WO2018135923A1

WO2018135923A1 - Radio map construction method

Info

Publication number: WO2018135923A1
Application number: PCT/KR2018/000984
Authority: WO
Inventors: 한동수
Original assignee: 한국과학기술원
Priority date: 2017-01-23
Filing date: 2018-01-23
Publication date: 2018-07-26

Abstract

According to the present invention, a radio map construction method uses a genetic algorithm and comprises the steps of: (a) generating a plurality of chromosomes, each including a set of pairs consisting of a fingerprint labeled with an address and a position selected within a region of the address; (b) generating a temporary radio map by using the pairs of the chromosomes; (c) arranging collected fingerprint sequences by using the temporary radio map; and (d) evaluating the arrangement of the fingerprint sequences.

Description

How to build a radio map

The present invention relates to a method of constructing a radio map, and more particularly, to a method of constructing a radio map for collecting a fingerprint for mapping an address and using the same to construct a radio map for accurate indoor location service.

Cloud sourcing (Crowdsourcing) techniques that collect Wi-Fi fingerprints from any device have recently been actively used to build the radio maps needed for Wi-Fi Positioning systems. In addition, various machine learning techniques have recently been utilized to position label cloud-sourced fingerprints from unspecified devices.

In particular, various attempts have been made to automatically construct radio maps using semi-supervised learning techniques and unsupervised learning techniques.

However, it is extremely rare to use cloud-sourced fingerprints to construct a real radio map. This is because there is no practical way to label cloud sourced fingerprints yet.

Although there is a method of manually collecting AP location and / or location labeled fingerprints and using them as a location reference point of a location-labeling machine learning technique, a location reference point as described above Gathering references is the biggest obstacle to the widespread use of location-labeling machine learning techniques.

Fully automatic labeling of cloud-sourced fingerprints allows for automatic building of radio maps of building units, city units, country units, and even global regions, ultimately leading to indoor (indoor) locations for global regions. The service becomes available.

For example, indoor location services at the city level may provide location services for users in each building and building in the city, which may improve the life of a city person and provide various advantages. Urban location services at the city level may be linked to or configured with outdoor location services, and enable more effective location services in areas such as police, firefighting and rescue. The city radio map should model the Wi-Fi signal pattern for each location in each building in the city. City radio maps can be built by hand, but this can be virtually impossible because of the large amount of manpower and time required.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problem, and an object thereof is to provide a radio map construction method for automatically constructing a radio map using a cloud sourced fingerprint.

In addition, the present invention provides a method of building a radio map that enables a specific application to collect a fingerprint mapped to an address and to build a radio map from the collected address-fingerprint according to the classified region type. The purpose is.

It is also an object of the present invention to provide a method for constructing a radio map from an address-fingerprint and a fingerprint sequence collected using a quasi-supervised learning technique.

In addition, an object of the present invention is to provide a radio map construction method that can map the best fingerprint by using a cloud-sourced fingerprint to each location where the radio map is to be built using a genetic algorithm.

The technical problems to be achieved in the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned above will be clearly understood by those skilled in the art from the following description. Could be.

Radio map construction method according to an aspect of the present invention, using a genetic algorithm, (a) a plurality of chromosomes comprising a set of pairs consisting of a fingerprint labeled with an address and a predetermined position in the region of the address. Generating; (b) generating a temporary radio map using the pair of chromosomes; (c) placing a collected fingerprint sequence using the temporary radio map; And (d) evaluating the placement of the fingerprint sequence.

In the radio map construction method as described above, when the step (a) is first executed to form the first generation, the predetermined position is randomly selected within the area of the corresponding address, and the steps (b) to (d). ) May be performed sequentially on each of a plurality of chromosomes.

In the radio map construction method, the step (d) may evaluate the placement of the fingerprint sequence based on the generated temporary radio map using a loss function reflecting the mobility of the user in the fingerprint sequence.

In the radio map construction method described above, the radio map construction method includes: filtering a portion of the plurality of chromosomes according to the evaluation result in the step (d) and including the filtered portion of the chromosomes and a new chromosome. It can be configured to repeat the step (b) to the step (d).

In the radio map construction method, before the step (a), further comprising the step of modeling the internal structure of the building on which the radio map is to be built in a topology according to the Hidden Markov model (HMM); Each location within a building can be represented by a state and the geographic accessibility between the locations can be represented by a transition.

In the above-described radio map construction method, the temporary radio map may be generated through a process of generating a fingerprint corresponding to each location within a building using the set of waste fish and using an interpolation technique.

In the radio map construction method, the radio map construction method includes the steps (b) to (d) through generation evolution when there is no more gain as a result of the evaluation using the loss function compared to the previous generation. The repeated execution of can be terminated.

In the radio map construction method, before the step (a), collecting fingerprints corresponding to the designated address through the specified app; And classifying the fingerprints addressed through the collection according to addresses, wherein the designated app may be a payment app or a resource usage metering app.

In the radio map construction method, steps (a) to (d) may be performed on fingerprints of addresses classified as commercial areas.

In the radio map construction method, when the classified address is a residential area, a radio map of the residential area may be further constructed through an interpolation technique for fingerprints corresponding to the address of the residential area.

The method of building a radio map, the method comprising: clustering fingerprints collected corresponding to a designated address into a plurality of groups and determining a fingerprint to be labeled with the address using fingerprints of the clustered group; Can be.

In the method of building a radio map, before the step (a), the method may further include collecting a fingerprint sequence including a fingerprint and a series of collection times using a designated app.

In the radio map construction method, the collecting fingerprints corresponding to the designated address may include collecting the fingerprints corresponding to the designated address during a designated time before, after, or before and after a transaction occurs by the designated app. And collecting fingerprints including the identifier of the wireless AP and the pair of signal strengths.

In the radio map construction method performed by the radio map construction apparatus according to an aspect of the present invention, a fingerprint labeled with an address is collected and accumulated through cloud sourcing, and a finger labeled with the address is generated using a genetic algorithm. For each of the prints, determining an optimal position within the area of the corresponding address.

In the above-described radio map construction method, a set of pairs consisting of a fingerprint labeled with the address and a predetermined position within a region of the corresponding address may be used as a chromosome in the genetic algorithm.

In the radio map construction method, the method may include collecting an unlabeled fingerprint sequence, arranging the collected fingerprint sequence based on a temporary radio map corresponding to the chromosome and evaluating the placement. .

In the radio map construction method described above, the result of the evaluation may be used to determine the chromosome to be used in the next generation and to determine whether to end the evolution.

The radio map construction method according to the present invention as described above has an effect of automatically building a radio map using a cloud sourced fingerprint.

In addition, the above-described radio map construction method according to the present invention utilizes a specific application application to collect a fingerprint labeled with an address, and to generate a radio map of an arbitrary region from address-fingerprints collected according to the classified region types within the region. It has the effect of being able to build.

In addition, the radio map construction method according to the present invention as described above is able to map the fingerprint to the correct location using a quasi-map learning technique using the collected address-fingerprint and location-unlabeled fingerprints as input. It works.

In addition, the radio map construction method according to the present invention as described above has an effect of automatically and quickly building a radio map from cloud-sourced fingerprints by using genetic algorithms that evolve to multiple generations.

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned above may be clearly understood by those skilled in the art from the following description. will be.

1 is a diagram illustrating an example of a connection configuration between devices configured in a radio map construction system.

2 is a diagram illustrating an exemplary block diagram of an apparatus for constructing a radio map.

3 is a diagram illustrating an exemplary representative flow for building a radio map.

FIG. 4 is a diagram illustrating a specific example flow for constructing a radio map from a fingerprint addressed for a commercial area.

FIG. 5 is a diagram illustrating an example of constructing a radio map through HMM modeling using a fingerprint sourced in a commercial area.

The above objects, features, and advantages will become more apparent from the detailed description given hereinafter with reference to the accompanying drawings, and accordingly, those skilled in the art to which the present invention pertains may share the technical idea of the present invention. It will be easy to implement. In addition, in describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

As in the example of FIG. 1, the radio map construction system according to the present invention includes a radio map construction apparatus 100, one or more portable terminals 200, and a communication network 300.

Referring briefly to the components of FIG. 1, the apparatus 100 for constructing a radio map 100 is an apparatus for constructing a radio map using a fingerprint collected through cloud sourcing. The radio map building apparatus 100 utilizes one or more portable terminals 200 to collect fingerprints and map the collected fingerprints to corresponding addresses obtained in connection with a transaction performed by the portable terminal 200. The collected fingerprint is labeled with an address.

The radio map building apparatus 100 constructs a radio map that can be used in an apartment, a building, a building, a city, a country, and the like by using a fingerprint labeled as an address, and then uses the generated radio map as a portable terminal 200 or another device (server). ) And the like. The radio map construction apparatus 100 will be described in detail later with reference to FIG. 2.

The portable terminal 200 is a terminal that a user can carry and use. The portable terminal 200 may be a smartphone, a tablet PC, a notebook, a dedicated portable device, or the like.

The portable terminal 200 includes at least a communication interface configured to transmit and receive radio signals of a wireless packet through a processor (for example, an application processor), a storage medium (for example, a nonvolatile memory, a hard disk, etc.), and a communication network 300. do. The communication interface includes, for example, one or more communication chipset (s) and an antenna or the like for connecting to a wireless communication network and / or a mobile communication network such as Wi-Fi.

The portable terminal 200 stores various programs in a storage medium (for example, nonvolatile memory). For example, the portable terminal 200 stores a mobile app that can be executed in a processor and performs various designated functions in a storage medium.

The mobile app used in the present invention is configured to collect a fingerprint that can be labeled with an address. For example, the payment app of the portable terminal 200 collects the WLAN fingerprint for a predetermined time (1 minute, etc.) before, after, or before and after the payment transaction. Alternatively, the usage metering app configured to read electricity, gas, water, and the like and transmit the meter reading collects the WLAN fingerprint for a predetermined time before, after, or before and after the meter reading.

The collected Wi-Fi fingerprint includes an identifier (eg, SSID, BSSID, etc.) of a wireless AP that can be connected or recognized, and a signal strength of a wireless signal from the wireless AP. The portable terminal 200 may collect one or more pairs of the wireless AP identifier and the signal strength during the execution of a designated function (payment, meter reading, etc.) and transmit them to the radio map building apparatus 100 through the communication network 300.

This type of mobile app can match at least the collected fingerprint to an address. For example, the fingerprint collected at the time of payment, such as a credit card through the payment app, may be matched to that collected at a point where a payment is made (shop, restaurant, etc.). Fingerprints collected during the meter reading by the meter, such as electricity and gas, can be matched to the address (apartment, house, villa, shop, restaurant, etc.) where the meter is made.

In the case of apartments, 'address' includes information that can distinguish each family, and in the case of a building, 'address' includes information that can distinguish each store. For example, apartments can include lakes and buildings, including floors and rooms.

In addition to the payment app or the usage metering app, a mobile app having other functions may be used to map an address to a fingerprint. When a transaction occurs through such a mobile app, the corresponding address must be specified or extracted. For example, the address of the card merchant where payment is made may be specified or extracted using a payment server, etc., and the address where the meter reading is also specified or extracted using a meter reading server.

Meanwhile, the portable terminal 200 may collect a fingerprint sequence through a mobile app such as a payment app or a metering app, or through another type of mobile app. For example, the portable terminal 200 may collect a fingerprint sequence consisting of a series of fingerprints in a building using any app frequently used indoors, and transmit the fingerprint sequence to the radio map construction apparatus 100.

The fingerprint sequence includes a plurality of radio AP identifiers and signal strength pairs and collection times, and the radio map building apparatus 100 may also use the fingerprint sequence in a radio map building method.

As such, the portable terminal 200 may collect a fingerprint to be used to construct a radio map and transmit the fingerprint to the radio map construction apparatus 100. Fingerprints to be used for building the radio map can be collected through an app utilized by the general user, and these fingerprints can be collected in a wide range of regions. As a result, cloud sourcing of fingerprints allows for more realistic radio maps for indoor location services.

The communication network 300 connects the radio map building apparatus 100, any server, and / or the portable terminal 200 with each other, and provides a transmission / reception path for various data. The communication network 300 includes a mobile communication network and / or an internet network provided by a mobile communication provider. The portable terminal 200 is connected to the communication network 300 with or without a wireless AP or the like to enable data communication with the radio map building apparatus 100 and / or any server.

2 is a diagram illustrating an exemplary block diagram of the apparatus 100 for constructing a radio map.

According to FIG. 2, the radio map building apparatus 100 includes a communication unit 101, a storage unit 103, a connection unit 105, and a control unit 107. 2 preferably shows a functional block diagram and each functional block has a corresponding hardware block. The radio map construction apparatus 100 of FIG. 2 may be configured using various hardware. For example, the radio map building apparatus 100 may be configured using one or more PCs, workstations, and / or main frames. Other blocks not shown in FIG. 2 may be further included in this block diagram according to design variations.

The apparatus 100 for constructing a radio map may further perform other functions as well as constructing a radio map. For example, the radio map building apparatus 100 may further perform a function as a server for a location server or an app (payment app, metering app, etc.) that provides a location service using a radio map. According to a design example, such a location server or an app server may be separated from the radio map building apparatus 100 (server).

Referring to the radio map construction device 100 through FIG. 2, the communication unit 101 is an interface for transmitting and receiving data with the communication network 300. The communication unit 101 is configured to transmit and receive data with the portable terminal 200 and any other server.

For example, the communication unit 101 may receive a fingerprint collected by the portable terminal 200 through the communication network 300, and may transmit all or part of the constructed radio map, location information estimated from the fingerprint, and the like. The indicated data can be transmitted to the portable terminal 200 or the like.

The storage unit 103 stores various data. The storage unit 103 includes a storage medium of a large capacity such as a volatile memory, a nonvolatile memory, and / or a hard disk to store various data and a program driven by the control unit 107 or the like.

The storage unit 103 stores various fingerprints to be used for constructing a radio map. For example, the storage unit 103 stores a fingerprint that can be collected and mapped to an address through a payment app or a metering app. Fingerprints collected at one address may be different from each other. For example, a plurality of fingerprints having different wireless AP identifiers or signal strengths for the same address are collected through a payment app or a metering app.

The storage unit 103 also stores fingerprints that are not labeled with an address. For example, the storage unit 103 may store fingerprint sequences collected through a payment app, a metering app, or the like through a road guidance app or a schedule management app. Such fingerprint sequences include fingerprints collected at each of a series of times within a particular region. The area may be extracted or recognized from a road guidance app or a schedule management app.

The storage unit 103 also stores various DBs. For example, the storage unit 103 stores an address-labeling fingerprint DB including fingerprints labeled with an address, and stores a radio map DB constructed for a specific building, a building, an area, a city, and the like.

The address-labeling fingerprint DB is composed of a plurality of fingerprint items, and each fingerprint item stores an address (identifier) from which the fingerprint is collected and a fingerprint corresponding to the address (identifier), and corresponds to one address. The fingerprint may include a plurality of fingerprints collected.

The radio map DB is composed of a plurality of radio map items, and each radio map item stores a 'location' and a fingerprint corresponding to the location. In each item of the radio map DB, the location and the fingerprint are mapped, not the address, and the location represents the exact physical point in the building, for example, an identifier that identifies the floor of the building and It can be specified by the two-dimensional coordinate of. The construction of the address-labeling fingerprint DB and the radio map DB will be described in detail with reference to FIG. 3.

The connection unit 105 transmits and receives data between blocks in the radio map construction apparatus 100. The connection unit 105 is configured using a local area network, a parallel bus, a serial bus, or the like.

The control unit 107 includes one or more execution units to load the program stored in the storage unit 103 and execute the command code of the program through the execution unit to control the radio map building apparatus 100.

For example, the control unit 107 collects a fingerprint mapped to a specific address in association with the portable terminal 200 connected to the communication network 300 and when sufficient fingerprints are collected for each address or for an arbitrary address. In addition, an address-labeling fingerprint DB is constructed or updated using the fingerprint mapped to the address.

In addition, the control unit 107 has an address-labeling fingerprint DB and a fingerprint sequence when an address-labeling fingerprint DB has been established and has sufficiently collected fingerprint sequences (which may be fingerprints not labeled with an address). Build a radio map DB using

The control unit 107 may provide all or a part of the constructed radio map DB to another server or the portable terminal 200 or may directly use it for various location services (route guidance, indoor route guidance, etc.). Various control flows performed by the control unit 107 to build a DB will be described in detail with reference to FIG. 3.

The radio map construction method of FIG. 3 is performed by the radio map construction apparatus 100 and preferably the control unit 107 of the radio map construction apparatus 100 loads the program of the storage unit 103 and any other portable type. In conjunction with the terminals 200 to build a radio map.

First, as the radio map building apparatus 100 is driven or starts to be executed according to a program (S100), the radio map building apparatus 100 displays a fingerprint corresponding to a designated address in a specific app (eg, the portable terminal 200). For example, the collection through the payment app, resource usage metering app (S101).

For example, the portable terminal 200 may use a payment transaction through a payment app or a (cumulative) usage measurement transaction through a metering app for a predetermined time before, after, or before a time point (for example, 30 seconds, 1 minute, 2). The WLAN fingerprint is collected through the Wi-Fi interface provided in the network. The fingerprint includes one or more pairs of signal strengths measured by the portable terminal 200 for the identifier of the wireless AP and the strength of the wireless signal output from the wireless AP. A transaction may be regarded as a functional unit for completing a specific function with a server (a payment server or a metering server), and the portable terminal 200 and the metering server may transmit and receive various communication packets within a transaction.

As such, the portable terminal 200 collects the WLAN fingerprints for a predetermined time before, after, or after a transaction, and the collected WLAN fingerprints are transmitted via a server or directly to the radio map building apparatus 100 ( 300). In this case, the collected WLAN fingerprint and address information corresponding to the WLAN fingerprint may be provided together, or address information corresponding to the WLAN fingerprint may be provided from a subject providing a corresponding service. The address information may be a combination of letters, numbers, and the like (eg, an affiliate store number) that may represent the address itself or identify the address. The address information may at least specify where the transaction (payment or resource usage measurement) occurred. The radio map building apparatus 100 collects and temporarily stores one or more (preferably multiple) WLAN fingerprints corresponding to each address (identifier).

Thereafter, the apparatus 100 for constructing a radio map determines a fingerprint to be labeled with each address by using a fingerprint for each address that is temporarily stored (S103). The determination process of the fingerprint to be labeled with the address may be performed, for example, when the number of fingerprints collected for each address exceeds a set number or a collection period exceeds.

Multiple fingerprints can be collected for a single address. The apparatus 100 for constructing the radio map clusters the collected fingerprints into a plurality of groups corresponding to the designated address, and determines the fingerprint to be labeled with the designated address using the fingerprints of each clustered group. In detail, the apparatus 100 for constructing a radio map clusters a plurality of fingerprints collected corresponding to a single address according to similarities between fingerprints and labels the addresses using fingerprints of the group having the largest number among the clustered groups. Determine the fingerprint to be. According to an embodiment, the fingerprint to be labeled may be composed of an identifier of a common wireless AP and a signal strength value pair according to a signal strength average value or probability corresponding to the common AP identifier.

The radio map building apparatus 100 configures an address (identifier) and a fingerprint to be labeled as a fingerprint item and stores the fingerprint item in the address-labeling fingerprint DB.

Thereafter, the radio map building apparatus 100 classifies the address-labeled fingerprint according to the address. In detail, the radio map building apparatus 100 groups the fingerprints labeled with each address by region (S105). For example, the radio map construction apparatus 100 groups the addresses of the same or adjacent locations by the building name, area name, or area unit to which each address belongs. Within a group, a plurality of fingerprints labeled with an address are included (ie, a plurality of fingerprints respectively corresponding to a group of addresses), and each fingerprint is regarded as a fingerprint collected at a corresponding address.

The radio map building apparatus 100 classifies each address group according to the region type of the address group (S107). For example, the radio map building apparatus 100 classifies each group of address groups into a residential area, a public area, or a commercial area.

Thereafter, the radio map building apparatus 100 constructs a radio map according to the classification type for each address group.

For each address group, the radio map building apparatus 100 determines whether the corresponding address group is classified as a residential area (S109), and if the address area is a residential area, targets fingerprints of the address group corresponding to the address of the residential area. The radio map of this residential area is further constructed using an interpolation technique (S111).

Meter users use metering apps to measure their energy usage by visiting their homes, and each home is typically at a distance from each other (e.g. apartments or single-family homes) to respond to fingerprints collected through metering apps. By setting the center of the address and the fingerprint of the address where the fingerprint is not collected, it is possible to construct a radio map of the corresponding residential area by calculating using a known interpolation technique.

The radio map building apparatus 100 determines whether the corresponding address group is a public area (S113), and in the case of a public area, constructs a radio map of the corresponding area through an unsupervised learning technique (S115).

In the case of public areas, there is a limit to the collection of fingerprints that can specify an address through a payment app or a metering app due to a small number of shops or energy consumption points where payment is made. Thus, for example, a radio map can be created through self-learning techniques. It may be preferable to build.

The radio map building apparatus 100 determines whether the corresponding address group is a commercial area (S117), and when the commercial area is a commercial area, the radio map building apparatus 100 generates a radio map of the commercial area through a semi-supervised learning technique according to the present invention. Build (S119).

In commercial areas, there are many considerations for the construction of radio maps, such as the uncertainty of the location of fingerprint collection collected through the payment app, the complexity of the indoor space, and specific problems such as corridors. In consideration of these problems, the present invention constructs a radio map capable of accurate location guidance through a semi-supervised learning technique.

A detailed plan of the radio map construction in the commercial area will be described in more detail with reference to FIG. 4. As the radio map for each region is completed, the flow of FIG. 3 ends (S150).

4 illustrates a flow for constructing a radio map from an address-labeled fingerprint for a commercial area, and FIG. 5 illustrates an exemplary image related to the flow of FIG. 4.

The flow of FIG. 4 may be performed by the control unit 107 which is performed by the radio map construction apparatus 100 and preferably by loading and executing the program of the storage unit 103.

First, the radio map building apparatus 100 collects a fingerprint sequence through a designated app of the portable terminal 200 separately or together with a fingerprint used for address-labeling (S201) and temporarily stores the storage unit 103. do. This process may be performed together in step S101 of FIG. 3 or before and after step S101. The fingerprint sequence includes a series of fingerprints and the time at which each fingerprint was collected.

The fingerprint sequence is collected from the portable terminal 200 of the unspecified user and collected through the designated app. The fingerprint sequence may further include a commercial area identifier indicating the collected region.

Fingerprint sequences collected by unspecified persons are used to construct a radio map of a commercial area. A large number of fingerprint sequences can be collected for a particular commercial area and these fingerprint sequences are preferably used for evaluating temporarily constructed radio maps.

As the fingerprint sequence of a specific commercial area is sufficiently secured or as a specified period elapses, the radio map building apparatus 100 constructs a radio map for the commercial area.

In detail, first, the topology of the corresponding commercial area is modeled (S203). The radio map constructing apparatus 100 generates a topology representing an internal structure (plan view, etc.) of a commercial area (building, etc.). For example, the radio map construction apparatus 100 expresses the topology in a topology by HMM (Hidden Markov Model) state modeling representing the internal structure of a commercial area (see Step 1 of FIG. 5).

The internal structure is transformed into an HMM topology where location is represented by state and geographic accessibility between locations within the HMM topology is represented by transitions.

After topology modeling, the radio map construction apparatus 100 includes a chromosome including a set of pairs consisting of a fingerprint labeled with an address using a genetic algorithm and a randomly selected position in a region of the address. A chromosome (chromosome) is generated (S205), but a plurality of such chromosomes (for example, 100) are generated to form, for example, 100 populations (see Step 2 of FIG. 5).

Each chromosome contains one or more pairs (preferably plural) of a fingerprint labeled with the address of the commercial area in the address-labeling fingerprint DB for each address and a predetermined location within the area of that address. do. When constituting at least the first generation, the predetermined location is randomly selected within the range of addresses.

The radio map construction apparatus 100 is evolved by repeatedly performing each generation using a genetic algorithm. Some of the chromosomes of each generation are filtered and used as the chromosomes of the next generation. Since it is generated and filled, the positions selected corresponding to the same fingerprint may be different from the positions selected in the previous generation.

An address can generally be represented by a particular area (a location range area in the form of a square or rectangular or straight or curved area) and the predetermined location can be any location within that location area range.

The predetermined location may indicate a specific location where the fingerprint is collected and may be a randomly selected location. The selection position may be a position that is randomly selected differently for each chromosome or generation for the same address.

Each chromosome may include a plurality of pairs of fingerprints and selected locations labeled with addresses, corresponding to the addresses of each shop in a specific commercial area (e.g., a building or a specific floor of a building) on which a radio map is to be constructed. have.

Thereafter, the radio map construction apparatus 100 generates a temporary radio map by using a pair of fingerprints and corresponding selected positions for each chromosome generated in one generation (S207).

As a location corresponding to each address (for example, a shop or a restaurant) is selected, a radio map is also constructed for an area other than the internal area corresponding to each address. For example, a radido map can also be constructed for corridors outside shops and restaurants.

The radio map construction apparatus 100 uses a pair of fingerprints and corresponding selected positions already configured through the generation of chromosomes, for example, at each position of a commercial area (building) that requires a fingerprint through an interpolation technique. It is possible to generate a fingerprint for a radio map having a location and a pair of fingerprints for each location in a commercial area.

The interpolation technique used to generate the radio map may be a known technique, for example, an inverse distance weighting (IDW) technique.

Thereafter, the radio map construction apparatus 100 places (S209) the fingerprint sequence collected in the corresponding commercial area. This fingerprint sequence can be a cloud sourced and unlabeled address.

The radio map constructing apparatus 100 arranges a cloud sourced fingerprint sequence in a radio map generated for each chromosome (see Step 3 of FIG. 5).

The radio map constructing apparatus 100 evaluates all the batches for each chromosome and compares them with other batches so that the radio map constructing apparatus 100 finds the best results within that chromosome. An algorithm for finding the best result for one chromosome is referred to as a local optimization algorithm hereinafter.

The local optimization algorithm estimates the model parameters of the HMM using an EM-Expimation (EM) algorithm given an unlabeled fingerprint sequence. Likelihood function,

Can be used to perform an evaluation of an unlabeled fingerprint sequence with a set of estimated model parameters. U represents a set of fingerprint sequences, P represents a placement and λ represents a set of estimated HMM model parameters.

The product of the appearance probabilities of all fingerprint placements on the fingerprint sequence is used to evaluate the placement. The similarity of the fingerprint, the mobility of the user and the structure of the building may be factors that determine the probability of appearance. The similarity of the fingerprint can be calculated through well-known techniques (eg Euclidean distance, cosine distance, etc.) and the user's mobility and building structure can be reflected as a transition of states in the HMM.

As such, the radio map constructing apparatus 100 may determine an optimized arrangement through a local optimization algorithm for a single chromosome. For 100 chromosomes, batches based on 100 temporary radio maps are repeated.

Thereafter, the radio map construction apparatus 100 evaluates each fingerprint sequence arrangement based on the temporary radio map corresponding to each of the chromosomes of the generation (S211).

In operation S209, the radio map construction apparatus 100 performs local optimization based on each temporary radio map corresponding to each chromosome. In step S211, in order to compare and evaluate each temporary radio map (that is, each chromosome), a batch based on each temporary radio map is evaluated (see Step 4 of FIG. 5). If 100 chromosomes are generated in step S205, relative evaluation of 100 temporary radio maps (chromosomes) is performed in step S211.

Criteria for selecting a temporary radio map (chromosome) from the entire temporary radio map (chromosome) will be included in the next generation. Earlier, the appearance probability used in the local optimization algorithm of step S209 may be used.

Preferably, the present invention utilizes a loss function that fully reflects the user's mobility to evaluate each radio map (placement of fingerprint sequences based on). Continuously collected fingerprints may be considered to be collected from near states. In fact, the probability that two consecutively collected fingerprints are far from each other is low and the user's mobility can be used for evaluation regardless of the number of training samples.

The fingerprint sequence collected according to the present invention includes a collection time of each fingerprint, and the loss function is

It is expressed as

The loss function represented by the above equation is detected as the moving distance accompanying the change from one fingerprint to another in the batches of fingerprint sequences is greater than or equal to the maximum movable distance. Each time there is a penalty for that temporary radio map (chromosome)

) Is defined to accumulate.

The radio map building apparatus 100 may select a radio map corresponding to one generation of chromosomes to be evaluated and having the lowest loss function value among the temporary radio maps arranged as a temporary radio map having the best arrangement in the current generation. have.

Thereafter, the radio map construction apparatus 100 configures the next generation according to the batch evaluation result for each of the plurality of chromosomes, and generates the next generation through a genetic algorithm. In this case, the next generation may be generated by using selection, crossover, and mutation according to a genetic algorithm. Then, the above steps S205 to S211 are repeated for the next generation.

And the radio map construction apparatus 100 compares the arrangement in the current generation and the arrangement in the previous (previous) generation, but if there is no more gain in the loss function described above (Evolution) process (steps S207 to) After the repetition of the step S211), the radio map to be used for the actual location service is finally constructed (S213).

If gain exists in the arrangement of the current generation compared to the arrangement of the previous generation, the radio map constructing apparatus 100 generates the chromosomes of the next generation (S205) and generates a temporary radio map corresponding to each of the generated chromosomes ( In operation S207, the unlabeled fingerprint sequence may be disposed in each of the generated radio maps (S209), and the arrangement may be evaluated (S211).

In evolution, some of the chromosomes of the current generation are filtered out (reselected) and the remaining chromosomes can be newly generated to make up the entire chromosome. For example, 30 chromosomes with a small loss according to the loss function among 100 chromosomes may be selected and the remaining 70 may be newly generated to constitute the total 100 chromosomes of the next generation.

The number of generations required for convergence of the genetic algorithm depends on the number of chromosomes, the size of the commercial area (building, etc.), the number of HMM states, the number of training samples (unlabeled fingerprint sequences), and the like.

The genetic algorithm according to the present invention can construct a radio map that can converge faster and provide higher accuracy than the self-learning technique by utilizing address-labeled fingerprints while covering a wider range of radio maps. have.

As evolution ends through evaluation and one particular chromosome is selected, the radio map building apparatus 100 constructs a radio map for the actual location service (S213).

For example, the radio map construction apparatus 100 divides a commercial area where a radio map is to be constructed into partitions and the like, and uses each of the partitions (e.g., not yet mapped) using a labeled fingerprint and a selected position of the last selected chromosome. Calculate the fingerprint to be mapped to the central location of the passage section, address area that cannot be collected, etc.).

The radio map building apparatus 100 may further construct a radio map for another commercial area.

The constructed radio map may be provided to the portable terminal 200 or another server, and the portable terminal 200 or the server may provide a location service for the portable terminal 200 using the provided radio map.

The radio map construction method according to the present invention collects fingerprints that can be labeled with a specific address through crowd sourcing utilizing a transaction that can be mapped. In addition, the radio map method according to the present invention enables the accurate collection position of the address-labeled fingerprint to be determined using a genetic algorithm and also a fingerprint sequence collected through cloud sourcing, thereby building a map.

Accordingly, the radio map constructed in accordance with the present invention can construct an accurate radio map substantially faster.

As a result of the experiment, the radio map constructed according to the present invention was able to confirm through the experiment that the indoor road guidance service having a higher accuracy than the existing radio map is possible.

Although the accuracy may vary depending on the characteristics of the commercial area, experiments at Lotte World Mall, COEX, and Times Square in Korea show that the location prediction accuracy is improved by about 30% compared with the existing radio map. .

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

Claims

A radio map building method performed by a radio map building device,

(a) generating a plurality of chromosomes comprising a set of pairs consisting of a fingerprint labeled with an address and a predetermined location within a region of the address;

(b) generating a temporary radio map using the pair of chromosomes;

(c) placing a collected fingerprint sequence using the temporary radio map; And

(d) evaluating the placement of the fingerprint sequence;

Radio map construction method using genetic algorithm.
The method of claim 1,

The predetermined position is randomly selected within the area of the corresponding address when at least performing step (a) for the first time to construct the first generation,

Step (b) to (d) is performed sequentially for each of the plurality of chromosomes,

How to build a radio map.
The method of claim 2,

In step (d), evaluating the placement of the fingerprint sequence based on the temporary radio map generated using the loss function reflecting the user's mobility in the fingerprint sequence.

How to build a radio map.
The method of claim 3,

The radio map construction method,

Filtering a portion of the plurality of chromosomes according to the evaluation result in step (d) and constructing a next generation including the filtered portion of the chromosomes and a new chromosome, repeating steps (b) to (d) To perform,

How to build a radio map.
The method of claim 1,

Before the step (a), further comprising the step of modeling the internal structure of the building on which the radio map is to be built in a topology according to the Hidden Markov model (HMM),

The totology represents each location in the building as a state and represents the geographical accessibility between the locations as a transition.

How to build a radio map.
The method of claim 5,

The temporary radio map is generated through the process of generating a fingerprint corresponding to each location within a building using the set of waste fish and using an interpolation technique.

How to build a radio map.
The method of claim 4, wherein

The radio map construction method,

If there is no more gain as a result of the evaluation using the loss function compared to the previous generation, terminating the repetition of the steps (b) to (d) through generation evolution,

How to build a radio map.
The method of claim 1,

Before step (a) above,

Collecting fingerprints corresponding to the designated address through the designated app; And

Classifying the address-labeled fingerprints according to an address through the collection;

The specified app is a payment app or a resource usage metering app,

How to build a radio map.
The method of claim 8,

Steps (a) to (d) are performed for fingerprints of addresses classified as commercial areas,

How to build a radio map.
The method of claim 9,

If the classified address is a residential area, further constructing a radio map of the residential area through an interpolation technique for fingerprints corresponding to the address of the residential area,

How to build a radio map.
The method of claim 8,

Clustering the collected fingerprints into a plurality of groups corresponding to the designated address and determining a fingerprint to be labeled with the address using the fingerprints of the clustered group;

How to build a radio map.
The method of claim 8,

Before step (a) above,

Collecting a fingerprint sequence including a fingerprint and a series of collection times using a designated app; further comprising:

How to build a radio map.
The method of claim 8,

The collecting of fingerprints corresponding to the designated address may include determining an identifier and signal strength of one or more wireless APs collected by the designated app during a designated time before, after, or before or after a transaction occurs by the designated app. Collecting fingerprints including a pair;

How to build a radio map.
A radio map building method performed by a radio map building device,

Collect and accumulate fingerprints labeled with addresses through cloud sourcing,

Determining an optimal position in the region of the address for each of the fingerprints labeled with the address using a genetic algorithm,

How to build a radio map.
The method of claim 14,

Using a set of pairs consisting of a fingerprint labeled with the address and a predetermined location within a region of the address as a chromosome in the genetic algorithm,

How to build a radio map.
The method of claim 15,

Collecting an unlabeled fingerprint sequence, placing the collected fingerprint sequence based on a temporary radio map corresponding to the chromosome and evaluating the placement;

How to build a radio map.
The method of claim 16,

By using the results of the evaluation to determine the chromosome to be used in the next generation and to determine whether the evolution (Evolution),

How to build a radio map.