JP7394695B2 - robot control device - Google Patents

Description

The present invention relates to a robot control device and a picking system.

In recent years, the decreasing working population has become a problem, and there is a need to eliminate labor shortages and improve productivity in the manufacturing industry. For this reason, there are high expectations for the use of autonomous robot systems in which robots can recognize, make decisions, and operate themselves.

Patent Document 1 discloses a technique for calculating an appropriate stacking pattern of articles. For this purpose, Patent Document 1 states, "A first physical model of the article is created based on the dimensions of each type of article, and a first physical model of the storage area is created based on the dimensions of the storage area in which a plurality of articles are stacked." a model creation unit 12 that creates a second physical model; a placement unit 14 that arranges the first physical model within the second physical model in order of priority predetermined for each type of article; Each time a model of and a physical calculation unit 16 that calculates changes in the position or orientation of the physical model.''

JP2019-89160A

In Patent Document 1, an appropriate stacking pattern is calculated by determining a place position within a container in consideration of the weight, shape, and load capacity of the articles. Patent Document 1 does not consider the moment balance of the entire container. For this reason, in Patent Document 1, for example, when a heavy item is placed unevenly in a container, the moment balance of the entire container is likely to be lost, making it difficult for a worker or a robot to stably grip the container.

The present invention has been made in view of the above-mentioned problems, and provides a robot control device and a picking system that are capable of storing a plurality of articles in a container while taking into account the moment balance of the container.

In order to solve the above problems, a robot control device according to one aspect of the present invention is a robot control device that controls a robot that stores articles in a container, and which stores information about a plurality of articles to be stored in the container. a container determination unit that determines a container based on the list; a position and orientation candidate extraction unit that extracts candidate patterns of positions and orientations of the plurality of articles in the container based on a first characteristic of the plurality of articles; a moment balance determination unit that determines a candidate pattern in which the balance of moments acting on the container by a plurality of articles falls within a predetermined range from among the candidate patterns extracted by the position and orientation candidate extraction unit and selects the candidate pattern as a final pattern; A control signal that determines a predetermined order for accommodating the plurality of articles in the container according to the final pattern selected by the determination unit, and causes the robot to output a control signal for accommodating the plurality of articles in the container according to the predetermined order. An output section.

According to the present invention, a plurality of articles can be accommodated in a container such that the moment balance of the container containing the plurality of articles falls within a predetermined range.

An overall schematic diagram of a picking system including a robot control device. FIG. 1 is a configuration explanatory diagram of a picking system including a robot control device. A functional explanatory diagram of a picking system including a robot control device. Example of article characteristics database. An example of a container database. Example of cushioning material database. 5 is a flowchart showing the overall processing of the robot control device. An explanatory diagram for calculating accommodation in a container by approximating articles and cushioning materials with polygons. FIG. 3 is a diagram illustrating patterns of positions that an article can take within a container. An explanatory diagram showing patterns of articles that can be stored in a container. An explanatory diagram showing an example of an article whose arrangement direction is determined. FIG. 3 is a diagram illustrating a state in which another article is placed on the article, the load capacity of which exceeds that of a certain article. An explanatory diagram of a pattern of articles that cannot be adjacent to each other. An explanatory diagram for calculating the moment of a container containing an article. An explanatory diagram for balancing the moment of a container containing articles. FIG. 7 is a functional explanatory diagram of a picking system including a robot control device according to a second embodiment. FIG. 7 is a functional explanatory diagram of a picking system including a robot control device according to a third embodiment. FIG. 7 is an explanatory diagram of a picking system including a robot control device according to a fourth embodiment. State transition diagram of reinforcement learning. 12 is a flowchart showing a process for determining the housing order according to the fifth embodiment.

Embodiments of the present invention will be described below based on the drawings. In this embodiment, a plurality of articles 7 are accommodated in the container 6 such that the moment balance of the container 6 containing the plurality of articles 7 falls within a predetermined range.

The plurality of articles 7 can include articles of different types. Different types of articles are, for example, articles having different shapes, weights, hardnesses (ease of deformation), uses, contents, raw materials, and the like. Examples include tools, machine parts, drinking water, alcohol, canned goods, retort food, confectionery, detergents, deodorants, toilet cleaning tools, kitchen utensils, electrical products, and medicines.

When the type of article to be accommodated in a container is determined, there is little need to consider the moment balance when the article is accommodated in the container. However, when miscellaneous articles of different types are housed in one container, each article has a different weight, shape, and center of gravity position, and also has various positions and postures within the container. Therefore, it is necessary to accommodate a plurality of different types of articles in a container, taking into consideration the moment balance of the entire container containing the articles. If this is not done, it will be difficult to reliably send containers containing articles to the next process. In particular, when an arm robot or a worker grasps a container containing items and sends it to the next process, if the moment of the entire container is not balanced, the container may tilt more than a predetermined value during transportation. It is also possible.

Therefore, in the present embodiment, in pick-and-place for a plurality of types of articles (workpieces), the placement location (place position) and posture of the article are determined in consideration of the moment balance. As a result, according to the present embodiment, heavy objects are prevented from being disposed unevenly in the container, causing the moment balance of the entire container to be lost, and from falling when a robot or worker holds the container. be able to.

As shown in FIGS. 1 to 3, the robot control device 1 of the present embodiment includes, for example, an article characteristic database 122 that compares a list L1 of articles 7 to be packed with the characteristics of a group of articles, and , a container determining unit 131 that determines the type of container 7 to be packed based on the volume, and candidates for the placement position and orientation of the group of articles based on at least one of the article characteristics of weight, volume, shape, and load capacity. Based on the weight of the article characteristics and the place position of the group of articles, the place position/posture candidate extracting unit 151 determines the vertical direction on each surface when the bottom surface of the container is divided at predetermined boundaries among the candidates. A moment balance determining unit 171 selects items for which the difference in moments is equal to or less than a predetermined value, and a pick-and-place unit picks and places a group of articles in a container so that they have the placement position and orientation selected by the moment balance determining unit 171. A place section 181 is provided.

Thereby, according to the present embodiment, as described above, the position and orientation of the place can be determined in consideration of the moment balance of the entire container, so that work efficiency and reliability can be improved.

A first embodiment will be described using FIGS. 1 to 15. FIG. 1 is an explanatory diagram showing the overall configuration of a picking system 100 including a robot control device 1. As shown in FIG.

The picking system 100 stores a predetermined article 7 carried into the picking station PS1 in an empty container 6 carried into the picking station PS1, and sends the container to the next process.

The picking system 100 includes, for example, a robot control device 1, a command device 2, an arm robot 3, a camera 4, an article shelf 5, an automatic transport vehicle 51, a container carrying-in conveyor device 52, and a container carrying-out conveyor device 53. Equipped with.

The robot control device 1 controls the operation of the arm robot 3. The robot control device 1 controls the arm robot 3 according to the article list L1 commanded by the command device 2 to store a specified group of articles into the empty container 6. The robot control device 1 can detect, for example, the shape of the article 7, the type of the article 7, the attitude of the article 7, and the position of the article 7 using a sensor 4 such as a camera.

The sensor 4 is not limited to a camera, and may be, for example, a weight sensor or a tactile sensor. The weight sensor can detect the weight of each article and the weight of the entire container. With the tactile sensor, it can be determined whether the container 6 or the article 7 is gripped. A plurality of sensors and a plurality of types of sensors may be used. Note that a configuration may also be adopted in which a wireless IC tag is attached to the article 7 and the contents of the wireless ID tag can be read by an IC tag reader. If the sensor 4 is not required, there is no need to provide the sensor.

The arm robot 3 includes, for example, a horizontally rotatable base 30, an arm 31 rotatably provided on the base 30, and a tip portion 32 provided at the tip of the arm 31. A robot other than the arm robot 3 may be used.

The command device 2 transmits information on the articles 7 to be accommodated in the container 6 (information on the group of articles to be accommodated) to the robot control device 1 based on the input article list L1. Furthermore, the command device 2 may also send commands to an article shelf station (not shown) that stores articles on the article shelf 5 and each conveyor device 51, 52, 53.

The automatic transport vehicle 51 transports the article shelf 5 containing the articles 7 at the article shelf station to the picking station PS1. As mentioned above, the articles 7 are various, such as mechanical parts, electrical products, foods, drinking water, and detergents. The automatic transport vehicle 51 constitutes an example of an "article conveyance device" together with the article shelf 5. The automatic transport vehicle 51 may be integrated with the article shelf 5, or may be a separate body. The automatic transport vehicle 51 and the goods shelf 5 may be configured separately, and an empty automatic transport vehicle 51 may be allocated to the goods shelf 5 as needed. That is, the article shelf 5 is coupled to the automatic transport vehicle 51 as needed or at all times, and is movable according to instructions.

The container carry-in conveyor device 52, which is an example of a “container carry-in device”, is a device that carries empty containers 6 to the picking station PS1. A container unloading conveyor device 53, which is an example of a "container unloading device", transports containers 6 containing a plurality of articles 7 and cushioning materials 8 (see FIG. 8) at a picking station PS1 to the next process such as a shipping process. It is a sending device. Below, a plurality of articles housed in the container 6 may be referred to as an article group.

In this embodiment, a picking station PS1 is provided between the container carrying-in conveyor device 52 and the container carrying-out conveyor device 53. That is, the carrying in of the empty container 6 and the carrying out of the container 6 loaded with the articles 7 are not continuous.

Furthermore, the heightwise position of the picking station PS1 and the heightwise position of the container unloading conveyor device 53 differ by a dimension H. Since there is a step in the height direction between the picking station PS1 and the container delivery conveyor device 53, the arm robot 3 needs to stably grip and move the container 6 containing the article 7.

However, even if there is no step in the height direction between the picking station PS1 and the container unloading conveyor device 53, this embodiment works effectively if the picking station PS1 and the container unloading conveyor device 53 are separated. .

FIG. 2 is an explanatory diagram of the configuration of the picking system 100 including the robot control device 1. As shown in FIG. The robot control device 1 is, for example, a computer including an arithmetic device 11, a communication device 12, a storage device 13, and an input/output device 14.

Communication device 12 communicates with command device 2 . The communication device 12 receives a command based on the article list L1 from the communication device 22 of the command device 2.

The storage device 13 stores a predetermined computer program and data for making the computer function as the robot control device 1. The arithmetic device 11 loads and executes a predetermined computer program to move the plurality of articles 7 into the container as instructed by the command device 2 so that the moment of the entire container is balanced and a predetermined condition is satisfied. 6 in a predetermined order.

The input/output device 14 is connected to the arm robot 3 and the sensor 4. Control commands issued from the arithmetic device 11 are transmitted to the arm robot 3 via the input/output device 14. The signal detected by the sensor 4 is passed to the robot control device 1 via the input/output device 14.

The command device 2 is, for example, a computer including an arithmetic device 21, a communication device 22, and a storage device 23.

The storage device 23 stores other computer programs and data for causing the computer to function as a command device. The arithmetic device 21 reads and executes another predetermined computer program, thereby issuing a command based on the description in the article list L1.

The communication device 22 communicates with the communication device 12 of the robot control device 1 and transmits the command issued from the arithmetic device 21 to the robot control device 1 . The communication device 22 receives the notification of completion of loading from the robot control device 1. The notification of completion of loading is a notification indicating that a plurality of ordered items have been stored in the container 6.

FIG. 3 is a functional explanatory diagram of the picking system 100 including the robot control device 1. The robot control device 1 includes, for example, an article list receiving section 111, an article characteristic reading section 121, a container determining section 131, a cushioning material determining section 141, and a position and posture candidate extracting section 151 (hereinafter, candidate extracting section 151). ), an adjacency determination unit 161, a moment balance determination unit 171, a pick and place unit 181, and a completion processing unit 191 as functions. Furthermore, the robot control device 1 also includes, for example, an article characteristic database 122, a container database 132, and a cushioning material database 142. In the figure, the database is abbreviated as DB.

The article list receiving section 111, the article characteristic reading section 121, the container determining section 131, the cushioning material determining section 141, the position and orientation candidate extracting section 151, the adjacency determining section 161, the moment balance determining section 171, and the pick and place section 181. , are connected in this order.

Although details will be described later, the article list receiving section 111 receives a command (a command based on the article list L1) transmitted from the command transmitting section 230 of the command device 2. The article characteristic reading unit 121 refers to the article characteristic database 122 for each article listed in the designated article list L1, and acquires the characteristics of each article.

The container determining unit 131 selects the type of container that can accommodate each specified item by referring to the container database 132 based on the size (including volume) of each specified item. The cushioning material determining unit 141 calculates the volume of the gap when each article is accommodated in the container based on each specified article and the selected container type, and refers to the cushioning material database 142 to calculate the volume of the gap. Determine the cushioning material to fill the volume of.

The candidate extracting unit 151 extracts a candidate pattern of the position and orientation of each article in the container 6 based on the first characteristic of the specified article characteristics of each article. The first characteristics are physical characteristics of the article 7, such as shape (dimensions of each part), weight, volume, load capacity, material, restrictions on arrangement direction, and the like.

The adjacency determination unit 161 determines the adjacency of containers based on the second characteristic set for each designated item among the candidate patterns extracted by the candidate extraction unit 151. For example, if food and toilet cleaning supplies are placed next to each other in a container, buyers may be offended. Therefore, in this embodiment, the types of articles that should not be adjacent to each other in the container are preset based on the sensory characteristics as the second characteristics, that is, based on the likes and dislikes of a general user. The method of setting the likes and dislikes of general users does not matter. For example, the administrator of the picking system 100 may set it manually, or it may be set using machine learning by text analysis of posts on social networking services, etc.

The moment balance determination unit 171 determines, for each candidate pattern, whether the moment of the entire container containing each specified article is balanced within a predetermined range.

The pick-and-place unit 181, which is an example of a “control signal output unit”, controls the arm robot 3 according to the final pattern selected by the moment balance determination unit 171 to select the cushioning material and each specified article. Store in a container.

The completion processing unit 191 determines whether each specified article and cushioning material can be accommodated in the container based on the detection signal from the sensor 4 such as a camera and the signal from the arm robot 3. When the completion processing section 191 determines that each article specified in the article list L1 has been stored in the container, it transmits the determination result to the completion notification receiving section 240 of the command device 2.

The command device 2 includes, for example, an article list database 210, an article list reading section 220, a command transmitting section 230, and a completion notification receiving section 240.

For example, an article list L1 issued from an order management device (not shown) or the like is registered in the article list database 210. The article list reading unit 220 reads out one unprocessed article list L1 from the article list database 210 and issues a command for storing the articles 7 listed in the article list L1 into the container 6. The issued command is transmitted from the command transmitting section 230 to the article list receiving section 111 of the robot control device 1.

On the other hand, when the completion notification receiving unit 240 receives a notification from the robot control device 1 that the storage of each item into the container 6 has been completed, the completion notification receiving unit 240 notifies the item list reading unit 220. The article list reading unit 220 that has received this notification reads the next article list L1 from the database 210.

The configuration of the picking system 100 will be explained in more detail. The robot control device 1, the arm robot 3, and the sensor 4 are each connected. A robot control device 1 and a command device 2 are also connected.

The connection here may be a wired connection or a wireless connection. As described above, the sensor 4 is a tactile sensor that determines whether the article 7 is touching the tip of the arm, a camera that recognizes surrounding images, or the like. A tactile sensor or other sensor other than a camera may be used, or a plurality of sensors may be used. If it is not necessary, the sensor may not be provided. The sensor 4 may be provided in the arm robot 3 or may be provided in a location other than the arm robot 3 (for example, on the ceiling or wall around the arm robot 3).

Note that the pick and place performed by the picking system 100 refers to the work of placing a plurality of items 7 specified in advance into one container 6 such as a cardboard or a container.

The arm robot 3 is a device that moves one or more articles 7 from the article shelf 5 to the container 6 under the control of the robot control device 1 . The arm robot 3 implements pick-and-place using, for example, a suction cup, a hand, or other tip portion 32 suitable for pick-and-place.

The pick-and-place execution command and the information on the list L1 of the plurality of article groups to be packed into the container 6 are transmitted from the command device 2 to the robot control device 1. When the arm robot 3 has finished filling the containers 6 with all the ordered articles, the arm robot 3 places the containers containing the articles on the container delivery conveyor device 53.

The container unloading conveyor device 53 transports the loaded containers 6 to a predetermined transport location. Container carry-in conveyor device 52 conveys empty containers 6 to picking station PS1. The robot control device 1 executes pick-and-place by using, for example, sensing data such as an RGB image, a distance image, barcode data, and the contents of a wireless ID tag from the sensor 4.

The robot control device 1 may be placed outside the arm robot 3 as a separate device from the arm robot 3, or may be provided within the arm robot 3. Furthermore, the robot control device 1 may be provided in a server on a communication network, and may control the arm robot 3 via the communication network. The robot control device 1 and the command device 2 may be integrated or may be configured as separate devices.

Next, the flow of processing will be explained. First, the processing of the command device 2 will be explained. The article list reading unit 220 of the command device 2 acquires the data of the article list L1 from the article list database 210 in accordance with a trigger such as an external command or the elapse of a predetermined time. The article list database 210 is a database that stores a list L1 of articles to be packed in a certain container.

Then, the command transmitter 230 of the command device 2 transmits the data of the list L1 read by the article list reader 220 and the pick-and-place execution command to the article list receiver 111.

The processing of the robot control device 1 will be explained. Upon receiving the command from the command device 2, the article list receiving section 111 of the robot control device 1 causes the article characteristic reading section 121 to read characteristic data of each specified article from the article characteristic database 122.

The article characteristics database 122 stores in advance the characteristics of each article 7 that may be included in the article list L1, such as weight, volume, shape, load capacity, articles that cannot be adjacent to each other, directions that cannot be placed, and materials. It is a database.

The article characteristic reading section 121 transmits the characteristic data read from the article characteristic database 122 to the container determining section 131. The container determination unit 131 determines the total volume of all the articles to be picked and placed based on the volumes of the articles in the characteristic data.

Further, the container determining unit 131 determines the container 6 by referring to the container database 132 based on the calculated total volume. Then, the container determining unit 131 transmits data indicating the determined type of container to the cushioning material determining unit 141.

The cushioning material determining unit 141 reads cushioning material data from the cushioning material database 142. The cushioning material database 142 stores recommended types of cushioning materials for each article that can be picked and placed. The cushioning material determining unit 141 determines the type of cushioning material based on data on the group of articles to be picked and placed. Then, the cushioning material determining unit 141 calculates the difference between the total volume of the article group and the volume of the container as the volume of the necessary cushioning material. The cushioning material determining unit 141 transmits the determined type of cushioning material, its volume, and the type of container to the position and orientation candidate extraction unit 151.

The position and orientation candidate extraction unit 151 selects items in the container based on the characteristics of the group of items to be picked and placed, such as weight, volume, shape, load capacity, and direction in which placement is prohibited, the type of container, and the amount of cushioning material. A candidate pattern group is determined to determine the best position and posture for arranging the cushioning material and the cushioning material. The details will be described later with reference to FIG. The candidate extraction unit 151 transmits a plurality of pick-and-place candidate patterns to the adjacency determination unit 161.

Among the extracted candidate patterns, the adjacency determining unit 161 eliminates candidate patterns in which articles that should not be adjacent are adjacent to each other based on the data of the adjoining items among the characteristics of the article group. , and transmits the remaining candidate patterns to the moment balance determination section 171. Details will be described later.

The moment balance determination unit 171 determines, from among the group of articles, the one with the best moment balance for the entire container as the final pattern (the finally adopted pick-and-place pattern). Details will be described later.

The pick and place unit 181 outputs a control signal (or control command) to the arm robot 3 so that the position and orientation of the article group are determined by the moment balance determination unit 171. The pick-and-place unit 181 uses sensing data from the sensor 4 when controlling the arm robot 3. Then, when the arm robot 3 completes all the pick-and-place operations, the completion processing unit 191 detects the completion of the work and sends it to the completion notification receiving unit 240.

Upon receiving the notification from the completion processing unit 191, the completion notification receiving unit 240 of the command device 2 confirms that the pick and place has been completed. The completion notification receiving unit 240 issues a command to the article list reading unit 220 to move on to the next task.

According to the picking system 100 of this embodiment, when picking and placing multiple types of articles, physical characteristics such as weight, volume, shape, load capacity, and appropriate placement direction, and emotional characteristics such as adjacency, etc. The position and orientation of the place can be determined by comprehensively considering the moment balance of the entire container 6 and the arrangement of the cushioning material. Therefore, the picking system 100 of the present embodiment can perform the packing work of the articles 7 into the containers 6 and the carrying out work from the picking station PS1 with high reliability, and the work efficiency is also improved.

FIG. 4 shows an example of the article characteristics database 122. The article characteristics database 122 is a database that stores characteristics of articles that may be picked and placed. In FIG. 4, characteristics such as weight 1222, volume 1223, shape 1224, load capacity 1225, non-adjacent product 1226, non-arrangement direction 1227, and material 1228 are stored for each product name 1221 in FIG. The shape 1224 may be the actual dimensions and shape of the article 6, or may be a polygon shape that approximates the actual shape of the article 6. In this embodiment, a case will be described in which the shape of a polygon is stored.

Properties other than those illustrated may be added to database 122, and some of the illustrated properties may be omitted from database 122. By using the article characteristic database 122, the robot control device 1 can grasp the characteristics of each article group to be picked and placed.

FIG. 5 is an example of the container database 132. The container database 132 stores, for example, a total volume 1321 of a group of articles to be picked and placed and a container 1322 recommended for storing the group of articles in association with each other. Thereby, the container determining unit 131 can determine a container suitable for storing a group of articles based on the total volume of the group of articles to be picked and placed.

An example of the cushioning material database 142 is shown using FIG. The cushioning material database 142 stores, for example, names 1421 of articles that can be candidates for a group of articles to be picked and placed, and cushioning materials 1422 recommended for these article names 1421 in association with each other. The cushioning material 1422 can include not only the type of cushioning material but also a polygon shape that approximates the shape of the cushioning material. By using the cushioning material database 142, the cushioning material determining unit 141 can determine recommended cushioning materials based on the type of the article group to be picked and placed.

The internal processing of the position and orientation candidate extraction unit 151 will be explained using the flowchart of FIG.

The candidate extracting unit 151 reads the characteristics of a plurality of articles (group of articles) to be stored (S11), and comprehensively extracts candidate patterns indicating positions and orientations where the group of articles and the cushioning material can be stored in the container. (S12).

The candidate extraction unit 151 excludes patterns that include inappropriate postures of the article from among the extracted candidate patterns (S13). Furthermore, the candidate extracting unit 151 eliminates patterns in which a weight exceeding the load capacity acts on the upper part of the article from among the candidate patterns that have passed step S13 (S14). The above procedure determines the position and orientation of the articles when picking and placing the articles into the container. Note that the order of the process of excluding extracted candidate patterns based on conditions is not limited to the example shown in FIG. 7.

Thereafter, the adjacency determination unit 161 excludes candidate patterns that include items for which adjacent placement is prohibited (S15), and the moment balance determination unit 171 determines the moment balance of the entire container containing the group of items, and determines the A well-balanced pattern is selected (S16).

An example of a method of enumerating candidate patterns indicating positions and orientations in which a group of articles and a cushioning material can be stored in a container will be described with reference to FIGS. 8, 9, and 10.

As shown in FIG. 8, an article 7 to be packed in a container 6 is handled by a simplified polygon 71 closest to the article 6. The cushioning material 8 provided in the gap around the article 7 in the container 6 is also treated as a polygon. The polygons of the buffer material 8 are omitted from illustration.

That is, the article characteristic database 122 stores the article 7 and the cushioning material 8 in advance as, for example, rectangular parallelepiped polygons. This saves the effort of calculating the polygon closest to the article, and allows the position and orientation pattern to be quickly calculated.

The cushioning material determining unit 141 calculates the amount of cushioning material 8 necessary to fill the gap created when the group of articles is accommodated in the container, which is calculated from the volume of the group of articles and the volume of the container 6. In step S12 described above, for example, candidate patterns of arrangement that can accommodate the group of articles and the group of cushioning materials without overflowing from the container 6 are listed.

FIG. 9 shows how to enumerate the patterns in which articles can be present in the container 6. The group of articles to be packed in the container 6 is recognized as a rectangular parallelepiped polygon. The shortest side of each side of the polygon 71 is defined as the minimum side length W1 (see FIG. 9(3)).

FIG. 9(1) shows a state in which the polygon 71 of the article 7 is present at the corner of the container 6. FIG. 9(2) is a pattern in which the polygon 71 of the article 7 is moved upward (in the z direction) by the minimum side length W1 from the position in FIG. 9(1) using the minimum side length W1 as a unit of movement. .

9(3), FIG. 9(4), and FIG. 9(5) are patterns in which the polygon 71 of the article 7 is moved by the minimum side length W1 in the x direction from the position shown in FIG. 9(1). FIG. 9(6) is a pattern in which the polygon 71 of the article 7 is moved in the y direction from the position shown in FIG. 9(1).

In this way, all patterns in which the polygon 71 of the article 7 can exist in the container 6 are enumerated. When enumerating patterns, the arrangement positions are shifted using the minimum side length W1 as the minimum unit. Although not shown, the candidate extraction unit 151 also lists patterns in which the posture of the polygon 71 of the article 7 is changed at every predetermined angle for each position that the polygon 71 of the article 7 can take within the container 6. The candidate extraction unit 151 performs the above procedure for all the article groups, and enumerates patterns in which each of the article groups can exist in the container 6.

FIG. 10 is an example of candidate patterns extracted by the candidate extraction unit 151. Even when the same group of articles is housed in the same container, a large number of candidate patterns are calculated based on combinations of the positions and postures of each article.

FIG. 11 shows an example of an article in an inappropriate posture in relation to step S13 in FIG. The upside-down article 7(1) needs to be placed facing upward as shown on the left side of FIG. As shown on the right side of FIG. 11, the article 7(2) that is upside down is in an inappropriate position. Candidate patterns in which even one item in the container is in an inappropriate posture are eliminated.

FIG. 12 relates to step S14 in FIG. 7 and shows an example of a pattern in which a weight exceeding the load capacity is placed on the top of the article. In FIG. 12, the polygon 71(2) of the other article is placed on the polygon 71(1) of one article, and a part of the weight of the other article acts on the one article.

The weight added from the polygon 71 (2) of the other article to the polygon 71 (1) of one article is calculated by the area where the polygon 71 (2) of the other article is in contact with the polygon 71 (2) of one article. do. For example, if 40% of the surface of polygon 71(2) that is in contact with polygon 71(1) is in contact with polygon 71(1), 40% of the weight of the other item is It can be thought that it depends on the goods.

In this way, the weight added to each article in the container 6 is calculated, and candidate patterns in which weight exceeding the load capacity of the article is added are eliminated. Although FIG. 12 illustrates a case where two articles partially overlap, in the case where three or more articles overlap, it is similarly calculated whether the load applied to the article exceeds the load capacity of the article.

FIG. 13 is a diagram showing the adjacency of articles in relation to step S15 in FIG. In FIG. 13(1), when a polygon 71(3) of a certain article contacts a polygon of another article, all the articles whose polygons are in contact are determined to be adjacent. In FIG. 13(2), articles adjacent to a certain article 7(4) are defined by the distance from the center of the article 7(4). That is, if the distance from the center of one article 7(4) to the center of another article is less than or equal to the predetermined adjacent radius r1, it is determined that those articles are adjacent.

In this way, for each candidate pattern, adjacent articles are detected in the container 6, and if the adjacent articles are adjacent additional articles, that candidate pattern is excluded. For example, when it can be assumed that a general user emotionally dislikes having items adjacent to each other, such as food and toilet cleaning supplies, these items are registered in the item characteristics database 122 as non-adjacent items 1226. Alternatively, product value may decrease due to liquid leakage during transportation, such as items with liquid contents (juice, soy sauce, liquid detergent, etc.) and items that are damaged if they get wet with liquid (tissues, notebooks, electronic products, etc.) As in the case where there is a risk, articles with a high risk during transportation are also registered in the article characteristic database 122 as articles 1226 that cannot be adjacent to each other.

By eliminating candidate patterns that include items that cannot be adjacent to each other, it is possible to prevent items that should not be placed adjacent to each other, such as food and detergent that may leak, and to improve transportation. It is possible to prevent the quality of the items inside the container from deteriorating due to carelessness in handling the contents. That is, reliability when transporting a container packed with a plurality of articles can be improved.

An example of a method for determining the moment balance of the entire container will be explained using FIGS. 14 and 15. This description relates to step S16 in FIG. In this embodiment, candidate patterns in which the moment balance of the entire container is worse than a predetermined threshold value are excluded.

FIG. 14 is an explanatory diagram showing how the moment of the article existing in the container 6 is determined. FIG. 14(1) shows a perspective view of the container 6. The container bottom surface 61 in FIG. 14(2) is the container 6 shown in FIG. 14(1) viewed from the xy plane in the figure.

The center point P1 is the center point of the bottom surface 61 of the container. The mass point P2 is a point on which the weight of a certain article of the article group is applied. It is assumed that a mass point P2 is located at the center of the article. The center point-to-mass point distance L2 is the distance between the center point P1 and the mass point P2. The moment due to the article is obtained by integrating the weight applied to the mass point P2 and the distance L2 between the center point and the mass point. By the above method, all moments exerted on the container 6 by the group of articles are determined.

FIG. 15 shows a procedure for determining the moment balance of the entire container 6 based on the moments of each of the article groups determined in the procedure of FIG.

The container side surface 62 is the container 6 viewed from the xz plane in the figure. Centerline 64 is the center line of container side 62. The container side surface 63 is the container 6 viewed from the yz plane in the figure. The center line 65 is the center line of the container side surface 63. When the container side surface 62 is viewed from the direction shown in the figure, the moments applied in the left and right portions divided by the center line 64 are respectively summed. If the difference between the respective summed moments is larger than a predetermined threshold, which is stored in a database, for example, the candidate pattern is excluded. In a similar manner, it may be determined whether the moment balance exceeds a predetermined threshold value based on the moment balance between the container side surface 63 and the center line 65.

In the case of the picking system 100 to which this embodiment is not applied, heavy items may be disposed unevenly within the container, causing an imbalance in the moment of the entire container. Therefore, if a worker or robot grasps a container with poor moment balance, the possibility of the container falling increases. On the other hand, the robot control device 1 according to the present embodiment determines the position and orientation of the place by taking into account the balance of moments, so it is possible to realize highly reliable pick-and-place.

A second example will be described using FIG. 16. In each of the following embodiments including this embodiment, differences from the first embodiment will be mainly described. This embodiment shows an example in which it is possible to deal with articles that are not registered in the article characteristic database 122.

A picking system 100A of this embodiment adds a characteristic determining section 123 to the picking system 100 of the first embodiment.

The functional configuration of the picking system 100A will be explained using FIG. 16. Between the article list receiving section 111 and the container determining section 131, a characteristic determining section 123 is provided in parallel with the article characteristic reading section 121. When an unregistered article that does not exist in the article characteristic database 220 is included in the article group described in the article list, the characteristic determining unit 123 determines the characteristic of the unregistered article.

A method for determining the characteristics of unregistered articles on the spot will be explained. The characteristic determination unit 123 can, for example, cause the arm robot 3 to hold an unregistered article and estimate the weight, which is one of the characteristics of the unregistered article, from the force applied to the arm robot 3 in the vertical direction. Alternatively, the characteristic determining unit 123 acquires RGB image data and distance image data of the unregistered article using the sensor 4, which is a camera, and performs image processing on the image data to recognize the shape and shape of the unregistered article. Volume can also be detected. Furthermore, the characteristic determination unit 123 can also detect the type of article, material, and direction in which the article cannot be placed by reading a character string or barcode printed or pasted on the unregistered article. For example, a product label formed as a print medium or an electronic recording medium may include information such as the product type, material, orientation, etc.

If there is an item that is difficult to determine for an unregistered article among the items managed in the article characteristic database 122 described in FIG. 4, the characteristic determining section 123 may set an initial value prepared in advance. .

This embodiment configured in this manner also provides the same effects as the first embodiment. Furthermore, in this embodiment, even if the article list includes an article that is not registered in the article characteristic database 122, the characteristic determining section 123 can determine the characteristics of the unregistered article without interrupting the work. , the picking system can be operated.

A third example will be described using FIG. 17. In this embodiment, the arm robot 3 does not place the cushioning material into the container 6, but another robot or operator puts the cushioning material into the container.

FIG. 17 shows the functional configuration of a picking system 100B according to this embodiment. Comparing the picking system 100B with the picking system 100 described in FIG. 9 have been newly added. In the robot control device 1 of the picking system 100B, the container determining section 131 and the candidate extracting section 151 are connected to exchange data.

The cushioning material input section 9 receives gap capacity data from the moment balance determination section 171. The gap volume data is data indicating the volume of a gap that occurs in the container 6 when a group of articles is stored in the container. The cushioning material input unit 9 inputs the amount of cushioning material required to fill the gap into the container 6 based on the gap capacity data. The cushioning material input unit 9 is, for example, a device or a robot that inputs granular cushioning material or bulk cushioning material into the container 6. The buffer material input unit 9 may be operated by a worker. The amount of cushioning material may be displayed on a display provided at the picking station, and the operator may manually add the required amount of cushioning material into the container.

This embodiment configured in this manner also provides the same effects as the first embodiment. Furthermore, in this embodiment, since the cushioning material is loaded into the container by the cushioning material input unit 9 that is separate from the arm robot 3, the work efficiency is improved compared to the case where both the article and the cushioning material are placed in the container by the arm robot 3. can be improved. For example, in this embodiment, when using fine cushioning material such as bulk cushioning material, work efficiency is improved.

A fourth example will be described using FIGS. 18 and 19. In this embodiment, the order in which a group of articles is placed in a container (place order) is determined using so-called artificial intelligence technology.

In the picking system 100C of this embodiment, a play order determining section 150 is added to the picking system 100 described in the first embodiment.

FIG. 18 shows a functional block diagram of the picking system 100C. This picking system 100C has a place order determining unit 150 added to the picking system 100 described in FIG. 3.

The place order determining unit 150 receives and calculates the data on the article group, the data on the characteristics of each article, and the data on the cushioning material from the cushioning material determining unit 141, so as to determine which of the article group and the cushioning material. It is determined in which order to put them into the container 6.

The place order determining unit 150 can determine the order of places using, for example, artificial intelligence technology. The place order determining unit 150 determines the order of places that is most likely to be appropriate, for example, by using machine learning such as reinforcement learning. An example of determining the order of places by reinforcement learning will be described later with reference to FIG. When the place order is determined by the place order determining unit 150, the candidate extracting unit 151 lists the candidate patterns based on the place order. By determining the order of places, the number of candidate patterns to be listed is reduced, and effects such as preventing explosion of combinations, reducing the load on the computer, and improving calculation speed can be obtained.

A method in which the place order determination unit 150 determines the order of places by reinforcement learning will be described using FIG. 19. FIG. 19 is a state transition diagram in reinforcement learning.

A case will be described in which three groups of articles, article A, article B, and article C, are stored in the container 6. At this time, as shown in the state transition diagram, the patterns of the first item, second item, and third item can be enumerated, and the state transition can be expressed. By using reinforcement learning to learn the rewards that can be obtained by choosing which item to put next in each state, it becomes possible to determine the order of placement that is most likely to be appropriate. A reward is something that you receive more when you make a choice that ultimately results in an order that is close to the optimal. Although we used an example that uses reinforcement learning, other types of machine learning such as supervised learning and unsupervised learning may also be used.

This embodiment configured in this manner also has the same effects as the first embodiment for hemorrhoids. Furthermore, in this embodiment, the optimal order of places is determined by machine learning or the like, and candidate patterns are listed in accordance with this order. In this embodiment, since the order of places is determined, the number of candidate patterns to be listed can be reduced, and the number of combinations can be prevented from increasing explosively. Therefore, according to this embodiment, it is possible to obtain the effect of reducing the load on the computer or improving the calculation speed.

A fifth example will be described using FIG. 20. In this embodiment, the past article list closest to the article list to be processed this time is detected, and the pick-and-place order used in the past article list is used.

FIG. 20 is a flowchart of the play order determination process. The play order determining unit 150 calculates the feature amount of the article list to be processed (S21), and searches for previously processed article lists based on the calculated feature amount (S22). Although not shown, in this example, information for specifying the article list, the type of container when the article group described in the article list is housed in the container, the type and amount of cushioning material, and the order in which the article group is accommodated. is assumed to be saved.

The play order determining unit 150 determines whether a past article list similar to the feature amount of the article list to be processed has been detected (S23). When a past article list with the most similar feature values to the article list to be processed is detected (S23: YES), the play order determining unit 150 reads out the past article list (S24). The play order determination unit 150 determines the play order of the article list to be processed based on the play order when the article groups described in the past article list were stored in containers (S25).

On the other hand, if no past item list with similar feature values to the item list to be processed is detected (S23: NO), the place order determining unit 150 determines the place order by the machine learning described in FIG. 19. (S26). The characteristic amounts of the article lists are similar when, for example, the difference in the characteristic amounts of the article lists to be compared is within a predetermined value. The characteristic amounts of the article lists are not similar when the difference in the characteristic amounts of the article lists to be compared exceeds a predetermined value.

This embodiment configured in this way has the same effects as the fourth embodiment. Furthermore, in this embodiment, the article list closest to the article list to be processed is detected from among the past article lists, and the places related to the article list to be processed are determined based on the order of places regarding the most similar article list. Decide on the order. Therefore, in this embodiment, the calculation time can be shortened by using the past processing contents.

Note that the present invention is not limited to the above-described embodiments, and includes various modifications. The above embodiments have been described in detail to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment. Moreover, the configuration of another embodiment can be added to the configuration of one embodiment. Further, other configurations can be added, deleted, or replaced with a part of the configuration of each embodiment.

Each component of the present invention can be selected arbitrarily, and inventions having selected configurations are also included in the present invention. Furthermore, the configurations described in the claims can be combined in combinations other than those specified in the claims.

1: robot control device, 2: command device, 3: arm robot, 4: sensor, 5: article shelf, 6: container, 7: article, 8: cushioning material, 9: cushioning material input unit, 100: picking system, 111: Article list receiving unit, 121: Article characteristic reading unit, 122: Article characteristic database, 123: Characteristic determining unit, 131: Container determining unit, 132: Container database, 141: Cushioning material determining unit, 142: Cushioning material database, 150: Place order determination unit, 151: Position and orientation candidate extraction unit, 161: Adjacency determination unit, 171: Moment balance determination unit, 181: Pick and place unit, 191: Completion processing unit

Claims (7)

A robot control device that controls a robot that stores articles in a container,
a container determining unit that determines a container based on an article list storing information on a plurality of articles to be accommodated in the container;
a position and posture candidate extraction unit that extracts candidate patterns of positions and postures of the plurality of articles in the container based on first characteristics of the plurality of articles;
a moment balance determination unit that determines, among the candidate patterns extracted by the position and orientation candidate extraction unit, a candidate pattern in which the balance of moments acting on the container by the plurality of articles falls within a predetermined range, and selects the candidate pattern as a final pattern; ,
determining a predetermined order for accommodating the plurality of articles in the container according to the final pattern selected by the moment balance determining section, and generating a control signal for accommodating the plurality of articles in the container according to the predetermined order. a control signal output unit that outputs a control signal to the robot;
an adjacency determination unit that determines whether or not an article is adjacent when stored in the container, based on a second characteristic set to the plurality of articles stored in the container;
The first property is a physical property of the article,
The second characteristic is preset as a type of article that should not be adjacent to each other in the container, based on the likes and dislikes of a general user.
Robot control device. The adjacency determination unit excludes, from among the candidate patterns, candidate patterns in which articles whose adjacency is prohibited are adjacent.
The robot control device according to claim 1 . The moment balance determination unit selects, as the final pattern, a candidate pattern in which a difference in moment acting in the vertical direction is equal to or less than a predetermined value between regions formed by dividing the bottom surface of the container.
The robot control device according to claim 1. further comprising a cushioning material determination unit that determines a cushioning material to be provided in a gap between the containers in which the plurality of articles are housed;
The robot control device according to claim 1. The control signal output unit further outputs a control signal for causing the determined cushioning material to be provided in a gap between the plurality of articles.
The robot control device according to claim 4 . further comprising a characteristic determination unit that determines a first characteristic of at least one article among the plurality of articles based on information from the sensor;
The robot control device according to claim 1. The plurality of articles include articles of different types,
The robot control device according to claim 1.

Images