JP7207704B2 - Learning system and robot positioning system - Google Patents

Description

The present invention relates to a learning system and a robot position adjustment system, and more particularly to technology for adjusting the working position of a robot.

Conventionally, as shown in FIG. 14, in a factory having a plurality of (production) lines, when performing work such as processing and assembling parts or products in each line at different time slots, workers should: They often move between lines to perform work on those lines.

On the other hand, in recent years, in a (production) line in a factory, a mobile robot with a working robot mounted on an unmanned guided vehicle is stopped in front of each piece of equipment on the line while traveling along the travel path, automatically A robot system designed to perform operations such as processing and assembly has appeared (see, for example, Patent Document 1).

JP-A-2002-73171

However, in the robot system as disclosed in Patent Document 1, the stop position of the automatic guided vehicle deviates from the predetermined stop position, which causes the (working) position of the working robot to deviate. In some cases, work troubles such as stoppage of work by the work robot and failure of work such as processing and assembly by the work robot may occur.

In particular, when the above-mentioned mobile robot is used in common for multiple lines in a factory, the number and number of facilities that serve as stopping positions for automated guided vehicles on the line is larger than when it is used for only one line. Due to the large number of types, there is a high possibility that the stop position of the automatic guided vehicle will deviate from the predetermined stop position, which is likely to cause troubles in the work of the above-mentioned work robot.

The above problem is solved by having workers stop in front of each piece of equipment (at each work position) on the line while running the trolley equipped with the work robot along the travel path, and perform processing, assembly, and other work. The same is true when

The present invention is intended to solve the above-mentioned problems, and it is possible to prevent troubles during work by a working robot even when the stop position of an automatic guided vehicle or cart deviates from a predetermined stop position. It is an object of the present invention to provide an efficient learning system and a robot position adjustment system.

In order to solve the above-mentioned problems, a learning system according to a first aspect of the present invention comprises a mobile robot having a working robot in a factory line, which has a camera, and an automatic guided vehicle for moving the working robot; Positional accuracy indicating whether or not the positional deviation of the work robot due to deviation of the stop position of the automatic guided vehicle on the line from a predetermined stop position is at a problematic level. A learning system comprising a position correct/incorrect information input unit for inputting information and a computer communicating with the mobile robot, wherein the mobile robot is photographed by the camera at each stop position of the automatic guided vehicle. a robot-side transmission unit configured to transmit the image data received from the robot-side transmission unit to the computer, and the computer determines in advance the stop position of the automatic guided vehicle on the line based on the image data transmitted from the robot-side transmission unit. a neural network storage unit for storing a neural network for determining whether or not positional displacement of the working robot due to deviation from a predetermined stop position is at a problematic level; and the robot side. A machine learning unit that performs machine learning of the neural network based on the image data transmitted from the transmission unit and the position correctness/incorrectness information input by the user from the position correctness/incorrectness information input unit.

In this learning system, the mobile robot is used in common on a plurality of lines in the factory, and the machine learning unit determines when the mobile robot is positioned on any one of the plurality of lines. Also, machine learning of the neural network may be performed based on the image data transmitted from the robot-side transmission unit and the position correct/incorrect information input by the user from the position correct/incorrect information input unit. .

A robot positioning system according to a second aspect of the present invention comprises a working robot on a factory line having a camera, a mobile robot equipped with an automated guided vehicle for moving the working robot, and for inputting positional correctness/incorrectness information indicating whether or not the positional deviation of the work robot due to deviation of the stop position of the transport vehicle from a predetermined stop position is at a problematic level; A robot position adjustment system comprising a position correct/incorrect information input unit, a robot control unit that controls the working robot, and a computer that communicates with the mobile robot, wherein the mobile robot is an automatic guided vehicle. a robot-side transmission unit for transmitting image data captured by the camera at each stop position to the computer, the robot control unit moving the working robot based on the instruction command received from the computer; The computer uses a computer-side transmission unit that transmits the instruction command to the robot control unit, and a trained neural network, based on the image data transmitted from the robot-side transmission unit, to perform the above-mentioned command on the line. a position correctness/incorrectness determination unit that determines whether or not positional deviation of the working robot due to deviation of the stop position of the automated guided vehicle from a predetermined stop position is at a problematic level ; When the positional deviation of the work robot is at a problematic level as a result of the determination by the position correctness determination unit, the position of the work robot is sent to the robot control unit using the computer-side transmission unit. a command transmission control unit for controlling to transmit an instruction command for adjustment , wherein the trained neural network receives the image data transmitted from the robot-side transmission unit and the position correctness/incorrectness information input unit. It is a neural network machine-learned based on the position correct/incorrect information input by the user.

In this robot position adjustment system, the mobile robot is used in common on a plurality of lines in the factory, and the position correctness/incorrectness determination unit determines whether the mobile robot is positioned on any of the plurality of lines. Also when doing so, it may be determined based on the image data transmitted from the robot-side transmission unit whether or not the positional deviation of the working robot is at a problematic level .

In this robot position adjustment system, the position correctness/incorrectness determination unit determines whether the positional deviation of the arm of the working robot is at a problem level based on the image data transmitted from the robot-side transmission unit. may be determined, and the instruction command transmitted by the command transmission control unit may be an instruction command for adjusting the position of the arm of the working robot.

A robot positioning system according to a third aspect of the present invention comprises a working robot on a factory line having a camera, a mobile robot equipped with an unmanned guided vehicle for moving the working robot, and the unmanned robot on the line. A position for inputting position correctness/incorrectness information indicating whether the positional deviation of the work robot due to deviation of the stop position of the transport vehicle from each predetermined stop position is at a problematic level. A robot position adjustment system comprising a correct/incorrect information input unit and a robot control unit that controls the working robot, wherein the robot control unit uses a trained neural network to determine each position of the automatic guided vehicle. Positional deviation of the working robot caused by deviation of the stop position of the automatic guided vehicle on the line from each of the predetermined stop positions based on the image data captured by the camera at the stop position. and a position correctness/incorrectness determination unit for determining whether or not the level is at a problematic level, and as a result of determination by the position correctness/incorrectness determination unit, when the positional deviation of the work robot is at a problematic level, the work The trained neural network receives image data captured by the camera at each stop position of the automatic guided vehicle and input by the user from the position correct/incorrect information input unit. It is a neural network that is machine-learned based on the position correct/incorrect information.

In this robot position adjustment system, the mobile robot is used in common on a plurality of lines in the factory, and the position correctness/incorrectness determination unit determines whether the mobile robot is positioned on any of the plurality of lines. Also when doing so, it is determined whether or not the positional deviation of the working robot is at a problem level based on the image data captured by the camera at each stop position of the automatic guided vehicle. good too.

According to the learning system according to the first aspect of the present invention, the image data transmitted from the robot side transmission unit (image data captured by the camera at each stop position of the automatic guided vehicle) and the user input, ( positional correctness information indicating whether the positional deviation of the working robot caused by the deviation of the stop position of the automatic guided vehicle from the predetermined stop position on the line is at a problematic level ; can be used for machine learning of neural networks. Then, the learned neural network after this machine learning detects that the stop position of the automated guided vehicle on the line deviates from the predetermined stop position based on the image data transmitted from the robot-side transmission unit. It is possible to determine whether or not the resulting positional deviation of the work robot is at a problematic level . As a result, even if the stop position of the automatic guided vehicle deviates from the predetermined stop position, it is possible to prevent troubles in the work by the working robot based on the above determination result.

According to the robot position adjustment system according to the second aspect of the present invention, the computer uses a trained neural network to determine the stop position of the automatic guided vehicle on the line based on the image data transmitted from the robot-side transmitter. It is determined whether or not the positional deviation of the working robot caused by the deviation from the predetermined stop position is at a problematic level. is at a problematic level, control is performed to transmit an instruction command for adjusting the position of the working robot to the robot control unit. Then, the robot control unit moves the working robot based on the instruction command received from the computer to adjust the position of the working robot. As a result, even if the stop position of the automatic guided vehicle deviates from the predetermined stop position, the position of the working robot can be adjusted, preventing troubles in the work caused by the working robot. be able to.

According to the robot position adjustment system according to the third aspect of the present invention, the robot control unit uses a learned neural network to determine the position on the line based on the image data captured by the camera at each stop position of the automatic guided vehicle. It is determined whether or not the positional deviation of the working robot due to the deviation of the stop position of the automatic guided vehicle from the predetermined stop position is at a problematic level, and as a result of this determination, To control to adjust the position of a working robot when the positional deviation of the working robot is at a problematic level . As a result, even if the stop position of the automatic guided vehicle deviates from the predetermined stop position, the position of the working robot can be adjusted, preventing troubles in the work caused by the working robot. be able to.

FIG. 2 is a side view of the mobile robot in the learning/robot position adjustment system of the first embodiment of the present invention; (a) and (b) are, respectively, an electrical block configuration diagram of an automatic guided vehicle of the mobile robot and an electrical block configuration diagram of a rotary laser radar of the automatic guided vehicle. FIG. 2 is an electrical block diagram of the learning/robot position adjustment system. FIG. 4 is a functional block configuration diagram of a CPU of the server in FIG. 3; The figure which shows the state of the work of the robot in the equipment (machining center) arrange|positioned at the line of a factory. FIG. 4 is an explanatory diagram of a case where the mobile robot is commonly used in a plurality of lines in a factory; FIG. 4 is an explanatory diagram of a case where the above-mentioned mobile robot is commonly used for a plurality of lines in which facilities are arranged on both sides of a traveling path; FIG. 2 is an electrical block configuration diagram of a robot position adjustment system in which a learned arm position correct/wrong judgment NN and an arm position adjustment program are stored in a memory of a centralized control operation panel; FIG. 2 is an electrical block configuration diagram of a robot position adjustment system in which a learned arm position correctness determination NN and an arm position adjustment program are stored in a memory of a robot controller; FIG. 2 is an electrical block configuration diagram of a robot position adjustment system in which a learned arm position correctness determination NN and an arm position adjustment program are stored in a memory of a robot controller mounted on the robot. FIG. 11 is a side view of a robot with a cart in a learning/robot position adjustment system according to a second embodiment of the present invention; FIG. 2 is an electrical block configuration diagram of the same learning/robot position adjustment system. FIG. 4 is an explanatory diagram of a case where the robot with a cart is commonly used in a plurality of lines in a factory; Explanatory diagram of work in a factory having a plurality of conventional lines.

A learning system and a robot position adjustment system according to embodiments embodying the present invention will be described below with reference to the drawings.

First, a learning system according to a first embodiment of the present invention and a mobile robot in a robot positioning system will be described with reference to FIG. A mobile robot 1 includes a robot 2 that is a working robot in a factory line and an unmanned guided vehicle 3 that moves the robot 2 . As shown in FIG. 1, the robot 2 is mounted on a pedestal 15 of the automatic guided vehicle 3. As the automatic guided vehicle 3 moves and stops, each work position on the line (the position of each piece of equipment on the line) changes. ) and stop. The robot 2 described above includes a camera 4 , an arm 5 and a hand 6 . The camera 4 has a built-in image sensor, and captures images of the surroundings of the robot 2 moment by moment. Send to 8. The arm 5 is, for example, a vertically articulated robot arm or a horizontally articulated robot arm.

The camera 4 in the mobile robot 1 may be attached to the side of the hand 6 as indicated by the solid line in FIG. 1, or may be attached to the base 2a of the robot 2 as indicated by the broken line in FIG. Alternatively, it may be attached to the pole 16 installed on the pedestal 15 of the automatic guided vehicle 3 . In FIG. 1, the camera mounted on the base 2a of the robot 2 is denoted by 4', and the camera mounted on the pole 16 is denoted by 4''.

The automatic guided vehicle 3 includes a rotary laser radar 11 , front wheels 13 a , rear wheels 13 b , motor units 12 a and 12 b with encoders for the front and rear wheels, a handle 14 and a base 15 . The number of front wheels 13a is one or two, and the number of rear wheels 13b is two. The handle 14 is held by the user when moving the mobile robot 1 from one line to another (see FIG. 6). The pedestal 15 described above is a stand for placing the robot 2 thereon.

Next, referring to FIGS. 2(a) and 2(b) in addition to FIG. 1, the electrical block configuration of the automatic guided vehicle 3 will be described. As shown in FIG. 2B, the rotary laser radar 11 includes a laser projector 11a that emits a laser beam, and a reflector R (see FIG. 2A) for the laser beam emitted from the laser projector 11a. a laser receiver 11b composed of a photodiode for receiving the reflected light, a turntable 11d rotatably supporting the radar main body 11c including the laser projector 11a and the laser receiver 11b, and rotating the turntable 11d and a motor unit 11e with an encoder for.

Further, the automatic guided vehicle 3 includes an AGV controller 20 that controls and calculates the entire device. The AGV controller 20 controls, for example, the rotation of the turntable 11d by the motor unit 11e with the encoder, the control of the laser projector 11a and the laser receiver 11b, the drive circuits 21a and 21b and the steering circuits 22a and 22b, which will be described later. and detection (calculation) of the distance and direction to each reflector R, which will be described later.

The automatic guided vehicle 3 (the AGV controller 20 thereof) is arranged from the rotary laser radar 11 to the traveling path (see 76a to 76c in FIG. 6, etc.) of the factory line using the rotary laser radar 11 described above. The distance and direction to each reflector R are detected. The AGV controller 20 obtains the distance to each reflector R based on the time from the emission of the laser beam by the laser projector 11a to the reception of the reflected light from each reflector R by the laser receiver 11b. Further, the AGV controller 20 obtains the rotation angle of the turntable 11d based on the output value from the encoder built in the motor unit 11e with the encoder, thereby determining the rotation angle of the turntable 11d from each reflector R by the laser receiver 11b. The direction of each reflector R is detected when reflected light is received.

After arranging each reflector R along the traveling path (of the unmanned guided vehicle 3) in the factory line, the AGV controller 20 detects the distance and direction to each reflector R, thereby enabling unmanned transportation. The car 3 can be made to travel at an appropriate position on the travel road.

The motor units 12a and 12b with encoders shown in FIGS. 1 and 2(a) drive and steer the front wheels 13a and the rear wheels 13b, respectively. A drive circuit 21a and a steering circuit 22a shown in FIG. 2A are circuits for driving and steering the front wheels 13a, and a drive circuit 21b and a steering circuit 22b are circuits for driving and steering the rear wheels 13b. circuit.

Next, referring to FIG. 3, the electrical block configuration of the learning/robot position adjustment system 10 according to the first embodiment will be described. This learning/robot position adjustment system 10 includes the components of both the learning system in claims 1 and 2 and the robot position adjustment system in claims 5 to 7 . That is, the learning/robot position adjustment system 10 corresponds to both the learning system in claims 1 and 2 and the robot position adjustment system in claims 5 to 7. As shown in FIG. 3, the learning/robot position adjustment system 10 includes the above mobile robot 1, a robot controller 9 (robot control unit) which is a controller for controlling the robot 2 in the mobile robot 1, and a centralized control operation panel. 7, a cloud-based server 8 (a “computer” in the claims) that communicates with the centralized control panel 7 via a network (Internet 17), and an auxiliary operation panel 18 .

The robot 2 of the mobile robot 1 includes a transmission/reception line 31 (robot-side transmission section) in addition to the camera 4, arm 5, and hand 6 described above. The transmission/reception line 31 is a circuit for wired or wireless communication with the robot controller 9 . This transmission/reception line 31 transmits image data captured by the camera 4 at each stop position of the automatic guided vehicle 3 to the server 8 via the robot controller 9 and the central control operation panel 7 .

The robot controller 9 described above is configured using a microcontroller, and controls the operations of the arm 5 and hand 6 of the robot 2 using a program for robot work. Also, the robot controller 9 moves the robot 2 based on instruction commands received from the server 8 via the Internet 17 and the central control operation panel 7 . The robot controller 9 has a CPU 33 , a transmission/reception line 34 and a memory 35 . The CPU 33 performs overall control and calculation of the robot controller 9 . The transmission/reception line 34 is a circuit for wired or wireless communication with (the transmission/reception line 31 of) the robot 2 and (the transmission/reception line 42 of) the central control operation panel 7 . The memory 35 stores data such as programs for robot work.

The centralized control operation panel 7 is used for controlling various devices in the line and inputting instructions to the various devices. The centralized control operation panel 7 includes a CPU 41 , a transmission/reception line 42 , a memory 43 , an operation section 44 (a position correct/incorrect information input section in claims), and a monitor 45 . The operation unit 44 of the centralized control operation panel 7 is used to perform input instruction operations to the robot controller 9 for setting the operation of the robot 2, and an arm position correctness determination neural network (hereinafter referred to as "arm position correctness determination" on the server 8 side to be described later). NN") 55 is used to input position correctness information for creating a training (learning) data set for machine learning. This position correct/incorrect information indicates whether or not there is a problem with the position of the robot 2 (more precisely, the position of the arm 5 of the robot 2) when the robot 2 is working in each piece of equipment on the line. This is information (so-called annotation data) to indicate.

An operator at the centralized control operation panel 7 watches on a monitor 45 images of the surroundings of the robot 2 during work, sent from the camera 4 of the robot 2 at each stop position of the automatic guided vehicle 3 (each work position of the robot 2). When it is determined that there is no problem with the position of the robot 2 (more precisely, the position of the arm 5 of the robot 2), the operating unit 44 is used to indicate that there is no problem with the position of the robot 2. When correct/incorrect information is input and it is determined that there is a problem with the position of the robot 2, the operation unit 44 is used to enter position correct/incorrect information indicating that there is a problem with the position of the robot 2. FIG. The above-mentioned "there is a problem with the position of the robot 2" means that the position of the robot 2 (more precisely, the position of the arm 5 of the robot 2) has deviated, causing the robot 2 to stop working or It means the case where a work trouble such as a failure of the work such as processing or assembly by the robot 2 actually occurs, and the case where these troubles could have occurred.

The server 8 includes a CPU 51, a transmission/reception line 52 (a "computer-side transmission unit" in the claims), a hard disk 53 (a "neural network storage unit" in the claims), an operation unit 58, and a memory 59. there is The CPU 51 performs control and calculation of the server 8 as a whole. The hard disk 53 stores a training data set 54 , an arm position correct/wrong decision NN 55 , a learning program 56 and an arm position adjusting program 57 .

The arm position correct/wrong determination NN55 is image data captured by the camera 4 of the robot 2 at each stop position of the automatic guided vehicle 3 (each work position of the robot 2), which is transmitted from (the transmission/reception line 31 of) the robot 2. It is a neural network that determines whether there is a problem with the (current) position of the arm 5 of the robot 2 based on. The arm position correct/wrong determination NN 55 is a type of neural network that performs object recognition (image class classification), and is preferably based on, for example, a convolutional neural network.

The above-mentioned training data set 54 is an image when the above-mentioned positional correctness/incorrectness information is input, among the images of the surroundings of the robot 2 in operation, which are sent from the camera 4 of the robot 2 at each stop position of the automatic guided vehicle 3. data (image data) and position correct/incorrect information corresponding to the image data. The learning program 56 is a program for performing machine learning of the arm position correct/wrong determination NN 55 based on the image data and position correct/wrong information stored in the training data set 54 . The arm position adjustment program 57 is a program for adjusting the position of the arm 5 of the robot 2 when there is a problem with the position of the arm 5 of the robot 2 as a result of the determination using the arm position correct/wrong determination NN55. is.

FIG. 4 shows functional blocks of the CPU 51 on the server 8 side. The CPU 51 on the server 8 side has a machine learning section 61 , a position correctness determination section 62 , and a command transmission control section 63 . The machine learning unit 61 uses the above image data and position correct/incorrect information stored in the training data set 54 (image data transmitted from the transmission/reception line 31 of the robot 2 and input by the user from the operation unit 44 of the central control operation panel 7). Machine learning of the arm position correctness/incorrectness determination NN55 is performed based on the position correctness/incorrectness information). The position correct/wrong determination unit 62 uses the learned arm position correct/wrong determination NN 55 to determine the image data (taken by the camera 4 at each stop position of the unmanned guided vehicle 3) transmitted from the transmission/reception line 31 of the robot 2. Based on this, it is determined whether or not there is a problem with the position of the arm 5 of the robot 2 . The command transmission control unit 63 uses the transmission/reception line 52 to notify the robot controller 9 of the position of the arm 5 of the robot 2 when there is a problem with the position of the arm 5 of the robot 2 as a result of the determination by the position correctness determination unit 62 . control to send an instruction command to adjust the

Next, the adjustment of the position of the arm 5 of the robot 2 will be further described with reference to FIG. 5 in addition to FIGS. FIG. 5 shows how the robot 2 works when the equipment arranged on the factory line is the machining center 71 . In FIG. 5, the robot 2 holds a work 72 in the hand 6, and the work 72 is fixed at a jig 73 of the machining center 71 when the arm 5 of the robot 2 is positioned correctly during work. (Position indicated by broken line in the figure) and cut by the cutting tool 74 . However, if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position and there is a problem with the position of the arm 5 of the robot 2 during work, the work of the robot 2 may be stopped or the robot There is a possibility of causing work troubles such as failure of work such as cutting by 2.

In the learning/robot position adjustment system 10 of the present embodiment, when learning the arm position correct/wrong determination NN55, as described above, if there is a problem with the position of the arm 5 of the robot 2 (if a trouble actually occurs during work) , and if these troubles could have occurred), the operator at the central control operation panel 7 uses the operation unit 44 to indicate that there is a problem with the position of the arm 5 of the robot 2. Enter your information. Conversely, when there is no problem with the position of the arm 5 of the robot 2 , the operator uses the operation unit 44 to input position correct/incorrect information indicating that there is no problem with the position of the arm 5 of the robot 2 . Then, the server 8 (the machine learning unit 61 thereof) receives the image data transmitted from the camera 4 of the robot 2 side and the operator ( Machine learning of the arm position correct/wrong determination NN55 is performed based on the position correct/wrong information input by the user. At this point, even if there is a problem with the position of the arm 5 of the robot 2, the position of the arm 5 of the robot 2 ( automatic) adjustment is not performed.

Then, when the machine learning of the arm position correct/wrong determination NN55 by the server 8 has progressed to some extent and the arm position correct/wrong determination NN55 can be regarded as a trained neural network, for example, the system administrator on the server 8 side , the operation unit 58 instructs to start (automatic) adjustment processing of the position of the arm 5 of the robot 2 using the arm position correct/wrong determination NN 55 on the server 8 side and the arm position adjustment program 57 .

When the (automatic) adjustment processing of the position of the arm 5 of the robot 2 described above is started, the CPU 51 of the server 8 (the position correctness determination unit 62 thereof) uses the learned arm position correctness determination NN 55 to determine the position of the robot 2. Based on image data transmitted from the transmission/reception line 31 (image data of the surroundings of the robot 2 transmitted from the camera 4 when the robot 2 is working), it is determined whether or not there is a problem with the position of the arm 5 of the robot 2. . Then, the CPU 51 (the command transmission control unit 63 thereof) of the server 8 uses the transmission/reception line 52 when there is a problem with the position of the arm 5 of the robot 2 as a result of the determination using the learned arm position correct/wrong determination NN55. Then, the robot controller 9 is controlled to transmit an instruction command for adjusting the position of the arm 5 of the robot 2 . Then, the robot controller 9 moves the robot 2 and adjusts the position of the arm 5 of the robot 2 based on the instruction command received from the server 8 . As a result, even if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 caused by the deviation of the stop position of the automatic guided vehicle 3 can cause a problem. It is determined whether there is a certain level (whether there is a problem with the position of the arm 5 of the robot 2), and as a result of this determination, if there is a problem with the position of the arm 5 of the robot 2, the arm 5 position can be adjusted. Therefore, even if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position, it is possible to prevent troubles in the operation of the robot 2 .

During the position adjustment process of the arm 5 of the robot 2, the learned arm position correctness determination NN 55 determines that there is no problem with the position of the arm 5 of the robot 2 based on the image data transmitted from the transmission/reception line 31 of the robot 2. The position of the arm 5 of the robot 2 is repeatedly adjusted until

In this embodiment, as shown in FIG. 6, mobile robot 1 is commonly used in a plurality of lines in a factory. Specifically, a worker grips the handle 14 of the mobile robot 1 shown in FIG. and the mobile robot 1 is caused to work on the line B. In this manner, the mobile robot 1 is alternately used on lines A to C according to the operating schedule of each line on that day.

In FIG. 6, reference numerals 79a to 79f denote facilities such as machining centers arranged on lines A to C, and reference numerals 76a to 76c denote running paths on lines A to C, respectively. Further, reference numerals 77a to 77c indicate parts feeding devices in lines A to C, respectively, reference numerals 78a to 78c indicate finished product discharge conveyors in lines A to C, respectively, and reference numerals 18a to 18c indicate , respectively show the auxiliary consoles in each of the lines A-C. The parts feeding devices 77 a to 77 c described above are devices for supplying workpieces to the mobile robot 1 . Finished product discharge conveyors 78a to 78c are conveyors for discharging finished products after processing or assembly process by facilities 79a to 79f such as machining centers.

In addition, (1) to (4) in FIG. 6 are respective stop positions of the automatic guided vehicle 3 on lines A to C. As shown in FIG. (1) above is the position where the robot 2 receives the workpiece (part), (2) and (3) are the positions where the robot 2 performs the first and second operations on each line, and ( 4) is the position where the robot 2 places the finished product on the finished product discharge conveyors 78a-78c (or in the cardboard boxes on the finished product discharge conveyors 78a-78c).

As shown in FIG. 6, when the mobile robot 1 is used in common for a plurality of lines in a factory, compared with the case where the mobile robot 1 is used for only one line, the stop position and Since the number and types of equipment (equipment 79a to 79f in FIG. 6) are large, there is a high possibility that the stop position of the automatic guided vehicle 3 will deviate from the predetermined stop position. The above problems can easily occur.

However, according to the learning/robot position adjustment system 10 of the present embodiment, when the machine learning of the arm position correct/wrong determination NN55 progresses to some extent and the arm position correct/wrong determination NN55 can be regarded as a learned neural network, The CPU 51 of the server 8 (the position correctness determination unit 62 thereof) uses the learned arm position correctness determination NN 55 to determine the position of the robot 2 when the mobile robot 1 is positioned on any of the plurality of lines. Based on the image data transmitted from the transmission/reception line 31, it is determined whether or not there is a problem with the position of the arm 5 of the robot 2. When the CPU 51 of the server 8 (the command transmission control unit 63 thereof) determines that there is a problem with the position of the arm 5 of the robot 2 as a result of determination using the arm position correctness determination NN 55 that has already been learned, the transmission/reception line 52 is used. Then, the robot controller 9 is controlled to transmit an instruction command for adjusting the position of the arm 5 of the robot 2 . Therefore, even when the mobile robot 1 is commonly used in a plurality of lines in a factory, the position of the arm 5 of the robot 2 can be adjusted if there is a problem with the position of the arm 5 of the robot 2. .

A factory line to which the learning/robot position adjustment system 10 of the present embodiment and its mobile robot 1 can be applied is a line in which facilities 79a to 79f are arranged on one side of traveling paths 76a to 76c as shown in FIG. However, the line may be a line in which facilities 79g to 79n are arranged on both sides of running paths 76d and 76e as shown in FIG.

In FIG. 7, reference numerals 79g to 79n denote facilities such as machining centers arranged on lines A and B, and reference numerals 76d and 76e denote running paths on lines A and B, respectively. Further, reference numerals 77d to 77g indicate parts feeding devices in lines A and B, respectively, reference numerals 78d to 78g indicate finished product discharge conveyors in lines A and B, respectively, and reference numerals 18d and 18e indicate , respectively show the auxiliary consoles in each of the lines A and B. FIG.

When the machine learning of the arm position correct/wrong determination NN55 by the server 8 progresses and the arm position correct/wrong determination NN55 can be regarded as a trained neural network, the learned The arm position correct/wrong determination NN81 may be stored in the memory 43 of the centralized control operation panel 7 or the memory 35 of the robot controller 9 together with the arm position adjustment program 57 . As shown in FIG. 8, in the robot position adjustment system 80 in which the memory 43 of the centralized control panel 7 stores the learned arm position correctness determination NN81 and the arm position adjustment program 57, the centralized control panel 7 corresponds to the "computer" of the robot positioning system in As shown in FIG. 9, the robot position adjustment system 83 is a robot position adjustment system in which the learned arm position correct/wrong determination NN81 and the arm position adjustment program 57 are stored in the memory 35 of the robot controller 9 . 8 and 9, electric blocks equivalent to those in FIG. 3 are denoted by the same reference numerals, and description thereof will be omitted.

10, the robot 2 itself is equipped with a robot controller 86 composed of a processor such as a microcontroller. The position correctness determination NN81 and the arm position adjustment program 57 may be stored. Also in FIG. 10, the same reference numerals are given to the same electric blocks as in FIG. 3, and the description thereof is omitted.

However, if there are many changes in the type, arrangement, number, etc. of equipment in the factory line to which the learning/robot position adjustment system 10 is applied, the arm position correct/wrong decision NN 55 stored in the hard disk 53 of the server 8 shown in FIG. , the arm position correctness determination NN 55 and the arm position adjustment program 57 of the server 8 are used to determine whether there is a problem with the position of the arm 5 of the robot 2 and to determine whether there is a problem with the position of the arm 5 of the robot 2. It is desirable to perform the adjustment processing of the position of 5.

As described above, according to the learning/robot position adjustment system 10 of the first embodiment, image data transmitted from the transmission/reception line 31 of the robot 2 (photographed by the camera 4 at each stop position of the automatic guided vehicle 3) machine learning of the arm position correct/wrong determination NN55 can be performed based on the position correct/wrong information input by the operator (user) from the operation unit 44 of the centralized control console 7 .

Further, according to the learning/robot position adjustment system 10 of the first embodiment, the machine learning of the arm position correct/wrong determination NN55 has progressed to some extent, and the arm position correct/wrong determination NN55 can be regarded as a trained neural network. After that, the server 8 determines whether there is a problem with the position of the arm 5 of the robot 2 based on the image data transmitted from the transmission/reception line 31 of the robot 2 using the learned arm position correct/wrong determination NN 55 . is determined, and as a result of this determination, when there is a problem with the position of the arm 5 of the robot 2, the robot controller 9 is controlled to transmit an instruction command for adjusting the position of the arm 5 of the robot 2. . Then, the robot controller 9 moves the robot 2 and adjusts the position of the arm 5 of the robot 2 based on the instruction command received from the server 8 . As a result, even if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 caused by the deviation of the stop position of the automatic guided vehicle 3 can cause a problem. It is determined whether there is a certain level (whether there is a problem with the position of the arm 5 of the robot 2), and as a result of this determination, if there is a problem with the position of the arm 5 of the robot 2, the arm 5 position can be adjusted. Therefore, it is possible to prevent troubles during work by the robot 2 .

8 and the arm position adjustment program 57 are stored in the memory 43 of the centralized control console 7, the arm position correct/wrong determination NN 55 is a learned neural network. It is possible to obtain the same actions and effects as those of the learning/robot position adjustment system 10 after reaching a state in which it can be considered to be. Specifically, the centralized control panel 7 uses the learned arm position correct/wrong judgment NN81 to determine whether there is a problem with the position of the arm 5 of the robot 2 based on the image data transmitted from the transmission/reception line 31 of the robot 2. If there is a problem with the position of the arm 5 of the robot 2 as a result of this determination, control is performed so that an instruction command for adjusting the position of the robot 2 is sent to the robot controller 9 . Then, the robot controller 9 moves the robot 2 and adjusts the position of the arm 5 of the robot 2 based on the instruction command received from the central control operation panel 7 . As a result, even if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 caused by the deviation of the stop position of the automatic guided vehicle 3 can cause a problem. It is determined whether or not there is a certain level, and if there is a problem with the position of the arm 5 of the robot 2 as a result of this determination, the position of the arm 5 of the robot 2 can be adjusted. Therefore, it is possible to prevent troubles during work by the robot 2 .

Also in the robot position adjustment system 83 shown in FIG. 9 and the robot position adjustment system 85 shown in FIG. Based on the image data captured by the camera 4 at each stop position, it is determined whether there is a problem with the position of the arm 5 of the robot 2, and as a result of this determination, there is a problem with the position of the arm 5 of the robot 2. At times, it controls to adjust the position of the arm 5 of the robot 2 . As a result, even if the stop position of the automatic guided vehicle 3 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 caused by the deviation of the stop position of the automatic guided vehicle 3 can cause a problem. It is possible to determine whether or not there is a certain level (whether or not there is a problem with the position of the robot 2), and as a result of this determination, if there is a problem with the position of the robot 2, the position of the robot 2 can be adjusted. . Therefore, it is possible to prevent troubles during work by the robot 2 .

Next, with reference to FIGS. 11 to 13, a learning/robot position adjustment system 95 according to a second embodiment of the present invention and a cart-equipped robot 91 in this learning/robot position adjustment system 95 will be described. In FIGS. 11 and 13, the same members as in FIGS. 1 and 6 are denoted by the same reference numerals, and descriptions thereof are omitted. Also, in FIG. 12, the same reference numerals are given to the same electric blocks as those in FIG. 3, and the description thereof will be omitted.

FIG. 11 shows a cart-equipped robot 91 in a learning/robot positioning system 95 of the second embodiment. The learning/robot position adjustment system 10 of the first embodiment includes the mobile robot 1 in which the robot 2 is mounted (mounted) on the pedestal 15 of the automatic guided vehicle 3. However, the learning/robot position adjustment system 10 of the second embodiment A robot positioning system 95 includes a cart-equipped robot 91 in which a robot 2 is attached (mounted) on a pedestal 96 of a cart 92 instead of the mobile robot 1 described above. Other configurations are the same between the learning/robot position adjustment system 10 of the first embodiment and the learning/robot position adjustment system 95 of the second embodiment.

The carriage 92 includes front and rear wheels 93 and a handle 97 in addition to a pedestal 96 to which the robot 2 is attached. Similar to the handle 14 of the automatic guided vehicle 3 in the first embodiment, the handle 97 is gripped by the user when the robot with cart 91 is moved from one line to another (see FIG. 13). be. In the second embodiment, as shown in FIG. 13, rails 94 (rails 94a to 94c in the example of FIG. 13) are provided on the running paths (running paths 76a to 76c in the example of FIG. 13) in each line of the factory. ) are arranged, and front and rear wheels 93 of the carriage 92 are fitted in the rails 94 and move on the rails 94 .

Also in this embodiment, as shown in FIG. 13, the robot 2 attached to the robot with a cart 91 is commonly used in a plurality of lines in the factory. Specifically, an operator grips the handle 97 of the robot with cart 91 shown in FIG. (the front and rear wheels 93 of the carriage 92 are removed from the rails 94a on the line A and fitted into the rails 94b on the line B), and the robot 2 of the robot 91 with the carriage is moved to each work position on the line B. The worker is moved to perform work using each facility of line B. In this manner, the cart-equipped robot 91 is alternately used on lines A to C according to the operating schedule of each line on that day.

FIG. 12 shows the electrical block configuration of the learning/robot position adjustment system 95 of the second embodiment. This electrical block configuration is the same as the electrical block configuration of the learning/robot position adjustment system 10 of the first embodiment shown in FIG.

According to the learning/robot position adjustment system 95 of the second embodiment, as in the learning/robot position adjustment system 10 of the first embodiment, image data transmitted from the transmission/reception line 31 of the robot 2 (the At each work position on the line, image data captured by the camera 4) and position correctness information input by the operator (user) from the operation unit 44 of the centralized control operation panel 7. Machine learning of NN55 can be performed.

Further, according to the learning/robot position adjustment system 95 of the second embodiment, the machine learning of the arm position correct/wrong determination NN55 has progressed to some extent, and the arm position correct/wrong determination NN55 can be regarded as a trained neural network. After that, the server 8 determines whether there is a problem with the position of the arm 5 of the robot 2 based on the image data transmitted from the transmission/reception line 31 of the robot 2 using the learned arm position correct/wrong determination NN 55 . is determined, and as a result of this determination, when there is a problem with the position of the arm 5 of the robot 2, the robot controller 9 is controlled to transmit an instruction command for adjusting the position of the arm 5 of the robot 2. . Then, the robot controller 9 moves the robot 2 and adjusts the position of the arm 5 of the robot 2 based on the instruction command received from the server 8 . As a result, even if the stop position of the carriage 92 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 due to the deviation of the stop position of the carriage 92 is at a problem level. (whether or not there is a problem with the position of the arm 5 of the robot 2). can be adjusted. Therefore, it is possible to prevent troubles during work by the robot 2 .

In the second embodiment as well, when machine learning of the arm position correctness determination NN55 by the server 8 has progressed to some extent and the arm position correctness determination NN55 can be regarded as a learned neural network, the first embodiment Similar to the robot position adjustment system 80 and the robot position adjustment system 83 (see FIGS. 8 and 9) in the embodiment, the learned arm position correct/wrong decision NN 81 and the arm position adjustment program 57 are stored in the memory 43 of the centralized control console 7, Alternatively, it may be stored in the memory 35 of the robot controller 9 . In addition, similar to the robot position adjustment system 85 (see FIG. 10) in the first embodiment, the robot 2 itself is equipped with a robot controller 86, and a memory 88 in the robot controller 86 stores the correct/wrong learned arm position. The judgment NN81 and the arm position adjustment program 57 may be stored.

Also in the second embodiment, similarly to the robot position adjustment system 80 in the first embodiment shown in FIG. 43, the same actions and effects as the learning/robot position adjustment system 95 can be obtained after the arm position correct/wrong determination NN 55 can be regarded as a trained neural network.

Also in the second embodiment, similarly to the robot position adjustment system 83 and the robot position adjustment system 85 in the first embodiment shown in FIGS. 57 is stored in the memory 35 of the robot controller 9 or the memory 88 of the robot controller 86, the robot controller 9 or the robot controller 86 uses the learned arm position correctness determination NN81 to detect the camera at each work position of the robot 2. Based on the image data photographed by 4, it is determined whether or not there is a problem with the position of the arm 5 of the robot 2. As a result of this determination, if there is a problem with the position of the arm 5 of the robot 2, the robot 2 to adjust the position of the arm 5. As a result, even if the stop position of the carriage 92 deviates from the predetermined stop position, the positional deviation of the arm 5 of the robot 2 due to the deviation of the stop position of the carriage 92 is at a problem level. (whether there is a problem with the position of the robot 2) is determined, and as a result of this determination, if there is a problem with the position of the robot 2, the position of the robot 2 can be adjusted. Therefore, it is possible to prevent troubles during work by the robot 2 .

Variant:
The present invention is not limited to the configuration of each of the above embodiments, and various modifications are possible without changing the gist of the invention. Next, modified examples of the present invention will be described.

Variant 1:
In each of the above embodiments, the image data transmitted from the transmission/reception line 31 on the robot 2 side and the position correct/incorrect information input by the operator from the operation unit 44 of the centralized control operation panel 7 are transmitted via the Internet 17 to Machine learning of the arm position correct/wrong determination NN 55 stored in the hard disk 53 of the server 8 is performed by transmitting to the server 8 on the cloud. However, a neural network that determines whether or not there is a problem with the position of the robot arm is stored in the hard disk of a computer located on the premises of the factory, and the neural network stored in the computer located on the premises of the factory is used. Machine learning of the network may be performed.

Modification 2:
Further, in each of the above-described embodiments, the "learned neural network" in the claims is a neural network that determines whether or not there is a problem with the position of the arm 5 of the robot 2, and this "learned neural network" ” shows an example of adjusting the position of the arm 5 of the robot 2 when there is a problem with the position of the arm 5 of the robot 2 . However, the learning system and the robot position adjustment system of the present invention are not limited to this. If the result of determination using this learned neural network is that there is a problem with the position of the robot's hand, the position of the robot's hand may be adjusted. In addition, the "learned neural network" is a neural network that determines whether there is a problem with the position of the tip, center, or rear end of the robot. As a result of determination using this trained neural network, When there is a problem with the position of the robot's leading end, center, or trailing end, the position of the robot's leading end, center, or trailing end may be adjusted.

Variant 3:
Further, in the first embodiment, an example in which the robot 2 is attached (mounted) on the pedestal 15 of the automatic guided vehicle 3 has been described, but the present invention is not limited to this. Alternatively, the automatic guided vehicle may move the robot by pushing the back of the robot. Alternatively, the robot may be attached to the front of the automated guided vehicle.

Variant 4:
Further, in the above-described first embodiment, an example in which the automatic guided vehicle 3 is a so-called laser-guided automatic guided vehicle has been described, but the automatic guided vehicle guidance method is not limited to this. A guidance system may be used, or a multi-guidance system using both magnetic guidance and laser guidance may be used.

1 mobile robot 2 robot (working robot)
3 Automatic guided vehicle 4 Camera 7 Central control operation panel (computer)
8 Server (computer)
9, 86 robot controller (robot control unit)
10, 95 Learning/robot positioning system (learning system and robot positioning system)
31 transmission/reception line (robot side transmitter)
44 operation unit (position correct/incorrect information input unit)
52 transmission/reception line (computer side transmission unit)
53 hard disk (neural network storage unit)
55 Arm position correct/wrong judgment NN (neural network for judging whether or not there is a problem with the position of the working robot, and trained neural network)
61 machine learning unit 62 position correct/wrong determination unit 63 command transmission control unit 80, 83, 85 robot position adjustment system 81 learned arm position correct/wrong determination NN (learned neural network)
92 Bogie A, B, C line

Claims (7)

a mobile robot comprising a working robot in a factory line having a camera and an automatic guided vehicle for moving the working robot;
Positional accuracy indicating whether or not the positional deviation of the work robot due to deviation of the stop position of the automatic guided vehicle on the line from a predetermined stop position is at a problematic level. a location accuracy information input unit for inputting information;
A learning system comprising a computer that communicates with the mobile robot,
The mobile robot includes a robot-side transmission unit that transmits image data captured by the camera at each stop position of the automatic guided vehicle to the computer,
The computer is
The position of the working robot caused by the shift of the stop position of the automatic guided vehicle on the line from a predetermined stop position based on the image data transmitted from the robot-side transmission unit. a neural network storage unit that stores a neural network that determines whether the deviation is at a problematic level ;
a machine learning unit that performs machine learning of the neural network based on the image data transmitted from the robot-side transmission unit and the position correctness/incorrectness information input by the user from the position correctness/incorrectness information input unit. system. The mobile robot is commonly used in a plurality of lines in the factory,
The machine learning unit receives the image data transmitted from the robot-side transmission unit and the position correctness/incorrectness information input unit from the user when the mobile robot is positioned on any one of the plurality of lines. 2. The learning system according to claim 1, wherein the machine learning of the neural network is performed based on the position correct/incorrect information input by. a mobile robot comprising a working robot in a factory line having a camera and an automatic guided vehicle for moving the working robot;
Positional accuracy indicating whether or not the positional deviation of the work robot due to deviation of the stop position of the automatic guided vehicle on the line from a predetermined stop position is at a problematic level. a location accuracy information input unit for inputting information;
a robot control unit that controls the working robot;
A robot positioning system comprising a computer that communicates with the mobile robot,
The mobile robot includes a robot-side transmission unit that transmits image data captured by the camera at each stop position of the automatic guided vehicle to the computer,
The robot control unit moves the working robot based on the instruction command received from the computer,
The computer is
a computer-side transmission unit that transmits the instruction command to the robot control unit;
Using a trained neural network, based on the image data transmitted from the robot-side transmission unit , the stop position of the automatic guided vehicle on the line deviates from a predetermined stop position. a position correctness/incorrectness determination unit that determines whether or not the positional deviation of the working robot is at a problematic level ;
When the positional deviation of the work robot is at a problematic level as a result of the determination by the position correctness determination unit, the position of the work robot is sent to the robot control unit using the computer-side transmission unit. a command transmission control unit that controls to transmit an instruction command for adjustment,
The trained neural network is a neural network that has undergone machine learning based on the image data transmitted from the robot-side transmission unit and the position correctness/incorrectness information input by the user from the position correctness/incorrectness information input unit. be
Robot positioning system. The mobile robot is commonly used in a plurality of lines in the factory,
The position correctness/incorrectness determination section determines the position of the work robot based on the image data transmitted from the robot-side transmission section when the mobile robot is positioned on any of the plurality of lines. 4. The robot positioning system according to claim 3 , wherein it is determined whether the displacement is at a problematic level . The position correctness/incorrectness determination unit determines whether or not the positional deviation of the arm of the working robot is at a problematic level based on the image data transmitted from the robot-side transmission unit,
5. The robot position according to claim 3 , wherein the instruction command that is controlled to be transmitted by the command transmission control unit is an instruction command for adjusting the position of the arm of the working robot. adjustment system. a mobile robot comprising a working robot in a factory line having a camera and an automatic guided vehicle for moving the working robot;
Positional correctness/incorrectness information indicating whether or not the positional deviation of the working robot due to deviation of the stop position of the automatic guided vehicle on the line from each predetermined stop position is at a problematic level. a location accuracy information input unit for inputting
A robot position adjustment system comprising a robot control unit that controls the working robot,
The robot control unit
Using a trained neural network, based on the image data captured by the camera at each stop position of the automatic guided vehicle, the stop position of the automatic guided vehicle on the line is determined from each of the predetermined stop positions. a position correctness/incorrectness determination unit that determines whether the positional deviation of the work robot due to the deviation is at a problematic level ;
controlling to adjust the position of the working robot when the positional deviation of the working robot is at a problematic level as a result of the determination by the position correctness determining unit ;
The trained neural network performs machine learning based on image data captured by the camera at each stop position of the automatic guided vehicle and the position correctness/incorrectness information input by the user from the position correctness/incorrectness information input unit. is a neural network with
Robot positioning system. The mobile robot is commonly used in a plurality of lines in the factory,
The position correctness/incorrectness determination unit, when the mobile robot is positioned on any one of the plurality of lines, based on the image data captured by the camera at each stop position of the automatic guided vehicle, 7. The robot position adjustment system according to claim 6 , wherein it is determined whether or not the positional deviation of the working robot is at a problematic level .

Images