WO2023119739A1

WO2023119739A1 - Optimization device and optimization method

Info

Publication number: WO2023119739A1
Application number: PCT/JP2022/033298
Authority: WO
Inventors: 文也工藤; 史子紅山; 契宇都木; 誠也伊藤; 統宙月舘
Original assignee: 株式会社日立製作所
Priority date: 2021-12-21
Filing date: 2022-09-05
Publication date: 2023-06-29
Also published as: JP2023092308A

Abstract

This optimization device, which optimizes an sequence of work performed using a plurality of autonomous bodies, comprises: an optimization information input unit for inputting optimization information that includes initial order information representing an initial value for a work sequence planned for the work, variation factor information representing a factor for a variation in the work sequence, and work environment information representing a work environment in which the work is performed; an optimization subject updating unit for determining, on the basis of the initial order information and the variation factor information, an optimization subject order representing a work sequence to be optimized; a search unit for searching for an optimal solution for the work sequence, on the basis of the work environment information and the optimization subject order; and an optimal solution output unit for outputting the optimal solution retrieved by the search unit.

Description

Optimization device, optimization method

The present invention relates to an apparatus and method for optimizing the order of work performed using multiple autonomous bodies.

In recent years, in warehouses and factories, in addition to human workers, robots (hereinafter collectively referred to as “autonomous bodies”) that operate autonomously and perform predetermined work in accordance with a work order set in advance ) is being promoted. In the operation of such an autonomous body, there is a need for technology that realizes operational efficiency in response to the growing needs for operational efficiency.

For example, the technology of Patent Document 1 is known for improving the operational efficiency of autonomous bodies. In Patent Document 1, attention is paid to the movement vector of the centroid vector of the group with high fitness in the solution vector set, and if the movement vectors point in the same direction, it is assumed that there is a solution vector with high fitness in that direction. A genetic algorithm that has excellent global solution update capability by making a decision and updating the solution vector group along the vector, and simultaneously optimizing the solution vector by recombination operation targeting the current solution vector set. An optimization adjustment method is disclosed in which the direction of the optimum solution can be estimated at high speed using the history of past solution vector updates while making use of the characteristics of .

Japanese Patent Application Laid-Open No. 2002-251600

In the technology of Patent Document 1, when the divergence between the work results of the autonomous body and the plan becomes large due to changes in the work environment and the plan needs to be revised, the plan is revised using the calculated optimal solution. is difficult to do. Therefore, it is necessary to re-estimate the optimum solution, and there is a problem that the work order cannot be efficiently optimized according to changes in the work environment.

The present invention has been made in view of the above, and its main purpose is to efficiently optimize the order of work performed using a plurality of autonomous bodies according to changes in the work environment.

An optimization device according to the present invention is a device for optimizing the order of work performed using a plurality of autonomous bodies, comprising: initial order information representing an initial value of the work order planned for the work; an optimization information input unit for inputting optimization information including variation factor information representing a variation factor of work order and work environment information representing a work environment in which the work is performed; the initial order information and the variation factor; an optimization target update unit that determines an optimization target order representing a work order to be optimized based on information; and a search that searches for an optimal solution for the work order based on the work environment information and the optimization target order. and an optimum solution output unit for outputting the optimum solution searched by the search unit.
An optimization method according to the present invention is a method for optimizing the order of work performed using a plurality of autonomous bodies, comprising: initial order information representing an initial value of the work order planned for the work; optimization information including variable factor information representing a variable factor of work order and work environment information representing a work environment in which the work is performed is input to a computer, and the computer processes the initial order information and the variable factor; determining an order to be optimized representing a work order to be optimized based on the information, searching for an optimal solution of the work order based on the work environment information and the order to be optimized by the computer; and outputting the optimum solution from the computer.

According to the present invention, the order of work performed using a plurality of autonomous bodies can be efficiently optimized according to changes in the work environment.

FIG. 1 is a block diagram showing the configuration of an optimization device according to one embodiment of the present invention. FIG. 2 is a diagram showing a specific example of initial order information. FIG. 3 is a diagram showing a specific example of optimization parameters. FIG. 4 is a diagram showing a specific example of warehouse static basic information. FIG. 5 is a flow chart showing the processing flow of the optimization device according to one embodiment of the present invention. FIG. 6 is a diagram showing a specific example of the optimization information input screen. FIG. 7 is a flowchart showing the flow of search processing. FIG. 8 is a flow chart showing the flow of the environmental variation calculation process. FIG. 9 is a diagram showing a specific example of the environmental variation visualization screen. FIG. 10 is a flowchart showing the flow of neighborhood solution generation processing. FIG. 11 is a diagram showing a specific example of the search status screen.

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an optimization device according to one embodiment of the present invention. The optimization device 100 according to this embodiment is a computer having a central processing unit 1, a main memory device 2, a secondary memory device 3, an input device 4 and an output device 5. These devices are connected via a bus 6. are connected to each other.

The central processing unit 1 is configured using, for example, a CPU (Central Processing Unit), and performs various processes and calculations for operating the optimization device 100 . The central processing unit 1 executes the programs stored in the main storage device 2 and the secondary storage device 3 to perform an optimization information input unit 11, an optimization target update unit 12, a search unit 13, and an optimal solution output unit 14. realize each functional block. Details of these functional blocks will be described later. Note that some or all of the functions of the central processing unit 1 may be implemented using a device other than the CPU, such as a GPU (Graphic Processing Unit), FPGA (Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit), etc. good.

The main storage device 2 is configured using a high-speed and volatile storage device such as a DRAM (Dynamic Random Access Memory), and is used as a work area when the central processing unit 1 executes programs.

The secondary storage device 3 is configured using, for example, a magnetic storage device such as a HDD (Hard Disk Drive) or a large-capacity and non-volatile storage device such as an SSD (Solid State Drive). and various information used in the processing of the central processing unit 1 are stored. The information stored in the secondary storage device 3 includes initial order information 31, work progress information 32, additional order information 33, calculation time specification information 34, optimization parameters 35, warehouse static basic information 36, calculation history information 37. , best order information 38 are included. Details of these pieces of information will be described later.

The input device 4 is a device that detects an input operation from a user who uses the optimization device 100, and is implemented using, for example, a keyboard, a mouse, and the like. The content of the user's input operation detected by the input device 4 is transmitted to the central processing unit 1 and reflected in the processing and calculations performed by the central processing unit 1 .

The output device 5 is a device that presents the results of processing and calculations performed by the optimization device 100 to the user who uses the optimization device 100, and is realized using, for example, a display device, a printer, or the like. The presentation contents on the output device 5 are controlled by the central processing unit 1 .

A terminal connected to the optimization device 100 via a network may be used instead of the input device 4 and the output device 5. In this case, a communication interface (not shown) of the optimization device 100 receives from the terminal communication information indicating the content of the input operation performed by the user using the terminal, and transmits the processing result of the optimization device 100 to the terminal. , can be output to a terminal and presented to the user.

Also, the configuration of the optimization device 100 shown in FIG. 1 may be physically constructed on one computer, or may be distributed and constructed on a plurality of computers. Furthermore, logical partitions may be physically configured on one or more computers, and each configuration of the optimization device 100 may be constructed on these logical partitions.

Next, the functional blocks of the optimization information input unit 11, the optimization target update unit 12, the search unit 13, and the optimal solution output unit 14 in the central processing unit 1 will be described. In the optimization device 100 of the present embodiment, the central processing unit 1 operates as these functional blocks, so that the order of work performed using a plurality of autonomous bodies in warehouses, factories, etc. is determined according to the work environment. Search for the optimum work order by changing. As a result, even when the work environment changes, the original work order can be used to efficiently obtain the optimum work order.

The optimization information input unit 11 inputs information necessary for the optimization device 100 to obtain the optimum work order according to the work environment. Specifically, initial order information 31, work progress information 32, additional order information 33, calculation time specification information 34, optimization parameters 35, warehouse static basic information 36, and calculation history stored in the secondary storage device 3. By reading each information of the information 37, these pieces of information are input. It is assumed that these pieces of information are pre-stored in the secondary storage device 3 before the optimization device 100 starts processing.

The optimization target update unit 12 updates the initial order information 31 representing the initial value of the work order planned in advance and the factors of fluctuation of the work order from the initial value among the information input by the optimization information input unit 11. Based on the work progress information 32 and the additional order information 33 respectively represented, an optimization target order representing the work order to be optimized is determined.

The search unit 13 retrieves the warehouse static basic information 36 representing the work environment in which the work is performed, among the information input by the optimization information input unit 11, and the optimization target determined by the optimization target updating unit 12. Search for the optimal work order solution based on the order. For example, a work order that maximizes the overall work efficiency of a plurality of autonomous bodies is searched for as an optimal solution.

The optimum solution output unit 14 outputs the optimum solution found by the search unit 13. The optimum solution output by the optimum solution output unit 14 is presented to the user by the output device 5 . In this way, it is possible to inform the user of the optimum work order according to the work environment after the change, and to provide useful information when the user reviews the work plan.

Next, the details of the information stored in the secondary storage device 3 will be explained. In the following, an example of optimizing the order of warehouse picking operations performed by a plurality of autonomous bodies will be described. Note that the autonomous body may be a robot or a human operator. Moreover, these may be mixed and operated.

The initial order information 31 is information representing the initial value of the work order planned in advance. FIG. 2 is a diagram showing a specific example of the initial order information 31. As shown in FIG. As shown in FIG. 2, the initial order information 31 is represented by table data combining a plurality of records having, for example, shipping date/time 311, SKU (Stock Keeping Unit) 312, and quantity 313 fields. A record of the initial order information 31 is set for each work unit.

The shipping date and time 311 stores information on the scheduled time for shipping the product from the warehouse. Each autonomous body picks up and ships a designated product from a shelf in the warehouse according to the order of times indicated by the shipping date and time 311 . The SKU 312 stores a unique ID that identifies the type of each product for inventory management. By storing the product ID in the SKU 312, the product to be shipped is designated. The number 313 stores the number of products to be shipped.

The configuration of the initial order information 31 shown in FIG. 2 is an example, and other information such as the ID of the autonomous body (robot or worker) in charge of each task may be included in the initial order information 31. Arbitrary information can be included in the initial order information 31 according to the operation mode of the warehouse, the size of the products to be handled, and the like. Each autonomous body performs warehouse picking work according to the initial order information 31 or an order list containing the same content.

The work progress information 32 is information representing the current progress of work with respect to the initial value of the work order represented by the initial order information 31, that is, the work order at the time of planning. The additional order information 33 is information representing work added after the initial value of the work order represented by the initial order information 31 . All of these pieces of information are information representing factors of variation in the work order, and hence are sometimes collectively referred to as "variation factor information" below.

The work progress information 32 and the additional order information 33 can be represented by table data similar to the initial order information 31 shown in FIG. 2, for example. Specifically, the work progress information 32 can be represented by extracting, for example, records corresponding to work that has already been performed from among the plurality of records that constitute the initial order information 31 . Further, the additional order information 33 can be represented by setting a record for each added work and setting the same fields as the initial order information 31 in each record.

It should be noted that the initial order information 31, the work progress information 32, and the additional order information 33 do not necessarily need to be expressed in the above table format. For example, these information may be described on a text basis, or may be expressed in other formats.

The calculation time specification information 34 is information representing the calculation time that can be used by the optimization device 100 to find the optimum solution for the work order. For this calculation time, for example, a value specified in advance by the user is set.

The optimization parameter 35 is information representing an optimization method and its parameter values that can be used when the optimization device 100 finds the optimum solution for the work sequence. FIG. 3 is a diagram showing a specific example of the optimization parameters 35. As shown in FIG. As shown in FIG. 3, the optimization parameter 35 is configured by combining, for example, a technique 351 and parameter information 352 consisting of a plurality of items.

The method 351 is information representing the type of optimization method. Here, information representing various well-known optimization methods such as "GA" representing genetic algorithm, "BO" representing Bayesian optimization, and "RL" representing reinforcement learning can be recorded. The parameter information 352 is information representing parameter values used in the optimization technique specified in the technique 351 . In the example of FIG. 3, the parameter information 352 includes parameter values of a crossover rate 3521, a mutation rate 3522, and a selection method 3523 as parameter values when using a genetic algorithm (GA).

The contents of the parameter information 352 are not limited to those shown in FIG. 3, and various parameter values can be included in the parameter information 352 according to the type of optimization method. For example, when the optimization method is Bayesian optimization (BO), it is conceivable to set the number of search epochs, the acquisition function, and the like in the parameter information 352 . Further, each parameter value of the parameter information 352 can be individually specified in detail by the user, or can be set to be automatically selected.

The warehouse static basic information 36 is information representing the work environment of a warehouse where picking work is carried out by multiple autonomous bodies. FIG. 4 is a diagram showing a specific example of the warehouse static basic information 36. As shown in FIG. As shown in FIG. 4, the warehouse static basic information 36 includes, for example, table data of an autonomous body parameter 361, product placement 362, product master 363, and warehouse layout 364. FIG.

The autonomous body parameter 361 is information about the performance of the autonomous body that performs the picking work, and is configured by combining each parameter of the number of autonomous bodies 3611, maximum speed 3612, and rotation speed 3613, for example. The number of autonomous bodies 3611 represents the number of autonomous bodies, the maximum speed 3612 represents the maximum speed that the autonomous body can generate during work, and the rotation speed 3613 represents the rotation speed when the autonomous body rotates during work. there is Note that the contents of the autonomous body parameters 361 are not limited to those shown in FIG. 4, and various parameter values can be included in the autonomous body parameters 361 .

The product placement 362 is information that represents the placement of each product picked by the autonomous body in the warehouse. The product arrangement 362 is represented by table data combining a plurality of records having fields of shelf ID 3621, column ID 3622, stage ID 3623, and SKU 3624, for example. A record of product placement 362 is set for each type of product for inventory management.

The shelf ID 3621 stores an ID that identifies the shelf on which each product is placed. The row ID 3622 and the stage ID 3623 store IDs that specify the row and stage of the shelf on which each product is placed. The SKU 3624 stores an ID that identifies the type of each product. That is, in the product arrangement 362, the location of each product is specified by combining the values of the shelf ID 3621, row ID 3622 and stage ID 3623, and the product type (SKU) is specified by the value of SKU 3624.

The product master 363 is information representing attributes of all products stored in the warehouse. The product master 363 is represented by table data combining a plurality of records having SKU 3631, size 3632, weight 3633, and shipping frequency 3634 fields, for example. A record of the product master 363 is set for each type of product for inventory management.

The SKU 3631 stores an ID that identifies the type of each product, similar to the SKU 3624 of the product placement 362 . The size 3632, weight 3633, and shipping frequency 3634 store the size, weight, and shipping frequency of each product as information representing attributes of each product.

The warehouse layout 364 is information representing the layout such as the arrangement of shelves in the warehouse. The warehouse layout 364 is represented by table data combining a plurality of records having, for example, x-coordinate 3641, y-coordinate 3642, status 3643, and object ID 3644 fields. A record of the warehouse layout 364 is set for each layout element that defines the layout in the warehouse.

The x-coordinate 3641 and y-coordinate 3642 store coordinate values representing the position of each layout element in the warehouse. The status 3643 stores information representing the attributes of each layout element such as "passage", "shelf", "moving object", and "exit". The object ID 3644 stores a unique ID that identifies each layout element.

It should be noted that the static warehouse basic information 36 shown in FIG. 4 is an example, and the static warehouse basic information 36 can be configured by combining arbitrary information other than this. In addition, in this embodiment, an example of optimizing the order of warehouse picking work performed by a plurality of autonomous bodies is described. However, when optimizing the work order for other types of work, instead of the warehouse static basic information 36, the work environment can be represented by information according to the work in question. is preferred. That is, the warehouse static basic information 36 of the present embodiment corresponds to an example of work environment information representing the work environment in which work is performed, and various other work environment information is used to optimize the work order. It can represent the work environment in which the work to be done is performed.

The calculation history information 37 is information representing the calculation history when the optimization device 100 obtained the optimum work order in the past. This includes the contents of the past optimization parameters 35 and warehouse static basic information 36, and the past evaluation values of each optimum solution candidate calculated when searching for the optimum solution using these.

The optimum order information 38 is information representing the optimum work order solution obtained by the optimization device 100 .

In the optimization device 100 of the present embodiment, the initial value of the work order of the picking work represented by the initial order information 31 is changed based on the above-described variation factor information (work progress information 32 and additional order information 33). Thus, the work order to be optimized is determined. Then, based on the work environment represented by the warehouse static basic information 36, search processing is performed according to the calculation time specification information 34 and the optimization parameters 35, thereby searching for the optimum work order solution, and obtaining the optimum order information. 38 is determined. This optimizes the order of warehouse picking operations performed by a plurality of autonomous bodies.

Next, the processing contents of the optimization device 100 will be described below with reference to FIGS. 5 to 11. FIG. FIG. 5 is a flow chart showing the processing flow of the optimization device 100 according to one embodiment of the present invention. The optimization device 100 executes the processing shown in the flowchart of FIG. 5 at predetermined intervals by executing a predetermined program in the central processing unit 1 .

In step S10, the optimization information input unit 11 of the central processing unit 1 inputs optimization information for obtaining the optimum solution. Here, as the optimization information, the above-mentioned information stored in the secondary storage device 3, that is, the initial order information 31, the work progress information 32, the additional order information 33, the calculation time specification information 34, the optimization parameters 35, the warehouse The static basic information 36 and the calculation history information 37 are read from the secondary storage device 3 and stored in the main storage device 2 .

In step S20, the optimization target updating unit 12 of the central processing unit 1 determines whether the work progress and additional orders are empty based on the optimization information input in step S10. The work progress indicated by the work progress information 32 is empty, that is, the work progress is not registered in the work progress information 32, and the additional order indicated by the additional order information 33 is empty, that is, the additional work is not performed in the additional order information 33. If not registered, the initial value of the work order represented by the initial order information 31 is directly determined as the work order to be optimized, and its contents are output as the order to be optimized 41, and the process proceeds to step S40. On the other hand, if at least one of the work progress and the additional order is not empty, the process proceeds to step S30.

In step S30, the optimization target update unit 12 of the central processing unit 1 updates the optimization target work order based on the optimization information input in step S10. Specifically, with respect to the initial value of the work order represented by the initial order information 31, the variable factor of the work order represented by the variable factor information included in the optimization information, that is, the work progress represented by the work progress information 32, and the additional By reflecting the content of the additional order indicated by the order information 33, the work order is changed from the initial value. Then, the work order after the change is determined as the work order to be optimized, and its content is output as the optimization target order 41 . As a result, the initially planned work order can be updated in consideration of the post-planning fluctuation factors, and the updated work order can be used as an optimization target for subsequent search processing. Note that the order of work to be optimized represented by the order to be optimized 41 determined here may be any one in which the variable factor represented by the variable factor information is reflected in the initial value of the work order according to a predetermined rule. Efficiency need not be considered.

In step S40, the search unit 13 of the central processing unit 1, based on the optimization information input in step S10 and the optimization target order 41 output from the optimization target update unit 12 in step S20 or S30, A search process is performed to search for the optimum solution for the work order. Details of the search processing performed in step S40 will be described later.

In step S50, the optimal solution output unit 14 of the central processing unit 1 outputs the optimal solution found by the search process in step S40. Here, optimum order information 38 is generated based on the content of the optimum solution found, and the generated optimum order information 38 is stored in the secondary storage device 3 and output to the output device 5 for presentation to the user.

After executing the process of step S50, the central processing unit 1 ends the process shown in the flowchart of FIG.

FIG. 6 is a diagram showing a specific example of the optimization information input screen displayed on the output device 5 when inputting the optimization information in step S10 of FIG. The optimization information input screen 110 shown in FIG. 6 includes an optimization work registration field 111, a set value input field 112, and an execution button 113. As shown in FIG.

The optimization work registration field 111 is a part for the user to specify the type of work to be optimized. For example, when the work order is optimized and the work efficiency is improved, the user checks the “work order” column in the optimized work registration column 111 . In addition, product placement, travel layout, patrol routes of autonomous bodies, and the like can be designated as optimization targets in the optimization work registration field 111 .

The setting value input field 112 is a part for specifying the calculation time value when performing optimization, the target improvement rate for determining the end of calculation, the presence or absence of progress orders and additional orders, and the like. A user who intends to optimize work in a warehouse or factory can use an input screen such as that shown in FIG. 6 to input information necessary for optimization into the optimization device 100 and execute optimization. The contents designated here are reflected in the calculation time designation information 34 and the optimization parameters 35 .

The execution button 113 is a part for reflecting the contents specified in the optimization job registration field 111 and the setting value input field 112 in the optimization information and starting execution of optimization. By operating the execution button 113 on the optimization information input screen 110, the user confirms the contents of the optimization information input to the optimization device 100 in step S10, and causes the optimization device 100 to perform the processing after step S20. can be executed.

In the process of step S10 in FIG. 5, the user can determine the content of the optimization information to be input via the optimization information input screen 110 in FIG. 6, for example. However, it is not always necessary to use a visualized input screen like the optimization information input screen 110 when inputting the optimization information. For example, it is also possible to input the optimization information by inputting the information into a CUI (Character User Interface) console, or by filling in a setting file. In addition to this, the optimization information can be input by any method in step S10.

FIG. 7 is a flowchart showing the flow of search processing executed in step S40 of FIG.

In step S410, the search unit 13 uses the optimization parameter 35, the static warehouse basic information 36, and the calculation history information 37 included in the optimization information input in step S10 of FIG. conduct. In this environmental change amount calculation process, the difference between the contents of the past optimization parameters 35 and the static warehouse basic information 36 represented by the calculation history information 37 and the contents of the current optimization parameters 35 and the contents of the static basic warehouse information 36 is calculated. By doing so, the amount of change in the work environment based on a certain point in the past is obtained, and the result of the calculation is output as the amount of environmental change 42 . Details of the environmental variation calculation process will be described later.

In step S420, the search unit 13 visualizes the environmental change amount based on the environmental change amount 42 calculated by the environmental change amount calculation process in step S410. The visualized amount of environmental change is presented to the user by being output to the output device 5 . An example of the display screen of the output device 5 at this time will be described later.

At step S430, the search unit 13 determines whether or not it is necessary to search again for the neighborhood solution based on the value of the environmental change amount 42 calculated at step S410. Here, for example, the value of the environmental change amount 42 is compared with a predetermined reference value, and if the value of the environmental change amount 42 is less than the reference value, it is determined that re-searching of the neighborhood solution is unnecessary, and the process proceeds to step S440. On the other hand, if the value of the environmental change amount 42 is equal to or greater than the reference value, it is determined that it is necessary to re-search the neighborhood solution for the optimization target order 41 set in step S20 or S30 of FIG. Proceed to S450.

At step S440, the search unit 13 generates an initial solution for the work order. Here, for example, by setting a random work order regardless of the optimization target order 41, an initial solution for the work order is generated. Note that any method can be used to perform the processing of step S440 as long as the initial solution of the work order can be generated.

In step S450, the search unit 13 performs neighborhood solution generation processing using the environment change amount 42 calculated by the environment change amount calculation processing in step S410 and the optimization target order 41. In this neighborhood solution generation process, by changing the order of work expressed by the optimization target order 41 based on the environment change amount 42, a neighborhood solution 43 for the optimization target order 41 is generated and output. Details of the neighborhood solution generation process will be described later.

After the processing of step S440 or S450 is executed, the initial solution generated in step S440 or the neighborhood solution 43 generated in step S450 is set as the optimum solution candidate, and the loop processing of steps S460 to S500 described below is performed.

In step S460, the search unit 13 calculates the evaluation value of the optimum solution candidate based on the optimization parameters 35 and the warehouse static basic information 36. Here, for example, a simulator is used to calculate the evaluation value by calculating a throughput value indicating the degree of work efficiency when the optimum solution candidate is applied based on the optimization parameter 35 and the warehouse static basic information 36. conduct. The evaluation value calculated here is a value that indicates the degree of superiority or inferiority of the optimum solution candidate with respect to the objective of optimization, and the content differs depending on the type of task designated as the optimization target. For example, when performing optimization processing with the objective of minimizing the total moving distance of the autonomous body, the total moving distance of the autonomous body may be calculated as the evaluation value. In addition to this, any physical quantity or information quantity can be calculated as the evaluation value of the optimum solution candidate.

In step S470, the search unit 13 secondary stores the optimization parameters 35 and the warehouse static basic information 36 used when calculating the evaluation value in step S460, and the calculated evaluation value as the calculation history information 37. Store in device 3. The calculation history information 37 saved here is used in the environmental variation calculation process in step S410 as described above in subsequent processes.

In step S480, the search unit 13 selects one of the optimal solution candidates generated so far as an excellent solution. Here, based on the evaluation value calculated in step S460, the optimum solution candidate having the evaluation value that best matches the optimization objective is selected as the excellent solution. When the loop processing of steps S460 to S500 is executed first, the optimal solution candidate whose evaluation value is calculated in step S460 may be selected as the excellent solution regardless of the evaluation value.

At step S490, the search unit 13 generates a variant solution based on the good solution selected at step S460. Here, for example, a variant solution can be generated by partially changing the work order selected as a good solution. Note that any method can be used to generate a mutant solution as long as the mutant solution can be generated based on the good solution.

In step S500, the search unit 13 determines whether or not to end the search for the optimum solution. Here, it is determined whether or not the search end condition specified by the calculation time specification information 34 and the optimization parameter 35 is satisfied.If the search end condition is not satisfied, the process returns to step S460 and the loop processing of steps S460 to S500 is performed. repeat. At this time, the search unit 13 calculates an evaluation value in step S460 for the mutant solution generated in step S490 of the previous loop processing, and compares the evaluation value with the evaluation value of the excellent solution selected in step S480 of the previous loop processing. Compare and select one of them as the new good solution. On the other hand, if the search end condition is satisfied, the search processing shown in the flowchart of FIG. 7 is ended, and the process proceeds to step S50 of FIG. At this time, in step S50, as a result of the loop processing of steps S460 to S500 being executed multiple times in the search processing of FIG. 7, the excellent solution finally selected in step S480 is output as the optimum solution.

In the search process of FIG. 7, by repeating the loop process of S460 to S500 as described above, a plurality of variant solutions are generated for the optimum solution candidate (initial solution or neighboring solution), and evaluation values are calculated for each of them. , either the optimal solution candidate or multiple variant solutions can be selected as the optimal solution.

FIG. 8 is a flow chart showing the flow of the environmental variation calculation process executed in step S410 of FIG.

In step S411, the search unit 13 uses the optimization parameters 35, the warehouse static basic information 36, and the calculation history information 37 included in the optimization information input in step S10 of FIG. Calculate the similarity between the working environment and the current working environment. Here, the difference between the contents of the past optimization parameter 35 and the static basic warehouse information 36 represented by the calculation history information 37 and the current optimization parameter 35 and the static basic warehouse information 36 is obtained, and the similarity is calculated from the difference value. Calculate degrees. Specifically, for example, the content of the product placement 362 in the past static warehouse basic information 36 and the content of the product placement 362 in the current warehouse static basic information 36 are regarded as arrays, respectively, and between these arrays Calculate the rank correlation coefficient of This makes it possible to quantify the degree of change in work environment between the past and present, that is, the similarity. It should be noted that the degree of similarity calculated here may quantitatively represent the degree of change in the work environment, and any calculation method can be used to perform the processing of step S411. For example, for changes in environmental variables such as the speed of an autonomous body, difference calculation methods such as taking the absolute value of the difference, the square of the difference, and the logarithm of the absolute value of the difference are conceivable.

In step S412, the search unit 13 calculates an environment change function based on the similarity calculated in step S411. Here, in addition to the similarity calculation result, other information such as the optimization parameter 35 and the warehouse static basic information 36 are used to apply a predetermined environment change function to obtain a comprehensive environment Calculate the amount of change. For example, a linear calculation method that weights the difference between each parameter value between the past and present and sums it up, or a non-linear calculation method that calculates the product of the difference of each parameter value, etc., to obtain the amount of environmental change. be able to. The scalar value calculated in this manner is output as the environmental change amount 42 .

After executing the process of step S412 and outputting the environmental change amount 42, the search unit 13 ends the environmental change amount calculation process shown in the flowchart of FIG. 8, and proceeds to step S420 of FIG.

FIG. 9 is a diagram showing a specific example of the environmental variation visualization screen displayed on the output device 5 when visualizing the environmental variation in step S420 of FIG. The environment change amount visualization screen 120 shown in FIG. 9 includes an environment change amount graph 121 indicated by a solid line and a productivity change graph 122 indicated by a broken line. In these graphs, the horizontal axis represents time, and the vertical axis represents the environmental variation 42 and the evaluation value of the optimum solution. As described above, the information on the evaluation value of the optimum solution is included in the calculation history information 37 saved in step S470 of FIG.

The environmental change graph 121 represents time-series changes in the environment change amount 42, and the productivity change graph 122 represents time-series information of the productivity represented by the evaluation value of the obtained optimal solution, for example, the degree of work efficiency. represent. By comparing these graphs, the user can visually grasp the correspondence relationship between the amount of environmental change and the optimum productivity, and can determine whether the amount of environmental change appropriately captures the intended change in productivity. can judge. As a result, in the neighborhood solution generation process executed in step S450 of FIG. 7, it is possible to consider how to reflect the environmental variation in the parameters generated in step S451, which will be described later, and make adjustments as necessary. Become.

FIG. 10 is a flow chart showing the flow of the neighborhood solution generation process executed in step S450 of FIG.

In step S451, the searching unit 13 generates a parameter reflecting the current environmental change amount based on a certain point in the past based on the environmental change amount 42 calculated in step S410 of FIG. Here, for example, the value of the environmental change amount 42 is reflected to generate a parameter that determines the magnitude of the perturbation to be applied to the optimization target order 41 in step S452, which will be described later.

In step S452, the search unit 13 applies perturbation to the aforementioned optimization target order 41 based on the parameters generated in step S451. Here, for example, the order of each work in the work order represented by the order to be optimized 41 is digitized, the width of the probability distribution is determined according to the parameter, and each numerical value is varied according to the width of the probability distribution, Perturbation according to parameters can be added to the optimization target order 41 . For the probability distribution at this time, for example, a normal distribution or various other distributions can be used. For example, when a normal distribution is used, by generating the variance according to the value of the environmental change amount 42 as a parameter in step S451, the order of each work of the optimization target order 41 digitized in step S451 is calculated as follows. It is possible to vary and apply perturbations according to the generated dispersion. Even when other probability distributions are used, by applying the value of the environmental change amount 42 to the parameter representing the spread of the distribution, the perturbation to the optimization target order 41 can be similarly treated as a probability distribution.

In step S453, the search unit 13 rearranges the optimization target orders 41 to which the perturbation is applied in step S452. Here, for example, by sorting numerical values of the optimization target orders 41 after applying the perturbation in ascending order, the optimization target orders 41 that reflect the environmental change amount 42 are rearranged.

In step S454, the search unit 13 re-integers each numerical value of the optimization target orders 41 rearranged in step S453, thereby determining the rearranged work order reflecting the environmental variation 42. Then, the determined work order after rearrangement is output as the neighborhood solution 43, the neighborhood solution generation process shown in the flowchart of FIG. 10 is terminated, and the process proceeds to step S460 of FIG.

In the neighborhood solution generation process, by thus perturbing the optimization target order 41 according to the environmental variation 42, it is possible to generate a neighborhood solution 43 in which the optimization target order 41 is changed. At this time, various similarities can be expressed by changing the type of probability distribution assigned to each numerical value of the work order represented by the optimization target order 41 and its parameters. Note that the number of neighborhood solutions 43 output in step S454 may be one, or a plurality of them. It is also possible to output as

FIG. 11 shows a search situation screen displayed on the output device 5 when optimization processing is executed by the optimization device 100 of the present embodiment and then optimization processing is executed again in response to a change in the working environment. It is a figure which shows a specific example. In the search status screen 130 shown in FIG. 11, a broken-line graph 131 indicates the past search results, and a solid-line graph 132 indicates the current search results. In these graphs, the horizontal axis represents the calculation time required for the search, and the vertical axis represents the productivity (for example, the throughput value) of the warehouse work obtained as the evaluation value of the optimum solution. Note that the search situation screen 130 in FIG. 11 shows an example of using a genetic algorithm (GA) as an optimization method, but a similar screen can be used with Bayesian optimization (BO) or other optimization methods. can indicate the search status.

In graph 131, optimization information including warehouse static basic information 36 and initial order information 31 according to the work environment is input to the optimization device 100 of the present embodiment as a result of past search from a state without prior knowledge. , shows the state when searching for a certain period of time using a genetic algorithm (GA). On the other hand, the graph 132 shows the result of re-searching in a changed working environment by utilizing the information of the past search result indicated by the graph 131 as the current search result.

As described above, in the optimization device 100 of the present embodiment, the search unit 13 executes the environmental change amount calculation process shown in FIG. An environmental variation 42 representing the degree can be calculated, and a neighborhood solution 43 can be generated based on the environmental variation 42 by executing the neighborhood solution generation process shown in FIG. Then, by executing the search processing shown in FIG. 7, a plurality of mutant solutions based on the neighborhood solution are generated, evaluation values of the neighborhood solution and the plurality of mutation solutions are calculated, and based on these evaluation values, the neighborhood is generated. Either the solution or multiple variant solutions can be selected as the optimal solution. As a result, past search results can be efficiently updated according to changes in the work environment, and the optimum solution for the current work environment can be obtained.

A curve 133 in FIG. 11 indicates the distribution of perturbations applied to the optimization target order 41 when the neighborhood solution 43 is generated. The distribution of perturbations represented by this curve 133 conceptually shows the extent to which the past search results are reflected in the current search, taking into account the degree of change in the work environment between the past and this time. In the search process shown in FIG. 7, a neighborhood solution 43 is generated from the optimum solution based on the past search results represented by the calculation history information 37, and an optimization method such as a genetic algorithm (GA) is used to generate a new optimum solution. explore. As a result, compared to the case of re-executing the search from the beginning, it is possible to utilize the past search result information while reflecting changes in the environment, and to re-execute the search efficiently.

According to the embodiment of the present invention described above, the following effects are obtained.

(1) The optimization device 100 is a device for optimizing the order of work performed using a plurality of autonomous bodies, and initial order information 31 representing the initial value of the work order planned for the work, Optimization information including variable factor information (work progress information 32 and additional order information 33) representing variable factors of the work order and work environment information (warehouse static basic information 36) representing the work environment in which the work is performed. an optimization information input unit 11 for inputting an optimization target update unit 12 for determining an optimization target order 41 representing a work order to be optimized based on initial order information 31 and variation factor information; and work environment information and a search unit 13 for searching for the optimum solution of the work order based on the optimization target order 41, and an optimum solution output unit 14 for outputting the optimum solution searched by the search unit 13. Since this is done, the order of work performed using a plurality of autonomous bodies can be efficiently optimized according to changes in the work environment.

(2) The search unit 13 generates a neighborhood solution 43 in which the optimization target order 41 is changed according to changes in the work environment based on the work environment information (step S450), and searches for the optimum solution using the neighborhood solution 43. (Steps S460-S500). By doing so, the optimization target order 41 can be used to generate the neighborhood solution 43 according to changes in the working environment, and the neighborhood solution 43 can be used to efficiently search for the optimum solution.

(3) Based on the work environment information, the search unit 13 calculates the environment change amount 42 representing the work environment change amount (step S410), and generates the neighborhood solution 43 based on the environment change amount 42 (step S450). ). Specifically, the neighborhood solution 43 is generated by changing the optimization target order by applying a perturbation according to the environmental change amount 42 to the optimization target order 41 (steps S451 to S454). Since this is done, it is possible to generate the neighborhood solution 43 that appropriately reflects the amount of change in the working environment with respect to the optimization target order 41 .

(4) The search unit 13 visualizes the environmental variation 42 and presents it to the user (step S420). By doing so, it is possible to inform the user how much the current work environment has changed with reference to a certain point in the past.

(5) The search unit 13 determines whether or not to generate the neighborhood solution 43 based on the environmental change amount 42 (step S430). If it is determined not to generate the neighborhood solution 43 (step S430: NO), The initial solution set irrespective of the optimization target order 41 is used instead of the neighborhood solution 43 to search for the optimum solution (steps S440, S460 to S500). Since this is done in this way, if the change in the current work environment with respect to a certain point in the past is small, and therefore even if the neighborhood solution 43 is generated, the difference from the original optimization target order 41 is considered to be small, the neighborhood solution 43 By using the initial solution instead of , the optimal solution can be reliably searched.

(6) The search unit 13 generates a plurality of mutant solutions based on the neighborhood solution 43 (step S490), calculates the evaluation values of the neighborhood solution 43 and the plurality of mutation solutions (step S460), and uses these evaluation values as Based on this, either the nearest neighbor solution 43 or a plurality of mutant solutions is selected as the optimum solution (step S480). In this way, the most appropriate order of operations for the purpose of optimization can be obtained as the optimal solution.

(7) The optimum solution output unit 14 displays, for example, the search situation screen 130 of FIG. Present. Since this is done, the user can be notified of how the evaluation value of the optimum solution has changed according to the change in the work environment.

It should be noted that the present invention is not limited to the above-described embodiments and modifications, and can be implemented using arbitrary constituent elements within the scope of the gist of the present invention. Moreover, it is also possible to arbitrarily combine each embodiment and modifications.

The above embodiments and modifications are merely examples, and the present invention is not limited to these contents as long as the features of the invention are not impaired. Moreover, although various embodiments and modifications have been described above, the present invention is not limited to these contents. Other aspects conceivable within the scope of the technical idea of the present invention are also included in the scope of the present invention.

DESCRIPTION OF SYMBOLS 1...Central processing unit, 2...Main storage device, 3...Secondary storage device, 4...Input device, 5...Output device, 6...Bus, 11...Optimization information input unit, 12...Optimization target updating unit, 13 ... search unit, 14 ... optimum solution output unit, 31 ... initial order information, 32 ... work progress information, 33 ... additional order information, 34 ... calculation time designation information, 35 ... optimization parameters, 36 ... warehouse static basic information, 37... Calculation history information, 38... Optimal order information, 41... Optimization target order, 42... Environmental change amount, 43... Neighborhood solution, 100... Optimization device

Claims

An apparatus for optimizing the order of work performed with a plurality of autonomous bodies, comprising:
initial order information representing an initial value of the work order planned for the work; variation factor information representing a variation factor of the work order; and work environment information representing a work environment in which the work is performed. an optimization information input unit for inputting optimization information;
an optimization target update unit that determines an optimization target order representing a work order to be optimized based on the initial order information and the variation factor information;
a search unit that searches for an optimal solution for the work order based on the work environment information and the order to be optimized;
an optimum solution output unit that outputs the optimum solution searched by the search unit.
The optimization device of claim 1, wherein
The optimization device, wherein the search unit generates a neighborhood solution in which the optimization target order is changed according to a change in the work environment based on the work environment information, and searches for the optimum solution using the neighborhood solution.
The optimization device of claim 2, wherein
The optimization device, wherein the search unit calculates an environment change amount representing a change amount of the work environment based on the work environment information, and generates the neighborhood solution based on the environment change amount.
4. The optimization device of claim 3, wherein
The optimization device, wherein the search unit generates the neighborhood solution by changing the optimization target order by applying a perturbation according to the environmental change amount to the optimization target order.
In the optimization device according to claim 3 or 4,
The optimization device, wherein the search unit visualizes the environmental variation and presents it to a user.
In the optimization device according to claim 3 or 4,
The search unit determines whether or not to generate the neighborhood solution based on the environmental change amount, and if it is determined not to generate the neighborhood solution, an initial solution set regardless of the optimization target order instead of the neighborhood solution to search for the optimum solution.
The optimization device of claim 2, wherein
The search unit generates a plurality of mutant solutions based on the neighborhood solution, calculates evaluation values of the neighborhood solution and the plurality of mutant solutions, and calculates the neighborhood solution or the plurality of mutant solutions based on the evaluation values. as the optimum solution.
The optimization device according to claim 7, wherein
The optimization device, wherein the optimum solution output unit visualizes and presents to a user how the evaluation value of the optimum solution changes according to changes in the working environment.
A method for optimizing the order of work performed with multiple autonomies, comprising:
initial order information representing an initial value of the work order planned for the work; variation factor information representing a variation factor of the work order; and work environment information representing a work environment in which the work is performed. Enter the optimization information into the computer,
determining, by the computer, an order to be optimized representing a work order to be optimized based on the initial order information and the variable factor information;
using the computer to search for an optimal solution for the work sequence based on the work environment information and the order to be optimized;
An optimization method, wherein the searched optimal solution is output from the computer.