Information processing device, information processing method, and information processing program

Abstract

This information processing device comprises: a learning data acquisition unit that acquires, as learning data, commodity information that is information on a commodity and a cargo handling attribute that is classified according to the cargo handling property of the commodity; and an inference model generation unit that generates an inference model for inferring the cargo handling attribute according to the cargo handling property of a target commodity by means of machine learning with the learning data.

Description

Information processing device, information processing method, and information processing program

[0001] The technology disclosed herein relates to an information processing device, an information processing method, and an information processing program.

[0002] A technique for predicting work time from process information such as product materials and production quantity using machine learning has been proposed as a technique for estimating unknown work time (see, for example, Non-Patent Document 1).

[0003] Hiroshi Yamashiro and Hirofumi Nonaka, "Working Time Estimation Using Machine Learning for Optimizing Parallel Machine Scheduling," Proceedings of the 83rd National Conference, 2021 (1), pp. 507-508, 2021-03-04

[0004] Various delivery services are provided in public, commercial, and office districts, and in recent years, delivery services using autonomous mobile robots have become increasingly popular. Prior to delivery, such services require the store to prepare the products and perform loading and unloading tasks, such as loading them onto delivery resources like trucks and robots. However, the tasks required for loading and unloading vary widely depending on the product, and each product has different handling characteristics. Therefore, it is desirable to consider the handling characteristics of each product.

[0005] The technology disclosed herein has been made in consideration of the above points, and aims to provide an apparatus, method, and program that can contribute to providing information that takes into account the loading and unloading characteristics of products.

[0006] A first aspect of the present disclosure is an information processing device that includes a learning data acquisition unit that acquires product information, which is information about a product, and handling attributes classified according to the handling properties of the product as learning data, and an inference model generation unit that generates an inference model that infers handling attributes according to the handling properties of a target product through machine learning using the learning data.

[0007] A second aspect of the present disclosure is an information processing device that includes a product information input unit that accepts input of product information, which is information about the target product, an estimation unit that estimates loading attributes according to the loading characteristics of the product using an inference model generated by machine learning, and a loading attribute output unit that outputs the loading attributes.

[0008] A third aspect of the present disclosure is an information processing device that includes a loading attribute acquisition unit that acquires loading attributes according to the loading characteristics of the target product, a prediction unit that predicts the loading time required to load the product based on the loading attributes, and a loading time output unit that outputs information regarding the loading time.

[0009] A fourth aspect of the present disclosure is an information processing method, in which a computer acquires handling attributes corresponding to the handling characteristics of a target product, predicts the handling time required to handle the product based on the handling attributes, and outputs information related to the handling time.

[0010] The technology disclosed herein has the effect of contributing to the provision of information that takes into account the ease of handling of goods.

[0011] FIG. 1 is a schematic diagram showing an example of a processing flow for predicting loading and unloading time. FIG. 2 is a block diagram showing an example of a hardware configuration of the information processing device 100 according to the present embodiment. FIG. 3 is a block diagram showing an example of a functional configuration of the information processing device 100 according to the present embodiment. FIG. 4 is a diagram showing an example of product information to be processed by the information processing device 100 according to the present embodiment. FIG. 5 is a diagram for explaining loading and unloading attributes to be processed by the information processing device 100 according to the present embodiment. FIG. 6 is a diagram showing an example of delivery information to be processed by the information processing device 100 according to the present embodiment. A flowchart showing an example of a processing flow when the information processing device 100 according to the present embodiment functions as a learning device. A flowchart showing an example of a processing flow when the information processing device 100 according to the present embodiment functions as an estimation device. A flowchart showing an example of a processing flow when the information processing device 100 according to the present embodiment functions as a prediction device.

[0012] An example of an embodiment of the disclosed technology will be described below with reference to the drawings. Note that in each drawing, the same or equivalent components and parts are given the same reference numerals. Also, the dimensional proportions in the drawings are exaggerated for the convenience of explanation and may differ from the actual proportions.

[0013] Figure 1 is a schematic diagram showing an example of the processing flow for predicting loading and unloading times. "Loading and unloading" is a general term for work related to handling products in the logistics process. Here, "loading and unloading" refers to all work that occurs prior to delivering products to customers using delivery resources, and in particular includes work such as picking and packing products to be delivered from stores or warehouses.

[0014] This figure shows an example in which two models, an inference model 10 and a prediction model 50, are used to predict the loading and unloading time required for such loading and unloading. The inference model 10 is a model that infers loading and unloading attributes from product information. The prediction model 50 is a model that predicts the loading and unloading time, which is the objective variable, using delivery information including loading and unloading attributes as explanatory variables.

[0015] The details of these two models and the various types of information will be described later. The information processing apparatus 100 according to this embodiment executes at least a part of the processing flow shown in this figure, for example.

[0016] 2 is a block diagram showing an example of the hardware configuration of the information processing device 100 according to this embodiment. As shown in FIG. 2, the information processing device 100 includes a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, a storage 104, an input unit 105, a display unit 106, and a communication interface (I / F) 107. Each component is connected to each other via a bus 109 so as to be able to communicate with each other.

[0017] The CPU 101 is a central processing unit that executes various programs and controls each component. The ROM 102 stores various programs and data. The RAM 103 temporarily stores programs or data as a working area. The storage 104 is configured with a storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) and stores various programs including an operating system and various data.

[0018] The CPU 101 reads a program from the ROM 102 or the storage 104 and executes the program using the RAM 103 as a work area. The CPU 101 controls the above components and performs various arithmetic processing in accordance with the program stored in the ROM 102 or the storage 104.

[0019] From here on, we will explain as an example a case where ROM 102 or storage 104 stores three programs: a learning program that generates an inference model 10, an estimation program that estimates cargo handling attributes, and a prediction program that predicts cargo handling time.

[0020] However, only one of the three programs may be stored in the ROM 102 or the storage 104, or two of the three programs may be stored in the ROM 102 or the storage 104. Furthermore, two or more of the three programs may be stored as an integrated program.

[0021] The input unit 105 includes a pointing device such as a mouse and a keyboard, and is used to input various types of information.

[0022] The display unit 106 is, for example, a liquid crystal display, and displays various information. The display unit 106 may be a touch panel type and function as the input unit 105.

[0023] The communication interface 107 is an interface for communicating with other devices, and the communication may use, for example, a wired communication standard such as Ethernet (registered trademark) or FDDI, or a wireless communication standard such as 4G, 5G, or Wi-Fi (registered trademark).

[0024] Next, the functional configuration of the information processing device 100 according to this embodiment will be described.

[0025] Fig. 3 is a block diagram showing an example of the functional configuration of the information processing device 100 according to this embodiment. As shown in Fig. 3, the information processing device 100 according to this embodiment may be configured with a learning block 110, an estimation block 130, and a prediction block 150. Below, a case where the information processing device 100 is configured with three blocks will be described as an example.

[0026] However, the information processing device 100 may be configured with only one of the three blocks. When the information processing device 100 is configured with only the learning block 110, the information processing device 100 may function as a learning device. Similarly, when the information processing device 100 is configured with only the estimation block 130, the information processing device 100 may function as an estimation device. Similarly, when the information processing device 100 is configured with only the prediction block 150, the information processing device 100 may function as a prediction device.

[0027] Furthermore, although the above description indicates that the information processing device 100 may be configured with only one block, the information processing device 100 may be configured with two of the three blocks. That is, the information processing device 100 may function as a learning device and an estimation device, or as an estimation device and a prediction device, or as a learning device and a prediction device.

[0028] The learning block 110 includes a learning data acquisition unit 111 and an inference model generation unit 112. These functional components are realized by the CPU 101 reading out a learning program from the ROM 102 or the storage 104, expanding it into the RAM 103, and executing it.

[0029] The learning data acquisition unit 111 acquires, as learning data, product information, which is information about the product, and handling attributes, which are classified according to the handling characteristics of the product.

[0030] The inference model generation unit 112 generates an inference model 10 that infers cargo handling attributes according to the cargo handling characteristics of the target product through machine learning using the training data. The generated inference model 10 may be stored in the learning block 110 as shown in the figure, or may be stored in another functional unit or another device.

[0031] The estimation block 130 includes a product information input unit 131, an estimation unit 132, and a cargo handling attribute output unit 133. These functional components are realized by the CPU 101 reading out an estimation program from the ROM 102 or the storage 104, expanding it into the RAM 103, and executing it.

[0032] The product information input unit 131 accepts input of product information, which is information on the target product.

[0033] The estimation unit 132 uses an inference model 10 generated by machine learning to estimate handling attributes according to the handling characteristics of the product for which product information has been input.

[0034] The cargo handling attribute output unit 133 outputs the estimated cargo handling attributes to other functional units or other devices.

[0035] The prediction block 150 includes a delivery information collection unit 151, a prediction model generation unit 152, a cargo handling attribute acquisition unit 153, a prediction unit 154, and a cargo handling time output unit 155. These functional configurations are realized by the CPU 101 reading out a prediction program from the ROM 102 or the storage 104, expanding it into the RAM 103, and executing it.

[0036] The delivery information collection unit 151 collects delivery information indicating delivery information.

[0037] The prediction model generation unit 152 generates a prediction model 50 that predicts the target variable, the loading and unloading time, through machine learning using delivery information including loading and unloading attributes as explanatory variables. The generated prediction model 50 may be stored in the prediction block 150 as shown in the figure, or may be stored in another functional unit or another device.

[0038] The cargo handling attribute acquisition unit 153 acquires cargo handling attributes according to the cargo handling characteristics of the target product.

[0039] The prediction unit 154 predicts the handling time required for handling the product based on the handling attributes. In this figure, an example is shown in which the prediction unit 154 predicts the handling time using a prediction model 50 generated by machine learning, with delivery information including the handling attributes as explanatory variables.

[0040] The loading / unloading time output unit 155 outputs information relating to the predicted loading / unloading time to other functional units or other devices.

[0041] Before describing the processing performed by the information processing device 100 having such functional units, various types of information that are processed by the information processing device 100 will be described first.

[0042] 4 is a diagram showing an example of product information to be processed by the information processing device 100 according to this embodiment. Product information is information indicating product details. If a product can be ordered online using an app, the product information may be information presented to a customer on the app. As an example, the product information may include items such as the product name, product description, price, and manufacturer, as shown in this figure.

[0043] The product name is the name of the product. For example, the product name may be "curry bread," "plum rice ball," "grilled rice ball," "blend coffee," "natural water," etc.

[0044] The product description is a description of the product. For example, if the product name is "Curry Bread," the product description may include information such as "Enjoy the deliciousness of freshly baked bread at home."

[0045] Also, if the product name is "Ume Onigiri," the product description may include information such as "Store away from direct sunlight and high temperatures and humidity. Pull the red tape on the top of the triangle and pull the film from the left and right to easily remove the onigiri."

[0046] Also, if the product name is "yaki onigiri," the product description may include information such as "Store at -18°C or below. Place the frozen product upright on a plate and heat in the microwave without covering it with plastic wrap."

[0047] Also, if the product name is "Blend Coffee," the product description may include information such as "The cup is embossed, but it may become hot. It may leak if left for a long time. The lid is easy to come off. Please be careful."

[0048] Also, if the product name is "Mineral Water," the product description may include information such as "Set of 8 bottles of 2 liters. Made from 100% recycled PET bottle material. Can also be used as emergency water."

[0049] The price is the price of the product. For example, if the product name is "curry bread," the price may be "138 yen / piece (149 yen / piece including tax)."

[0050] A manufacturer is a person who produces a product. For example, if the product name is "curry bread," the manufacturer may be "the bakery corner of a supermarket."

[0051] The following describes an example in which the information processing device 100 uses the product name and product description items as product information, but is not limited to this. In cases where careful handling is required for products priced above a threshold, or where special handling is required for products from a specific manufacturer, the information processing device 100 may use other items such as price and manufacturer as product information.

[0052] 5 is a diagram for explaining cargo handling attributes to be processed by the information processing device 100 according to this embodiment. The left side of the figure shows the product name included in the product information, and the right side of the figure shows the cargo handling attributes. The cargo handling attributes are attribute information according to the cargo handling characteristics of the product.

[0053] Such cargo handling attributes may be attributes classified according to the necessity of work prior to delivery, where the work may be work related to at least one of picking and packing.

[0054] For example, if "curry bread" is bread that is displayed on a tray in a bakery section, it needs to be individually wrapped before delivery. To indicate such handling characteristics, the handling attribute of "curry bread" may be "unwrapped bread."

[0055] Also, if "ume rice ball" is wrapped in a triangular film and can be stored at room temperature, there is no need to wrap it or keep it refrigerated before delivery. To indicate such handling characteristics, the handling attribute of "ume rice ball" may be "triangular rice ball."

[0056] Furthermore, if the "yaki onigiri" is a frozen food, it needs to be packaged with ice packs before delivery. In order to indicate such handling characteristics, the handling attribute of the "yaki onigiri" may be "frozen cooked rice."

[0057] Furthermore, if "Blend Coffee" is a beverage served in an embossed cup, a holder must be prepared to prevent the beverage from tipping over before delivery. To indicate such handling characteristics, the handling attribute of "Blend Coffee" may be "Beverage served in an embossed cup."

[0058] Furthermore, if "mineral water" is a beverage sold in bulk in plastic bottles, it is not displayed in the store and must be picked from a back yard such as a warehouse. To indicate such handling characteristics, the handling attribute of "mineral water" may be "plastic bottled beverage sold in boxes."

[0059] Such handling attributes are similar but not the same as product attributes used in product marketing. For example, even if two products are classified in the same category as blended coffee in terms of product attributes used in product marketing, coffee in an embossed scoop requires a holder to prevent it from tipping over, while a PET bottled drink does not. Therefore, blended coffee in an embossed scoop and blended coffee in a PET bottle are classified in different categories in terms of handling attributes. On the other hand, tea and coffee in PET bottled drinks are classified in different categories in terms of product attributes used in product marketing, but are classified in the same category in terms of handling attributes.

[0060] Handling operations include the work of the store preparing products to be delivered and the work of loading the products onto delivery resources such as trucks and robots. The work of the store preparing products to be delivered refers, for example, to the work of workers picking products to be delivered from the store or warehouse, bagging them, and preparing them for delivery to customers. The work time for such operations varies greatly depending on whether the ordered products are on display in the store, whether they need to be bagged, and other factors. Therefore, assigning such handling attributes to products in order to take into account the different handling characteristics of each product is an important factor in accurately predicting handling times.

[0061] In this disclosure, we focus on the fact that factors that affect loading and unloading time differ for each product, and define these as loading and unloading attributes. As described above, these loading and unloading attributes are different from product attributes used in conventional product marketing, and are a completely new category that is classified according to whether or not there are characteristics specific to loading and unloading.

[0062] FIG. 6 is a diagram showing an example of delivery information to be processed by the information processing device 100 according to this embodiment. The delivery information is information indicating details of delivery. The delivery information includes handling attributes. Such handling attributes may be obtained by converting the product name of the ordered product into handling attributes using a pre-stored conversion table. If the product name of the ordered product is not included in the conversion table, the handling attributes may be obtained from the estimation block 130 that estimates handling attributes.

[0063] In addition to cargo handling attributes, the delivery information may further include items such as predicted values ​​of a congestion prediction model, order time, order month, order day, number of items purchased, and store name.

[0064] The predicted value of the congestion prediction model is a predicted value generated by the congestion prediction model of the route cost prediction model. The order time is the time the order was placed. The order month is the month the order was placed. The order day of the week is the day the order was placed. The number of items purchased is the total number of items purchased. The store name is the name of the store where the items were purchased. In addition to the above information, the delivery information may further include information such as the delivery destination, the number of customers in the store at the time of delivery, the degree of congestion in the block, the weather, temperature, etc.

[0065] The prediction model 50 may be generated by machine learning using such delivery information, for example. More specifically, the CPU 101, as the prediction model generation unit 152, may generate the prediction model 50 that predicts the loading and unloading time, which is the objective variable, using a Light Gradient Boosting Machine (LightGBM) with such delivery information as the explanatory variable. In this case, the CPU 101 may convert the explanatory variables acquired as categorical data into dummy variables for use.

[0066] Note that the generation of a model using LightGBM is one example. Alternatively, the CPU 101 may generate the prediction model 50 that predicts the loading and unloading time using another machine learning algorithm, such as XGboost (Extreme Gradient Boosting).

[0067] What is noteworthy here is that the delivery information, which serves as an explanatory variable, includes handling attributes. When predicting handling times in a delivery service, it is desirable to consider handling attributes according to the product's handling characteristics. However, the tasks required for handling vary depending on the product, such as whether the product requires bagging, refrigeration or warmth, measures to prevent tipping, or picking from the back yard. Therefore, it takes a lot of man-hours to assign handling attributes to each individual product, which has a wide variety of handling characteristics.

[0068] Therefore, in the first aspect, the information processing device 100 may function as a learning device that generates an inference model 10 that infers cargo handling attributes.

[0069] 7 is a flowchart showing an example of a processing flow when the information processing device 100 according to this embodiment functions as a learning device. This flow may be executed by the CPU 101 reading out a learning program from the ROM 102 or the storage 104, expanding it in the RAM 103, and executing it.

[0070] In step S210, the CPU 101 arbitrarily samples products with clear handling attributes from the products that are the target of the delivery service. At this time, the products may be sampled manually.

[0071] Sampling is only necessary when setting cargo handling attributes for the first time or when adding a new product that is not thought to belong to the existing cargo handling attributes, and is not a necessary task every time the model is retrained.

[0072] In step S220, the CPU 101 acquires, as the learning data acquisition unit 111, product information, which is information about the product, and handling attributes, which are classified according to the handling characteristics of the product, as learning data. For example, the CPU 101 acquires, as a character string, product information, including the product name and product description, of the product sampled in step S210 as having clear handling attributes, and vectorizes this.

[0073] In step S230, the CPU 101 acquires product information of products that were not sampled in step S210 as the learning data acquisition unit 111. For example, the CPU 101 uses the products sampled in step S210 as samples in kNN (k-Nearest Neighbor), acquires product information including the product name and product description of the products that were not sampled as character strings, and vectorizes the character strings.

[0074] In steps S220 and S230, the CPU 101 may acquire the product information and cargo handling attributes through user input, through an auxiliary storage device such as a flash memory, optical disk, or magnetic disk, or from another device through a communication interface.

[0075] In step S240, the CPU 101, as the inference model generation unit 112, generates an inference model 10 that infers cargo handling attributes according to the cargo handling characteristics of a target product through machine learning using the training data. For example, the CPU 101 clusters products that are not sampled using the kNN algorithm. In this case, the CPU 101 may select one nearest sample with k=1.

[0076] Clustering using kNN is just one example. Alternatively, the CPU 101 may generate the inference model 10 that infers cargo handling attributes using other machine learning algorithms such as LightGBM or XGboost, with product information as explanatory variables.

[0077] In step S250, the information processing device 100 saves the inference model 10 generated in step S240. The saved inference model 10 may be used to obtain the handling attributes of the ordered product each time an order is placed.

[0078] In this way, the information processing device 100 functions as a learning device, and generates an inference model 10 that uses product information and handling attributes as learning data to infer handling attributes according to the handling characteristics of a target product. As a result, the information processing device 100 provides an environment in which handling attributes can be automatically assigned to a target product, thereby contributing to the provision of information regarding the handling characteristics of products.

[0079] In this case, the information processing device 100 arbitrarily samples products with clear handling attributes from the products that are the target of the delivery service, and generates the inference model 10 using the product information and handling attributes of the sampled products as training data. As a result, the information processing device 100 can classify products of unknown category into the same category as products that are similar to products that have significant handling characteristics.

[0080] Assigning handling attributes to each product to be delivered requires a lot of man-hours, and manually assigning handling attributes each time a new product is added makes operation cumbersome. In contrast, the information processing device 100 provides an environment in which handling attributes can be mechanically determined and automatically assigned, thereby reducing man-hours and simplifying operation.

[0081] 8 is a flowchart showing an example of a processing flow when the information processing device 100 according to this embodiment functions as an estimation device. This flow may be executed by the CPU 101 reading out an estimation program from the ROM 102 or the storage 104, expanding it in the RAM 103, and executing it.

[0082] In step S310, the CPU 101 accepts input of product information, which is information about the target product, as the product information input unit 131. At this time, the CPU 101 may, for example, accept input of product information indicating information about the ordered product from the cargo handling attribute acquisition unit 153 in the prediction block 150.

[0083] In step S320, the CPU 101, as the estimation unit 132, estimates handling attributes according to the handling characteristics of the product whose product information was input in step S310, using the inference model 10 generated by machine learning. More specifically, step S320 may include substeps of steps S321 and S322.

[0084] In step S321, the CPU 101, as the estimation unit 132, inputs the product information input in step S310 to the inference model 10. In response to this, the inference model 10 infers handling attributes according to the handling characteristics of the target product.

[0085] As an example, if product information is input in which the product name is "Salmon Onigiri" and the product description includes information that the product is packaged in film, the inference model 10 infers the loading attribute to be "Triangular Onigiri" and outputs this.

[0086] In step S322, the CPU 101, as the estimation unit 132, acquires the cargo handling attributes output by the inference model 10 in response to the input of the product information in step S321. The CPU 101 may, for example, use the inference model 10 generated by machine learning in this manner to estimate the cargo handling attributes according to the cargo handling characteristics of the product.

[0087] In step S330, the CPU 101, as the cargo handling attribute output unit 133, outputs the cargo handling attributes estimated in step S320 to other functional units or other devices through various interfaces. At this time, as one example, the CPU 101 may output the estimated cargo handling attributes to the cargo handling attribute acquisition unit 153 in the prediction block 150 that has accepted the input of the product information.

[0088] In this way, the information processing device 100 functions as an estimation device, accepting input of product information, which is information about the target product, and estimating and outputting the handling attributes of the input product using the inference model 10. As a result, the information processing device 100 outputs the handling attributes to be assigned to the target product, thereby contributing to the provision of information about the handling characteristics of the product. In particular, the information processing device 100 can automatically assign handling attributes to the product, thereby reducing the amount of work required to predict the handling time.

[0089] 9 is a flowchart showing an example of a processing flow when the information processing device 100 according to this embodiment functions as a prediction device. This flow may be executed by the CPU 101 reading a prediction program from the ROM 102 or the storage 104, expanding it in the RAM 103, and executing it.

[0090] In step S410, the CPU 101, as the delivery information collection unit 151, collects delivery information indicating delivery information. For example, when an order is placed for a delivery service, the CPU 101 may collect delivery information for each ordered product, such as that shown in FIG. 6, from information related to the delivery and order and environmental information. As an example, when an order is placed for product A, product B, and product C, the delivery information collection unit 151 may collect delivery information for product A, delivery information for product B, and delivery information for product C, respectively. At this point, the handling attributes are unknown, so the product name may be entered in the handling attribute field included in the delivery information.

[0091] In step S420, the CPU 101, as the handling attribute acquisition unit 153, acquires handling attributes corresponding to the handling properties of the target product. At this time, the CPU 101 may acquire the handling attributes by converting the product name of the ordered product using a pre-stored conversion table. As an example, assume that the product name of product A among the ordered products is included in the conversion table. In this case, the CPU 101 may convert the product name of product A into a handling attribute Ca corresponding to the handling properties of product A according to the conversion table.

[0092] On the other hand, if the product names of the ordered products are not included in the conversion table, the CPU 101 may obtain the handling attributes from the estimation block 130, which estimates the handling attributes. As an example, assume that the product names of product B and product C, among the ordered products, are not included in the conversion table. In this case, the CPU 101 may input the product information of product B to the product information input unit 131 in the estimation block 130. In response, the CPU 101 may obtain the handling attribute Cb corresponding to the handling characteristics of product B estimated in the estimation block 130 from the handling attribute output unit 133 in the estimation block 130. Similarly, the CPU 101 may input the product information of product C to the product information input unit 131 in the estimation block 130. In response, the CPU 101 may obtain the handling attribute Cc corresponding to the handling characteristics of product C estimated in the estimation block 130 from the handling attribute output unit 133 in the estimation block 130.

[0093] When cargo handling attributes are acquired from the estimation block 130 in this way, the CPU 101 updates the conversion table. As an example, when cargo handling attribute Cb is acquired, the CPU 101 associates the product name of product B with the cargo handling attribute Cb and adds them to the conversion table. Similarly, when cargo handling attribute Cc is acquired, the CPU 101 associates the product name of product C with the cargo handling attribute Cc and adds them to the conversion table. This makes it possible to omit acquiring the cargo handling attributes estimated in the estimation block 130 when an order for product B or product C is subsequently placed.

[0094] When the cargo handling attributes Ca, Cb, and Cc are acquired, the CPU 101 updates the cargo handling attribute items included in the delivery information from the product names of product A, product B, and product C to cargo handling attributes Ca, Cb, and Cc, respectively.

[0095] In step S430, the CPU 101 predicts the handling time required for handling the product based on the handling attributes as the prediction unit 154. More specifically, step S430 may include sub-steps of steps S431 and S432.

[0096] In step S431, the CPU 101, as the prediction unit 154, inputs the delivery information acquired in step S410, in which the handling attribute items have been updated in step S420, to the prediction model 50. In response, the prediction model 50 predicts the handling time, which is the objective variable, using the delivery information including the handling attributes as explanatory variables. As an example, the CPU 101 may input the delivery information for product A, the delivery information for product B, and the delivery information for product C to the prediction model 50. In response, the prediction model 50 may predict and output the handling time Ta for product A, the handling time Tb for product B, and the handling time Tc for product C.

[0097] In step S432, the CPU 101, as the prediction unit 154, acquires the handling times output by the prediction model 50 in response to the input of the delivery information in step S431. As an example, the CPU 101 may acquire the handling time Ta for product A, the handling time Tb for product B, and the handling time Tc for product C.

[0098] Then, the CPU 101 calculates the handling time for each product and predicts the handling time for the entire order. As an example, the CPU 101 may predict the handling time T for the entire order by adding up the handling time Ta for product A, the handling time Tb for product B, and the handling time Tc for product C.

[0099] In step S440, the CPU 101, as the loading / unloading time output unit 155, outputs information relating to the loading / unloading time T predicted in step S430 to other functional units or other devices via various interfaces. As an example, the CPU 101 may output the predicted loading / unloading time T itself. Alternatively, the CPU 101 may output some information calculated using the predicted loading / unloading time T, for example, the predicted arrival time of the product obtained by adding the delivery time to the loading / unloading time T, as information relating to the loading / unloading time.

[0100] In this way, the information processing device 100 functions as a prediction device, acquires handling attributes according to the handling characteristics of the target product, and predicts the handling time based on the handling attributes. As a result, the information processing device 100 outputs information about the handling time predicted in consideration of the handling attributes according to the handling characteristics of the product, thereby contributing to the provision of information about the handling characteristics of the product.

[0101] In particular, when predicting handling times using machine learning, the information processing device 100 uses handling attributes as explanatory variables, thereby enabling predictions that take into account information related to the handling of products, such as whether the ordered products need to be packaged or whether they need to be removed from the backroom. In this way, the information processing device 100 can accurately estimate the time required for the products to arrive by accurately predicting handling times using handling attributes. Therefore, the information processing device 100 can prevent inefficient use of delivery resources, such as robots, due to excessive handling time estimates.

[0102] Furthermore, the information processing device 100 does not use the product information as it is to predict the loading and unloading time, but uses the loading and unloading attributes to predict the loading and unloading time. As a result, the information processing device 100 can prevent the curse of dimensionality that occurs when the number of explanatory variables increases.

[0103] The generated model is used each time an order is placed to operate the delivery service, but as operation time passes, environmental changes such as seasonal changes and changes in worker proficiency are likely to occur. This can cause drift and reduce the accuracy of the model. Therefore, in order to operate a machine learning model in a production environment, it is necessary to address drift.

[0104] There are several possible methods for dealing with drift, but the information processing device 100 adopts a method in which the model is adapted to a new environment by periodically re-training the model using new data during predetermined time periods such as late at night or early in the morning when delivery services are suspended. In this case, realizing automatic classification of cargo handling attributes can reduce the cost of creating training data and also make it easier to re-train the model.

[0105] In addition, the various processes executed by the CPU after reading software (programs) in each of the above embodiments may be executed by various processors other than the CPU. Examples of processors in this case include PLDs (Programmable Logic Devices) whose circuit configuration can be changed after manufacture, such as FPGAs (Field-Programmable Gate Arrays), and dedicated electrical circuits, such as ASICs (Application Specific Integrated Circuits), which are processors having a circuit configuration designed specifically to execute specific processes. Furthermore, various processes may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor elements.

[0106] In addition, in each of the above embodiments, various processing programs are described as being stored (installed) in advance in the ROM 102 or the storage 104, but this is not limiting. The programs may be provided in a form stored in a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The programs may also be downloaded from an external device via a network.

[0107] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[0108] The present disclosure also includes the following: (Supplementary Note 1) An information processing device comprising a processor, the processor acquiring handling attributes according to the handling properties of a target product, predicting the handling time required to handle the product based on the handling attributes, and outputting information related to the handling time. (Supplementary Note 2) A non-transitory computer-readable medium having recorded thereon an information processing program, the processor acquiring handling attributes according to the handling properties of a target product, predicting the handling time required to handle the product based on the handling attributes, and outputting information related to the handling time.

Claims

1. A learning data acquisition unit that acquires, as learning data, product information that is information about a product and a handling attribute classified according to the handling property of the product; and an inference model generation unit that generates an inference model for inferring the handling attribute according to the handling property of a target product by machine learning using the learning data. An information processing apparatus comprising:

2. A product information input unit that receives an input of product information that is information about a target product; an estimation unit that estimates a handling attribute according to the handling property of the product using an inference model generated by machine learning; and a handling attribute output unit that outputs the handling attribute. An information processing apparatus comprising:

3. A handling attribute acquisition unit that acquires a handling attribute according to the handling property of a target product; a prediction unit that predicts the handling time required for handling the product based on the handling attribute; and a handling time output unit that outputs information about the handling time. An information processing apparatus comprising:

4. The information processing apparatus according to claim 3, wherein the prediction unit predicts the handling time using a prediction model generated by machine learning, with the delivery information including the handling attribute as an explanatory variable.

5. The information processing apparatus according to any one of claims 1 to 4, wherein the handling attribute is an attribute classified according to the necessity of work prior to delivery.

6. The information processing apparatus according to claim 5, wherein the work is work related to at least one of picking and packing.

7. An information processing method, wherein a computer: acquires a handling attribute according to the handling property of a target product; predicts the handling time required for handling the product based on the handling attribute; and outputs information about the handling time.

8. An information processing program for executing the information processing method according to claim 7.

Images