WO2021256848A1 - Method and system for predicting congestion level for each subway cabin on basis of short-range wireless communication signal - Google Patents

Abstract

One aspect of the present invention discloses a method by which a server device predicts a congestion level for each subway cabin. The method comprises the steps of: receiving channel information from a first access point (AP) through a network (wherein the channel information includes a link metric measured on the basis of a communication signal with a second AP, and the first AP is provided in a first subway cabin and the second AP is provided in a second subway cabin that is adjacent to the first subway cabin); calculating the congestion level of the first subway cabin or the second subway cabin on the basis of the received channel information (wherein the congestion level indicates the level related to the number of passengers present in the subway cabin); and providing the calculated congestion level of the subway cabin to a user terminal.

Description

Method and system for predicting congestion level by subway room based on short-distance wireless communication signal

The present invention relates to a method of predicting subway congestion, and more particularly, to a method for more accurately calculating the number of passengers in a subway.

Recently, a bus information/management system (BIS/BMS) that provides bus operation information in a jurisdictional area is being built around not only large cities such as metropolitan cities but also regional large cities. This bus information guide system collects the location of the running bus in real time from the traffic information center and processes it into traffic information such as the expected arrival time at the stop, the current bus location by route, the distance to the vehicle behind, and transfer information, etc. It is a system that provides information to users of personal terminals through information provided at each bus stop or through the Internet network, etc. Currently, even the congestion level of individual buses is provided.

However, office workers in the metropolitan area often use the subway to adhere to their commuting hours, but feel uncomfortable because the subway information system does not provide a congestion information function. In addition, since each subway carriage is connected to the subway, there is a technical problem in that it is difficult to adopt the information guide system of the bus as it is.

In particular, the subway is formed in a structure in which a plurality of carriages are connected to each other, and a space for accommodating passengers is formed in the interior of the carriage, which is automatically opened and closed when stopped at a subway station. Passengers wait for the subway at the platform of the subway station, and when the door is opened after the car arrives at the designated location, they get on the car. A number of chairs and handles are provided inside the carriage for passengers to sit on and for passengers to move while standing, but the number of chairs and handles is limited, providing the same convenience to passengers when carrying an unlimited number of passengers There is a problem that cannot be done. In particular, during rush hour, such as rush hour, the number of subway users increases rapidly, while the number of subway trains is limited, so the number of passengers in one carriage increases. When more passengers board a single carriage, the distance between the passengers decreases, increasing the discomfort of the passengers, and there is a risk that passengers exceeding the number of chairs and handles may be exposed to safety accidents.

In particular, a relatively large number of passengers are repeatedly boarded only in a specific carriage due to a phenomenon in which passengers are flocked to carriages that are stopped close to the stairs at the platform entrance, so there is a problem in that the load applied to the carriage increases and the wear and tear is biased. .

An object according to an aspect of the present invention for solving the above-mentioned problems is to predict the number of passengers in the cabin using an access point (AP) for wireless short-range communication installed in the subway train cabin based on a short-distance wireless communication signal. It is to provide a method and system for predicting congestion for each subway room.

According to an aspect of the present invention for achieving the above object, a method of predicting congestion for each subway cabin in a server device includes the steps of receiving channel information from a first access point (AP) through a network (The channel information includes a link metric measured based on a communication signal with a second access point, and the first access point is an access point installed in a first subway cabin, and the second access point is an access point installed in a second subway cabin adjacent to the first subway cabin), calculating a congestion degree of the first subway cabin or the second subway cabin based on the received channel information (the congestion degree is the second subway cabin) 1 indicating a level related to the number of passengers present in the subway cabin) and providing the calculated congestion degree of the first subway cabin to the user terminal.

The link metric may include at least one of Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and Compressed Beamforming Feedback (CBF) of a wireless short-range communication signal.

The received channel information includes (i) the link metric, (ii) at least one of identification information of the first subway cabin and identification information of a first access point, (iii) identification information of the second subway cabin and a second At least one of identification information of the access point, (iv) timestamp information indicating a time when data is collected, and (v) scanned identification information of neighboring access points may be included.

Calculating the congestion level of the first subway cabin or the second subway cabin based on the received channel information includes determining whether the first subway cabin is moving and using the channel information based on the determination result to calculate the degree of congestion of the first subway cabin or the second subway cabin.

The step of determining whether the first subway cabin is moving may include parsing identification information of neighboring access points from the channel information, and determining by comparing the parsed information with information of pre-stored subway station fixed access points. can

The parsed information includes identification information of a neighboring access point of the first access point, and in response to that identification information of the neighboring access point does not exist in the pre-stored information of the subway station fixed access point, currently the second It is determined that the first subway cabin is moving, and based on the determination that the first subway cabin is moving, calculation of the congestion level of the first subway cabin may be initiated (initiating).

Calculating the congestion level of the first subway cabin or the second subway cabin by using the channel information may include mapping the link metric to a congestion level using an artificial intelligence (AI) model. It may include calculating the congestion level of the subway cabin or the second subway cabin.

The AI model may be a machine learning model learned by using, as a training data set, level information of an actual passenger in an actual subway cabin and a link metric corresponding thereto.

The method further includes receiving a congestion level request from the user terminal, wherein the congestion level request includes identification information of the first subway cabin, and in response to the congestion level request, the congestion level of the first subway cabin is displayed It can be provided to a user terminal.

The method may further include verifying an API parameter of the congestion request and verifying the validity of an API security token associated with the congestion request.

The method further comprises, after receiving the channel information from a first access point (AP), verifying the validity of the received channel information, wherein the validity of the channel information is (i) the verified by verifying that the first subway cabin or the first access point is registered, and (ii) verifying the length of the channel information.

The method further comprises, based on a determination that the received channel information is valid, storing the channel information, wherein the channel information may be stored together with a corresponding passenger level result.

According to another aspect of the present invention for achieving the above object, an apparatus for predicting congestion for each subway cabin is an AP module (the channel) for receiving channel information from a first access point (AP) through a network. The information includes a link metric measured based on a communication signal with and from a second access point, the first access point being an access point installed in a first subway cabin, and the second access point being the second access point. 1 is an access point installed in the second subway cabin adjacent to the first subway cabin), a processor for calculating the congestion degree of the first subway cabin or the second subway cabin based on the received channel information (the congestion degree is in the first subway cabin) and an application module that provides a level (level) related to the number of existing passengers) and the calculated congestion level of the first subway cabin to the user terminal.

According to another aspect of the present invention for achieving the above object, a first access point device for predicting congestion for each subway cabin measures a channel state based on a communication signal with a second access point, and the measured An access point scan module for transmitting channel information including a channel state to a server through a network, and a memory for storing the channel information, wherein the channel state includes a link metric, and the second access point includes the first It may be an access point installed in the second subway cabin adjacent to the first subway cabin in which the access point device is installed.

The link metric may include at least one of Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and Compressed Beamforming Feedback (CBF) of a wireless short-range communication signal.

The access point scan module may generate the channel information by collecting at least one link metrics for a preset period.

When a plurality of link metrics are collected for the preset period, the channel information may be generated by calculating the aggregated data for the preset period by calculating and aggregating an average value of the link metrics.

The channel information may include (i) the link metric, (ii) at least one of identification information of the first subway cabin and identification information of a first access point, (iii) identification information of the second subway cabin and a second access point. It may include at least one of identification information of (iv) timestamp information indicating a time when data is collected, and (v) identification information of scanned neighboring access points.

A period for transmitting the channel information to the server may be longer than a period for measuring the link metric.

The access point scan module may store a current link metric measurement result and prepare for a next link metric measurement based on receiving a response from the server.

According to the method and system for predicting the congestion level for each subway cabin based on the short-range wireless communication signal of the present invention, the congestion level for each subway room is analyzed in real time and provided to passengers. It has the effect of making this possible and has the advantage of being easy to respond to an accident.

1 is a conceptual diagram schematically showing a system for executing a method for predicting congestion for each subway cabin according to an embodiment of the present invention;

Fig. 2 is a detailed block diagram showing the system of Fig. 1 in more detail;

3 is a flowchart showing the connection operation of the subway access point, the server device, and the AI model;

4 is a flowchart showing a link operation between a user terminal and a server device;

5 is an ER diagram (Entity-Relationship Diagram) showing the structure and relationship of data managed by the database;

6 is a flowchart illustrating a learning and testing process of an algorithm for predicting the number of passengers in a cabin of an AI model.

7 is a flowchart illustrating a learning process of a prediction algorithm of an AI model according to another embodiment of the present invention.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. and/or includes a combination of a plurality of related listed items or any of a plurality of related listed items.

When an element is referred to as being “connected” or “connected” to another element, it is understood that it may be directly connected or connected to the other element, but other elements may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. In describing the present invention, in order to facilitate the overall understanding, the same reference numerals are used for the same components in the drawings, and duplicate descriptions of the same components are omitted.

1 is a conceptual diagram schematically illustrating a system for executing a method of predicting congestion for each subway cabin according to an embodiment of the present invention. As shown in FIG. 1 , the congestion prediction system for each subway cabin according to an embodiment of the present invention includes access points 110-1 and 110-2, servers 100 and client terminals 130-1 and 130-2. ) may be included.

Referring to FIG. 1 , access points 110 - 1 and 110 - 2 are devices that support a wireless device to connect to a network, and may be installed in subway cabins 112-1 to 112-N. It may be installed per subway cabin in one or more channels. The access point 110-1 may be installed in the subway cabin 112-1, and the access point 110-2 may be installed in the subway cabin 112-2 adjacent to the previous subway cabin 112-1. Although expressed as "subway" in this specification, the part expressed as subway is replaced by other means of transportation such as general trains, trains, trains, express trains, and electric trains that perform similar functions It will be apparent to those of ordinary skill in the art that the congestion degree of the present invention can be calculated even though the present invention pertains. In addition, the term "cabin" may be used interchangeably with terms such as a carriage, a car, and the like in this specification.

Meanwhile, the access points 110 - 1 and 110 - 2 may be implemented as wireless short-range communication routers. For example, it may be implemented as a Wi-Fi (Wi-Fi) router. However, the wireless communication method is not necessarily limited to Wi-Fi. In the present specification, other wireless communication methods such as Zigbee and Bluetooth may be used in the part where the term "Wi-Fi" is used, and the access points 110-1 and 110-2 are also Wi-Fi. It can be implemented as a router that supports other wireless communication methods other than PI.

The access points 110-1 and 110-2 perform a function of measuring a Wi-Fi channel state in a subway cabin. These may be referred to as link metrics. The link metric may include at least one of Receiver Signal Strength Indicator (RSSI), Channel State Information (CSI), and Compressed Beamforming Feedback (CBF). In order to obtain the Wi-Fi link metric in the guest room, the access point 110-1 becomes a monitoring router and must be able to monitor another Wi-Fi router 110-2 installed in the adjacent guest room 112-2. That is, the access point 110-1 transmits a link metric measurement result based on a signal communicated with the access point 110-2 installed in the adjacent subway car 112-2 to the server 120 through the Internet connection. do.

The server 120 stores link metrics received from a plurality of access points 110-1 to 110-2 through the Internet, and performs a logical operation to predict and classify the passenger level of the subway train cabin. do. The server 120 may be implemented as a computing device, and is preferably implemented as a server-class computing device capable of processing large amounts of information. For example, the server 120 may be a dedicated physical server or a virtual cloud server such as Amazon Web Services (AWS), Alibaba Cloud, or Google Cloud Platform.

The server 120 may infer the passenger level corresponding to the link metric measurement result by using a pre-trained AI model. That is, based on the link metric information received from the access points 110-1 to 110-2, the level of the passenger in the subway cabin in which each of the access points 110-1 to 110-2 is installed is individually determined. can be inferred as

Meanwhile, before storing in the database, the server 120 first verifies the validity of the link metric received from the access points 110-1 to 110-2. As a result of validation, only when it is valid, the server 120 stores the corresponding link metric data in the database. Then, as described above, using the validly received data as an input function, it is input to the AI model, and the congestion level of the subway cabin is obtained as an output thereof. This may be output as a specific number, but may be derived as a grouped result called a level. For example, you could classify 100 people or less as "Free" level, 100-200 people as "Normal" level, 200-250 people as "Congested" level, and 250 or more people as "Very congested" level. have. In this case, the threshold value for classifying each level may be changed through user setting. Such level classification results are stored in the database so that the client terminals 130-1 to 130-2 can check them.

The server 120 responds to all data requests of the app of the client terminals 130-1 to 130-2 with respect to the passenger level for each subway cabin. The client terminals 130-1 to 130-2 are entities that directly interact with the user. The client terminals 130-1 to 130-2 may be portable terminals (eg, smart phones, laptops, PDAs, etc.), and fixed terminals (eg, electronic billboards, public display devices, PCs, and control devices). system, etc.). A client application (this may be referred to as a “client app”) may be installed in the client terminals 130-1 and 130-2 to request subway congestion information from the server 120 and receive the result. The client app sorts and filters data required by the user and sends a request to the server 120 . Then, after receiving a response to the request from the server 120, the data is directly displayed and visualized to the user. The client app may be implemented as a mobile application or a web app. In addition, in the case of a fixed terminal, such as a public display installed on a subway station platform, the result received from the server 120 can be displayed and visualized without an active request toward the server 120 .

The client terminals 130-1 to 130-2, like the access points 110-1 to 110-2, also need Internet access to communicate with the server 120. The access points 110-1 to 110-2 and the client terminals 130-1 to 130-2 communicate with the server 120 through the Restful API defined on the server 120 side. Restful API is an application program interface based on REST (Representational State Transfer). REST is a software architectural style that defines a set of constraints to be used to create a web service, and a web service conforming to the REST architectural style called RESTful Web Service provides interoperability between computer systems on the Internet. Restful API uses HTTP methodology defined in RFC 2616 protocol. In other words, Restful API uses GET method to retrieve resource, PUT method to change state or update resource, POST method to create resource, and remove it. To do this, use the DELETE method. Since the call (request) is stateless, REST is useful in cloud applications.

FIG. 2 is a detailed block diagram illustrating the system of FIG. 1 in more detail. As shown in FIG. 2 , the access points 210-1 to 210-2 may be referred to as a module (“access point scan module”) executing a Wi-Fi access point scan program (Wi-Fi AP Scan Program). ) and a Wi-Fi driver (Wi-Fi driver), and the server includes a backend server (220: backend server) and a database server (225: database server).

Referring to FIG. 2 , the access point 210 - 1 must be able to monitor another access point 210 - 2 in an adjacent cell in order to obtain a link metric. The link metric is very important to classify the passenger level of the subway cabin. The Wi-Fi link metric may be a set of RSSI and/or CSI/CBF. RSSI is an indicator indicating the strength of a received signal. It can be viewed as an estimated measure of the power level that a wireless device receives from another wireless device. If the wireless device is at a great distance, the signal will be weakened and the wireless data rate will be slower, lowering the overall data throughput. In general, RSSI information can be obtained relatively easily from a Wi-Fi router. Almost all Wi-Fi drivers provide a function to obtain RSSI information. One of them may be an IW tool provided by the Linux operating system. In addition, CSI or CBF information indicates a predefined channel attribute of a communication link. This information describes how the signal propagates from the transmitter to the receiver and exhibits combined effects, such as, for example, scattering, fading, and power attenuation with distance. The Wi-Fi driver may perform the function of monitoring such information.

In one embodiment of the present invention, the access point 210 - 1 transmits the previously and periodically measured link metrics to the server 220 . The measured link metric is a measurement result of RSSI and/or CSI/CBF between two access points 210 - 1 and 210 - 2 installed in train cabins adjacent to each other. When no passenger is present, RSSI and/or CSI/CBF between the two access points 210 - 1 and 210 - 2 may be measured as a signal in a state without interference. However, when there are some passengers, the RSSI and/or CSI/CBF measurement results between the two access points 210 - 1 and 210 - 2 are different from when there are no passengers due to interference and distortion depending on the passenger and the terminal possessed by the passenger. can vary a lot. According to an embodiment of the present invention, by learning the relationship between the number of passengers and the radio signal that is affected by them, that is, the relationship between the number of passengers and the link metric, the two access points 210 - 1 and 210 - 2 . It is possible to infer the number of passengers in the cabin with the result of the link metric between them.

Meanwhile, it is preferable that the period (period) for transmitting the link metric to the server 220 is longer than the period (period) for measuring the link metric between the two access points 210 - 1 and 210 - 2 . That is, one link metric data transmitted to the server 220 may include several link metric measurement results. The measurement results of the plurality of link metrics are used as input to the AI model for classifying the passenger level in the room between the two access points 210 - 1 and 210 - 2 .

According to the embodiment of the present invention, in the access point 210-1, more precisely, in the access point scan module, a program for performing the above operation, that is, a Wi-Fi AP scan program is executed. The program periodically calls APIs to the Wi-Fi driver to perform RSSI and/or CSI/CBF measurement functions. That is, the link metric (RSSI/CSI/CBF) is measured periodically (every second or another period) between itself and another access point 210 - 2 installed in an adjacent room. Then, a certain number of samples of RSSI/CSI/CBF are collected within a certain period. Then, the measurement result is stored in the internal memory of the access point 210-1. The measurement result is added to the internal memory together with the previous measurement result. After storing several link metric measurement results, the program transmits the collected data to the server 220 by making a stable API request to the server 220 in a specified format. At this time, data is transmitted to the cloud server through an HTTP request. In the process of sending data to the server 220 , the program synchronously waits for a response from the server 220 . Then, when receiving a response from the server 220, the program stores the previous measurement result and secures an internal memory used to perform the next task.

Meanwhile, the data transmitted to the server 220 by the Wi-Fi scan program includes the following.

1. Sample Collected Link Metrics (RSSI/CSI/CBF)

2. Room ID or identifier (MAC address) of access point 210-1 (this is unique information)

3. Identifier (MAC address) of the neighboring access point (210-2: monitoring target device)

4. Timestamp at the time of data collection

5. Scanned list of MAC addresses of neighboring APs

In relation to the server, according to the embodiment of FIG. 2 , at least two types of servers, that is, the back-end server 220 and the database server 225, work together to perform an in-cabin passenger level estimation operation. The backend server 220 is an AP module operating in conjunction with the access point 210-1, an AI module for estimating the passenger level in connection with the AI model, and operating in conjunction with the client terminal. Includes an application module. Each module may be implemented by one or more processors. Alternatively, a plurality of modules may be combined to be implemented as one processor.

The AP module is a module that interacts with the access point 210 - 1 , receives the link metric transmitted from the access point 21 - 1 , verifies validity, and transmits it to the database server 225 . That is, RSSI and/or CSI/CBF are stored together with identification information and time information of access points.

The AI module calls the AI model to input link metric information, and executes the AI model to obtain passenger level information as an output. Although the drawing shows that the AI model exists outside the server device, it may be a program external to the server, or in some cases, a program installed inside the server.

The application module responds to all data requests from the client terminal (client app) regarding the subway or each cabin passenger level.

Meanwhile, the backend server 220 continuously communicates with the database server 220 to continuously store the link metric transmitted by the access point 210 - 1 in the database server 225 . The database server 225 continuously serves as a storage medium for all collected data. The stability and performance of the servers 220 and 225 are important points of the system in order to handle a large number of requests from clients and access points 210-1. Therefore, high reliability can be obtained with two approaches: vertical and horizontal scaling. Vertical expansion means adding more resources (CPU/RAM/DISK) to the servers 220 and 225 . Vertical scaling usually means upgrading the server hardware. Some reasons to scale vertically include increasing input/output operations (IOPS), increasing CPU/RAM capacity, and increasing disk capacity. Horizontal scaling includes adding more processing units or physical systems to the backend server 220 or database server 225 . This could include increasing the number of nodes in the cluster, reducing the responsibilities of each member node, and providing additional endpoints for client connections.

3 is a flowchart illustrating a connection operation between a subway access point, a server device, and an AI model.

Referring to FIG. 3 , first, an indoor access point (Subway-AP) measures a link metric for a specific period (eg, 10 seconds). Then, after collecting link metric data for a certain period of time, the data is transmitted to the backend server through an HTTP request. The data transmitted to the server may include a measured link metric and identifier information of a room or an access point in which the link metric is measured. Before sending to the server, the measured link metrics for each cycle are aggregated. This data aggregation aims to always keep the length of the link metric sent to the server constant. Data aggregation is performed when an access point receives two or more link metric data for a specific period (eg, 1 second). In this case, by calculating the average value of the data and aggregating the data, the aggregated data for 1 second is calculated. In addition to identifier and link metric information, scanned neighboring AP list information is also collected to detect fixed access points fixedly installed on subway station platforms. The list of nearby APs is useful for detecting the movement status of the subway. After this collection, all the information is put together to make an API request to the server using the POST method with the URL and parameters defined in JSON format. In order to perform the next operation from the point of view of the access point, it must first wait for a response from the server. However, if there is no response from the server for the specified timeout period, the access point may perform the following operation.

When the link metric measurement result is transmitted from the access point, the backend server (more specifically, the AP module) first verifies the API request. The first validation process is to validate the received API parameters. The second verification process is to verify the room identifier or access point identifier. First, the server extracts identifier information from the API request body, and then performs a database query to check whether the identifier is previously registered in the system. If registered, the server performs the following process. The third validation process is to check the length of the link metric data. The last step is to check the train movement status.

The verification process is a very important part of the passenger level classification process. In particular, this is because the passenger level classification process is not valid in situations where a large number of passengers enter and exit the cabin through the cabin entrance between the platform and the cabin (ie, while the train is stopped). In this way, passengers entering and leaving the train can greatly affect the Wi-Fi signal in the train and make it unstable. This may result in inaccurate classification results. Therefore, it is preferable that the server performs the classification procedure only when the subway is moving.

The above train movement-related verification process is performed by checking the list of neighboring APs transmitted by the access point. That is, the list information of APs fixedly installed on the subway station platform in advance is stored in the database, and whether it is moving by comparing it with the scanned neighboring AP list included in the data from the access point (Subway-AP) installed in the subway cabin. to judge That is, if the identification information of the fixed AP stored in the database exists in the neighboring AP list, it is determined that it is not valid by determining that it is still present on or near the subway station platform. In this case, the data is discarded and not used in the classification process. On the other hand, when the identification information of the AP in the scanned neighboring AP list transmitted by the access point (Subway-AP) is not found or is not found in the database, it is determined that the subway is in a moving state and is sufficiently far from the subway station. In this case, the AP module of the backend server stores the data in the database and passes the link metric data to the AI model for the classification process.

After processing the access point's API request, the server delivers a response to the access point as soon as possible. The response contains a status code indicating the validation result of the server-side function. The status code can be used to determine what action the access point should take after receiving this response. The status code and the operation of the access point accordingly may be defined as follows.

200: Successful responses

Upon receiving such a response, the Wi-Fi AP scan program of the access point performs a new RSSI/CSI/CBF measurement operation.

400: Client Error - invalid syntax/format of request data

Upon receiving such a response, the Wi-Fi AP scan program of the access point discards the request and performs a new RSSI/CSI/CBF measurement operation.

500: Server Error - internal server error

The Wi-Fi AP scan program retransmits this request. When the Wi-Fi AP scan program receives this 500 response code twice in a row, it discards the request and starts a new RSSI/CSI/CBF measurement.

Referring to the operation in the AI module of FIG. 3 , when the AI module of the backend server receives the link metric that has passed the validation process in the previous step from the AP module, it starts the passenger level classification process. First, the AI module loads the AI model object. Here, the AI model may be generated as a result of a machine learning process. According to an embodiment of the present invention, the AI model may be at least one of a neural network model, a clustering model, and a K Nearest Neighbor (KNN) model. The AI model includes a black box function that can map link metrics to the number or level of passengers. When the AI model is successfully loaded, the AI module provides link metrics data as input to the AI model. In this case, multidimensional data (eg, a heatmap) extracted from a link metric between two access point devices may be used as an input of the AI model. Then, the classification process is executed based on the input link metric data, and when the classification process is completed, the AI model returns the passenger level to the server. The server stores the returned results in the database in association with the link metric data used as input to the AI model.

4 is a flowchart illustrating a link operation between a user terminal and a server device.

The passenger-level classification results of subway train cabins/vehicles can be accessed from external client apps in a web-based mobile application, and can also be accessed on public display devices of subway platforms through open APIs. In particular, at this time, methods for accessing passenger level information are largely divided into PULL and PUSH methods. Both methods use the Restful API as a communication medium with the server, but the basic differences between PULL and PUSH are as follows. The PULL method indicates how the client makes a request to the server.

4A is a flowchart illustrating a method in which a client app installed in a client terminal receives a passenger level result by a PULL method. Referring to FIG. 4A , the application module of the backend server operates in conjunction with the client app installed in the client terminal. First, the client app sends an API request. In this case, the API request may include time information and subway room ID information indicating a measurement target. In another example, in the API request, the measurement target may be specified as at least one of a subway station, a subway ID, and a subway line. If the measurement time and subway station are specified in the API request, the server searches for a subway that enters the specified subway station at the specified measurement time, and retrieves passenger level information of all or at least some rooms included in the searched subway. It can act to return.

Meanwhile, the application module receives the API request, verifies API parameters, and validates the API security token. Then, the database is queried for information related to the API parameter (eg, passenger level information of the subway cabin ID to be measured), and the information returned by the query is delivered to the client app. The client app displays it through a display means so that the user can check it. Regarding the use of the PULL method, this is the most commonly used method in a client-server architecture. To update data, the client app must actively and periodically send API requests to the server. When the client app receives a response from the server, it updates the information and immediately displays the result to the user through the terminal's display means.

One example of the PULL method may be that a web browser requests a web page. According to an embodiment of the present invention, passenger level information of a subway cabin is provided in real time through a user interface provided by a server using a subway passenger level related application or a web page related thereto.

On the other hand, PUSH indicates that the server starts updating information on the client. 4B is a flowchart illustrating a method for a public display device to receive a passenger level result by a PUSH method. Referring to FIG. 4B , the PUSH method starts with establishing a web socket connection between a client and a server. In this case, for the web socket connection, identification information (eg, subway station ID) related to the public display device is provided to the server. After the connection is established, the server actively transmits the updated data to the client when there is new updated data with respect to the passenger-level data of the subway coming into the specified subway station. If nothing has been updated, no separate information is transmitted to the client. However, when the latest update is detected, the corresponding information is transmitted to the public display device so that the result can be displayed in real time.

One example of a PUSH-style application is an email server sending an email to an email client. This can be very effective for real-time monitoring systems as it greatly reduces the number of API requests that clients send to the server. Therefore, in the case of a real-time client app, such as a public display device on a subway platform, it is preferable to use the PUSH method to access the passenger level information of the server.

5 is an ER diagram (Entity-Relationship Diagram) showing the structure and relationship of data managed by a database.

Referring to FIG. 5 , in order to efficiently store information related to system operation of the present invention, a database structure may be required. These are mainly subway train information (Train), cabin-AP information (Cabin-AP), link metric information (Metric), platform-AP information (Platform-AP), subway station information (Subway Station), and user information (User). ) is included. In particular, such information can be divided into two categories based on the source providing it. It consists of external-provided data and system-generated data.

First, the externally provided information includes train information (Train), cabin-AP information (Cabin-AP) and platform-AP information (Platform-AP), and subway station information (Subway Station).

Train information (Train) includes a train id (train_id) and a line number (line number). It has the form of a static table. And, since one train has a plurality of cabins and an AP may be installed in each cabin, one train information (Train) corresponds to a plurality of cabin-AP information (Cabin-AP).

Cabin-AP information (Cabin-AP) includes information of an AP corresponding to a room in which an access point (Wi-Fi router) is installed, and has a fixed table form. This table includes cabin identifier information (cabin_id) for distinguishing the first and second cabins, access point device identifiers (MAC addresses) (device_mac), other MAC addresses (mac_addr_2, mac_addr_5), and current passenger level information ( current_crowd_level) may be included.

Platform-AP information (Platform-AP) includes information about the AP installed in the subway station. It is stored in the form of a fixed table. This table stores information of access points (APs) installed around subway station platforms. This information, as described above, can be used to detect the movement status of the subway. This table has several columns. This includes a fixed access point identifier (platform_ap_id) fixedly installed on the platform, and a MAC address (mac_addr) of the corresponding AP.

Subway station information (Subway Station) may correspond to a plurality of platform-AP information (Platform-AP). This is because a plurality of APs may be fixedly installed in one station. Such subway station information is stored in the form of a fixed table, in which station ID (station_id), station name (station_name), and line number information (line_number) to which the station belongs may be stored.

Next, look at the data generated when the system is operating.

First, the metric table (Metric) is a table for storing information about the link metrics collected by the room AP. This has a data structure in which a plurality of metric tables (Metric) correspond to one cabin-AP information (Cabin-AP) because one cabin AP continuously collects link metrics over time. As described above, this table stores a set of link metric data (RSSI and/or CSI/CBF) provided by the Wi-Fi AP scan program of the guest room AP. This collected data is used in the AI model to classify passenger levels in subway cabins. Since this table stores all link metric data for all operating subway rooms, it may be the largest table in the congestion prediction system according to an embodiment of the present invention. Over time, the data in this table continues to grow at a constant rate. This table has the following columns: This is a metric self identifier (metric_id), a timestamp when data is inserted into the database, three types of link metric data (RSSI/CSI/CBF), and passenger level information corresponding to the link metric data (crowd_level) includes

Next, take a look at the users table, which stores a set of information about system users, such as administrators or other users with arbitrary roles. This table also stores token access information for client apps that wish to obtain passenger level information via public APIs. This may include a user identifier (user_id), a user name (username), and a password (password). In addition, as personal information, it includes information (is_verified), email address information (email), role information (role), and API token access information (api_token) about whether a user is verified in relation to a specific authority.

6 is a flowchart illustrating a learning and testing process of an algorithm for predicting the number of passengers in a cabin of an AI model.

The algorithm implementation part of the AI model is divided into two stages: a learning stage and a testing stage. The training phase is a phase that aims to create an AI model to be used in the testing phase. In contrast, the test phase is the implementation of a real system in which the real-time link metric data of the access point device is classified at the subway cabin passenger level.

Referring to Figure 6 (a), the AI model training process during the training phase requires labeled link metric data as an input. The labeled data is the measured data and is a set of link metric data related to the actual passenger level.

Labeled data can be collected by performing data collection experiments directly in the train cabin. The data labels (passenger level) are manually classified into passenger levels by counting the number of passengers in the train and then using a predefined level-by-level threshold. The collected data is then processed and divided into two partitions: a training data set and a test data set. The training dataset is the dataset used in the training process of the AI model, and the test dataset is the dataset used to measure the accuracy of the AI model. In the AI model training process, balancing the data distribution for each passenger level is very important to avoid overfitting the AI model. During the training process, tuning of AI model parameters can also be performed to generate an AI model with the best performance. Once the training process is complete, the AI model can be used for testing steps integrated with the system.

Referring to FIG. 6B , in the test phase, the system provides link metric data measured from the train cabin to the access point device in real time. This data is used as input to an algorithm that maps the data to the passenger level. This phase includes two main functions: train movement status detection and passenger level classification.

Because the Wi-Fi signal is greatly affected by the movement of surrounding objects, passenger level classification can only be accurate when there is not much change/movement within the train. In this respect, the train movement status detection function is an important pre-processing function required before performing passenger level classification. The train's movement status is determined by scanning RF signals from fixed access point devices installed at subway stations. It is preferable that the passenger level classification proceeds only when the train is in motion.

The passenger level classification function utilizes an AI model to estimate the passenger level in the train cabin. AI models are created in the machine learning process during the learning phase. As previously described, the AI model performs a black box function that can map link metrics between two access point devices to the passenger level. Finally, the AI model returns the results (passenger level) to the system for use by the user.

7 is a flowchart illustrating a learning process of a prediction algorithm of an AI model according to another embodiment of the present invention.

Referring to FIG. 7 , before training the AI model, a feature extraction process may be performed. In training the AI model according to an embodiment of the present invention, the training data is passenger level and link metric data corresponding thereto, as described above. At this time, when the amount of training data is too large and overlapping is severe, characteristic data to be actually used is extracted using a feature extraction technique.

Also, it is important to properly process the feature extracted data before training the AI model properly. This is called data pre-processing. This may include steps of feature scaling and noise removal. Among the feature variables extracted from the learning data set, when the size of a specific feature variable is larger than the reference value and the data distribution is distorted, the data distribution can be made even by scaling the data. Noise removal removes data that is not necessary for learning and leaves only necessary data. After the preprocessing is complete, the AI model is trained.

The system or apparatus described above may be implemented as a hardware component, a software component, and/or a combination of the hardware component and the software component. For example, systems, devices, and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA). ), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions, may be implemented using one or more general purpose or special purpose computers. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that can include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

Software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or apparatus, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited embodiments and drawings, various modifications and variations are possible from the above description by those skilled in the art. For example, the described techniques are performed in an order different from the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (20)

1. In the method of predicting congestion for each subway cabin in the server device,

   Receiving channel information through a network from a first access point (AP), the channel information includes a link metric measured based on a communication signal with a second access point, the first the first access point is an access point installed in a first subway cabin, and the second access point is an access point installed in a second subway cabin adjacent to the first subway cabin;

   calculating a congestion degree of the first subway cabin or the second subway cabin based on the received channel information, wherein the congestion degree indicates a level related to the number of passengers present in the subway cabin; and

   Congestion prediction method for each subway room, comprising the step of providing the calculated congestion degree of the subway cabin to a user terminal.
2. The method of claim 1,

   The link metric includes at least one of Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and Compressed Beamforming Feedback (CBF) of a wireless local area communication signal.
3. 3. The method of claim 2,

   The received channel information includes (i) the link metric, (ii) at least one of identification information of the first subway cabin and identification information of a first access point, (iii) identification information of the second subway cabin and a second At least one of identification information of an access point, (iv) timestamp information indicating a time when data is collected, and (v) a method of predicting congestion for each subway cabin, including scanned identification information of neighboring access points .
4. The method of claim 1,

   Calculating the degree of congestion of the first subway cabin or the second subway cabin based on the received channel information comprises:

   determining whether the first subway cabin is moving; and

   and calculating a congestion degree of the first subway cabin or the second subway cabin using the channel information based on the determination result.
5. 5. The method of claim 4, wherein the step of determining whether the first subway cabin is moving,

   Parsing identification information of neighboring access points from the channel information, and determining by comparing the parsed information with pre-stored information of subway station fixed access points.
6. 6. The method of claim 5,

   The parsed information includes identification information of a neighboring access point of the first access point,

   In response to the identification information of the neighboring access point does not exist in the information of the pre-stored subway station fixed access point, it is determined that the first subway cabin is currently moving,

   Based on the determination that the first subway cabin is moving, initiating the congestion calculation of the first subway cabin, the congestion prediction method for each subway cabin.
7. 5. The method of claim 4,

   Calculating the congestion level of the first subway cabin or the second subway cabin by using the channel information includes mapping the link metric to a congestion level using an artificial intelligence (AI) model, whereby the first Congestion prediction method for each subway cabin, comprising the step of calculating the congestion level of the subway cabin or the second subway cabin.
8. 8. The method of claim 7,

   The AI model is a machine learning model learned by using, as a training data set, level information of real passengers in an actual subway cabin and a link metric corresponding thereto, a method of predicting congestion for each subway cabin.
9. The method of claim 1,

   Further comprising the step of receiving a congestion level request from the user terminal,

   The congestion level request includes identification information of the first subway cabin,

   In response to the congestion level request, the congestion level of the first subway cabin is provided to the user terminal, the congestion level prediction method for each subway cabin.
10. The method of claim 1,

    verifying the API parameter of the congestion request; and

    Congestion prediction method for each subway room, further comprising the step of verifying the validity of the API security token related to the congestion request.
11. The method of claim 1,

    After receiving the channel information from the first access point (AP: Access Point), further comprising the step of verifying the validity of the received channel information,

    The validity of the channel information is verified by (i) verifying that the first subway cabin or the first access point is registered, and (ii) verifying the length of the channel information. Prediction method.
12. 12. The method of claim 11,

    Based on the determination that the received channel information is valid, further comprising the step of storing the channel information,

    The channel information is stored together with the passenger level result corresponding thereto, the congestion prediction method for each subway cabin.
13. In the congestion prediction device for each subway cabin,

    An AP module for receiving channel information through a network from a first access point (AP), the channel information includes a link metric measured based on a communication signal with a second access point, the the first access point is an access point installed in a first subway cabin, and the second access point is an access point installed in a second subway cabin adjacent to the first subway cabin;

    a processor for calculating a congestion degree of the first subway cabin or the second subway cabin based on the received channel information, the congestion degree indicating a level related to the number of passengers present in the first subway cabin; and

    Congestion prediction apparatus for each subway cabin, including an application module for providing the calculated congestion degree of the first subway cabin to a user terminal.
14. In the first access point device for predicting the congestion level for each subway room,

    an access point scan module for measuring a channel state based on a communication signal with a second access point and transmitting channel information including the measured channel state to a server through a network; and

    A memory for storing the channel information,

    The channel state includes a link metric,

    The second access point is an access point installed in a second subway cabin adjacent to the first subway cabin in which the first access point device is installed.
15. 15. The method of claim 14,

    The link metric includes at least one of Received Signal Strength Indicator (RSSI), Channel State Information (CSI), and Compressed Beamforming Feedback (CBF) of a wireless short-range communication signal. A first access point device for predicting congestion for each subway cabin.
16. 15. The method of claim 14,

    The access point scan module collects at least one link metrics for a preset period to generate the channel information, a first access point device for predicting congestion for each subway cabin.
17. 17. The method of claim 16,

    When a plurality of link metrics are collected for the preset period, by calculating and aggregating the average value of the link metrics, calculating the aggregated data for the preset period to generate the channel information, for predicting congestion for each subway cabin A first access point device.
18. 17. The method of claim 16,

    The channel information may include (i) the link metric, (ii) at least one of identification information of the first subway cabin and identification information of a first access point, (iii) identification information of the second subway cabin and a second access point. At least one of the identification information of (iv) timestamp information indicating a time at the time of data collection, and (v) a method for predicting the congestion level for each subway room, including the scanned identification information of the neighboring access points 1 access point device.
19. 15. The method of claim 14,

    A first access point apparatus for estimating congestion for each subway cabin, wherein a period for transmitting the channel information to the server is longer than a period for measuring the link metric.
20. 15. The method of claim 14,

    The access point scan module stores a current link metric measurement result based on receiving a response from the server, and prepares for a next link metric measurement, a first access point device for predicting congestion for each subway room.

Images