Method, data processing apparatus, and positioning apparatus for generating fingerprint map for unexplored area

Generative models like GANs automate fingerprint map construction for indoor positioning, reducing time and cost by calculating reference points and access points, thereby improving indoor positioning efficiency.

JP2024160920AInactive Publication Date: 2024-11-15EWHA UNIV IND COLLABORATION FOUND
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
JP2023184711
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-02
Filing Date
2023-10-27
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Fingerprint-based indoor positioning requires extensive time and resources for building fingerprint maps, which is inefficient and costly.

Method used

Utilizing generative models, specifically GAN-based models, to automatically construct fingerprint maps by calculating reference points and access point locations, extracting feature vectors, and generating fingerprint maps using arbitrary noise as a condition.

Benefits of technology

Enables the creation of fingerprint maps without a wireless signal collection process, enhancing the versatility and universality of indoor positioning services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024160920000001_ABST
    Figure 2024160920000001_ABST
Patent Text Reader

Abstract

To provide a method for generating a fingerprint map for an unexplored area.SOLUTION: A method for generating a fingerprint map for an unexplored area includes the steps of accepting, by a data processing apparatus, input of locations of reference points and locations of a plurality of access points (APs) for a service area, extracting, by the data processing apparatus, feature vectors for each AP on a basis of the locations of the reference points and the locations of the APs, generating, by the data processing apparatus, a fingerprint map for each of the APs by inputting, as conditions, the feature vectors extracted for each corresponding AP into a previously trained generative model for each of the APs and inputting random noise, and generating, by the data processing apparatus, a final fingerprint map for the service area by combining the fingerprint maps generated for the APs.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The technology described below relates to techniques for automatically generating fingerprint maps.

[0002] This application is the result of research conducted with funding from the government (Ministry of Science and ICT) for fiscal year 2023 and support from the Korea Research Foundation (No. NRF-2021R1A2B5B01002906) and the Korea Institute for Information and Communications Technology Planning and Evaluation (No. RS-2022-00155966) and the Center for Development of Innovative Human Resources for Artificial Intelligence Fusion (Ewha Women's University). [Background technology]

[0003] Indoor positioning refers to positioning in an area where it is difficult to receive a signal from a satellite navigation system, such as indoors. Representative indoor positioning methods use wireless signals, such as Wi-Fi and Bluetooth (registered trademark). Furthermore, a representative example of wireless signal-based indoor positioning is fingerprint technology. Summary of the Invention [Problem to be solved by the invention]

[0004] The fingerprint-based indoor positioning requires collecting wireless signals from a service area in advance to build a fingerprint map, which has the disadvantage that it takes a lot of time and money to build the fingerprint map.

[0005] The techniques described below attempt to automatically build fingerprint maps using generative models. [Means for solving the problem]

[0006] A method for generating a fingerprint map for an unsearched area includes the steps of: a data processing device receiving input of a position of a reference point for a service area and the positions of a number of access points (APs); the data processing device extracting a feature vector for each of the APs based on the position of the reference point and the position of the AP; the data processing device inputting the extracted feature vector for each of the APs as a condition into a pre-trained generative model for each of the APs, and inputting the arbitrary noise to generate a fingerprint map for each of the APs; and the data processing device combining the fingerprint maps generated for the APs to generate a final fingerprint map for the service area.

[0007] A data processing device that generates a fingerprint map for an unsearched area includes an input device that receives input of the positions of reference points and APs for a service area, a storage device that stores a generative model that is input as conditions a fingerprint map of a specific area and feature vectors extracted based on APs at reference points in the specific area and is trained to generate a fingerprint map for the specific area, and a calculation device that extracts the distances from each of the reference points for the service area to the AP and the angles between each of the reference points and the AP and inputs them as conditions for the generative model, and inputs arbitrary noise into the generative model to generate a fingerprint map for the service area.

[0008] A positioning device that uses a virtual fingerprint map includes a communication device that receives a radio signal from an AP at a current location, a storage device that stores a virtual fingerprint map, and a calculation device that refers to the virtual fingerprint map and estimates the current location based on the strength of the radio signal from the AP. Effect of the Invention

[0009] The technology described below can automatically build a fingerprint map using only the positions of access points (APs) arranged in a service area without a wireless signal collection process. Therefore, the technology described below imparts versatility to a fingerprint-based positioning technique and contributes to the universalization of indoor positioning services. [Brief description of the drawings]

[0010] [Figure 1] 1 is an example of a fingerprint-based positioning system. [Diagram 2] 1 is an example of a general process for constructing a generative model. [Diagram 3] This is an example of a specific learning process for a generative model. [Figure 4] 1 is an example of a process in which a data processing device generates a fingerprint map using a generative model. [Diagram 5] 1 is an example of a data processing apparatus for generating a fingerprint map for an unsearched area. [Figure 6] 1 is an example for a positioning device using a virtual fingerprint map. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] The technology described below can be modified in various ways and can have various embodiments, so a specific embodiment will be illustrated in the drawings and described in detail. However, this is not intended to limit the technology described below to a specific embodiment, and it should be understood that the technology described below includes all modifications, equivalents, or alternatives included in the spirit and technical scope of the technology described below.

[0012] Terms such as "first," "second," "A," and "B" may be used to describe various components, but the components are not limited by these terms and are used only to distinguish one component from another. For example, a first component can be named a second component, and similarly, a second component can be named a first component, without departing from the scope of the technology described below. The term "and / or" includes any combination of two or more associated listed items or any of two or more associated listed items.

[0013] In the terms used in this specification, the singular expression should be understood to include the plural expressions unless the context clearly indicates otherwise, and the term "comprise" or the like should be understood to mean the presence of a stated feature, number, step, operation, component, part, or combination thereof, but not to exclude the presence or additional possibility of one or more other features or numbers, steps, operations, components, parts, or combinations thereof.

[0014] Before describing the drawings in detail, it is to be clear that the division of components in this specification is merely a division according to the main function of each component. That is, two or more components described below may be combined into one component, or one component may be divided into two or more components according to more specific functions. Each of the components described below may additionally perform part or all of the functions of the other components in addition to its own main function, and some of the main functions of each component may be exclusively performed by the other components.

[0015] Additionally, in performing a method or method of operation, the steps of the method may be performed in an order other than that stated, unless the context clearly dictates a particular order, i.e., the steps may be performed in the same order as stated, substantially simultaneously, or in reverse order.

[0016] The technology described below is a technique for generating a fingerprint map for positioning in a service area.

[0017] The service area refers to an area where it is difficult to receive satellite navigation signals. Therefore, the service area hereinafter is used to include indoor areas.

[0018] The techniques described below use generative models to generate fingerprint maps.

[0019] Various types of generative models are being researched. For example, generative models include autoencoders, GANs (generative adversarial networks), and diffusion models. As explained below, the researchers used GAN-based models. Therefore, the following explanation focuses on GAN models. However, the model for generating the fingerprint map may be any of various types of generative models.

[0020] Hereinafter, it will be explained that the learning device constructs a generative model using the learning data. The learning device is a device capable of data processing and model learning. The learning device can also construct the learning data. For example, the learning device can be realized by a PC, a server on a network, a smart device, a chipset with a dedicated program built in, etc.

[0021] Hereinafter, it will be described that the data processing device generates the fingerprint map using the generative model. For example, the data processing device can be realized by a PC, a server on a network, a smart device, a chipset with a dedicated program embedded, etc.

[0022] In the following description, the wireless signal may be any of various types of signals such as Wi-Fi, Bluetooth, etc. A terminal receives the wireless signal in a service area. The terminal may be any of various types of devices such as a smart device, a mobile device, a wearable device, a robot, etc.

[0023] In the following description, the positioning device performs positioning in the service area using the generated fingerprint map. The positioning device may be a device such as the terminal or a server.

[0024] FIG. 1 is an example of a fingerprint-based positioning system 100 .

[0025] Although the service area S is actually a three-dimensional space, the fingerprint map is generally composed of information on a two-dimensional plane. The service area S can be defined based on a certain coordinate system by the area (including the position of the reference point) and the position of the AP (access point). The AP is a device that transmits wireless signals such as Wi-Fi in the service area.

[0026] 1A shows a process in which a data processing device 50 generates a fingerprint map. The data processing device 50 extracts certain vector information from AP location information. The data processing device 50 inputs arbitrary noise and vector information into a pre-constructed generative model to generate a fingerprint map M for a service area S. The fingerprint map generally has wireless signal strength information for each rectangular section in the form of a grid map. The generative model construction process and a specific fingerprint map generation process will be described later.

[0027] In FIG. 1, (B) and (C) show a process in which a positioning device performs positioning in a service area using a fingerprint map.

[0028] A positioning device 110 such as a smart device can estimate its current location based on the strength of a wireless signal received from an AP at its current location in a service area S by referring to the fingerprint map.

[0029] The terminal 80 collects the strength of a wireless signal received from an AP at the current location of the service area S. The terminal 80 transmits the wireless signal strength information to a separate positioning device 120. The positioning device 120, such as a computer device or a server, can estimate the current location of the terminal 80 based on the strength of a wireless signal received from an AP at the current location of the service area S by referring to a fingerprint map. The positioning device 120 can transmit the positioning result to the terminal 80.

[0030] 2 is a schematic diagram of an example of a process 200 for constructing a generative model. It will be described that a learning device constructs the generative model. The learning device corresponds to a computer device that controls the data processing of the fingerprint and the learning process of the generative model.

[0031] The learning device collects pre-generated fingerprint maps (210). The fingerprint map used in the model building process is called the training fingerprint map. The training fingerprint map can be generated manually based on the strength of radio signals collected in a specific area according to conventional methods. The training fingerprint map can also be extracted from a public database (DB).

[0032] The learning device extracts a wireless signal feature vector from the training fingerprint map (220). The wireless signal feature vector is used as additional information (conditions) in the training process of the generative model.

[0033] The process by which the learning device extracts a wireless signal feature vector will now be described. The learning device extracts a wireless signal feature vector for each AP from the learning fingerprint map.

[0034] The learning fingerprint map stores the strength of wireless signals received at each reference point (or data point) in the service area. The learning fingerprint map generally has wireless signal strength information for each rectangular section in the form of a grid map. The following Equation 1 shows a fingerprint X(i) at the i-th reference point in the service area.

[0035]

number

[0036] X(i) is the fingerprint at the i-th reference point. s j is the wireless signal strength received from the jth AP (j is the AP index) at the ith reference point. N ap is the number of APs located in the service area. data is the number of reference points in the service area.

[0037] The learning device can extract a wireless signal feature vector for each AP. The learning device can define the feature vector for each AP in the fingerprint map as shown in Equation 2 below. X j (i) denotes the strength of the wireless signal received from APj at the i-th reference point.

[0038]

number

[0039] The researchers extracted (i) the distance to the AP and (ii) the angle with the AP as feature values ​​based on a reference point in the service area. That is, a wireless signal feature vector can be defined as (distance value, angle) for each reference point. The learning device can extract the wireless signal feature vector based on the position of the AP known in advance in the service area based on a specific coordinate system. The researchers used the azimuth angle (angle on a plane) between the reference point and the AP as the feature value.

[0040] Alternatively, the learning device can estimate the distance between a specific point and an AP based on the strength of the wireless signal at the specific point using a well-known wireless signal propagation model, such as the log-distance path loss model.

[0041] The learning device can extract a radio signal feature vector represented by the following Equation 3 from the fingerprint map. In other words, the learning device can generate a new fingerprint map (feature vector map) represented by the following Equation 3. X j represents the wireless signal feature vector for APj.

[0042]

number

[0043] d i is the distance from the i-th reference point to APj, and a i is the angle between the i-th reference point and APj.

[0044] A learning device trains 230 a generative model using the training fingerprint map and the wireless signal feature vector.

[0045] Figure 3 shows an example of a specific learning process for a generative model.

[0046] The researchers used a conditional GAN ​​model (cGAN), which consists of a mutually adversarial generator network (G) 310 and a discriminator network (D) 320.

[0047] The generator network (G) 310 is the actual fingerprint map x real We learn the distribution of x from arbitrary noise z to generate a fake fingerprint map x fake =G(z|c). At this time, the generator network (G) 310 receives the above-mentioned wireless signal feature vector as a condition. The wireless signal feature vector includes the distance and angle from the reference point to the AP. Figure 3 shows a condition c, which is a wireless signal feature vector for n reference points.

[0048] The discriminator network (D) 320 receives the fingerprint map G(z|c) generated by the generator network (G) 310 and the actual fingerprint map, and discriminates whether G(z|c) is real or fake. At this time, the discriminator network (D) 320 also receives a condition that is a wireless signal feature vector.

[0049] The more similar the distribution of the generated fingerprint map G(z|c) is to the distribution of the actual fingerprint map, the closer the discrimination value D(G(z|c) output by the discriminator network (D) 320 is to 1. The objective function of the cGAN 300 can be defined as Equation 4 below.

[0050]

number

[0051] p data (x real ) is the distribution of the actual fingerprint map, and z~p z (z) denotes the distribution of the generated fingerprint maps.

[0052] The researchers used AC-GAN (Auxiliary Classifier GAN) to prevent overfitting of the generator network (G) 310. AC-GAN includes a classifier (class loss) that predicts the condition (distance and angle) in addition to a discriminator (sample loss) that classifies whether the generated fingerprint map G(z|c) is real data. The cGAN 300 is trained in the discriminator network (D) 320 in a direction that minimizes both sample loss and class loss. The use of AC-GAN may be an optional configuration for generating fingerprint maps.

[0053] 4 is an example of a process 400 in which a data processing device generates a fingerprint map using a generative model. FIG 4 is an example of fingerprint map generation for an unknown service area in which wireless signal information has not been searched for.

[0054] The data processing device receives information about a service area for which a fingerprint map is to be generated (410). The data processing device can receive location information about the service area (the location of the reference area in the service area) and the locations of multiple APs arranged in the service area. It is assumed that there are n APs arranged in the service area.

[0055] First, the description will focus on the generation of the fingerprint map construction for AP1. The data processing device can generate the above-mentioned wireless signal feature vector based on the positions of the reference points in the service area and the position of AP1 (420). The wireless signal feature vector is composed of the distance and angle from each of the reference points to AP1. The wireless signal feature vector is information used as a condition in cGAN.

[0056] The data processing device inputs an arbitrary noise z to a generator network G (430). The generator network G is a network that has completed learning in the process of Figure 3. The generator network G generates a fingerprint map for AP1 (440). The left side of Figure 4 shows an example of generating a fingerprint map for one AP.

[0057] The data processing device repeats the same process for all APs in the service area to generate n fingerprint maps (450).

[0058] The data processor merges (460) the n fingerprint maps to generate a single fingerprint map that includes multiple wireless signal strength vectors.

[0059] The positioning device can then use the single fingerprint map generated for the service region to locate the terminal's position.

[0060] The researchers examined the performance of their generative model for calculating the fingerprint map.

[0061] We will briefly explain the model the researchers constructed and the process of generating the fingerprint map.

[0062] The researchers used TensorFlow 1.15.0. The generator network G and discriminator network D they built both contain one hidden layer with a ReLu activation function, and the discriminator network D has a sigmoid activation function at the output layer.

[0063] The researchers sampled 100 uniformly distributed variables in U[-1,1] and used them as input noise. They built a generative model using four real-world domain datasets. The four real-world domains are listed in Table 1 below.

[0064] [Table 1]

[0065] The researchers evaluated (a) the deviation in distance (m) between the estimated location using the generated fingerprint map and the actual location, and (b) the precision of the estimated location using the generated fingerprint map. Precision was determined with a margin of error of 1.5 m from the correct answer.

[0066] The researchers evaluated the performance of the generative model in estimating the fingerprints of unexplored points. The researchers built a generative model for each of the four domains (environments) and used the generative model to estimate the fingerprints of unexplored points in those domains. The researchers removed the value of an arbitrary point from the actual fingerprint map in each of the four environments and used the generative model to estimate the fingerprints of the removed points. Table 2 below shows the results of estimating the fingerprints of unexplored points. Table 2 shows the average values ​​of the estimation results. Fingerprint estimation of unexplored points can be called fingerprint map augmentation.

[0067] [Table 2]

[0068] In Table 2, Unaug. is the result of localization using a fingerprint map where the fingerprints of unexplored points are not augmented. As mentioned above, the researchers constructed two generative models. In Table 2, cGAN is a general conditional GAN, and AC-GAN is a GAN with an added class loss. When localization was performed with an augmented fingerprint map, the performance (distance deviation) improved to a maximum of 0.77m. Referring to Table 2, when the service area was complex and had many obstacles, the generative model showed higher performance.

[0069] The researchers evaluated the performance of the generative model in estimating fingerprint maps for areas where AP placement was changed. The researchers built a generative model for each of the four areas (environments), changed the placement of APs in the area, and then used the generative model to generate a fingerprint map. Table 3 below shows the results of evaluating the performance of the fingerprint map in areas where AP placement was changed.

[0070] [Table 3]

[0071] In Table 3, Upper is the performance when using the actual fingerprint map. Even in a complex environment with many obstacles (Ofc-HighObs), the generative model showed significant performance in positioning services.

[0072] The researchers evaluated the performance of generative models to estimate fingerprint maps in completely new domains. They built generative models for each of the four domains (environments) and used the generative models to generate fingerprint maps for the other domains. Table 4 below shows the results of evaluating the performance of the fingerprint maps generated in the new environments.

[0073] [Table 4]

[0074] In Table 4, Upper is the performance when using an actual fingerprint map. Referring to Tables 3 and 4, generating a new fingerprint map using a generative model built in the same environment (structure, obstacles or not) showed slightly higher performance. However, generating a fingerprint map for a completely new environment using a generative model (Table 4) also showed significant results in terms of positioning services.

[0075] FIG. 5 is an example of a data processing device 500 that generates a fingerprint map for an unsearched area. The data processing device 500 is a device that generates a fingerprint map for a new service area using the above-mentioned generation model. The data processing device 500 can be physically realized in various forms. For example, the data processing device 500 may have the form of a computer device such as a PC, a smart device, a network server, a chip set dedicated to data processing, etc.

[0076] The data processing device 500 may include a storage device 510 , a memory 520 , a processing device 530 , an interface device 540 , a communication device 550 , and an output device 560 .

[0077] The storage device 510 stores the above-mentioned generative model. The generative model is a model that has completed learning with the learning data through the above-mentioned process.

[0078] The storage device 510 can store information for the new service area. The service area information can include the location of the area, the locations of the reference points, the locations of each of the APs, and the like.

[0079] The storage device 510 can store a program that controls the process of generating a fingerprint map using a generative model.

[0080] The storage device 510 can store the generated fingerprint map.

[0081] The memory 520 may store data and information generated in the process of the data processing device 500 generating a fingerprint map.

[0082] The interface device 540 is a device that receives certain commands and data input from the outside.

[0083] The interface device 540 can receive an input of the generative model from a physically connected input device or an external storage device. The interface device 540 can receive an input of information on the service area from a physically connected input device or an external storage device.

[0084] The interface unit 540 may also communicate the generated fingerprint map to an external object.

[0085] The interface device 540 is meant to include a configuration for transmitting data or information received by the communication device 550 to the inside of the data processing device 500 .

[0086] The communication device 550 refers to a configuration for receiving and transmitting certain information via a wired or wireless network.

[0087] The communication device 550 can receive a generative model from an external object.

[0088] The communication device 550 can receive service area information from an external object.

[0089] The communication device 550 may also transmit the generated fingerprint map to an external object, such as a positioning device.

[0090] The output device 560 is a device that outputs certain information. The output device 560 can output an interface, a fingerprint map, etc., required for the data processing process.

[0091] The computing device 530 uses the information of the service area to calculate the distance and angle between each reference point of the service area and the AP. The computing device 530 generates the feature vector described above. The computing device 530 calculates the feature vector d for the reference point i. i , a iThe computing device 530 generates a feature vector set for n reference points located throughout the service area. The computing device 530 computes a feature vector set for each AP.

[0092] The computing device 530 generates a fingerprint map for each AP arranged in the service area. The computing device 530 inputs a feature vector extracted under any noise and conditions to the generation model to generate a fingerprint map. The computing device 530 generates a fingerprint map for each AP. The computing device 530 combines the fingerprint maps of all the APs to generate a final fingerprint map having strength information for multiple wireless signals. This process is as described in FIG. 4.

[0093] The computing device 530 may be a device such as a processor, AP, or a programmed chip that processes data and performs certain operations.

[0094] In some cases, the data processing device 500 may be a positioning device. In this case, the data processing device 500 can perform positioning based on wireless signal information received from a service area by using the generated fingerprint map.

[0095] FIG. 6 is an example for a positioning device 600 using a virtual fingerprint map.

[0096] The positioning device 600 may collect information on a service area in which it is located. The positioning device 600 may transmit the service area information to the data processing device. The positioning device 600 may receive a fingerprint map for the service area generated by the data processing device using a generative model. The fingerprint map generated by the data processing device is not a map generated by collecting actual wireless signal strength, and therefore may be called a virtual fingerprint map. In the following description, it is assumed that the positioning device 600 has a virtual fingerprint map for the service area in which it is currently located.

[0097] The positioning device 600 may be any of a variety of types of devices. For example, the positioning device 600 may be any of a variety of types of devices, such as a smart device, a wearable device, a robot, a vacuum cleaner, etc.

[0098] The positioning device 600 may include a storage device 610 , a memory 620 , a processing device 630 , an interface device 640 and a communication device 650 .

[0099] The storage device 610 stores the virtual fingerprint map described above.

[0100] The storage device 610 can store a program for estimating a current location using the fingerprint map. For example, the program can be a program for estimating a location based on a Nearest Neighbor (NN) algorithm.

[0101] The memory 620 can store data and information generated in the process of the positioning device 600 performing positioning using the fingerprint map.

[0102] The interface device 640 is a device that receives certain commands and data input from the outside.

[0103] The interface device 640 can receive the virtual fingerprint map input from a physically connected input device or an external storage device.

[0104] The interface device 640 is meant to include a configuration for transmitting data or information received by the communication device 650 to the inside of the positioning device 600 .

[0105] The communication device 650 refers to a configuration that receives and transmits certain information via a wired or wireless network.

[0106] The communication device 650 can receive a fingerprint map from an external object.

[0107] The communication device 650 can receive wireless signals transmitted by APs in the service area.

[0108] The computing device 630 can determine the strength of the wireless signal transmitted by the AP at the current location.

[0109] The computing device 630 can estimate the current location by comparing the virtual fingerprint map with the wireless signal strength of the current location, which can include strength information for each of multiple wireless signals.

[0110] The computing device 630 may be a device such as a processor, AP, or a programmed chip that processes data and performs certain operations.

[0111] In addition, the above-mentioned generative model construction method, generative model-based fingerprint map generation method, or positioning method can be realized by a program (or application) including an executable algorithm that can be executed by a computer. The program can be provided by being stored in a transitory or non-transitory computer readable medium.

[0112] The non-transitory readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short moment, such as a register, cache, memory, etc. Specifically, the above-mentioned various applications or programs can be stored and provided in a non-transitory readable medium, such as a CD, a DVD, a hard disk, a Blu-ray (registered trademark) disk, a USB, a memory card, a ROM (read-only memory), a PROM (programmable read only memory), an EPROM (Erasable PROM, EPROM), an EEPROM (Electrically EPROM), or a flash memory.

[0113] The temporary readable medium refers to various types of RAM, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synclink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM).

[0114] It should be noted that the present embodiment and the drawings attached to this specification merely clearly illustrate a portion of the technical ideas contained in the above-mentioned technology, and that any modifications and specific embodiments that a person skilled in the art can easily deduce within the scope of the technical ideas contained in the specification and drawings of the above-mentioned technology are all obviously included in the scope of the above-mentioned technology.

Claims

1. receiving input of a location of a reference point and a plurality of access points (APs) for a service area by a data processing device; The data processing device extracts a feature vector for each of the APs based on the position of a reference point and the position of the AP; The data processing device inputs the feature vector extracted for each of the APs as a condition into a generative model previously trained for each of the APs, and inputs the arbitrary noise to generate a fingerprint map for each of the APs; combining the fingerprint maps generated by the data processing devices for the APs to generate a final fingerprint map for the service region.

2. The method for generating a fingerprint map for an unexplored region according to claim 1 , wherein the generative model is a conditional generative adversarial network (GAN) or an AC-GAN (Auxiliary Classifier GAN).

3. The method for generating a fingerprint map for an unsearched area as described in claim 1, wherein the feature vector for any of the APs includes, at each of the reference points, a distance between one of the reference points and any of the APs, and an angle between the one of the reference points and any of the APs.

4. receiving input of reference point locations and access point (AP) locations for a service area by a data processing device; The data processing device extracts a distance from each of the reference points to the AP and an angle between each of the reference points and the AP; The data processing device inputs arbitrary noise into a pre-trained generative model, and generates a fingerprint map for the service area by inputting the distance and the angle at the reference point as conditions; A method for generating a fingerprint map for an unsearched area, wherein the generative model is a model trained to generate a fingerprint map for the specific area by inputting a fingerprint map of a specific area and a feature vector extracted based on an AP at a reference point of the specific area as conditions.

5. an input device for receiving input of a reference point location and an access point (AP) location for a service area; a storage device that stores a generative model that is input as conditions based on a fingerprint map of a specific region and a feature vector extracted based on an AP at a reference point of the specific region and that is trained to generate a fingerprint map for the specific region; A data processing device that generates a fingerprint map for an unsearched area, comprising: a calculation device that extracts the distance from each of the reference points for the service area to the AP, and the angle between each of the reference points and the AP, and inputs them as conditions for the generative model, and inputs arbitrary noise into the generative model to generate a fingerprint map for the service area.

6. A data processing device for generating a fingerprint map for an unsearched area as described in claim 5, wherein the feature vector includes a distance between one reference point and the AP of the specific area for each of the reference points of the specific area, and an angle between the one reference point and the AP of the specific area.

7. 6. The data processing apparatus for generating a fingerprint map for an unsearched region according to claim 5, wherein the generative model is a conditional generative adversarial network (GAN) or an auxiliary classifier GAN (AC-GAN).

8. A communication device that receives a wireless signal from an access point (AP) at a current location; a storage device for storing the virtual fingerprint map; A computing device that refers to the virtual fingerprint map and estimates a current location based on the strength of the wireless signal of the AP, A positioning device that uses a virtual fingerprint map, wherein the virtual fingerprint map is generated by a generative model that is trained to generate a fingerprint map for a specific area by inputting a fingerprint map of a specific area and a feature vector extracted based on an AP at a reference point of the specific area as conditions.

9. The positioning device using a virtual fingerprint map as described in claim 8, wherein the feature vector includes a distance between one reference point and the AP of the specific area for each of the reference points of the specific area, and an angle between the one reference point and the AP of the specific area.

Citation Information

Patent Citations

  • System, method, and program for locating article

    JP2013053930A

  • System, method, and computer program for dynamically creating wireless maps

    JP2015531053A

  • Systems And Methods For Localization

    US20160183050A1

  • Predicting received signal strength in a telecommunication network using deep neural networks

    US20190150006A1

  • Network locationing RF planner

    US20210263128A1