JP7084098B1 - Forklift and cargo handling system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a forklift capable of calling attention to an operator even when the operator is in a blind spot area. An image data generation unit 203, a learning model unit 204 that generates a safety score, a processing unit 205 that acquires a safety score from a learning model unit 204, a voice output unit 201 that outputs a melody, and a melody. The map generation unit 207 includes a voice control unit 208 that adjusts the output state of the device, a position estimation unit 206 that estimates the self-position, and a map generation unit 207 that generates an environment map showing the distribution of the safety score. The safety score obtained from the processing unit 205 is used as the current score, the safety score at the self-position in the environment map is used as the past score, the two are compared, and the safety score with the lower safety is output to the voice control unit 208. However, the voice control unit 208 is characterized in that the output state of the melody is adjusted based on the safety score acquired from the map generation unit 207. [Selection diagram] Fig. 1

Description

The present invention relates to forklifts and cargo handling systems.

Conventionally, forklifts that run while outputting a melody have been known for the purpose of calling attention to workers in the workplace (see, for example, Patent Document 1). The forklift described in Patent Document 1 has an image data generation unit that captures the front of the vehicle body and generates image data, and a learning model unit that generates a safety score indicating the degree of safety of the traveling route based on the image data. And adjust the output state of the melody according to the safety score.

For example, when there is an operator on the traveling route in the image data, the forklift determines that the safety is low and generates a safety score of a small numerical parameter. For this reason, it is preferable that a safety score with a small numerical parameter is always generated in an area where the operator frequently works.

However, when the worker is in the blind spot area (for example, the area hidden by the rack or luggage and cannot be seen from the shooting means) at the timing when the shooting means of the image data generation unit is shooting, the forklift has no worker. Will generate a safety score similar to. As a result, there arises a problem that it is not possible to call attention to the worker.

Japanese Patent Application No. 2019-219523

The present invention has been made in view of the above circumstances, and an object thereof is to provide a forklift and a cargo handling system capable of calling attention to an operator even when the operator is in a blind spot area. It is in.

In order to solve the above problems, the forklift according to the present invention
A forklift that performs cargo handling work in the workplace.
With the car body
An image data generation unit that captures an image including the front of the vehicle body and generates image data,
When the image data is input, a learning model unit machine-learned to generate a safety score indicating the degree of safety of the traveling route of the vehicle body by using a machine learning algorithm having predetermined parameters, and a learning model unit.
A processing unit that acquires the safety score from the learning model unit by inputting the image data into the learning model unit, and a processing unit.
An audio output unit that outputs a predetermined melody while driving,
A voice control unit that adjusts the output state of the melody,
The position estimation unit that estimates the self-position and the position estimation unit
A map generator that generates an environmental map showing the distribution of the safety score in the workplace is provided.
The map generator
The safety score obtained from the processing unit is used as the current score, the safety score at the self-position in the environment map is used as the past score, and the current score and the past score are compared and the safety is lower. The safety score of the above is output to the voice control unit,
The voice control unit
It is characterized in that the output state of the melody is adjusted based on the safety score obtained from the map generation unit.

In this configuration, the safety score generated in real time based on the image data is used as the current score, the safety score at the self-position in the environment map is used as the past score, and the melody is based on the safety score of the lower safety one. Adjust the output status of. For example, when a worker enters a blind spot area in an area where the worker frequently works, the past score has a lower safety score (score of a small numerical parameter) than the current score. Therefore, according to this configuration, it is possible to appropriately call attention to the worker in the blind spot area.

In the above forklift
The map generation unit can be configured to update the past score using the current score after outputting the safety score to the voice control unit.

In the above forklift
It is provided with an alarm sound output unit that outputs an alarm sound louder than the melody under the control of the voice control unit.
When the safety score is smaller than a predetermined threshold value, the voice control unit can be configured to output the alarm sound to the alarm sound output unit instead of outputting the melody to the voice output unit.

In order to solve the above problems, the cargo handling system according to the present invention is
A cargo handling system including a management device and a plurality of forklifts that perform cargo handling work in the workplace under the control of the management device.
The forklift is
With the car body
An image data generation unit that captures an image including the front of the vehicle body and generates image data,
When the image data is input, a learning model unit machine-learned to generate a safety score indicating the degree of safety of the traveling route of the vehicle body by using a machine learning algorithm having predetermined parameters, and a learning model unit.
A processing unit that acquires the safety score from the learning model unit by inputting the image data into the learning model unit, and a processing unit.
An audio output unit that outputs a predetermined melody while driving,
A voice control unit that adjusts the output state of the melody,
It is equipped with a position estimation unit that estimates its own position.
The management device is
A communication unit for acquiring the safety score and the self-position from the forklift,
A map generator that generates an environmental map showing the distribution of the safety score in the workplace is provided.
The map generator
The safety score obtained from the forklift is used as the current score, the safety score at the self-position of the forklift in the environment map is used as the past score, and the current score is compared with the past score for safety. The lower safety score is transmitted to the voice control unit of the forklift.
The voice control unit
It is characterized in that the output state of the melody is adjusted based on the safety score obtained from the map generation unit.

In the above cargo handling system
The processing unit transmits image data with a safety score in which the image data input to the learning model unit is associated with the safety score acquired from the learning model unit to the management device.
The management device is
A collection unit that collects image data with a safety score and generates learning data,
A parameter update unit that performs machine learning based on the learning data, generates update data for updating the parameters included in the learning model unit, and transmits the update data to the learning model unit.
Equipped with
The learning model unit can be configured to update the parameters based on the update data.

In the above cargo handling system
The map generation unit can be configured to update the past score using the current score after transmitting the safety score to the voice control unit.

According to the present invention, it is possible to provide a forklift and a cargo handling system capable of calling attention to an operator even when the operator is in a blind spot area.

It is a block diagram of the cargo handling system which concerns on 1st Embodiment. It is a side view of the forklift which concerns on 1st Embodiment. It is a top view of the cargo handling system which concerns on 1st Embodiment. (A) It is a model diagram displayed on the management apparatus which concerns on 1st Embodiment. (B) It is a model diagram (environmental map) generated by the map generation part of the forklift which concerns on 1st Embodiment. It is a block diagram of the cargo handling system which concerns on 2nd Embodiment.

Hereinafter, embodiments of the forklift and cargo handling system according to the present invention will be described with reference to the accompanying drawings.

[First Embodiment]
FIG. 1 shows a block diagram of a cargo handling system 1A according to the first embodiment of the present invention. The cargo handling system 1A includes a management device 100A and a plurality of laser-guided unmanned forklifts 200A (corresponding to the "forklift" of the present invention).

The management device 100A includes a communication unit 101, a general control unit 102, and a display unit 103. As shown in FIG. 3, the management device 100A may be provided outside the work place 2 (in this embodiment, a warehouse having a plurality of racks 3) on which the unmanned forklift 200A travels, or may be provided inside the work place 2. May be good.

The communication unit 101 is configured to perform wireless communication with an unmanned forklift 200A registered in advance in the management device 100A.

The integrated control unit 102 is configured to manage the traveling and cargo handling work of the unmanned forklift 200A. For example, the overall control unit 102 creates a cargo handling work schedule for the unmanned forklift 200A and determines a traveling route for smoothly performing the cargo handling work. The integrated control unit 102 notifies the unmanned forklift 200A of the traveling route via the communication unit 101.

The display unit 103 is composed of, for example, a liquid crystal display. On the display unit 103, traveling information and cargo handling work information of the unmanned forklift 200A are displayed. As shown in FIG. 4A, the display unit 103 may display the model diagram M1 of the work place 2 and display the current location Ma of the unmanned forklift 200A, the travel route Ra, and the like as travel information in the model diagram M1.

The unmanned forklift 200A is configured to perform traveling and cargo handling work under the control of the management device 100A. As shown in FIG. 2, the unmanned forklift 200A includes a vehicle body 210, a cargo handling device 211, and a laser scanner 212 provided on the upper portion of the vehicle body 210. The cargo handling device 211 includes a mast provided on the front and rear of the vehicle body 210 so as to be horizontally movable, and a fork provided on the mast so as to be able to move up and down.

As shown in FIG. 3, the laser scanner 212 projects the laser L around while rotating the laser light source, and detects the reflected light L'from a plurality of reflectors 4 arranged in the work place 2. The calculation unit of the laser scanner 212 stores the position of the reflector 4 on a predetermined map, and calculates the current location (self-position) of the vehicle body 210 based on the principle of triangulation. In this way, the unmanned forklift 200A travels according to the notified travel route while acquiring the current location information regarding the current location of the vehicle body 210.

With reference to FIG. 1 again, the unmanned forklift 200A includes an audio output unit 201, an alarm sound output unit 202, an image data generation unit 203, a learning model unit 204, a processing unit 205, and a position estimation unit 206. It includes a map generation unit 207 and a voice control unit 208.

The voice output unit 201 is composed of, for example, at least one speaker, and outputs a predetermined melody while the vehicle body 210 is traveling under the control of the voice control unit 208. The output melody is stored in the voice control unit 208.

The alarm sound output unit 202 is composed of, for example, at least one speaker, and outputs an alarm sound having a volume higher than that of the melody under the control of the voice control unit 208. Further, when the obstacle sensor is provided in the front part of the vehicle body 210, an alarm is given when the obstacle sensor detects an obstacle existing in front of the vehicle body 210 (for example, a luggage placed on a traveling route). The sound output unit 202 may output an alarm sound.

The image data generation unit 203 is configured to include an photographing means and an image processing means. The photographing means captures an image (still image and / or moving image) including the front of the vehicle body 210 and outputs the image to the image processing means. The image processing means generates image data that can be input to the learning model unit 204 based on the image.

When the image data generated by the image data generation unit 203 is input, the learning model unit 204 uses a machine learning algorithm such as a neural network having predetermined parameters to indicate the degree of safety of the travel route. A trained model machine-learned to generate scores.

In machine learning of a trained model, a large amount of teacher data is input by using a machine learning algorithm such as a neural network as described above. The teacher data includes data in which a predetermined safety score is associated with image data obtained by photographing a traveling route in the workplace 2. As the safety score, a numerical parameter (numerical parameter of 1 to 5 in this embodiment) is used. Numerical parameters are set by a person or a computer. For example, when traveling near the rack 3 and there is a person on the traveling route, it is judged that the safety is low and a small numerical parameter is set. On the other hand, if there are no people, it is judged that the safety is high and a large numerical parameter is set. It can be inferred that there is a certain relationship such as a correlation between the condition of the traveling route shown by the image data and the degree of safety.

The processing unit 205 is configured to acquire a safety score from the learning model unit 204 by inputting the image data acquired from the image data generation unit 203 into the learning model unit 204. The processing unit 205 outputs the acquired safety score to the map generation unit 207. Further, the processing unit 205 associates the image data input to the learning model unit 204 with the safety score acquired from the learning model unit 204 to generate image data with a safety score, and the image data with the safety score. May be transmitted to the management device 100A.

The position estimation unit 206 is configured to recognize the current location (self-position) of the vehicle body 210 and acquire the current location information regarding the current location of the vehicle body 210. In the present embodiment, the position estimation unit 206 corresponds to the laser scanner 212 and its calculation unit. The position estimation unit 206 outputs the acquired current location information to the map generation unit 207.

The map generation unit 207 is configured to store and generate an environmental map showing the distribution of safety scores in the workplace 2. As shown in FIG. 4B, the environment map is a model diagram M2 of the workplace 2 in which the safety score is associated with each predetermined area. The safety score will be updated as appropriate as described below.

When the safety score is input from the processing unit 205 and the current location information (self-position) is input from the position estimation unit 206, the map generation unit 207 uses the safety score acquired from the processing unit 205 as the current score and sets the environment. The safety score at the self-position in the map is used as the past score, and the current score and the past score are compared. After comparison, the map generation unit 207 outputs the safety score (safety score of a small numerical parameter) having the lower safety to the voice control unit 208.

The map generation unit 207 is configured to output the safety score to the voice control unit 208 and then update the past score using the current score. As an update method, the map generation unit 207 may simply replace the current score with the past score, or calculate the average value of the past score and the current score for at least one time and replace the past score with the average value. May be good.

The processing unit 205, the map generation unit 207, and the voice control unit 208 are composed of, for example, at least one microcomputer, and the CPU of the microcomputer executes a predetermined program to execute the processing unit 205, the map generation unit 207, and the voice control unit 208. Various functions of 208 are realized.

The voice control unit 208 is configured to adjust the output state of the melody and the alarm sound according to the safety score acquired from the map generation unit 207. The voice control unit 208 stores data in which the relationship between the safety score and the output state of the melody and the alarm sound is defined as shown in Table 1.

When the numerical parameter of the safety score is in the range of 5 to 2, the voice control unit 208 outputs the melody so that the volume of the melody becomes louder as the numerical parameter of the safety score becomes smaller (the safety becomes lower). Adjust the condition.

When the numerical parameter of the safety score is 1, the voice control unit 208 causes the voice output unit 201 to stop the output of the melody, while the alarm sound output unit 202 outputs the alarm sound. Instead of stopping the output of the melody, the voice control unit 208 may output the alarm sound to the alarm sound output unit 202 in a state where the volume of the melody is lower than the volume 4.

When the melody is changed in the direction in which the numerical parameter of the safety score becomes larger (safer becomes higher), the voice control unit 208 adjusts the previous output state in order to prevent the melody output state from changing frequently. It may be changed after a predetermined time (for example, several seconds) has elapsed from.

As described above, in the unmanned forklift 200A according to the present embodiment, the safety score generated in real time based on the image data is used as the current score, and the safety score at the self-position in the environment map is used as the past score. Adjust the output state of the melody based on the lower safety score.

For example, if a worker enters the blind spot area in an area where the worker is frequently working, the past score will be a less safe score (small numerical parameter score) than the current score, so an unmanned forklift. The 200A adjusts the output state of the melody based on the past score. Therefore, according to the unmanned forklift 200A according to the present embodiment, it is possible to appropriately call attention to a worker in a blind spot area (for example, an area hidden by a rack or luggage and not visible from the photographing means). ..

[Second Embodiment]
FIG. 5 shows a block diagram of the cargo handling system 1B according to the second embodiment of the present invention. The cargo handling system 1B includes a management device 100B and a plurality of laser-guided unmanned forklifts 200B. The unmanned forklift 200B is common to the first embodiment except that the map generation unit 207 is not provided.

The management device 100B manages the traveling and cargo handling work of a plurality of unmanned forklifts 200B. The management device 100B is common to the first embodiment except that it includes a map generation unit 104, a collection unit 105, and a parameter update unit 106.

The map generation unit 104 has the same configuration as the map generation unit 207 included in the unmanned forklift 200A of the first embodiment. That is, when the safety score and the current location information (self-position) are input from the unmanned forklift 200B, the map generation unit 104 uses the safety score acquired from the unmanned forklift 200B as the current score, and the safety at the self-position in the environment map. The sex score is used as the past score, and the current score and the past score are compared. After the comparison, the map generation unit 104 transmits the safety score (safety score of a small numerical parameter) having the lower safety to the voice control unit 208 of the unmanned forklift 200B.

The collecting unit 105 is configured to collect image data with a safety score from the unmanned forklift 200B and generate learning data that can be input to the parameter updating unit 106 based on the image data with the safety score. ..

Further, the collecting unit 105 collects image data (image data without a safety score generated by the image data generation unit 203) from the unmanned forklift 200B, and links the image data with the safety score for learning. Data may be generated. In this case, the learning data can be generated in the same manner as the teacher data of the first embodiment. That is, the collecting unit 105 uses numerical parameters (numerical parameters 1 to 5 in this embodiment) as the safety score, and sets the safety score for each image data. For example, when traveling near the rack 3 and there is a person on the traveling route, it is judged that the safety is low and a small numerical parameter is set. On the other hand, if there are no people, it is judged that the safety is high and a large numerical parameter is set. It can be inferred that there is a certain relationship such as a correlation between the condition of the traveling route shown by the image data and the degree of safety.

The parameter update unit 106 generates update data for updating the parameters of the machine learning algorithm (for example, the weighting parameters of the neural network) included in the learning model unit 204 of the unmanned forklift 200B. The parameter update unit 106 performs machine learning based on the learning data by using a machine learning algorithm such as a neural network in order to generate the update data. The parameter update unit 106 transmits the generated update data to a plurality of unmanned forklifts 200B via the communication unit 101.

The learning model unit 204 of the unmanned forklift 200B updates the parameters of the machine learning algorithm based on the received update data. This allows the learning model unit 204 to generate a safety score based on a machine learning algorithm with updated parameters.

[Modification example]
Although the embodiment of the forklift and the cargo handling system according to the present invention has been described above, the present invention is not limited to the above embodiment.

The forklift of the present invention uses a vehicle body, an image data generation unit that captures an image including the front of the vehicle body to generate image data, and a machine learning algorithm having predetermined parameters when the image data is input. A learning model unit that has been machine-learned to generate a safety score that indicates the degree of safety of the vehicle's travel route, and a process that acquires a safety score from the learning model unit by inputting image data into the learning model unit. A unit, a voice output unit that outputs a predetermined melody while driving, a voice control unit that adjusts the output state of the melody, a position estimation unit that estimates the self-position, and an environment that shows the distribution of safety scores in the workplace. It is equipped with a map generation unit that generates a map, and the map generation unit uses the safety score obtained from the processing unit as the current score, the safety score at its own position in the environment map as the past score, and the current score and the past score. If the safety score of the lower safety is output to the voice control unit in comparison with, and the voice control unit adjusts the output state of the melody based on the safety score obtained from the map generation unit. The configuration can be changed as appropriate.

The forklift of the present invention is not limited to the laser-guided unmanned forklift, and may be an unmanned forklift of another type or a manned unmanned forklift capable of switching between manned operation and unmanned operation.

The forklift of the present invention preferably includes an autonomous traveling mechanism. Autonomous traveling mechanisms include, for example, a laser guidance mechanism or a SLAM guidance mechanism. As shown in the above embodiment, the laser guidance mechanism is a mechanism for autonomously traveling based on a self-position and a posture (posture angle) specified by using a laser scanner and a reflector provided on a wall or the like. .. The SLAM guidance mechanism is a mechanism for autonomously traveling based on the self-position and posture specified by SLAM, which estimates the self-position and creates an environmental map.

In the first embodiment, the numerical parameter of the safety score is set in the range of 1 to 5, but the range can be changed as appropriate. Further, the voice control unit 208 adjusts the output state of the melody so that the volume of the melody becomes louder as the numerical parameter of the safety score becomes smaller. The playback speed and the like may be changed.

The collecting unit 105 and the parameter updating unit 106 according to the second embodiment may be provided in the unmanned forklift 200B instead of the management device 100B. However, from the viewpoint of being able to acquire a large amount of data, it is preferable that the management device 100B includes a collecting unit 105 and a parameter updating unit 106.

1A, 1B Cargo handling system 2 Workplace 3 Rack 4 Reflector 100A, 100B Management device 101 Communication unit 102 General control unit 103 Display unit 104 Map generation unit 105 Collection unit 106 Parameter update unit 200A, 200B Unmanned forklift 201 Voice output unit 202 Alarm sound Output unit 203 Image data generation unit 204 Learning model unit 205 Processing unit 206 Position estimation unit 207 Map generation unit 208 Voice control unit 210 Body 211 Cargo handling device 212 Laser scanner

Claims (6)

A forklift that performs cargo handling work in the workplace.
With the car body
An image data generation unit that captures an image including the front of the vehicle body and generates image data,
When the image data is input, a learning model unit machine-learned to generate a safety score indicating the degree of safety of the traveling route of the vehicle body by using a machine learning algorithm having predetermined parameters, and a learning model unit.
A processing unit that acquires the safety score from the learning model unit by inputting the image data into the learning model unit, and a processing unit.
An audio output unit that outputs a predetermined melody while driving,
A voice control unit that adjusts the output state of the melody,
The position estimation unit that estimates the self-position and the position estimation unit
A map generator that generates an environmental map showing the distribution of the safety score in the workplace is provided.
The map generator
The safety score obtained from the processing unit is used as the current score, the safety score at the self-position in the environment map is used as the past score, and the current score and the past score are compared and the safety is lower. The safety score of the above is output to the voice control unit,
The voice control unit
A forklift characterized by adjusting the output state of the melody based on the safety score obtained from the map generation unit. The forklift according to claim 1, wherein the map generation unit updates the past score by using the current score after outputting the safety score to the voice control unit. It is provided with an alarm sound output unit that outputs an alarm sound louder than the melody under the control of the voice control unit.
The voice control unit is characterized in that, when the safety score is smaller than a predetermined threshold value, the voice control unit outputs the alarm sound to the alarm sound output unit instead of outputting the melody to the voice output unit. Item 2. The fork lift according to Item 1 or 2. A cargo handling system including a management device and a plurality of forklifts that perform cargo handling work in the workplace under the control of the management device.
The forklift is
With the car body
An image data generation unit that captures an image including the front of the vehicle body and generates image data,
When the image data is input, a learning model unit machine-learned to generate a safety score indicating the degree of safety of the traveling route of the vehicle body by using a machine learning algorithm having predetermined parameters, and a learning model unit.
A processing unit that acquires the safety score from the learning model unit by inputting the image data into the learning model unit, and a processing unit.
An audio output unit that outputs a predetermined melody while driving,
A voice control unit that adjusts the output state of the melody,
It is equipped with a position estimation unit that estimates its own position.
The management device is
A communication unit for acquiring the safety score and the self-position from the forklift,
A map generator that generates an environmental map showing the distribution of the safety score in the workplace is provided.
The map generator
The safety score obtained from the forklift is used as the current score, the safety score at the self-position of the forklift in the environment map is used as the past score, and the current score is compared with the past score for safety. The lower safety score is transmitted to the voice control unit of the forklift.
The voice control unit
A cargo handling system characterized in that the output state of the melody is adjusted based on the safety score acquired from the map generation unit. The processing unit transmits image data with a safety score in which the image data input to the learning model unit is associated with the safety score acquired from the learning model unit to the management device.
The management device is
A collection unit that collects image data with a safety score and generates learning data,
A parameter update unit that performs machine learning based on the learning data, generates update data for updating the parameters included in the learning model unit, and transmits the update data to the learning model unit.
Equipped with
The cargo handling system according to claim 4, wherein the learning model unit updates the parameters based on the update data. The fourth or fifth aspect of the present invention, wherein the map generation unit updates the past score by using the current score after transmitting the safety score to the voice control unit. Cargo handling system.

Images