Control method and control device

The system corrects and predicts congestion situations using machine learning to adjust notifications, effectively managing user distribution and reducing stress by converging to a target state.

JP2024166805AInactive Publication Date: 2024-11-29DENSO TEN LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023083154
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-19
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing systems struggle to accurately predict and control user behavior in response to congestion situations, often leading to users focusing on less crowded areas rather than achieving the desired distribution of users across multiple locations.

Method used

A system that reads the detected congestion situation, corrects it based on a target state using machine learning to determine the appropriate notification to users, and performs feedback control to adjust the congestion situation to the target state.

Benefits of technology

The system effectively adjusts congestion by predicting user behavior changes and converging to a target state, reducing stress and optimizing user distribution across locations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024166805000001_ABST
    Figure 2024166805000001_ABST
Patent Text Reader

Abstract

To provide a technology for flexibly performing notification of a congestion state of a monitoring target according to the monitoring target.SOLUTION: In a control method according to the present technology, a computer repeatedly executes feed back control for reading a detected congestion state of a monitoring target from a storage device, correcting the read congestion state on the basis of a difference between the congestion state and a targeted state, and outputting a notification instruction signal of congestion state after the correction to a notification device for which a user of the monitoring target can make confirmation.SELECTED DRAWING: Figure 6
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a control method and a control device for reporting a congestion state. [Background technology]

[0002] Conventionally, a toilet management method for managing the congestion status of multiple toilet rooms has been proposed (for example, Patent Document 1). This technology includes a usage information acquisition process for acquiring usage information related to the usage status of multiple toilet rooms, a congestion status determination process for determining the actual congestion status of the multiple toilet rooms based on the acquired usage information, a congestion status correction process for correcting the actual congestion status determined in the congestion status determination process, and a transmission process for transmitting the correction result of the congestion status corrected in the congestion status correction process to an output terminal. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2022-039679 A Summary of the Invention [Problem to be solved by the invention]

[0004] When the congestion status is presented to users, there is a possibility that users will end up concentrating in areas that are relatively less crowded. In addition, it is difficult to predict the behavior of users who have been presented with the congestion status, and even if the congestion status is corrected, there is a possibility that people's movements will not be controlled as intended. Therefore, the purpose of this technology is to provide a technology that can flexibly notify the congestion status of a monitored object according to the monitored object. [Means for solving the problem]

[0005] The control method of the present disclosure involves a computer performing feedback control in which the detected congestion status of the monitored object is read from a storage device, correction is made to the read congestion status based on the difference between the congestion status and a target state, and a notification instruction signal for the corrected congestion status is output to an alarm device that can be confirmed by users of the monitored object. Effect of the Invention

[0006] According to the present technology, it is possible to flexibly notify the congestion status of a monitoring target depending on the monitoring target. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram illustrating an example of a system configuration. [Diagram 2] FIG. 2 is a block diagram illustrating an example of a system. [Diagram 3] FIG. 3 is a process flow diagram illustrating an example of the learning process. [Figure 4] FIG. 4 is a diagram illustrating an example of a change in the retention number. [Diagram 5] FIG. 5 is a diagram for explaining the correction rule. [Figure 6] FIG. 6 is a process flow diagram illustrating an example of the operation process. [Figure 7] FIG. 7 is a diagram illustrating an example of a system configuration. [Figure 8] FIG. 8 is a diagram for explaining an example of a notification to an administrator. [Figure 9] FIG. 9 is a diagram for explaining another example of the change in the staying number. [Figure 10] FIG. 10 is a diagram for explaining another example of the change in the staying number. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] <Embodiment 1> Hereinafter, an embodiment will be described with reference to the drawings. FIG. 1 is a diagram for explaining an example of the system configuration. The system 1 detects a congestion state in some monitored object and performs a predetermined notification process for users. The monitored object is a place where a queue or traffic congestion occurs, such as a store, a counter, an attraction such as an amusement park, a public toilet, a parking lot, a road section, etc. The congestion state is expressed, for example, by the number of people or the number of vehicles forming a queue (called the "staying number"). In addition, the system 1 notifies the user of the congestion state corrected based on the difference between the latest congestion state and the target state, thereby converging to the target state. The correction according to this embodiment is a process of learning the magnitude of change in the behavior of a user who receives feedback on the congestion state, and changing the actual congestion state for notification to the user. The correction amount is determined, for example, by machine learning, so as to converge to the target state early. In other words, the correction amount that takes into account the degree to which the behavior of a user who receives feedback on the congestion state will change is used to perform a correction that approaches the target, and then feedback of the predicted information is performed. Learning is a process of creating a trained model that indicates how much the actual number of people waiting should be changed (i.e., increased or decreased) to bring the number of people waiting closer to a target state, by machine learning through repeated feedback. That is, the relationship between the number of people notified and the change in the number of people waiting after the notification is learned by a method such as regression analysis. Learning may be performed by distinguishing between seasons, days of the week, time periods, etc. Furthermore, a trained model expressed by a regression formula, for example, can be used as a correction rule for determining the number of people to be notified based on the actual number of people waiting and the target state, or in other words, it can be said to be a prediction model that predicts the number of people waiting after the notification of the number of people.

[0009] The system 1 includes a sensor system 2 installed in a monitoring target 20, which is, for example, a store, a display device (also called an "alarm device") 3 including a user terminal 3A carried by a user 30 or a digital signage 3B installed in the monitoring target 20, and a server 4. In the example of FIG. 1, multiple combinations of the monitoring target 20 and the sensor system 2 are included. Also, multiple combinations of the user 30 and the user terminal 3A are included. The sensor system 2, the display device 3, and the server 4 are connected to be able to communicate with each other via a network 5. The network 5 includes an IP (Internet Protocol) network, and the devices connected to the network 5 The devices can communicate based on a predetermined protocol. Part of the network 5 is a telephone network (fixed telephone network or mobile communication network), an ad hoc network, an intranet, a VPN (Virtual Private Network), a LAN (Local Area Network), etc. , Wireless LAN, Wide Area Network (WAN), Internet, etc. It is also possible.

[0010] The sensor system 2 detects, for example, users 30 who are queuing up at the monitoring target 20 and notifies the server 4 of the number of people waiting to enter the store. The server 4 is a control device for correcting the number of people waiting to enter so as to eliminate the difference between the number of people waiting to enter and the target situation, and notifying the user 30 of the number of people waiting to enter as a predicted result of the crowded situation. The target situation may be determined relatively, for example, so that the number of people waiting to enter the multiple monitoring targets 20 is equal. For the correction, a predetermined correction amount is applied based on the difference between the most recent number of people waiting to enter and the target situation. For example, the notification is made via at least one of the user terminal 3A and the digital signage 3B (i.e., the display device 3). Note that the sensor system 2 grasps the crowded situation only as an example, and the server 4 may grasp the crowded situation by communicating with the user terminal 3A carried by the user 30, for example.

[0011] FIG. 2 is a block diagram showing an example of the system 1. The sensor system 2 is a device that calculates the number of staying people based on predetermined sensing data and transmits the calculated number to the server 4. The sensor system 2 includes a processor 21, a storage device 22, a communication interface (I / F) 23, and a sensor 24. These components are connected via a bus. The processor 21 is an arithmetic processing device such as a CPU (Central Processing Unit). The processor 21 calculates the number of staying objects based on the data output by the sensor 24 according to a predetermined program, for example, and transmits the calculated number of staying objects to the server 4 via the communication I / F 23. The storage device 22 is a main storage device such as a random access memory (RAM) or a read only memory (ROM), and an auxiliary storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The main storage device temporarily stores the program read by the processor 21 and information transmitted to and received from other computers, and secures a working area for the processor 21. The auxiliary storage device stores the program executed by the processor 21 and information transmitted to and received from other computers. The storage device 22 may be a cache memory provided in the processor 21. The communication I / F 23 is, for example, a network card or a communication module, and communicates with other computers via the network 5 based on a predetermined protocol. The sensor 24 is, for example, a camera, but may also be a human sensor that detects a person using a predetermined method, a ticket issuing machine for numbered tickets, a communication device that detects the user terminal 3A using a BLE (Bluetooth (registered trademark) Low Energy) beacon, or the like. When the sensor 24 is a camera, the processor 21 may count the number of people using, for example, a person recognition technique using machine learning.

[0012] The display device 3 receives information on the number of staying objects 20 from the server 4 and displays it on a display. The display device 3 includes a processor 31, a storage device 32, a communication I / F 33, and a user interface (UI) 34. These components are connected via a bus. The processor 31 is an arithmetic processing device such as a CPU. The processor 31 stores information on the number of staying objects received via the communication I / F 33 in the storage device 32 or outputs it to the UI 34 according to a predetermined program, for example. The storage device 32 is a main storage device such as a RAM or a ROM, and an auxiliary storage device such as an HDD, an SSD, or a flash memory. The main storage device temporarily stores programs read by the processor 31 and information transmitted and received between other computers, and secures a working area for the processor 31. The auxiliary storage device stores programs executed by the processor 31, information transmitted and received between other computers, and the like. The storage device 32 may be a cache memory provided in the processor 31. The communication I / F 33 is, for example, a network card or a communication module, and communicates with other computers via the network 5 based on a predetermined protocol. The UI 34 includes a display and displays predetermined information under the control of the processor 31 .

[0013] The server 4 is a server computer. The server 4 may be a virtual server provided by one or more cloud servers. The server 4 includes a processor 41, a storage device 42, and a communication I / F 43. These components are connected via a bus. The processor 41 is an arithmetic processing device such as a CPU, and executes a program to perform each process according to the present embodiment. In FIG. 2, functional blocks are shown in the processor 41. That is, the processor 41 executes a predetermined program to function as a receiving unit 411, a correcting unit 412, a notifying unit 413, and a learning unit 414. The receiving unit 411 acquires information from the sensor system 2 via the communication I / F 43 and stores the information in the storage device 42. The receiving unit 411 receives, for example, information indicating the number of retentions for the monitoring target 20. The correcting unit 412 corrects the retention number using, for example, a predetermined learned model. The learned model is, for example, a correction rule that learns the relationship between the retention number to be notified by correcting the actual retention number and the change in the retention number after the notification. The correction unit 412 uses the trained model to find the corrected number of people to be notified in order to bring the actual staying number closer to a predetermined target. The notification unit 413 transmits the corrected number of people to the display device 3 via the communication I / F 43. The learning unit 414 learns the relationship between the staying number to be notified by correcting the actual staying number and the change in the staying number after notification, and creates a trained model, or modifies the trained model to improve prediction accuracy. The storage device 42 is a main storage device such as RAM or ROM, and an auxiliary storage device such as HDD, SSD, or flash memory. The main storage device stores programs read by the processor 41 and other computers. The auxiliary storage device temporarily stores information transmitted to and received from other computers and secures a working area for the processor 41. The auxiliary storage device stores programs executed by the processor 41 and information transmitted to and received from other computers. The storage device 42 may be a cache memory provided within the processor 41. The communication I / F 43 is, for example, a network card or a communication module, and communicates with other computers via the network 5 based on a predetermined protocol.

[0014] <Learning process> FIG. 3 is a process flow diagram showing an example of the learning process. In the learning process, the actual congestion situation and the corrected congestion situation are fed back to the user, and the relationship between the correction amount when notifying the user of the congestion situation and the change in the congestion situation after the notification is learned. The receiving unit 411 of the server 4 is assumed to continuously perform a process of acquiring the staying number from the sensor system 2 and storing it in the storage device 42. Also, it is assumed that the target state of the correction is defined in advance. The target state may be determined based on the relative congestion situation of the multiple monitoring targets 20, such as a state in which the staying numbers are uniform for the multiple monitoring targets 20 or a state in which the staying numbers are at a predetermined ratio. Also, a specific numerical value that is a target of the staying number may be set for each monitoring target 20. The specific numerical value may be set for each predetermined period, such as a season or a time period.

[0015] First, the notification unit 413 of the server 4 reads the most recent number of users waiting for the monitoring target 20 from the storage device 42 and notifies the user 30 (FIG. 3: S1). For example, the notification unit 413 reads the number of users waiting for each of the multiple monitoring targets 20 and notifies the user 30 without correcting the read number of users. FIG. 4 is a diagram for explaining an example of a change in the number of users waiting. For example, at the time shown in FIG. 4 (1), information is read out indicating that four users 30 are waiting at the monitoring target 20A (referred to as "store A") and one user 30 is waiting at the monitoring target 20B (referred to as "store B"). The user terminal 3A is also notified that the number of users waiting at store A is four and the number of users waiting at store B is one.

[0016] Next, after a predetermined time has elapsed since S1, the correction unit 412 reads out the most recent number of people staying for the monitoring target 20 from the storage device 42, and determines an initial correction rule (FIG. 3: S2). The predetermined time is set in advance, for example, 30 minutes. In this step, a correction rule is determined for correcting the actual number of people staying to a predicted value of the congestion state to be notified to the user. In this embodiment, in order to bring the number of people staying closer to the target state, the user is notified of a predicted value of the congestion state that is increased or decreased from the actual state. For example, when the most recent number of people staying is greater than the target state, the user is notified of a corrected predicted value that is increased from the actual number of people staying in order to encourage the number of people staying to decrease. On the other hand, when the most recent number of people staying is less than the target state, the user is notified of a corrected predicted value that is decreased from the actual number of people staying in order to encourage the number of people staying to increase. Initially, a correction rule is created such that, for example, the greater the difference between the actual number of remnants and the target value, the greater the correction amount, and when the difference between the actual number of remnants and the target value is small (in other words, when the actual number of remnants has converged to the target state), the correction amount is relatively small or zero.

[0017] The degree of correction in the correction rule may be determined according to the degree of change in the number of people staying in the period from S1 to S2. The degree of change in the number of people staying in the period from S1 to S2 represents the result of the user changing his / her behavior in consideration of the congestion situation. Therefore, if the number of people staying in the period from S1 to S2 changes so as to approach the target more than a predetermined standard, the correction rule may be created so as to adjust the amount of correction relatively small. If the number of people staying in the period from S1 to S2 does not approach the target as much as the predetermined standard, the correction rule may be created so as to adjust the amount of correction relatively large. In other words, the correction rule can be said to be a prediction model that learns the relationship between the amount of correction when correcting the actual congestion situation for notification and the change in the congestion situation after the congestion situation is notified.

[0018] The correction rule may be a coefficient for correcting the actual number of people staying (the most recent number of people staying obtained from the sensor system 2) to determine a predicted value of the crowded state (the number of people to be notified). FIG. 5 is a diagram for explaining the correction rule. In FIG. 5(A), the horizontal axis represents the actual number of people staying, and the vertical axis represents the number of people to be notified. When the target state is defined by a specific numerical value (fixed value), the correction rule can be defined by a curve representing the relationship between the actual number of people staying (in other words, the difference between the actual number of people staying and the target value) and the number of people to be notified on a plane (quadrant) as shown in FIG. 5(A). Such a correction rule may be created for each monitored object 20 such as a store. In addition, the correction rule may be created for each type of situation, such as whether the situation of the monitored object 20 is in a busy time period or not.

[0019] The correction rule may be a model that has been machine-learned to determine the relationship between the difference between the actual number of people staying and the target value, and the number of people to be notified. That is, the machine learning may be a prediction model that calculates a predicted value of the number of people to be notified to approach the target state based on the difference between the actual number of people staying and the target value. Existing methods such as regression analysis and learning methods using neural networks can be used. (B) of FIG. 5 shows a three-dimensional diagram including an axis that represents the actual number of people staying, an axis that represents the difference from the target, and an axis that represents the number of people to be notified (the number of people after correction). In such a space (quadrant), the correction rule may be defined by a multiple regression plane created by regression analysis, for example. In addition to regression analysis, a correction rule that is expressed by a curved surface in a space such as that shown in (B) of FIG. 5 may be defined. In addition, a parameter that represents a state such as whether it is a busy time period or not, a parameter that represents the congestion state of other nearby stores, and the like may be used to perform regression analysis, for example, and define the correction rule by a hyperplane. In S2 of FIG. 3, an initial correction rule is created.

[0020] Then, the correction unit 412 corrects the number of staying users read in S2 and notifies the user 30 of the number of staying users as a predicted result of the crowded state (FIG. 3: S3). In this step, the correction unit 412 performs correction based on the correction rule created in S2, and the notification unit 413 transmits the number of users after correction to the display device 3. For example, at the time point shown in FIG. 4 (2), information indicating that four users 30 are staying in store A and two users 30 are staying in store B is read. In addition, when the number of staying users in each store is set to be the same (in other words, the ratio of the number of staying users is 1:1) as the target state, the number of users calculated by dividing the number of staying users in all stores in proportion to the ratio of each store is calculated as a specific target value. For example, the number of staying users in store A and store B is calculated to be 3.0 people as a specific target value. In addition, according to a predetermined correction rule, for example, for store A where the number of staying users is more than the target value, the number of users after correction is calculated to be, for example, "6 people" which is higher than the actual number, in order to encourage a decrease in the number of staying users. Similarly, for store B, where the number of people staying is less than the target value, in order to encourage an increase in the number of people staying, the corrected number of people is calculated to be, for example, "1 person," which is less than the actual number. Then, the calculated corrected number of people is notified to the user terminal 3A.

[0021] After that, the correction unit 412 judges whether the staying number has converged to a target state (FIG. 3: S4). In this step, the correction unit 412 obtains the staying number after a predetermined time from S2. Then, the correction unit 412 calculates a target value at that time. That is, as a specific target value, the number of people calculated by apportioning the staying numbers at all stores at that time according to the ratio of each store is calculated. Furthermore, the correction unit 412 judges that the number has converged if the difference between the target value and the actual staying number is within a predetermined threshold, for example.

[0022] If it is determined that the number of remaining nodes has not converged (S4: NO), the process returns to S3, and the process of further correcting and notifying the number of remaining nodes acquired in S4 is repeated. On the other hand, if it is determined that the number of remaining nodes has converged (S4: YES), the learning unit 414 of the server 4 performs a learning process of the correction rule (FIG. 3: S5). In this step, the correction rule is updated so that the time required for convergence is shortened, for example. For example, the amount of correction may be increased to strongly encourage the user to change his / her behavior, or may be decreased to suppress the user from changing his / her behavior.

[0023] Also, according to a predetermined correction rule, for example, when the number of staying users converges to a target state, the correction amount is reduced or set to zero, and the correction unit 412 calculates a predicted value of the crowded state, and the notification unit 413 transmits it to the display device 3. At the time point shown in (3) of FIG. 4, information indicating that three users 30 are staying in store A and two users 30 are staying in store B is read out. As specific target values, the numbers of users staying in store A and store B are calculated to be 2.5 people each. Then, for example, when the difference between the actual number of users staying and the target value is within one person (threshold value), it is determined that the number of users staying is converged, and the actual number of users staying is notified to the user terminal 3A.

[0024] <Operational processing> FIG. 6 is a process flow diagram showing an example of the operation process. In the operation process, the congestion state corrected using the correction rule created in the learning process is fed back to the user. In addition, if the congestion state does not converge to the target state within the expected period, the relationship between the correction amount when notifying the user of the congestion state and the change in the congestion state after the notification is re-learned as necessary. In addition, if business hours are set for the monitored object, the operation process may be started when the business starts, or may be executed only during the busy period of the monitored object. In addition, the receiving unit 411 of the server 4 continuously performs the process of acquiring the number of staying people from the sensor system 2 and storing it in the storage device 42. In addition, the state that is the target of the correction is defined in advance.

[0025] The correction unit 412 reads out the most recent number of people staying for the monitoring target 20 from the storage device 42 (S11 in FIG. 6). Then, the correction unit 412 corrects the number of people staying read out in S11 and notifies the user 30 of the result of the predicted congestion situation (S12 in FIG. 6). In this step, the correction unit 412 performs correction based on a correction rule created in advance, and the notification unit 413 transmits the corrected number of people to the display device 3. That is, based on the predetermined correction rule, for example, the larger the difference between the actual number of people staying and the target value, the larger the correction amount is, and when the difference between the actual number of people staying and the target value is small, the correction amount is relatively small or zero.

[0026] The learning unit 414 also judges whether the accuracy of the correction rule is sufficient (FIG. 6: S13). For example, even if a difference occurs between the staying number and the target value that exceeds a predetermined threshold, if the staying number converges to the target state within an expected time, the learning unit 414 judges that the accuracy of the correction rule is sufficient. For example, the judgment of convergence is performed after a predetermined time from an arbitrary starting point after the notification. Furthermore, within the predetermined time, the notification of the congestion state may be further repeated. The learning unit 414 may receive an input of the judgment result of whether the accuracy is sufficient or not from the user, and make a judgment according to the input information. If the staying number does not converge to the target state within an expected time (S13: NO), the learning process of the correction rule is performed (FIG. 6: S14). This step is the same as S5 in FIG. 3. Furthermore, if it is judged that the accuracy is sufficient in S13 (S13: YES), or after S14, the correction unit 412 judges whether to end the operation process (FIG. 6: S15). For example, if the monitored object 20 is a store, the operation process may be continuously executed during business hours or during a specified busy period, and may be terminated when business hours or the specified busy period ends. If it is determined that the operation process should be terminated (S15: YES), the process of FIG. 6 is terminated. On the other hand, if it is determined that the cloud-like process should not be terminated (S15: NO), the process returns to S11 and is repeated.

[0027] In this embodiment, by performing so-called feedback control, in which the notification of the congestion state to users and the learning process of S14 are repeated, the accuracy of the correction rule can be improved and appropriate people flow control can be performed. An example has been shown in which, when the number is large, the user is notified of a predicted value that is corrected to be greater than the actual number of waiting states in order to encourage a decrease in the number of waiting states, and, when the actual number of waiting states is smaller than the target state, the user is notified of a predicted value that is corrected to be less than the actual number of waiting states in order to encourage an increase in the number of waiting states, but the final correction rule is not limited to such examples.

[0028] That is, if there is a tendency that the number of staying people easily reverses among multiple monitoring targets 20 simply by notifying the actual number of staying people, it is possible that the learning process results in a correction to make the difference smaller than the actual number of staying people and notifying the result. FIG. 9 is a diagram for explaining another example of the change in the number of staying people. For example, in S1 of FIG. 3, it is assumed that information indicating that four users 30 are staying at monitoring target 20A (referred to as "store A") and one user 30 is staying at monitoring target 20B (referred to as "store B") is read out as shown in (1) of FIG. 9. At this time, similar to the example of FIG. 4, the user terminal 3A is notified that the number of people waiting at store A is four and the number of people waiting at store B is one.

[0029] At the time shown in FIG. 9 (2), information indicating that two users 30 are staying at store A and four users 30 are staying at store B is read out. That is, when there is a difference between the actual number of users staying at two stores, the number of users staying at each store is reversed after the actual number of users is notified in FIG. 9 (1). The target state is set to be the same number of users staying at each store, and the number of users staying at each store A and store B is calculated to be 3.0 people as a specific target value. In this case, if the actual number of users staying is notified in FIG. 9 (2), the number of users staying at each store may be reversed again. In addition, this may reduce the reliability of the notification for users 30. Therefore, for store A, which has a smaller number of users staying than the target value, a smaller number than the actual number is calculated as the number of users staying at store A, for example, "3 people" in order to suppress an increase in the number of users staying at store B. For store B, which has a larger number of users staying than the target value, a larger number than the actual number is calculated as the number of users staying at store B, for example, "3 people" in order to suppress a decrease in the number of users staying at store B.

[0030] At the time point shown in FIG. 9(3), information is read out indicating that three users 30 are staying in store A and two users 30 are staying in store B. As specific target values, the numbers of users staying in store A and store B are calculated to be 2.5 people each. Then, for example, when the difference between the actual number of users staying and the target value is within 1 person (threshold), it is determined that convergence has occurred, and the actual number of users staying is notified to user terminal 3A.

[0031] In addition, depending on the monitoring target 20, users may not flow to vacant places, but may instead concentrate on restaurants with long queues. In such cases, the feedback control according to the present embodiment can correct the correction rule to get as close as possible to the target state, instead of performing uniform control. FIG. 10 is a diagram for explaining another example of the change in the number of staying users. For example, as shown in (1) of FIG. 10, information is read out indicating that four users 30 are staying in the monitoring target 20A (referred to as "store A"), two users 30 are staying in the monitoring target 20B (referred to as "store B"), and eight users 30 are staying in the monitoring target 20C (referred to as "store C"). In addition, the target state is set to have the same number of staying users in each store, and the specific target values ​​are calculated to be 4.66··· people for each of stores A to C. In FIG. 10 (1), for stores with more waiting times than the target value, a corrected number of people greater than the actual number is notified in order to prevent the waiting times from increasing, and for stores with fewer waiting times than the target value, a corrected number of people greater than the actual number is notified in order to prevent the waiting times from decreasing.

[0032] In addition, in (2) of FIG. 10, information is read out indicating that three users 30 are staying at store A, two users 30 are staying at store B, and eight users 30 are staying at store C. In this case, the number of people waiting at each of stores A to C is calculated as 4.33··· people as a specific target value. In (2) of FIG. 10, for a store with a higher number of people waiting than the target value, a smaller number of people than the actual number is notified as the corrected number of people, and for a store with a lower number of people waiting than the target value, a larger number of people than the actual number is notified as the corrected number of people. However, as shown in (1) and (2) of FIG. 10, if there are regular customers at store C regardless of the feedback related to the present disclosure, the number of people waiting at store C may not increase or decrease. In such a case, the target state of store C may be set as an absolute number of people independent of stores A and B, rather than being expressed as a relative number of people with respect to stores A and B. For example, in (3) of FIG. 10, store C has a higher number of people waiting than stores A and B, but the number of people without correction is displayed as having converged to the target state overall.

[0033] As described above, the administrator who operates the monitored objects 20 can maintain the service level and reduce the opportunity loss of each monitored object 20. Furthermore, by improving the accuracy of the correction rule through the above-mentioned learning process, it is possible to reduce stress experienced by users of the monitored objects 20, for example, when users gather at a facility hoping for an empty space only to find that the facility is crowded contrary to their expectations.

[0034] The group of monitoring targets 20 for which the correction of the number of staying is performed is the monitoring targets 20 that are somehow related to each other. For example, it is preferable to target multiple monitoring targets 20 in the vicinity so that the users 30 can be dispersed. It is also preferable to perform the above-mentioned correction in a group of monitoring targets 20 that are the same type, such as a store, a counter, an attraction such as an amusement park, a public toilet, a parking lot, a road section, etc. In particular, it is preferable that the more detailed categories, such as the type of product or service or the business type, are the same or similar among the monitoring targets 20 of the same type. It is also possible to target multiple monitoring targets 20 that exist at a distance that can be reached by walking or car within a first threshold value of time or a first threshold value of distance. It is also preferable that the multiple monitoring targets 20 are separated by a second threshold value of time or a second threshold value of distance so that a user 30 standing next to one monitoring target 20 cannot be visually recognized by a user 30 standing next to the other monitoring target 20. Furthermore, when an obstacle such as a wall is present between users 30 standing next to multiple monitoring targets 20, the second threshold value of time and the second threshold value of distance may be corrected to be smaller than when no obstacle is present. In this way, the notification according to the present disclosure is more effective.

[0035] <Embodiment 2> In addition to notifying the user, a notification may also be sent to the administrator of the monitored object 20. The administrator may be, for example, a business operator or manager of a store, a parking lot, or the like. FIG. 7 is a diagram for explaining an example of the system configuration. Components common to the first embodiment are given the same reference numerals and will not be explained. The system of FIG. 7 further includes an administrator terminal 6 used by the administrator, and the administrator terminal 6 is also connected to the network 5. The administrator terminal 6 is a computer such as a PC (Personal Computer), and the device configuration is similar to that of the display device 3 shown in FIG. 2. do.

[0036] FIG. 8 is a diagram for explaining an example of a notification to an administrator. At the time shown in FIG. 8 (4), information is read out indicating that four users 30 are staying at store A and two users 30 are staying at store B. In addition, if the target state is set, for example, such that the number of users staying at each store is the same, a specific target value is calculated by apportioning the number of users staying at all stores according to the ratio of each store. For example, the specific target values ​​are calculated as 3.0 users for each of store A and store B. In addition, according to a predetermined correction rule, for example, for store A where the number of users staying is greater than the target value, a corrected number of users is calculated to be, for example, "6 users," which is greater than the actual number, in order to encourage a decrease in the number of users. Similarly, For store B, which has fewer staying people than the target value, the corrected number of people is calculated to be, for example, "1 person," which is less than the actual number, in order to encourage an increase in the staying number. The user terminal 3A is then notified of the calculated corrected number of people. Meanwhile, the notification unit 413 of the server 4 displays, for example, the actual staying number and the target state on the UI (user interface) of the administrator terminal 6. The target state can be said to be, in other words, a predicted value of a state that will converge after a predetermined time. The notification unit 413 of the server 4 may also display the corrected number of people on the administrator terminal 6. At the time point shown in (5) of FIG. 8, information indicating that three users 30 are staying in store A and two users 30 are staying in store B is read out. As a specific target value, the staying numbers of stores A and B are each calculated to be 2.5 people. Then, for example, when the difference between the actual staying number and the target value is within one person (threshold value), it is determined that the number of staying people has converged, and the actual staying number is notified to the user terminal 3A. At this time, the notification unit 413 of the server 4 displays, for example, the actual number of staying objects and the target state on the UI of the manager terminal 6. The manager terminal 6 may be further notified of information indicating that the target state has been converged to. Such a display is useful for the manager to grasp the status of the monitoring target 20.

[0037] <Modification> The created correction rule may be applied to other monitored objects 20. For example, a learned correction rule created for a certain food court, parking lot, etc. is applied to other monitored objects 20 as in so-called transfer learning, and further learning processing is performed in the operation processing of the other monitored objects 20. Transfer learning is a learning method in which a model that has been created and learned in advance with a certain monitored object 20 (or a group of monitored objects 20) as a learning object is reused for a learning processing of a different monitored object 20 (or a group of monitored objects 20). Transfer learning makes it possible to reuse a model with a small learning cost for optimizing it for a different monitored object 20. In transfer learning, the learned correction rule may be applied to monitored objects 20 with a similar configuration in particular. That is, a model may be reused for monitored objects 20 with the same type, such as a store, a counter, an attraction such as an amusement park, a public toilet, a parking lot, a road section, etc. Furthermore, among the monitored objects 20 of the same type, a model may be reused for monitored objects 20 with the same or similar more detailed categories, such as the type of product or service or the business type. In particular, it is suitable to reuse the model when the configurations are similar in food courts, parking lots, public restrooms, etc. In this way, the time required to create correction rules can be reduced.

[0038] Instead of the sensor system 2, the server 4 may acquire information from the user terminal 3A carried by the user 30, and the server 4 may calculate the staying number. For example, the sensor system 2 may transmit image data to the server 4, and the server 4 may detect the staying number by performing person recognition processing.

[0039] Also, instead of the sensor system 2, the user terminal 3A may calculate position information using, for example, a satellite positioning system (GNSS: Global Navigation Satellite System), and the server 4 may collect position information from each user terminal 3A. That is, a program installed in the user terminal 3A may obtain position information indicating the position of the user terminal 3A, and transmit the position information to the server 4, for example, while the congestion situation according to the above-mentioned embodiment is being displayed, or constantly in the background.

[0040] The monitoring target may be a single location. For example, a target congestion level may be set for each time period at one counter, and predicted information corrected for the actual congestion level may be notified to users. In this way, it is possible to distribute the usage time of users and alleviate congestion at the counter. On the other hand, if the method is applied to a set of multiple monitoring targets as described above, it is effective in that it is possible to distribute congestion among the multiple monitoring targets.

[0041] The congestion status may be converted into waiting time or the like instead of the number of people and notified.

[0042] <Other> The illustrated system and device configurations are merely examples, and are not limited to the above examples. For example, some of the functions of the server 4 may be provided by other computers. For example, the learning process, prediction process, and notification process may be performed by different devices. In addition, at least some of the illustrated functions may be shared and executed by multiple devices, or may be executed in parallel by multiple devices. Similarly, the processes shown in the processing flow may be executed in a different order or in parallel, as long as the results are the same.

[0043] The present technology also includes a computer program for executing the above-mentioned processing method, and a computer-readable recording medium on which the program is recorded. The above-mentioned processing is possible by reading the recording medium into a computer and executing the recorded program. A computer-readable recording medium is a recording medium that can store information such as data and programs by electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer. Among such recording media, those that can be removed from a computer include flexible disks, magneto-optical disks, optical disks, magnetic tapes, memory cards, etc. Furthermore, recording media fixed to a computer include hard disk drives and ROMs, etc.

[0044] The gist of the present disclosure is as follows. (Aspect 1) reading out the detected congestion status of the monitoring target from a storage device, correcting the read-out congestion status based on a difference between the detected congestion status and a target status, and outputting a notification instruction signal of the corrected congestion status to a notification device that can be confirmed by a user of the monitoring target; This is a control method in which a computer performs feedback control that repeats the above steps. (Aspect 2) The correction is performed using a prediction model that has learned in advance the relationship between the amount of correction when correcting the actual congestion status for notification and the change in the congestion status after the corrected congestion status is notified. A control method according to embodiment 1. (Aspect 3) The monitored object includes a plurality of areas, and the target state is determined based on the relative congestion status of the plurality of areas. A control method according to embodiment 2. (Aspect 4) The predictive model is created for each of the plurality of regions. A control method according to embodiment 3. (Aspect 5) The prediction model is created for each type of situation to be monitored. A control method according to any one of claims 2 to 4. (Aspect 6) The predictive model is created by transfer learning. A control method according to any one of aspects 2 to 5. (Aspect 7) If it is determined that the congestion state does not converge to the target state within a predetermined time after the output to the notification device, the relationship between the correction amount and the change in the congestion state is re-learned and the prediction model is updated. A control method according to any one of claims 2 to 6. (Aspect 8) The read congestion status and the target status are further notified to a manager of the monitored object. A notification method according to any one of aspects 1 to 7. (Aspect 9) reading out the detected congestion status of the monitoring target from a storage device, correcting the read-out congestion status based on a difference between the detected congestion status and a target status, and outputting a notification instruction signal of the corrected congestion status to a notification device that can be confirmed by a user of the monitoring target; The control device includes a control unit that executes feedback control by repeatedly performing the above steps. [Explanation of symbols]

[0045] 1: System 2: sensor system, 21: processor, 22: storage device, 23: communication I / F, 24: sensor 3: display device (3A: user terminal, 3B: digital signage), 31: processor, 32: storage device, 33: communication I / F 4: Server 41: Processor, 411: Receiving unit, 412: Correcting unit, 413: Notifying unit, 414: Learning unit, 42: Storage device, 43: Communication I / F 5: Network 6: Administrator terminal 20 (20A, 20B): Monitored 30: User

Claims

1. reading out the detected congestion status of the monitoring target from a storage device, correcting the read-out congestion status based on a difference between the detected congestion status and a target status, and outputting a notification instruction signal of the corrected congestion status to a notification device that can be confirmed by a user of the monitoring target; This is a control method in which a computer performs feedback control that repeats the above steps.

2. The correction is performed using a prediction model that has learned in advance the relationship between the amount of correction when correcting the actual congestion status for notification and the change in the congestion status after the corrected congestion status is notified. The control method according to claim 1 .

3. The monitored object includes a plurality of areas, and the target state is determined based on the relative congestion status of the plurality of areas. The control method according to claim 2 .

4. The predictive model is created for each of the plurality of regions. The control method according to claim 3.

5. The prediction model is created for each type of situation to be monitored. The control method according to claim 2 .

6. The predictive model is created by transfer learning. The control method according to claim 2 .

7. If it is determined that the congestion state does not converge to the target state within a predetermined time after the output to the notification device, the relationship between the correction amount and the change in the congestion state is re-learned and the prediction model is updated. The control method according to claim 2 .

8. The read congestion status and the target status are further notified to a manager of the monitored object. A control method according to any one of claims 1 to 7.

9. reading out the detected congestion status of the monitoring target from a storage device, correcting the read-out congestion status based on a difference between the detected congestion status and a target status, and outputting a notification instruction signal of the corrected congestion status to a notification device that can be confirmed by a user of the monitoring target; The control device includes a control unit that executes feedback control by repeatedly performing the above steps.

Citation Information

Patent Citations

  • Toilet management method, toilet management system, toilet management device, and toilet management program

    JP2022039679A