JP6904657B2 - Cargo handling system and control method - Google Patents

Description

The present invention relates to a cargo handling system and a control method including an unmanned forklift.

The unmanned forklift performs cargo handling work according to a preset route while recognizing its own position during traveling. For example, a laser unmanned forklift includes a laser scanner (see, for example, Patent Documents 1 and 2). The laser scanner transmits and receives the laser to and from the reflector while rotating the laser horizontally 360 degrees. The laser unmanned forklift recognizes a plurality of reflectors arranged along a traveling path in the warehouse with a laser scanner.

The reflector is fixed in the warehouse and its position is stored on the map. The laser unmanned forklift recognizes a plurality of reflectors with a laser scanner and calculates the current position based on the principle of triangulation. The laser unmanned forklift travels on a preset route based on the calculated current position.

Further, as disclosed in Patent Document 3, there is a shelf for storing luggage. Multiple shelves are installed in the warehouse to store multiple luggage. A laser unmanned forklift and a plurality of shelves are installed in the warehouse, and the laser unmanned forklift takes in and out luggage from the shelves.

By the way, it is desirable to carry out cargo handling work promptly. The unmanned forklift stops traveling at the cargo handling work position and then starts raising and lowering the fork. Therefore, when performing cargo handling work on a load at a high position on the rack, it may take a long time to raise and lower the fork. On the other hand, when the operator operates the manned forklift, the fork may be started to move up and down while the vehicle is running to reduce the fork lifting time at the cargo handling work position.

Japanese Unexamined Patent Publication No. 8-161039 Japanese Unexamined Patent Publication No. 8-166821 Japanese Unexamined Patent Publication No. 2003-20102

Therefore, the problem to be solved by the present invention is a cargo handling system and a control method for enabling an unmanned forklift to quickly perform cargo handling operations.

In order to solve the above problems, cargo handling system according to the present invention, a facility having a plurality of shelves, and unmanned forklift to perform traveling and load operations in a facility, Ru handling system der equipped with. The cargo handling system is a machine that collects teacher data based on the positional relationship between the vehicle body of the unmanned forklift and the fork and the load placed on the shelf, and machine learning from the teacher data collected in the collection unit. A learning model generation unit that generates and stores a learning model by learning, an acquisition unit that acquires the current positional relationship at predetermined time intervals, and a learning model generated by the learning model generation unit at the present time that is acquired from the acquisition unit. By applying it to the positional relationship, it includes a processing unit that determines the timing at which the fork starts to move up and down, and a control unit that executes predetermined control for raising and lowering the fork at the position of the vehicle body determined by the processing unit. The teacher data consists of input data and output data. The input data includes a horizontal distance D from the horizontal position of the vehicle body to the horizontal position of the load and a vertical distance H from the vertical position of the fork to the vertical position of the load. The output data includes a horizontal distance D1 from the horizontal position of the vehicle body that starts ascending the fork to the horizontal position of the load, and a horizontal distance D2 from the horizontal position of the vehicle body that starts descending the fork to the horizontal position of the load. ..

Preferably,
The input data further includes the traveling speed V of the vehicle body.
The output data further includes a velocity V1 for ascending the fork and a velocity V2 for descending the fork.

In order to solve the above problems, the control method of the cargo handling system according to the present invention is a control method of a cargo handling system including a facility having a plurality of shelves and an unmanned forklift that runs and handles cargo in the facility. Oh Ru. The control method is a collection step that collects teacher data based on the positional relationship between the vehicle body of the unmanned forklift and the fork and the load placed on the shelf, and machine learning from the teacher data collected in the collection step. A learning model generation step that generates and stores a learning model by learning, an acquisition step that acquires the current positional relationship at predetermined time intervals, and a learning model generated by the learning model generation unit are acquired in the acquisition step at the present time. By applying to the positional relationship, it includes a processing step for determining the timing at which the fork starts to move up and down, and a control step for executing predetermined control for raising and lowering the fork at the position of the vehicle body determined by the processing step. The teacher data consists of input data and output data. The input data includes a horizontal distance D from the horizontal position of the vehicle body to the horizontal position of the load and a vertical distance H from the vertical position of the fork to the vertical position of the load. The output data includes a horizontal distance D1 from the horizontal position of the vehicle body where the fork starts to rise to the horizontal position of the load, and a horizontal distance D2 from the horizontal position of the vehicle body where the fork starts to descend to the horizontal position of the load. ..

By providing the above-mentioned configuration, the cargo handling system and the control method according to the present invention allow an unmanned forklift to quickly perform cargo handling operations.

Top view showing the cargo handling system. A block diagram showing a cargo handling system. Top view for explaining teacher data. Side view for explaining teacher data. The flowchart which shows the control method of a cargo handling system.

Hereinafter, an embodiment of a cargo handling system and a control method according to the present invention will be described with reference to the drawings.

As shown in FIG. 1, the cargo handling system includes a facility 3 and a laser type unmanned forklift (hereinafter referred to as “forklift”) 2 that travels and handles cargo in the facility 3. In the present embodiment, the facility 3 is a warehouse, but may be a factory or the like. A plurality of shelves 1 are installed in the facility 3. A plurality of loads 10 are placed on the shelf 1. The forklift 2 travels in the facility 3 and performs cargo handling work of loading and unloading a predetermined load 10 from a predetermined shelf 1.

The forklift 2 operates automatically by laser guidance. The forklift 2 includes a laser scanner 20. A plurality of reflectors 21 are installed in the facility 3. The laser scanner 20 transmits and receives the laser L to and from the reflector 21 while rotating the laser L horizontally 360 degrees.

The forklift 2 recognizes a plurality of reflectors 21 arranged along the traveling path in the facility 3 by the laser scanner 20. The reflector 21 is fixed to the wall in the facility 3, and its position information is stored on the map. The forklift 2 recognizes the plurality of reflectors 21 with the laser scanner 20 and calculates the current position based on the principle of triangulation.

The laser scanner 20 detects the laser L reflected from the reflector 21 and calculates the angle (orientation) or distance between the laser scanner 20 and the reflector 21. When the reflector 21 is recognized by the laser scanner 20, the triangle specified by the reflector 21 is calculated based on the angle or distance between the laser scanner 20 and the reflector 21.

The current position of the forklift 2 is calculated by collating the angle or distance information recognized by the forklift 2 with the position information of the reflector 21 stored in advance.

The forklift 2 travels on a preset route based on the calculated current position. Further, the forklift 2 moves up and down and advances the fork 23 based on a preset horizontal position of the load 10 to perform cargo handling work.

As shown in FIG. 2, the cargo handling system includes a collecting unit 40 that collects teacher data 5. The teacher data 5 includes (1) a horizontal distance D on the path from the horizontal position of the vehicle body 22 to the horizontal position of the load 10 and (2) the vertical position of the load 10 from the vertical position of the fork 23 in the manned forklift operated by the operator. Vertical distance H, (3) Start ascending the fork 23 Horizontal distance (ascending timing) D1 on the path from the horizontal position of the vehicle body 22 to the horizontal position of the load 10, (4) Start descending the fork 23 Includes the horizontal distance (descending timing) D2 on the path from the horizontal position of the vehicle body 22 to the horizontal position of the load 10.

As shown in FIG. 3, (1) horizontal distance D is a horizontal distance on the route from the horizontal position (XY coordinates) of the vehicle body 22 to the horizontal position (XY coordinates) of the load 10. (3) The horizontal distance D1 is a horizontal distance (rising timing) on the route from the horizontal position (XY coordinates) of the vehicle body 22 where the fork 23 starts to rise to the horizontal position (XY coordinates) of the load 10. (4) The horizontal distance D2 is a horizontal distance (descending timing) on the path from the horizontal position (XY coordinates) of the vehicle body 22 where the fork 23 starts descending to the horizontal position (XY coordinates) of the load 10. As shown in FIG. 4, (2) the vertical distance H is the distance from the current vertical position (Z coordinate) of the fork 23 to the vertical position (Z coordinate) of the load 10.

The cargo handling system includes a learning model generation unit 41 that performs machine learning from the teacher data 5 ((1) to (4)) collected by the collection unit 40 and generates and stores a learning model by machine learning. The learning model generation unit 41 of the present embodiment carries out supervised learning. In supervised learning, a large amount of teacher data 5, that is, a set of input data ID and output data OD, is input to the learning model generation unit 41.

The input data ID includes (1) a horizontal distance D and (2) a vertical distance H. The output data OD includes (3) horizontal distance D1 and (4) horizontal distance D2. When the operator actually operates the manned forklift, in order to quickly perform the cargo handling work, the fork is started to be lifted and lowered while the vehicle is running to reduce the fork lifting time at the cargo handling work position. There is. Therefore, there is a certain relationship such as a correlation between the horizontal distances D, D1, and D2 between the vehicle body 22 and the load 10 and the raising and lowering of the fork 23 according to the vertical position of the load 10. Can be inferred.

The learning model generation unit 41 uses a machine learning algorithm such as a general neural network. The learning model generation unit 41 performs machine learning using the correlated input data ID and output data OD as teacher data 5, thereby estimating the output from the input (learning model), that is, the input data ID (1). When and (2) are input, a model that outputs the output data OD (3) and (4) is generated.

The cargo handling system includes an acquisition unit 45 that acquires the current input data ID at predetermined time intervals. As described above, the input data IDs are (1) horizontal distance D and (2) vertical distance H. The input data ID is acquired every predetermined time (for example, 1 minute).

The cargo handling system applies the learning model generated by the learning model generation unit 41 to the current input data ID acquired from the acquisition unit 45, so that (3) the horizontal position of the vehicle body 22 that starts ascending the fork 23. The process of determining the horizontal distance (ascending timing) D1 from the to the horizontal position of the load 10 and (4) the horizontal distance (descending timing) D2 from the horizontal position of the vehicle body 22 to the horizontal position of the load 10 when the fork 23 starts descending. A unit 42 is provided.

The cargo handling system includes a control unit 43 in which the forklift 2 executes predetermined control based on the output data OD determined by the processing unit 42. Then, the control unit 43 determines the timing at which the fork 23 starts to move up and down based on the horizontal distances D1 and D2.

As shown in FIG. 5, the above cargo handling system executes the following control method. In addition, in order to avoid duplicate explanation, the part already explained is omitted.
The teacher data 5 is collected by the collection unit 40 (collection step: S1). Then, the learning model generation unit 41 performs machine learning from the teacher data 5 collected in the collection unit 40 in the collection step S1, and generates and stores the learning model by machine learning (learning model generation step: S2). The acquisition unit 45 acquires the current input data ID at predetermined time intervals (acquisition step: S3).

By applying the learning model generated in the learning model generation step S2 to the current input data ID acquired in the acquisition step S3 by the processing unit 42, (3) horizontal distance D1 and (4) horizontal distance D2 can be obtained. Determine (prediction step: S4). The forklift 2 executes predetermined control by the control unit 43 based on the output data OD determined in the processing step S4 (control step: S5).

Although the preferred embodiments of the present invention have been described above, the configuration of the present invention is not limited to these embodiments.

In another embodiment, the input data ID may include (5) traveling speed V in addition to the above (1) and (2). Further, as the output data OD, in addition to the above (3) and (4), (6) the speed V1 for ascending the fork 23 and (7) the speed V2 for descending the fork 23 may be included. The teacher data 5 includes input data and output data having (5) to (7) in addition to (1) to (4). As a result, in addition to the timing of raising and lowering the fork 23, the fork 23 can be raised and lowered at a preferable speed.

In the above embodiment, the forklift 2 is laser-guided, but it may be electromagnetically guided by a magnetic tape or the like laid on the floor surface, or may be image-guided by a line drawn on the floor surface. ..

2 Unmanned forklift 3 Facility 40 Collection unit 41 Learning model generation unit 42 Processing unit 43 Control unit 45 Acquisition unit 5 Teacher data

Claims (4)

A cargo handling system including a facility having a plurality of shelves and an unmanned forklift that runs and handles cargo in the facility.
The cargo handling system
A collection unit that collects teacher data based on the positional relationship between the vehicle body and fork of the unmanned forklift and the load placed on the shelf.
A learning model generation unit that performs machine learning from the teacher data collected in the collection unit and generates and stores a learning model by the machine learning.
An acquisition unit that acquires the current positional relationship at predetermined time intervals,
By applying the learning model generated by the learning model generation unit to the current positional relationship acquired from the acquisition unit, a processing unit that determines the timing to start raising and lowering the fork, and a processing unit.
A control unit that executes predetermined control for raising and lowering the fork at the position of the vehicle body determined by the processing unit is provided .
The teacher data consists of input data and output data.
The input data is
The horizontal distance D from the horizontal position of the vehicle body to the horizontal position of the load and
Including the vertical distance H from the vertical position of the fork to the vertical position of the load.
The output data is
The horizontal distance D1 from the horizontal position of the vehicle body to the horizontal position of the load when the fork starts to rise,
A cargo handling system comprising a horizontal distance D2 from a horizontal position of the vehicle body to a horizontal position of the load at which the fork starts to descend. The input data further includes the traveling speed V of the vehicle body.
The cargo handling system according to claim 1 , wherein the output data further includes a speed V1 for raising the fork and a speed V2 for lowering the fork. A control method for a cargo handling system including a facility having a plurality of shelves and an unmanned forklift that runs and handles cargo in the facility.
The control method is
A collection step for collecting teacher data based on the positional relationship between the vehicle body and fork of the unmanned forklift and the load placed on the shelf, and
A learning model generation step in which machine learning is performed from the teacher data collected in the collection step and a learning model is generated and stored by the machine learning.
An acquisition step for acquiring the current positional relationship at predetermined time intervals, and
A processing step of determining the timing at which the fork starts to move up and down by applying the learning model generated by the learning model generation unit to the current positional relationship acquired in the acquisition step.
A control step for executing a predetermined control for raising and lowering the fork at the position of the vehicle body determined by the processing step is provided .
The teacher data consists of input data and output data.
The input data is
The horizontal distance D from the horizontal position of the vehicle body to the horizontal position of the load and
Including the vertical distance H from the vertical position of the fork to the vertical position of the load.
The output data is
The horizontal distance D1 from the horizontal position of the vehicle body to the horizontal position of the load when the fork starts to rise,
A method for controlling a cargo handling system , which comprises a horizontal distance D2 from a horizontal position of the vehicle body to a horizontal position of the load at which the fork starts to descend. The input data further includes the traveling speed V of the vehicle body.
The output data further includes a velocity V1 for ascending the fork and a velocity V2 for descending the fork.
The control method for a cargo handling system according to claim 3, wherein the cargo handling system is controlled.

Images