KR102439584B1 - Apparatus and method for managing the work plan of multiple autonomous robots - Google Patents

Abstract

A management device is provided. The management device may include: an acquisition unit configured to acquire an environment variable affecting a task performed by the autonomous robot; It may include; a processing unit that processes the obtained plurality of environment variables into an input data structure of the deep neural network model.

Description

Apparatus and method for managing the work plan of multiple autonomous robots}

The present invention relates to an apparatus and method for managing a work plan of a plurality of autonomous robots input to process a plurality of tasks.

Autonomous robots that exclude the intervention of managers as much as possible are being used in various fields.

As the field of application of autonomous robots expands, the cases in which a plurality of autonomous robots are put into one place at once are increasing.

When a plurality of autonomous robots are input to process a plurality of tasks, work efficiency may be significantly different depending on the order in which the plurality of autonomous robots perform tasks.

A means for determining the work plan of a plurality of autonomous robots put into the field is needed to improve work efficiency.

Korean Patent Application Laid-Open No. 2017-0095583 discloses an apparatus and method for adaptively generating a work plan of a robot.

Korean Patent Publication No. 2017-0095583

An object of the present invention is to provide an apparatus and method for managing a plurality of autonomous robots that are put into a job site and perform a job, and a job plan.

A management device of the present invention includes: an acquisition unit configured to acquire an environment variable affecting a task performed by an autonomous robot; It may include; a processing unit that processes the obtained plurality of environment variables into an input data structure of the deep neural network model.

The management method of the present invention includes an acquiring step of acquiring an environment variable affecting a task performed by an autonomous robot; a processing step of processing the obtained plurality of environment variables into an input data structure of a deep neural network model; a establishing step of establishing a work plan of the autonomous robot corresponding to an output result of the deep neural network model to which the environment variable of the input data structure is input; and an execution step of generating a work order of the autonomous robot according to the work plan and providing information on the work order to the autonomous robot.

In the present invention, in an environment in which a plurality of work robots are input to a work site in which a plurality of tasks exist sporadically, a work plan can be established to induce optimal work efficiency.

There may be various factors that affect work efficiency. Accordingly, as the number of factors used in the work plan increases, the accuracy of the work plan may be improved.

As such, if a plurality of tasks and a plurality of robots are mixed in addition to an environment requiring various kinds of factors, the number of factors affecting work efficiency may increase exponentially.

The management apparatus and management method of the present invention can present new factors that are helpful in improving actual work efficiency.

In addition, according to the present invention, an input data structure having a grid structure and a stack structure can be presented for management of factors that increase exponentially.

The present invention can propose a method of generating learning data having an input data structure having a grid structure and a stack structure.

According to the present invention, it is possible to provide an environment in which a work plan can be established within a set constant time regardless of a change in the number of autonomous robots input to a work site. In other words, according to the present invention, a work plan can be established within a set time regardless of an increase in the number of input factors.

1 is a block diagram showing a management device of the present invention.
2 is a schematic diagram showing the structure of input data described in the present invention.
3 is a flowchart illustrating a management method of the present invention.
4 is a diagram illustrating a computing device according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art to which the present invention pertains can easily implement them. However, the present invention may be embodied in several different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

In this specification, duplicate descriptions of the same components will be omitted.

Also, in this specification, when a component is referred to as 'connected' or 'connected' to another component, it may be directly connected or connected to the other component, but other components in the middle It should be understood that there may be On the other hand, in the present specification, when it is mentioned that a certain element is 'directly connected' or 'directly connected' to another element, it should be understood that the other element does not exist in the middle.

In addition, the terms used herein are used only to describe specific embodiments, and are not intended to limit the present invention.

Also in this specification, the singular expression may include the plural expression unless the context clearly dictates otherwise.

Also, in this specification, terms such as 'include' or 'have' are only intended to designate that the features, numbers, steps, operations, components, parts, or combinations thereof described in the specification exist, and one or more It should be understood that the existence or addition of other features, numbers, steps, acts, components, parts, or combinations thereof, is not precluded in advance.

Also in this specification, the term 'and/or' includes a combination of a plurality of listed items or any of a plurality of listed items. In this specification, 'A or B' may include 'A', 'B', or 'both A and B'.

Also, in this specification, detailed descriptions of well-known functions and configurations that may obscure the gist of the present invention will be omitted.

1 is a block diagram showing a management device of the present invention.

The management apparatus shown in FIG. 1 may include an acquisition unit 110 , a processing unit 130 , a establishment unit 150 , and an execution unit 170 .

The acquisition unit 110 may acquire an environment variable that affects the work performed by the autonomous robot 10 .

An autonomous robot is a robot that moves autonomously to execute a given task, and can mainly perform transport/delivery tasks of goods in a warehouse or indoor/outdoor environment.

When there are multiple autonomous robots and multiple task goals, it is difficult to systematically allocate the optimal conditions for all autonomous robots and tasks. With the existing technology, if the number of autonomous robots and the number of tasks increase, the complexity increases and the time required for planning increases exponentially.

In the present invention, an input data structure that can solve a plan for task assignment of each autonomous robot for a plurality of autonomous robots and a plurality of tasks using a deep neural network model and a training data set for learning are generated. suggest how to Using the deep neural network model generated through the present invention, it is possible to establish a work plan within a constant constant time regardless of the number of varying robots and the number of tasks.

When there are multiple autonomous robots and multiple work goals, a work plan is required for how to allocate them between the autonomous robot and each other, and the efficiency of the work plan is a cost function (cost function) can be set.

로 트리의 폭(b)과 깊이(d)에 비례하여 급격하게 증가한다. 따라서, 기존 PDDL 방안에 따르면, 자율 로봇과 작업의 수가 어느 정도 증가하게 되면, 설정 시간 내에 작업 계획을 얻을 수 없다.Currently, one of the methods of executing such a work plan is a method of outputting a work list planned by the PDDL planner by describing the space and problems in PDDL (Planning Domain Description Language). However, the PDDL planner takes the method of searching by constructing a tree data structure for multiple autonomous robots and multiple task goals, and mainly uses the A-Star(A ^* ) algorithm, which has high complexity.

It increases rapidly in proportion to the width (b) and depth (d) of the tree. Therefore, according to the existing PDDL scheme, if the number of autonomous robots and tasks increases to a certain extent, it is impossible to obtain a task plan within the set time.

정도의 복잡도를 가지므로, 상수 시간 또는 설정 시간 내에 작업 계획을 수립할 수 있다.In the present invention, a plan for task assignment that optimizes goals such as minimum time or autonomous robot performance remaining ratio for multiple autonomous robots and multiple tasks is to be solved using a deep neural network model, To this end, we propose a method of making autonomous robot and task information into an input data structure of a deep neural network model, and a method of generating a training data set using PDDL for supervised learning of a deep neural network model. Using a deep neural network model created and trained in this way,

Since it has a degree of complexity, it is possible to establish a work plan within a constant time or a set time.

As a deep neural network model, several known models (ResNet, EfficientNet, Xception, etc.) or a multi-layer fully connected neural network can be used.

Affecting the work may mean that the work efficiency can be changed by some factors. Working efficiency may mean an amount of work per unit time. Factors that change work efficiency may correspond to environmental variables. As an example, robot information and job information corresponding to environment variables may be, for example, as follows.

The types of information include the order number displayed for each position of multiple robots, the moving speed of the robot, the ratio of the remaining loading space of the robot, the remaining performance of the robot, the order number displayed by the starting position of multiple tasks, the order number displayed by the target position of the multiple tasks, There is a checkboard where the values of 0 and 1 are crossed.

Tasks referred to here spatially have a starting position and assumes that there is either a target position or an end position depending on the course of performing the task. Depending on the task, the starting position and the target position may be different or the same.

The management device of the present invention may be used to establish a work plan of the robot put into the work site. In order to establish a work plan for improving work efficiency, the obtaining unit 110 may obtain a first factor including at least one of a load space ratio of the robot and a residual value of the robot's performance.

Acquisition unit 110 includes at least one of the robot sequence number displayed for each position of the plurality of robots, the movement speed of the robot, the operation start sequence displayed for each starting position of the operation, the end sequence displayed for each target position of the operation, and a check board 2 additional factors can be obtained.

The processing unit 130 may process the plurality of environment variables acquired through the acquisition unit 110 into an input data structure of the deep neural network model. As an example, the processing unit 130 may make robot information and job information corresponding to environment variables into an input data structure of a deep neural network model.

The input data of the deep neural network model may have various structures.

2 is a schematic diagram showing the structure of input data described in the present invention.

The processing unit 130 of the present invention may process the array structure of the environment variables into a special structure that is easy to store, process, and operate regardless of the increase in the number of environment variables.

For example, when a plurality of types of environmental variables are acquired through the acquisition unit 110 , the processing unit 130 may express the same types of environmental variables as a grid on a two-dimensional plane composed of horizontally and vertically. The processing unit 130 may generate a structure in which a plurality of two-dimensional planes in which different types of environmental variables are expressed are stacked. The processing unit 130 may provide the corresponding structure as input data of the deep neural network model.

According to the present invention, input data of a deep neural network model can express a two-dimensional spatial plane composed of horizontal and vertical as a grid, and multiple information can be expressed by stacking several spatial planes in the depth direction.

According to the present invention, the input data of the deep neural network model has a structure in which a spatial plane is stacked in multiple layers, and a single spatial plane may include the same type of data.

The input data structure shown in FIG. 2 may be detailed as follows. In this case, the position of each layer representing different types of data may be changed.

The first plane (1st Layer) includes information on the order numbers of the plurality of robots, and the corresponding information may be expressed as in Equation 1.

V _{robot's index} is information recorded on the first plane and has a real value between 0 and 1. N _robot is the number of robots. INDEX _robot is the robot's sequence number and has an integer value from 1 to N _robot .

The second layer (2nd Layer) includes information on the movement speed of the robot, and the corresponding information can be expressed as Equation (2).

V _{robot's velocity} is information recorded on the second plane. It is a value obtained by dividing the movement speed of each robot by the maximum movement speed. It has a real value between 0 and 1.

The third plane 3rd includes information about the ratio of the remaining loading space of the robot, and the corresponding information may be expressed as Equation (3).

V _{robot's loading space ration} is information recorded on the third plane, and it is a value obtained by dividing the number of remaining loading spaces for each robot or the number of luggage compartments 11 by the maximum number of loading spaces, and has a real value between 0 and 1.

The fourth layer (4th Layer) includes information on the residual performance of the robot, and the information can be expressed by Equation (4).

V _{performance residual} is information recorded on the fourth plane, and the sum of the product of the robot's battery residual value (BATTERYLEVEL _robot ) multiplied by the first weight α and the robot's maintenance residual value (MAINTENANCELEVEL _robot ) multiplied by the second weight β to have a real value between 0 and 1. The maintenance residual value may indicate the remaining life of the robot, or may indicate a period in which the next maintenance is required.

The fifth plane (5th Layer) may be expressed as in Equation 5 as a sequence number of a plurality of tasks divided by start positions. For example,

는 다섯번째 평면에 기록되는 정보로, 0과 1 사이의 실수값을 갖는다. N_job은 작업의 개수이다. INDEX_job은 작업의 순번이며 작업의 우선 순위에 따라 1부터 N_job까지의 정수값을 갖는다.

is information recorded on the fifth plane, and has a real value between 0 and 1. N _job is the number of jobs. INDEX _job is the sequence number of the job and has an integer value from 1 to N _job according to the priority of the job.

The sixth plane (6th Layer) is expressed as in Equation (6) in the order of the plurality of tasks divided by target location.

는 여섯번째 평면에 기록되는 정보이며 0과 1 사이의 실수값을 갖는다. N_job은 작업의 개수이다. INDEX_job은 작업의 순번이며 작업의 우선 순위에 따라 1부터 N_job까지의 정수값을 갖는다.

is information recorded on the sixth plane and has a real value between 0 and 1. N _job is the number of jobs. INDEX _job is the sequence number of the job and has an integer value from 1 to N _job according to the priority of the job.

For example, when a robot puts a load on the conveyor at one side of the conveyor belt and unloads a load from the conveyor belt at the other side of the conveyor belt, the position of one side of the conveyor belt may correspond to the starting position related to Equation 5. The position of the other side of the conveyor belt for unloading the load from the conveyor belt may correspond to the target position or the end position related to Equation (6).

Each of the first to sixth planes above expresses information according to the position of the x-axis and y-axis of the robot or task, and in particular, the edge of the plane is filled with a value of 1. This is expressed as Equation 7 below.

The seventh layer may be expressed as in Equation (8).

은 각 x축과 y축에 대해, x축이 홀수일 때 y축이 짝수인 경우와 x축이 짝수일 때 y축이 홀수인 경우에는 실수 1의 값을 가지고 그 외에는 0의 값을 갖는다.

For each x-axis and y-axis, when the x-axis is odd, the y-axis is even, when the x-axis is even, the real number 1 when the y-axis is odd, and 0 otherwise.

The eighth plane (8th layer) can be expressed as in Equation (9).

는 각 x축과 y축에 대해, x축이 홀수일 때 y축이 홀수인 경우와 x축이 짝수일 때 y축이 짝수인 경우에는 실수 1의 값을 가지고 그 외에는 0의 값을 갖는다.

For each x-axis and y-axis, when the x-axis is odd, the y-axis is odd, when the x-axis is even, the real number 1 when the y-axis is even, and 0 otherwise.

The processing unit 130 may generate the input data structure of FIG. 2 by using the results of the mathematical models of Equations 1 to 9.

A generator for generating a set of learning data for supervised learning of a deep neural network model may be provided. The generator may virtually create an environment variable. The generating unit may be formed integrally with the obtaining unit 110 .

A Planning Domain Definition Language (PDDL) planner in which a problem of planning a plurality of robots and a plurality of tasks is defined may be provided.

The generator may generate output data while varying the number of robots and the number of tasks through the PDDL planner. The processing unit 130 may process the output data of the generator into an input data structure of the deep neural network model.

In order to train a deep neural network model that establishes a work plan by a method of supervised learning, a Training Dataset is required. PDDL is an abbreviation of Planning Domain Definition Language. With PDDL, various environments and problems to be solved can be defined, and the order of tasks can be determined and output with the PDDL Planner.

Therefore, by defining the problem of planning multiple robots and multiple tasks with PDDL, it generates various output data while varying the number of robots and the number of tasks through the PDDL planner, and processes it to create a training data set for a deep neural network model. can

A training data set may be generated according to the procedure below.

The generation unit includes serial numbers of a plurality of virtual robots arranged in a three-dimensional space, position coordinates of the virtual robots, moving speed of the virtual robots, information on the loading space of the virtual robots (maximum number of loads, the number of remaining loads or the number of empty luggage compartments), The performance residual value (battery residual value, maintenance residual value) of the virtual robot, the serial number of a plurality of virtual tasks existing in the three-dimensional space, the starting position of the virtual task, the target position of the virtual task are defined and the actual value is described. A Planning Domain Definition Language (PDDL) planner may be provided. The virtual robot may be a projection of a real robot on a virtual three-dimensional space, and the virtual task may be a projection of a real task on a virtual three-dimensional space.

Average information of the virtual robot may be defined in advance. For example, the coordinates of the robot are sequentially (0.5, 0.5, 0.0) in the space formed by the three orthogonal coordinate axes x, y, and z axes, the average speed is 1.0 m/s, and the maximum number of loads is 4 and the number of remaining loads is 3, and the residual performance may be set to 0.9, etc.

The generator may generate the number of virtual robots N _robot and the number of jobs N _job as random numbers, respectively.

The generator may assign serial numbers of the virtual robots as many as the number of virtual robots. As an example, the generator may assign serial numbers from 1 to N _robot to the N _robot virtual robots.

The generating unit generates load space information or performance residual value for the virtual robot as a value obtained by adding a normal distribution to the average information of a plurality of predefined virtual robots,

The generator may generate coordinates (x _robot , y _robot ) of the virtual robot as a value obtained by adding a uniform distribution to the average information of the virtual robot.

The generator may assign serial numbers of the virtual tasks as many as the number of virtual tasks. As an example, the generator may assign serial numbers from 1 to N _jobs to N _job virtual jobs.

The generator may generate a start position of a virtual job and a target position (x _job , y _job ) of the virtual job by substituting the values of the uniform distribution.

The generation unit may input data (serial numbers, locations, etc.) generated for the plurality of virtual robots and the plurality of virtual tasks to the PDDL planner.

The PDDL planner may output a work plan of the virtual robot based on the input data.

The processing unit 130 may generate a vector space in which the length of one side is the number of virtual robots and the length of the other side is expressed as the number of virtual tasks.

The processing unit 130 may encode and express the serial number of the virtual task selected for each virtual robot and the serial number of the selected robot for each virtual task on different coordinates of the vector space.

The processing unit 130 may convert the definitions of the plurality of virtual robots and the plurality of virtual tasks into an input data structure.

The processing unit 130 may output the vector space in which the serial number of the virtual work and the serial number of the virtual robot are encoded and expressed and the input data structure together as a file.

After that, it returns to the random number generation of the virtual robot and the random number generation of the virtual task, and the operation thereafter may be repeated as much as the size of the specified training data set.

The establishment unit 150 and the execution unit 170 may be used in the field.

Various environmental variables affecting work efficiency may be obtained by the obtaining unit 110 through various mathematical models that can be extracted at the work site. The corresponding environment variable may be converted into an input data structure of a deep neural network model that establishes a work plan by the processing unit 130 .

The establishment unit 150 may be provided with a deep neural network model in which learning is performed through a set of learning data generated by the generation unit.

The establishment unit 150 may generate and output a work plan applicable to the field through the deep neural network model to which the environmental variables of the field are input. The work plan may include the work sequence of a specific autonomous robot, the amount of work or load in the setting stage, and the like.

When the work plan is established by the establishment unit 150 , the execution unit 170 may divide and deliver the established work plan for each of a plurality of robots input to the work site. The serial number assigned to each autonomous robot may be used for distribution of each work plan. The corresponding serial number may correspond to the sequence number of one of the environment variables, and the like.

3 is a flowchart illustrating a management method of the present invention. The management method shown in FIG. 3 may be performed by the management device of FIG. 2 .

The management method of the present invention may include an acquisition step (S510), a processing step (S520), an establishment step (S530), and an execution step (S540).

The obtaining step S510 may obtain an environment variable that affects the work performed by the autonomous robot. The acquiring step ( S510 ) may be performed by the acquiring unit 110 .

In the processing step ( S520 ), the obtained plurality of environment variables may be processed into an input data structure of the deep neural network model. The processing step ( S520 ) may be performed by the processing unit 130 .

The establishment step ( S530 ) may establish a work plan of the autonomous robot corresponding to the output result of the deep neural network model to which the environmental variables of the input data structure are input. The establishment step ( S530 ) may be performed by the establishment unit 150 .

In the execution step S540, a work order of the autonomous robot may be generated according to the work plan, and information on the work order may be provided to the autonomous robot. The execution step ( S540 ) may be performed by the execution unit 170 .

4 is a diagram illustrating a computing device according to an embodiment of the present invention. The computing device TN100 of FIG. 4 may be a device (eg, a management device, etc.) described herein.

In the embodiment of FIG. 4 , the computing device TN100 may include at least one processor TN110 , a transceiver device TN120 , and a memory TN130 . Also, the computing device TN100 may further include a storage device TN140 , an input interface device TN150 , an output interface device TN160 , and the like. Components included in the computing device TN100 may be connected by a bus TN170 to communicate with each other.

The processor TN110 may execute a program command stored in at least one of the memory TN130 and the storage device TN140. The processor TN110 may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to an embodiment of the present invention are performed. The processor TN110 may be configured to implement procedures, functions, and methods described in connection with an embodiment of the present invention. The processor TN110 may control each component of the computing device TN100 .

Each of the memory TN130 and the storage device TN140 may store various information related to the operation of the processor TN110 . Each of the memory TN130 and the storage device TN140 may be configured as at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory TN130 may include at least one of a read only memory (ROM) and a random access memory (RAM).

The transceiver TN120 may transmit or receive a wired signal or a wireless signal. The transceiver TN120 may be connected to a network to perform communication.

On the other hand, the embodiment of the present invention is not implemented only through the apparatus and/or method described so far, and a program for realizing a function corresponding to the configuration of the embodiment of the present invention or a recording medium in which the program is recorded may be implemented. And, such an implementation can be easily implemented by those skilled in the art from the description of the above-described embodiments.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concept of the present invention as defined in the following claims are also present. It belongs to the scope of the invention.

10...Robot 11...Luggage compartment
110...Acquisition Department 130...Processing Department
150...Establishment Unit 170...Execution Unit

Claims (6)

an acquisition unit configured to acquire an environment variable affecting a task performed by the autonomous robot;
A processing unit that processes the obtained plurality of environment variables into an input data structure of a deep neural network model;
When a plurality of types of the environment variables are obtained through the obtaining unit, the processing unit expresses the same type of environment variables as a grid on a two-dimensional plane composed of horizontal and vertical,
The processing unit creates a structure in which a plurality of two-dimensional planes in which different types of environmental variables are expressed are stacked in a depth direction,
The processing unit provides the structure as input data of the deep neural network model,
By the processing unit, the input data has a structure in which the two-dimensional plane is stacked in a plurality of layers,
The single two-dimensional plane contains the same kind of environment variables,
Each stacked two-dimensional plane is a management device containing different kinds of environmental variables.
The method of claim 1,
For a work environment in which a plurality of robots process a plurality of tasks, the obtaining unit obtains a first factor including at least one of a loading space ratio of the robot and a residual performance value of the robot,
The acquisition unit includes at least one of a robot sequence number displayed for each position of the plurality of robots, a movement speed of the robot, a task start sequence displayed for each start position of the task, an end sequence displayed for each target position of the task, and a check board A management device for further acquiring a second factor.
The method of claim 1,
One of the plurality of two-dimensional planes includes information about the order number of the plurality of robots,
One of the plurality of two-dimensional planes includes information about the moving speed of the plurality of robots,
One of the plurality of two-dimensional planes includes information about the ratio of the remaining loading space of the plurality of robots,
One of the plurality of two-dimensional planes includes information about residual performance values of the plurality of robots,
One of the plurality of two-dimensional planes includes a sequence number of a plurality of tasks divided by start positions of the plurality of robots,
One of the plurality of two-dimensional planes is a management device including a sequence number of a plurality of tasks divided by target positions.
The method of claim 1,
A generator for generating a set of learning data for supervised learning of the deep neural network model;
A Planning Domain Definition Language (PDDL) planner in which the problem of planning multiple robots and multiple tasks is defined is prepared,
The generator generates output data while varying the number of robots and the number of tasks through the PDDL planner,
The processing unit is a management device for processing the output data of the generating unit into the input data structure of the deep neural network model.
The method of claim 1,
A generator for generating a set of learning data for supervised learning of the deep neural network model;
The generating unit includes serial numbers of a plurality of virtual robots arranged in a three-dimensional space, position coordinates of the virtual robot, moving speed of the virtual robot, loading space information of the virtual robot, performance residual value of the virtual robot, 3D A Planning Domain Definition Language (PDDL) planner in which serial numbers of a plurality of virtual tasks existing in space, a start position of the virtual task, and a target position of the virtual task are defined is provided,
The generator generates the number of virtual robots and the number of tasks as random numbers, respectively,
The generating unit gives the serial number of the virtual robot as much as the number of the virtual robot,
The generating unit generates load space information or a performance residual value for the virtual robot as a value obtained by adding a normal distribution to the predefined average information of the plurality of virtual robots,
The generating unit generates the coordinates of the virtual robot by adding a uniform distribution to the average information of the virtual robot,
The generating unit gives a serial number of the virtual task as much as the number of the virtual task,
The generating unit generates a start position of the virtual task and a target position of the virtual task by substituting the values of the uniform distribution,
The generating unit inputs the data generated for the plurality of virtual robots and the plurality of virtual tasks to the PDDL planner,
The PDDL planner outputs the work plan of the virtual robot based on the input data,
The processing unit generates a vector space in which the length of one side is the number of virtual robots and the length of the other side is expressed by the number of virtual tasks,
The processing unit encodes and expresses the serial number of the virtual task selected for each virtual robot and the serial number of the selected robot for each virtual task on different coordinates of the vector space,
The processing unit converts the definitions for the plurality of virtual robots and the plurality of virtual tasks into the input data structure,
The processing unit is a management device for outputting the vector space in which the serial number of the virtual work and the serial number of the virtual robot are encoded and expressed and the input data structure together as a file.
A management method performed by a management device, comprising:
an acquiring step of acquiring an environment variable that affects the work performed by the autonomous robot;
a processing step of processing the obtained plurality of environment variables into an input data structure of a deep neural network model;
a establishing step of establishing a work plan of the autonomous robot corresponding to an output result of the deep neural network model to which the environment variable of the input data structure is input;
an execution step of generating a work order of the autonomous robot according to the work plan and providing information on the work order to the autonomous robot;
The processing step is
When a plurality of types of the environment variables are obtained, the same types of environment variables are expressed as a grid on a two-dimensional plane composed of horizontally and vertically,
Creates a structure in which a plurality of two-dimensional planes in which different types of environment variables are expressed are stacked in the depth direction,
The input data processed through the processing step has a structure in which the two-dimensional plane is stacked in a plurality of layers,
The single two-dimensional plane includes the same kind of environmental variables, and each stacked two-dimensional plane includes different kinds of environmental variables.

Images