JP7044366B2 - Distributor, filling system and learning data generator - Google Patents

Description

The present invention relates to a distribution device, a filling system and a learning data generation device.

As a serving device for distributing cooked food and serving it in a container such as a lunch box, Patent Document 1 describes food in a hopper from a gate provided on the lower end side of the hopper. Disclosed is a structure in which food is dropped and served on a container passing under the gate.

However, in the filling device disclosed in Patent Document 1, for example, when highly viscous food such as spaghetti or cooked rice is put into the hopper, the food gradually adheres to the inner surface of the gate through which the food passes. There was a problem that the gate had to be cleaned frequently because the amount could not be measured accurately.

On the other hand, as a solution to the problem of the filling device using the hopper, Patent Document 2 discloses a filling device that grips and distributes food by a plurality of fingers.

However, in the filling device disclosed in Patent Document 2, although lumpy foods such as croquettes and hamburgers can be individually separated from the piled state, noodle-shaped, granular, thin plate-shaped, macaroni-shaped foods and the like can be separated. There is a problem that it is not possible to separate the noodles in an accurate amount when grasping and separating them in a predetermined amount.

JP-A-2004-196382 Japanese Patent Application Laid-Open No. 9-65383

Therefore, the present invention provides a distribution device capable of separating foods such as noodles, granules, thin plates, macaroni, etc. that need to be weighed in an accurate amount to some extent and easy to maintain. Is the main issue.

That is, the distribution device according to the present invention is a distribution device that grabs and distributes a part of foods such as noodles, granules, thin plates, macaroni, etc. from the piled state, and is a plurality of distribution devices from the piled foods. The multi-finger hand that grabs a part of food by opening and closing the fingers, the measuring instrument that measures the part of the food grabbed by the plurality of fingers, and the multi-finger hand are controlled under arbitrary control conditions. A learning work control unit that controls a series of weighing operations to grab a part of food from a pile of food and measure the mass while changing the control conditions, and a learning work control unit for each weighing operation. , Training data that generates training data in which the control conditions are associated with the mass of some foods grasped by the control conditions, and stores a plurality of training data generated by each weighing operation in the learning work as a training data set. It is characterized by including a storage unit and a distribution work control unit that controls to perform a distribution operation of grasping and distributing a part of food from the piled food based on the learning data set. It is something to do.

In such a case, since a multi-finger hand that opens and closes multiple fingers to grab the food is used, the food falls away from each finger when the multiple fingers are opened, and the food is dropped against each finger. It is relatively unlikely that food will adhere and remain, and maintenance will be easy. In addition, it is difficult to grab a specified amount of food such as noodles, granules, thin plates, macaroni, etc. with a multi-finger hand, but many foods to be distributed in advance are used for the distribution work. Based on the learning data generated by grasping and weighing while changing the control conditions with a finger hand, it becomes possible to execute distribution work, even for foods such as noodles, granules, thin plates, macaroni, etc. , You will be able to grasp the exact amount to some extent.

The control conditions are the opening width of each finger, the closing width of each finger, the closing speed of each finger, and each of the above, when a part of the food is grabbed from the piled food by the multi-finger hand. It is a gripping condition selected from the gripping force of the fingers and the insertion depth of each finger into the piled food.

Further, the distribution device further includes a height detecting unit for detecting the height of each coordinate with respect to the reference plane of the piled food, and the insertion depth of each finger into the piled food is determined. It may be set based on the height detected by the height detection unit.

Further, before the weighing operation of the learning work control unit, the opening width X of each finger and the insertion depth Y of each finger included in the control conditions of the weighing operation, and the coordinates detected by the height detecting unit. The filling shape data showing the filling shape of the food piled in the filling range of the diameter X and the depth Y centered on the highest coordinate which is the coordinate of the highest filling height of the piled food with reference to the height of the pile. The learning work control unit controls the multi-fingered hand under the control conditions, grabs the food from the highest coordinates of the filled food, and measures the mass. The operation may be executed, and the learning data storage unit may generate learning data by associating the control conditions with the mass and the peak shape data. In this case, the swelling shape data is a value represented by the ratio of the swelling height of the highest coordinates to the volume or cross-sectional area of the food squeezed in the swelling range.

In such a case, since the filling shape data is added to the learning data, it becomes possible to grasp foods such as noodle-shaped, granular, thin plate-shaped, and macaroni-shaped foods in a more accurate amount.

Further, in any of the distribution devices, the learning data storage unit selects learning data including a mass within a predetermined target mass range from the learning data set, and selects a plurality of selected learning data. It is stored as a target learning data set, and the distribution work control unit controls to perform a distribution work of grasping and distributing a part of food from the piled food based on the target learning data set. It may be a thing.

If this is the case, the number of learning data referenced in the distribution work control unit can be reduced in advance, thereby improving the processing speed of the distribution work in the distribution work control unit.

Further, in the distribution device that generates the target learning data set, the learning work control unit has a non-variable grip condition that does not change and a variable grip that changes from among a plurality of grip conditions included in the control condition of the multi-finger hand. A condition is selected, the unit learning work of repeating the series of weighing operations a specified number of times while changing the variable gripping condition is performed, and the non-variable gripping condition is added each time the unit learning work is repeated. The learning data storage unit generates the target learning data set for each unit learning work, and the target learning data set generated in any unit learning work in the learning work control unit. With reference to the fluctuation gripping conditions of each included learning data, an upper / lower limit value specifying unit for specifying the upper limit value and the lower limit value of the fluctuation gripping condition is further provided, and the upper limit of the fluctuation gripping condition specified by the upper / lower limit value specifying unit is further provided. The value and the lower limit value may be configured to be set as the upper limit value and the lower limit value of the fluctuation grasping condition selected in the next unit learning work in the learning work control unit.

With such a thing, it is possible to reduce the number of weighing operations required to grasp the relationship between the control condition of the multi-finger hand and the mass of some foods that can be grasped by the control condition. The time required for learning work can be shortened.

Further, when any one of the distribution work control units grabs and distributes a food having a predetermined set mass from the piled food, the set mass is obtained from each learning data stored in the learning data storage unit. The control condition of the learning data including the mass closest to the above is set as the control condition for distribution, and the control condition for distribution controls to grab a part of the food from the filled food and distribute it. May be good. Further, when the distribution work control unit grabs and distributes a food having a predetermined set mass from the piled food to be distributed, the above-mentioned learning data among the learning data stored in the learning data storage unit. With reference to each training data including the mass included in the predetermined error range centering on the set mass, the control conditions included in each training data are classified into multiple classes between the maximum value and the minimum value, and among all the classes. From each training data having the control condition in the class in which the most values are classified, the training data including the mass closest to the set mass is selected, and the control condition of the training data is set as the distribution control condition. Depending on the distribution control conditions, a part of the food is grabbed from the filled food to be distributed and controlled to be distributed, and the distribution work control unit controls the filled food to be distributed. When a food having a predetermined set mass is grabbed and distributed from the above, each learning including the mass included in a predetermined error range centering on the set mass among the learning data stored in the learning data storage unit. With reference to the data, the control conditions and the shape data included in each training data are classified into multiple classes between the maximum value and the minimum value, and the class with the largest number of values among all the classes is classified. Training data containing the mass closest to the set mass is selected from each training data having control conditions or ridge shape data, the control condition of the training data is set as the distribution control condition, and the distribution control condition determines the above. It may be controlled so as to grab a part of the food from the serving food to be distributed and distribute it.

If this is the case, for example, when it is necessary to serve a specified amount of food for each container to which it is distributed, such as a convenience store lunch, the amount specified to some extent accurately for each container. (Set mass) of food can be served.

Further, when the mass measured by the distribution work control unit by grasping a part of the food from the piled food according to the distribution control condition is outside the set mass range including the set mass, the distribution work control unit again. , The distribution control conditions may be used to control so as to grab a part of the food from the heaped food.

In such a case, if the mass of the food grabbed by the multi-finger hand deviates significantly from the set mass, or if the food grabbed by the multi-finger hand falls before being distributed to the distribution destination, It is possible to prevent the distribution work from being continued as it is.

Further, with a plurality of fingers configured so that the multi-finger hand of any of the above-mentioned distribution devices can be opened and closed inward and outward with one end side connected in a ring shape to the support block as a fulcrum. An outer elastic body connected to each finger and pulling each finger outward, an inner elastic body connected to each finger and pulling each finger inward, and the outer elastic body or It may be provided with a tension fluctuation mechanism that fluctuates the tension of at least one of the inner elastic bodies.

In such a case, since each finger is pulled outward and inward by an elastic body to adjust the gripping force, even a fragile food such as potato chips should be subjected to an unreasonable force. Can be grabbed without. The mode in which the elastic body is connected to the finger is not only the mode in which the elastic body is directly connected to the finger, but also the mode in which the elastic body is indirectly connected so that the tension of the elastic body is transmitted to the finger. Aspects are also included.

Further, in the serving system according to the present invention, the serving food is located in the vicinity of any of the distribution devices, the first supply means for supplying the serving food in the vicinity of the distribution device, and the serving food in the vicinity of the distribution device. It is provided with a second supply means for supplying a container for serving a part of the food grabbed from the above.

Further, the learning data generation device according to the present invention includes a multi-finger hand that grabs a part of food such as noodle-shaped, granular, thin plate-shaped, and macaroni-shaped food by opening and closing a plurality of fingers, and the plurality of foods. A series of weighings in which a measuring instrument that weighs a part of the food grabbed by a finger and the multi-fingered hand are controlled under arbitrary control conditions, and the part of the food is grabbed from the piled food and the mass is weighed. The learning work control unit that controls the operation to perform the learning work that repeats while changing the control condition, and the mass of a part of the food that grasps the control condition under the control condition for each weighing operation. It is characterized by including a learning data storage unit that generates associated learning data and stores the generated learning data as a learning data set.

With this configuration, it is possible to separate foods that need to be weighed, such as noodles, granules, thin plates, and macaroni, in an accurate amount to some extent, and maintenance is easy.

It is a schematic diagram which shows the filling system which concerns on Embodiment 1. It is explanatory drawing which shows the heap height of a heaped food. It is a flowchart explaining the learning work of the filling system which concerns on Embodiment 1. It is a flowchart explaining the transport operation of the filling system which concerns on Embodiment 1. FIG. It is a schematic diagram which shows the appearance of the multi-finger hand of the filling system which concerns on Embodiment 2. FIG. It is a schematic diagram which shows the internal structure of the multi-finger hand of the filling system which concerns on Embodiment 2. It is a schematic diagram which shows the multi-finger hand of the filling system which concerns on other embodiment. It is a schematic diagram which shows the multi-finger hand of the filling system which concerns on other embodiment. It is a schematic diagram explaining the filling range which concerns on other embodiment.

Hereinafter, the distribution device according to the present invention and the filling system using the transfer device will be described with reference to the drawings.

The distribution device according to the present invention is used to grab and distribute a predetermined amount of foods that need to be weighed, such as noodles, granules, thin plates, and macaroni. The distribution device according to the present invention can also be used as a serving system for grasping a predetermined amount of the served food and serving it in a container. In addition, noodle-shaped, granular, thin plate-shaped, macaroni-shaped foods, etc. are not one lump of foodstuffs, but are grasped as lumps of many foodstuffs. , Spaghetti, udon, buckwheat, etc., granular foods include rice, etc., and thin plate-shaped foods include potato chips, etc.

<Embodiment> As shown in FIG. 1, the serving system 100 according to the present embodiment has a distribution device DD for distributing the served food and a tray T containing the served food F in the vicinity of the distribution device DD. It is provided with a tray supply device SD1 for supplying and a container supply device SD2 for supplying a container B for serving a part of the served food F in the vicinity of the distribution device DD. The distribution device DD, the tray supply device SD1 and the container supply device SD2 are all connected to an information processing device (not shown) and are controlled by the information processing device. The tray supply device SD1 according to the present embodiment corresponds to the first supply means according to the claim, and the container supply device SD2 according to the present embodiment corresponds to the second supply means according to the claim.

The distribution device DD includes a hand support mechanism 10, a measuring instrument 20 connected to the tip of the hand support mechanism 10, a multi-fingered hand 30 connected to the measuring instrument 20, and a tray supply device SD1 in the vicinity of the distribution device DD. It is provided with four cameras C1 for photographing the tray T conveyed to the tray T from above, and one camera C2 for photographing the container B conveyed in the vicinity of the distribution device DD by the container supply device SD2 from above. ..

The hand support mechanism 10 is a 6-axis robot arm. Specifically, a support 12 that is rotatably attached to the base 11 and a three-joint arm 13 (first arm 13a, second arm 13b, third arm 13c) that is attached to the support 12 are not shown. It is equipped with an actuator (motor) for driving the 6 axes and various sensors for sensing the turning angle and joint angle of the 6 axes (not shown). The second arm 13b has a structure in which two parts are rotatably connected to each other, and a multi-finger hand 30 is attached to the third arm 13c located at the most tip of the three-joint arm 13 via a measuring instrument 20. It is connected so that it can turn. As a result, the hand support mechanism 10 can be moved by a total of 6 axes of 3 turning points and 3 joints.

The multi-finger hand 30 includes a support block 31 connected to the measuring instrument 20, three fingers 32 connected to the support block 31, an actuator for driving three fingers 32 (not shown), and three fingers (not shown). It is equipped with various sensors that sense the opening width, closing width, closing speed, gripping force, etc. of the finger. The three fingers 32 are configured to be openable and closable with respect to the support block 31 with one end side connected in a ring shape as a fulcrum.

The measuring instrument 20 is configured so that the mass of only the food f grasped by the multi-fingered hand 30 can be weighed in a state where the food f is grasped by the multi-fingered hand 30.

The camera C1 is for taking an image for detecting the height of the food F placed on the tray T supplied by the tray supply device SD1. The height is the height h from the reference surface reflected in the image to the outer surface of the entire food T, and for example, as shown in FIG. 2, the tray T reflected in the image. The height h from the bottom surface to the outer surface of the entire food T may be adopted. The image may be a still image or a moving image.

The camera C2 is for detecting the filling position in the container B supplied by the container supply device SD2.

The tray supply device SD1 and the container supply device SD2 are so-called belt conveyors. The tray supply device SD1 extends from, for example, a kitchen on the upstream side to an empty tray collection point on the downstream side. In this case, the tray supply device SD1 conveys the food F cooked in the kitchen to the vicinity of the distribution device DD in a state of being piled up on the tray T, and all of the food F piled up on the tray T is loaded by the distribution device DD. After being distributed to the plurality of containers B, the empty tray T will be transported to the tray collection point. Further, the container supply device SD2 extends from, for example, a container supply station on the upstream side to a container collection station on the downstream side. In this case, the container supply device SD2 conveys the container B supplied at the container supply station to the distribution device DD, and a part of the food f of the food F placed on the tray T is placed on the container B by the distribution device DD. After that, the container B on which the food f is served is transported to the container collection point.

The information processing device is a so-called computer having a CPU, an internal memory, an I / O buffer circuit, an AD converter, and the like. Then, by operating according to the control program stored in the predetermined area of the internal memory, the CPU and peripheral devices cooperate to operate, and the height detection unit, the learning work control unit, the learning data storage unit, the upper and lower limit value specifying units, It functions as a distribution work control unit, a tray supply control unit, and a container supply control unit.

The height detection unit is the height at each coordinate on the image plane of the food F placed on the tray T from the images obtained by taking the food F placed on the tray T with four cameras C1. It detects (altitude).

The learning work control unit controls and grabs a part of the food F piled up on the tray T supplied in the vicinity of the distribution device DD by the tray supply device SD1 by controlling the multi-finger hand 30 under arbitrary control conditions. A series of weighing operations for weighing a part of the food f that has been grasped is controlled so as to carry out a unit learning operation that repeats a predetermined number of times while changing the control conditions. More specifically, the learning work control unit selects a non-variable grip condition that does not change and a variable grip condition that changes from a plurality of grip conditions included in the control conditions of the multi-finger hand, and the above-mentioned. It is controlled so that the unit learning work of repeating a series of weighing operations a specified number of times while fluctuating the grasping conditions is performed a plurality of times. The learning work control unit adds a non-variable grasping condition every time the unit learning work is repeated, and performs a weighing operation. That is, the learning work control unit increases the non-variable grasping condition every time the unit learning work is repeated, and carries out the weighing operation.

The grasping conditions include the opening width of each finger 32, the closing width of each finger 32, the closing speed of each finger 32, and each finger 32 when a part of the food f is grasped from the food F piled up on the tray T. The gripping force of each finger 32 and the insertion depth of each finger 32 into the piled food F can be mentioned. The insertion depth of each finger 32 into the piled food F can be specified by the length of each finger inserted from the pile height detected by the height detection unit as a reference.

The learning data storage unit generates learning data in which the control conditions in the weighing operation and the mass of a part of the food grasped by the control conditions are associated with each weighing operation in the learning work control unit, and unit learning is performed. For each work, the generated plurality of learning data are stored as a learning data set. Further, the learning data storage unit selects learning data including a mass within a predetermined target mass range from the learning data set for each unit learning work, and selects a plurality of selected learning data as the target learning data set. Remember as.

The upper / lower limit value specifying unit specifies the upper limit value and the lower limit value from the fluctuation grasping conditions in each learning data included in the target learning data set generated in any unit learning work in the learning work control unit. Then, the upper limit value and the lower limit value of the specified variable gripping condition are set as the upper limit value and the lower limit value of the variable gripping condition selected in the next unit learning work in the learning work control unit. This control is performed when the same gripping condition is selected as the variable gripping condition in the previous unit learning work and the next unit learning work.

When the distribution work control unit grabs the food f having a predetermined set mass from the food F piled up in the tray T and distributes it to the container B, the target learning data set stored in the learning data storage unit is used. A learning data containing the mass closest to the set mass is selected from a plurality of included learning data, and based on the control conditions included in the selected learning data, some foods f are placed in the tray T from the foods F. It is intended to carry out the distribution work of grasping and distributing.

The tray supply control unit conveys the food F loaded on the tray T from the upstream side to the vicinity of the distribution device DD, and after the distribution of all the food F loaded on the tray T is completed, the tray T is emptied. Is executed to carry the control to the downstream side. Therefore, the tray supply control unit intermittently conveys the tray T.

The container supply control unit conveys the container B from the upstream side to the vicinity of the distribution device DD, and after the distribution of the food F loaded on the tray T to the container B is completed, the food f is loaded. The control for transporting the container B to the downstream side is executed.

Next, the operation when the learning work of the filling system 100 according to the present embodiment is executed will be described with reference to FIG.

When the execution signal of the learning work is input to the information processing device, the tray supply control unit executes an operation of transporting the food F loaded on the tray T by the tray supply device SD1 to the vicinity of the distribution device DD. After that, the learning work is executed.

Then, in the information processing device, the tray T is based on an image obtained by photographing the food F piled up on the tray T conveyed to the vicinity of the distribution device SD by the four cameras C1 in the height detection unit. The height at each coordinate on the image plane of the food F piled up in the above (hereinafter, also referred to as “learning height information”) is detected (step S1).

Next, the information processing apparatus selects an arbitrary height from the learning height information in the learning work control unit, and specifies the coordinates corresponding to the selected height (step S2). .. Further, the information processing apparatus selects one non-variable grip condition that does not change and a plurality of variable grip conditions that change from the plurality of grip conditions included in the control conditions in the learning work control unit (step 3). .. Then, the information processing apparatus sets control conditions, that is, each gripping condition in the learning work control unit (step 4).

Next, the information processing apparatus controls the multi-finger hand 30 according to the control conditions set in the learning work control unit, and picks up a part of the food from the specified coordinate portion in the food F placed on the tray T. After weighing with the measuring instrument 20, a series of weighing operations for returning a part of the food f to the tray T is performed (step S5). After that, the information processing device generates and stores the learning data in which the control condition is associated with the mass of the food grasped by the control condition in the learning data storage unit (step S6).

Then, the information processing apparatus repeats a series of weighing operations of steps S1 to S6 while changing each gripping condition included in the variable gripping condition, and determines whether or not the weighing operation has been performed a predetermined number of times (step S7). .. Then, if the specified number of times has not been carried out, the series of weighing operations of steps S1 to S6 are carried out again.

On the other hand, when the execution is performed a specified number of times, the unit learning work is completed (step S8). Then, the information processing apparatus stores a plurality of learning data generated in each measurement operation of the unit learning work as a learning data set in the learning data storage unit (step S9). Further, the information processing apparatus selects learning data including a mass within a predetermined target mass range from the learning data set in the learning data storage unit, and uses the selected learning data as the target learning data set. Store (step S10).

Next, the information treatment device determines whether or not the unit learning work has been performed a predetermined number of times (step 12). Then, when the learning work is carried out a specified number of times, the learning work is completed.

On the other hand, when the execution is not performed a specified number of times, the information processing device refers to a plurality of learning data included in the target learning data set stored in the learning data storage unit in the upper / lower limit value specifying unit, and each learning data. The largest upper limit value and the smallest lower limit value are specified from the fluctuation grasping conditions included in (step S12).

Next, the information processing apparatus selects the variable gripping condition and the non-variable gripping condition in the next unit learning work in the learning work control unit (step S14). In this case, at least one of the variable gripping conditions in the previous unit learning work is selected, and the selected variable gripping condition is added to the non-variable gripping condition. Therefore, the number of variable grip conditions decreases and the number of non-variable grip conditions increases.

Next, when the information processing device selects the upper and lower limit values of the variable gripping conditions specified in the upper and lower limit value specifying units as the variable gripping conditions in the next unit learning work, the variable gripping conditions are selected. It is set as the upper limit value and the lower limit value of the numerical range of (step S15).

Then, the information processing apparatus again carries out a series of unit learning operations of steps S1 to S10, and carries out an operation of repeating these operations.

Next, the operation when the distribution work of the filling system 100 according to the present embodiment is executed will be described with reference to FIG.

When the learning work is completed, the distribution work is started. Then, first, the information processing apparatus is loaded on the tray T based on the images obtained by photographing the food F loaded on the tray T to be distributed by the four cameras C1 in the height detection unit. The height at each coordinate on the image plane of the food F (hereinafter, also referred to as “distribution height information”) is detected (step S15).

Next, the information processing apparatus selects learning data including the mass closest to the preset set mass from the target learning data set generated in the final unit learning work in the distribution work control unit (step S16).

Next, the information processing apparatus determines in the distribution work control unit whether or not the insertion depth of each finger included in the selected learning data is longer than any of the heights included in the distribution height information. (Step S17).

If it is determined to be long, the information processing apparatus subsequently determines whether or not the determination has been continuously performed a predetermined number of times (step S18). Then, when it is determined that the tray T has continued for a predetermined number of times, it is determined that the tray T has become empty, and the distribution work is terminated. On the other hand, if it is determined that the food F has not been continuously used a specified number of times, the distribution device DD (multi-finger hand 30) is controlled so that the bulk (height) of the entire food F placed on the tray T increases. , The food F placed on the tray T is refilled, and the process is re-executed from step S15 (step S19). As the refilling operation, for example, an operation of lifting a part of the food F piled up with the multi-finger hand 30 and returning it again, or a movement of stirring the piled food F can be considered.

On the other hand, if it is determined that the information processing device is short, the information processing apparatus controls the multi-finger hand 30 according to the control conditions included in the selected learning data in the distribution work control unit, and one of the foods F loaded on the tray T is selected. The food f in the portion is picked up and the mass is measured by the measuring instrument 20 (step 20).

Next, the information processing apparatus determines in the distribution work control unit whether or not the mass of the weighed food f is within a predetermined set mass range (step S21). In addition, this set mass range is an error range that can be tolerated from the set mass. For example, when serving food f in a lunch box, if it is sufficient to serve more than the set mass, the set mass range is set. It becomes more than the mass.

Then, when the information processing apparatus determines that the mass is within the set mass range, the distribution work control unit identifies the serving position based on the image of the camera C2 that captures the container B, and moves to the serving position. Serve food f (step S23). After that, it is executed again from step 15.

On the other hand, when the information processing apparatus determines that the mass is not within the set mass range, the distribution work control unit returns a part of the food f grasped by the multi-finger hand to the tray T once, and executes the process again from step 16.

After the distribution work is completed for one tray T, the information processing apparatus conveys the next tray T to the vicinity of the distribution device DD and starts the distribution work again.

The learning operation may be performed again, for example, when the type of food loaded on the tray T supplied by the tray supply device SD1 is changed.

<Embodiment 2> This embodiment is a modified example of the multi-finger hand 30 according to the first embodiment. In FIG. 4, the tension adjusting mechanism is omitted. The multi-finger hand 30 is connected to the support box 31 by arranging and connecting to the support box 31 on a large diameter circle and six outer fingers 32a arranged and connected to the support box on a small diameter circle. Four inner fingers 32b, an outer elastic body 33 connected to the outside of each finger 32a, 32b, an inner elastic body 35 connected via an inner wire 34 of each finger 32a, 32b, and an inner elastic body. It is provided with a tension adjusting mechanism 36 for adjusting the tension of 35.

As shown in FIG. 4, the outer finger 32a and the inner finger 32b are arranged concentrically with respect to the outside of the bottom surface of the support box 31, respectively. The outer finger 32a and the inner finger 32b have a structure in which one end side is axially fixed to the support box 31 and the outer finger 32a and the inner finger 32b can swing inward or outward with the one end side as a fulcrum. Further, each of the fingers 32a and 32b is provided with a joint in the middle so that the angle on the tip side can be adjusted. Incidentally, the inner finger 32b is shorter than the outer finger 32a.

One end of the outer elastic body 33 is fixed to the outside of each finger 32a and 32b, and the other end is fixed to the outside of the bottom surface of the support box 31. As a result, the fingers 32a and 32b are always pulled outward by the outer elastic body 33 and are in an open state.

As shown in FIG. 5, the tension adjusting mechanism 36 is installed inside the support box 31. Specifically, the tension adjusting mechanism 36 includes a first regulating plate 36a fixed to the inside of the bottom surface of the support box 31 and a second regulating plate 36b fixed to the outside of the bottom surface of the support box 31. , A tube 36c spanned between the first regulating plate 36a and the second regulating plate 36b, a servomotor 36d fixed to the inside of the top surface of the support box 31, and an interval above the first regulating plate 36a. It is provided with a movable plate 36e, which is arranged apart from the servomotor 36d and fixed to the servomotor 36d. The distance between the movable plate 36e and the first specified plate 36a can be changed by the servomotor 36d. The movable plate 36e is connected to the first restricting plate 36a so that the number of inner elastic bodies 35 corresponding to the number of fingers 32 does not come into contact with the first restricting plate 36a.

The tube 36c is installed so as to connect the first regulation hole 36a'provided in the first regulation plate 36a and the second regulation hole 36b' provided in the second regulation plate 36b. One end of the tube 36c connected to the first regulation plate 36a is open toward the inner elastic body 35, and the other end of the tube 36c connected to the second regulation plate 36c is toward the finger 32. It is open. The wire 34 then extends from the inner elastic body 35 and is connected to the inside of the corresponding finger 32 through the tube 36c.

Explaining the operation of the multi-finger hand 30 according to the present embodiment, the movable plate 36e is moved upward by the servomotor 36b, in other words, in the direction away from each finger 32, thereby resisting the tension of the outer elastic body 33. The force (tension) that pulls each finger 32 inward becomes stronger. Then, when the force of pulling the fingers 32 outward by the outer elastic body 33 exceeds the force of pulling the fingers inward by the inner elastic body 35, each finger 32 closes inward. As a result, the food can be grasped by each finger 32. On the other hand, by moving the movable plate 36a downward by the servomotor 36b, in other words, in the direction approaching each finger 32, the force (tension) that pulls each finger 32 inward against the tension of the outer elastic body 33 is weak. Become. Then, when the force of pulling the fingers 32 outward by the outer elastic body 33 exceeds the force of pulling the fingers inward by the inner elastic body 35, the food grasped by each finger 32 can be released.

When adjusting the gripping force of each finger 32, the outer elastic body 33 and the inner elastic body 35 may be changed to those having different tensions. It can also be adjusted by adjusting the length of the wire 34 or adjusting the height position of the movable plate 36e.

With such a configuration, when the food is grasped with a multi-finger hand, each finger moves independently, so that a biased force acts on the food, and for example, a fragile food such as potato chips. It is possible to prevent damage to easily crushed foods such as tofu and tofu. Also, in this case, since each finger moves according to the shape of the food, the shape may change at any time when the food is grabbed together like noodles, granules, thin plates, macaroni, etc. Even if it is a food, each finger follows and moves according to the change in its shape, so that the food can be grasped without missing. In addition, when grasping fragile foods or crushable foods with a multi-finger hand, the gripping force (torque) per finger should be set weakly by increasing the number of fingers provided in the multi-finger hand. This makes it possible to prevent damage to each of the foods. The gripping force can be easily changed for each finger 32.

Further, since the wire 34 is guided by the tube 36c, even if the wire 34 slides on the tube 36c, no load is applied to the wire 34, and the wire 34 can be prevented from being broken.

In the present embodiment, the tension adjusting mechanism 36 is provided for the inner elastic body 35, but the tension adjusting mechanism 36 may be provided for the outer elastic body 33, and the tension adjusting mechanism 36 may be provided for both elastic bodies 35 and 36. A tension adjusting mechanism 36 may be provided.

<Embodiment 3> This embodiment is a modified example of the multi-finger hand 30 according to the first embodiment. The multi-finger hand 30 according to the present embodiment includes a support box 31, a plurality of fingers 32 arranged and connected to the support box 31 so as to face each other, and an elastic body 37 connected to one end of each finger 32. It is provided with an elevating mechanism 38 that elevates and elevates the elastic body 37 with respect to each finger 32.

The plurality of fingers 32 are arranged side by side so as to face the bottom surface of the support box 31, and are pivotally supported so that they can be opened and closed with one end side as a fulcrum. The plurality of fingers 32 project from the bottom surface of the support box 31 to the other end side for gripping food or the like.

The elastic body 37 is a so-called coil spring. The elastic body 37 is arranged inside the support box 31, one end of which is fixed to one end of each finger 32, and is connected to an elevating mechanism 38 arranged above each finger 32.

The elevating mechanism 38 is a so-called servomotor. The elevating mechanism 38 is fixed to the inner top surface of the support box 31 and operates so as to elevate each elastic body 37.

Explaining the operation of the multi-finger hand 30 according to the present embodiment, each elastic body 37 is lowered by the elevating mechanism 38, in other words, by moving one end of each finger 32 in the pushing direction, each finger 32 is moved to the upper end thereof. The side is pushed by each elastic body 37 and operates to open. In this case, the elastic body 37 transmits the pushing operation of the elevating mechanism 38 to the finger 32 while maintaining the contracted state.

On the other hand, by raising each elastic body 37 by the elevating mechanism 38, in other words, moving each finger 32 in the pulling direction, each finger 32 operates so that its upper end side is pulled by each elastic body 37 and closes to each other. .. In this case, the elastic body 37 maintains a contracted state until it is pulled by the elevating mechanism 38 by a predetermined distance, but gradually expands when it is pulled by the elevating mechanism 38 by a predetermined distance or more. When the elastic body 37 is in a stretched state, the force for pulling the fingers 32 increases or decreases depending on the degree of stretching, whereby the gripping force of each finger 32 can be adjusted. By adjusting the elastic force of each elastic body 37, the gripping force of each pair of fingers can be adjusted more finely.

Even with such a configuration, the same effect as that of the multi-finger hand 39 according to the second embodiment can be obtained.

<Embodiment 4> This embodiment is a modification of the information processing apparatus of the first embodiment. The information processing apparatus of the present embodiment is different from the information processing apparatus of the first embodiment in that the information processing apparatus of the present embodiment further includes a ridged data generation unit, and the learning work control unit of the information processing apparatus of the first embodiment. , The configuration of the learning data storage unit and the distribution work control unit is different. The differences between them will be described in detail below.

As shown in FIG. 9, the coordinate data generation unit includes the opening width X of each finger and the insertion depth Y of each finger included in the control conditions of the weighing operation before the weighing operation of the learning work control unit. With reference to the height of each coordinate detected by the height detection unit, the height of the diameter X and the depth Y centered on the highest coordinate P, which is the coordinate of the highest height of the food F, is set. It is used to generate the shape data indicating the shape of the food that has been piled up.

The filling shape data is, for example, a value represented by the ratio of the following two values in the food F filled before the weighing operation of the learning work control unit.
-The height of the highest coordinates-The cross-sectional area of the foods piled up in the piled range of the piled food F The cross-sectional area is the cross-sectional area of the cross-section that passes through the highest coordinates of the piled food F, and is one. It may be two or more. In the case of two or more, the cross-sectional area of two orthogonal cross sections passing through the highest coordinates may be used. In addition, instead of the cross-sectional area, the volume of the food loaded in the serving range of the serving food F may be used.

The learning work control unit of the present embodiment is different from the learning work control unit of the first embodiment in that when a part of the piled food F is grasped, a part of the food is grasped from the highest coordinates of the piled food F. It's different.

The learning data storage unit of the present embodiment is described above in that, for each weighing operation of the learning work control unit, the learning data is generated by associating the control conditions with the mass of the food grasped by the control conditions and the filling shape data. It is different from the learning data storage unit of the first embodiment.

The distribution work control unit of the present embodiment differs from the distribution work control unit of the first embodiment in the method of selecting the distribution control conditions. Specifically, when the distribution work control unit of the present embodiment grabs and distributes a food having a predetermined set mass from the piled food to be distributed, first, each stored in the learning data storage unit. Refer to the primary training data set including the mass included in the predetermined error range (tolerance range) centering on the set mass among the training data, and the control conditions and the shape data included in each training data of the primary training data set. Is classified into multiple classes between the maximum value and the minimum value.

Here, the term "multiple classes" refers to, for example, when the control condition is the opening width of each finger, refer to the maximum value and the minimum value of the opening width value of each finger included in each learning data of the primary learning data set. However, each range formed by equally dividing between the maximum value and the minimum value is shown.

Then, the distribution work control unit of the present embodiment refers to the number of values included in each class, and the set mass is set from each learning data having the control condition or the shape data in the class in which the most values are classified. The training data containing the mass closest to is selected, and the control conditions for the training data are set as the distribution control conditions.

That is, for example, when the learning data includes the value of the opening width of each finger and the value of the insertion depth of each finger, first, the value of the opening width of each finger included in each learning data is classified into a plurality of classes. , Classify the insertion depth value of each finger into multiple classes. At this time, it is preferable that the number of classes is the same under each control condition. Then, if the class containing the largest number of values among all the classes is a class in which the value of the opening width of each finger is classified, each learning having the opening width of each finger classified in this class is performed. Select the training data containing the mass closest to the set mass from the data.

It should be noted that the operation when the learning work of the filling system 100 according to the present embodiment is that the filling shape data generation unit generates the filling shape data before the weighing operation of the learning work control unit, and the learning work control unit It is the same as the operation of the first embodiment except that a part of the food is grabbed from the highest coordinates of the food F piled up in the weighing operation and the learning data generated by the learning data storage unit includes the filling shape data. is doing.

Further, the operation when the distribution work of the filling system 100 according to the present embodiment is executed is the same as the operation of the first embodiment except that the method of selecting the control conditions in the distribution work control unit is different.

<Other Embodiments> In the first embodiment, one distribution device DD is arranged with respect to the filling system 100, but a plurality of distribution device DDs may be arranged. In such a case, different foods can be sequentially arranged by each distribution device DD on the container B transported by the container transport device SD2, and the work efficiency is remarkably improved.

The distribution device DD according to the first embodiment can also be used as a learning data generation device by using it only for the learning work. In this case, the learning data generator creates a target learning data set for each type of food, and the target learning data set is used as a food by a robot arm equipped with a multi-finger hand having the same configuration as the learning data generator. Can be used for distribution.

In the first embodiment, the food in the state of being piled up on the tray T is distributed, but it is not always necessary to put on the tray T, and the food in the state of being piled up directly on the conveyor such as a belt conveyor is distributed. You may.

Further, in the first embodiment, the distribution device DD is used for the filling system 100, but the distribution device DD can be used not only for the filling system 100 but also for packing foods and the like.

In addition, the present invention is not limited to each of the above-described embodiments, and it goes without saying that various modifications can be made without departing from the spirit of the present invention.

100 Filling system DD distribution device SD1 tray supply device (first supply means)
SD2 container supply device (second supply means)
C1, C2 Camera 10 Hand support mechanism 20 Measuring instrument 30 Multi-finger hand 33 Outer elastic body 35 Inner elastic body 36 Tension adjustment mechanism

Claims (15)

It is a distribution device that grabs and distributes a part of food such as noodles, granules, thin plates, macaroni, etc. from the piled state.
A multi-fingered hand that grabs some foods by opening and closing multiple fingers from the piled foods,
A measuring instrument that weighs some of the foods grasped by the plurality of fingers,
A learning operation is performed in which the multi-fingered hand is controlled under arbitrary control conditions, and a series of weighing operations of grasping a part of the food from the piled food and measuring the mass is repeated while changing the control conditions. With the learning work control unit that controls
For each weighing operation, training data associated with the mass of a part of the food whose control condition is grasped under the control condition is generated, and a plurality of learning data generated by each weighing operation in the learning operation is collected as a learning data set. The learning data storage unit to be stored as
A distribution device including a distribution work control unit that controls to perform a distribution work of grasping and distributing a part of food from the piled food based on the learning data set. The control conditions are the opening width of each finger, the closing width of each finger, the closing speed of each finger, and the closing speed of each finger when a part of the food is grabbed from the piled food by the multi-finger hand. The distribution device according to claim 1, which is a gripping condition selected from the gripping force and the insertion depth of each finger into the piled food. Further, a height detecting unit for detecting the height of each coordinate with respect to the reference plane of the piled food is further provided.
The distribution device according to claim 2, wherein the insertion depth of each finger into the heaped food under the control conditions is set based on the heap height detected by the heap height detecting unit. Before the weighing operation of the learning work control unit, the opening width X of each finger and the insertion depth Y of each finger included in the control conditions of the weighing operation, and the height of each coordinate detected by the height detecting unit. With reference to the height, the filling shape data indicating the filling shape of the food piled in the filling range of the diameter X and the depth Y centered on the highest coordinate, which is the coordinate of the highest filling height of the piled food, is generated. It is further equipped with a coordinate data generation unit to be used.
The learning work control unit controls the multi-fingered hand under the control conditions, grabs the food from the highest coordinates of the piled food, and executes a weighing operation to measure the mass.
The distribution device according to claim 3, wherein the learning data storage unit associates the control conditions with the mass and the filling shape data to generate learning data. The distribution device according to claim 4, wherein the swelling shape data is a value represented by the ratio of the swelling height of the highest coordinates to the volume or cross-sectional area of the food squeezed in the swelling range. The learning data storage unit selects learning data including a mass within a predetermined target mass range from the learning data set, and stores the selected plurality of learning data as a target learning data set. ,
Any of claims 1 to 5, wherein the distribution work control unit controls to perform a distribution work of grasping and distributing a part of food from the piled food based on the target learning data set. Distributor according to. The learning work control unit selects a non-variable grip condition that does not change and a variable grip condition that changes from a plurality of grip conditions included in the control conditions of the multi-finger hand, and performs the series of weighing operations. The unit learning work that repeats a predetermined number of times while changing the variable gripping condition is performed a plurality of times, and the non-variable gripping condition is added each time the unit learning work is repeated.
The learning data storage unit generates the target learning data set for each unit learning operation.
The upper and lower limit values that specify the upper and lower limit values of the variable gripping conditions by referring to the variable gripping conditions of each learning data included in the target learning data set generated in the arbitrary unit learning work in the learning work control unit. Further equipped with a specific part,
The upper limit value and the lower limit value of the fluctuation gripping condition specified by the upper / lower limit value specifying unit are set as the upper limit value and the lower limit value of the fluctuation gripping condition selected in the next unit learning work in the learning work control unit. The distribution device according to claim 6, which is configured. When the distribution work control unit grabs and distributes a food having a predetermined set mass from the filled food to be distributed, the learning data stored in the learning data storage unit is the most suitable for the set mass. The control condition of the learning data including the close mass is set as the control condition for distribution, and the control condition for distribution controls to grab a part of the food from the filled food to be distributed and distribute it. The distribution device according to any one of claims 1 to 7. When the distribution work control unit grabs and distributes a food having a predetermined set mass from the piled food to be distributed, the set mass among the learning data stored in the learning data storage unit. Each learning data including the mass included in the predetermined error range is referred to, and the control conditions included in each learning data are classified into multiple classes between the maximum value and the minimum value, and the most among all the classes. From each learning data having the control condition in the class in which many values are classified, the training data including the mass closest to the set mass is selected, the control condition of the training data is set as the control condition for distribution, and the distribution thereof. The distribution device according to any one of claims 1 to 7, wherein a part of the food is grabbed and distributed according to the control conditions. When the distribution work control unit grabs and distributes a food having a predetermined set mass from a pile of food to be distributed, the set mass among the learning data stored in the learning data storage unit. With reference to each training data including mass included in a predetermined error range, the control conditions and ridge shape data included in each training data are classified into multiple classes between the maximum value and the minimum value, and all classes are classified. The training data containing the mass closest to the set mass is selected from each training data having the control condition or the shape data in the class in which the most values are classified, and the control condition of the training data is distributed. The distribution device according to claim 4, which is set as a control condition and is controlled so as to grab a part of the food from the serving food to be distributed and distribute the data according to the distribution control condition. When the mass measured by the distribution work control unit by grasping a part of the food from the piled food according to the distribution control condition is outside the set mass range including the set mass, the said again. The distribution device according to any one of claims 8 to 10, wherein the distribution device is controlled so as to grab a part of the food from the filled food according to the distribution control conditions. The multi-fingered hand according to any one of claims 1 to 11.
A plurality of fingers configured to open and close inwardly and outwardly with one end side connected in a ring shape as a fulcrum with respect to the support block.
An outer elastic body that is connected to each of the fingers and pulls each finger outward.
An inner elastic body that is connected to each of the fingers and pulls each finger inward.
A distribution device comprising a tension adjusting mechanism for adjusting the tension of at least one of the outer elastic body and the inner elastic body. The multi-fingered hand according to any one of claims 1 to 11.
Multiple fingers that are arranged facing each other or in a ring shape and open and close with the upper end side as a fulcrum,
An elastic body connected to the upper end of each finger,
It is provided above each finger and has an elevating mechanism for elevating and elevating the elastic body.
By lowering the elastic body by the elevating mechanism, the upper end of each finger operates so as to be pushed by the elastic body to open, and by raising the elastic body by the elevating mechanism, the upper end of each finger is raised. Is pulled by the elastic body to close, and by raising the elastic body by a predetermined distance or more by the elevating mechanism, the elastic body is in an extended state and the gripping force of each finger can be adjusted. Distributor configured. The distribution device according to any one of claims 1 to 13 and
A first supply means for supplying the served food in the vicinity of the distribution device,
A serving system comprising a second supply means for supplying a container for serving a part of the food grabbed from the serving food in the vicinity of the distribution device. A multi-fingered hand that grabs some foods by opening and closing multiple fingers from foods such as noodles, granules, thin plates, and macaroni .
A measuring instrument that weighs some foods grabbed by the multiple fingers,
A learning operation is performed in which the multi-fingered hand is controlled under arbitrary control conditions, and a series of weighing operations of grasping a part of the food from the piled food and measuring the mass is repeated while changing the control conditions. With the learning work control unit that controls
A learning data storage unit that generates learning data in which the control conditions are associated with the mass of a part of the foods grasped by the control conditions for each measurement operation, and stores the generated plurality of learning data as a learning data set. And, a learning data generator comprising.

Images