process, and to compare the first point group data with past point group data corresponding to the robot and past point group data corresponding to the surrounding objects to set the first point group data to the same classification as the point group data of the closer party.

3. The collision prediction method according to claim 2, wherein: The collision prediction method further comprises a determination step of determining whether the surrounding object has moved. In the classification step, when it is determined that the surrounding object is moving, the first classification method is used, and when it is determined that the surrounding object is not moving, the second classification method is used.

4. The collision prediction method according to claim 2, wherein: In the prediction process, at least one of the following prediction methods can be performed: A first prediction method compares the shortest distance between the point group data corresponding to the robot and the point group data corresponding to the surrounding objects with a predetermined threshold value, thereby predicting a collision; A second prediction method is to set at least one of the point group data corresponding to the robot and the point group data corresponding to the surrounding objects as voxels, and to determine overlap based on the voxels, thereby predicting a collision; as well as A third prediction method predicts a collision based on a difference in the number of points between current point group data corresponding to the robot and previous point group data corresponding to the robot.

5. The collision prediction method according to claim 1, wherein: The collision prediction method further comprises a pre-processing step of performing a predetermined pre-processing on the point group data acquired in the second acquisition step. The predetermined pre-processing includes a process of removing point group data located outside the movable range of the robot. The classification step classifies the point cloud data pre-processed in the pre-processing step.

6. The collision prediction method according to claim 1, wherein: The collision prediction method further comprises a setting step of setting a simulated point group for the point group data classified as point group data corresponding to the robot in the classification step. In the prediction step, collision prediction between the robot and the surrounding object is performed using point group data including the simulation point group set in the setting step, The pseudo point group is arranged in a straight line connecting the position of the front end of the robot and the position of the joint of the robot, or is arranged in a predetermined shape based on a position corresponding to a predetermined part of the robot.

7. The collision prediction method according to claim 1, wherein: The collision prediction method also comprises: a step of acquiring information on a moving direction of a predetermined position of the robot; and an extraction step, for the point group data corresponding to the robot, setting a predetermined range along the moving direction of the robot based on the predetermined position, and extracting the point group data corresponding to the surrounding objects contained in the range, In the prediction step, collision prediction between the robot and the surrounding objects is performed using the point group data corresponding to the surrounding objects and the point group data corresponding to the robot extracted in the extraction step.

8. The collision prediction method according to claim 1, wherein: The posture information includes at least one of a position and an angle of a front end portion of the robot and a position and an angle of a joint portion of the robot.

9. The collision prediction method according to claim 1, wherein: The structural information includes a three-dimensional model representing the shape of the robot.

10. A collision prediction device for predicting collisions between a multi-joint robot and surrounding objects, wherein: The collision prediction device comprises: A first acquisition mechanism, which acquires structural information and posture information of the robot; A second acquisition mechanism, which acquires point group data including the robot and the surrounding objects; a classification unit that classifies the point group data acquired by the second acquisition unit into point group data corresponding to the robot and point group data corresponding to the surrounding objects based on the structure information and posture information of the robot; as well as A prediction mechanism predicts a collision between the robot and the surrounding object based on the classification result of the classification mechanism.

11. A welding system, wherein: The welding system includes the collision prediction device according to claim 10 and a welding robot as the robot.

Images