Method of estimating a geographical position of a mobile device

Abstract

A method of estimating a geographical position of a mobile device configured to communicate using a telecommunication network comprising a plurality of cells is provided. The method comprises obtaining one or more data records from the mobile device, wherein each data record comprises a plurality of signal measurements and a respective cell identifier, wherein each signal measurement relates to a signal received by the mobile device from a cell of the plurality of cells attributed to the respective cell identifier. The method further comprises, for each data record, generating a ranked list of signal properties comprising the plurality of signal measurements, wherein each signal property of the ranked list of signal properties is related to a corresponding cell identifier assigned to the cell from which the signal was received based on an index of the signal property in the ranked list of signal properties. The method further comprises, for each data record, estimating the geographical position of the mobile device based on the ranked list of signal properties.

Description

Technical Field

[0001] This application relates to estimating the geographic location of a mobile device. It also relates to training models for estimating the geographic location of a mobile device. Furthermore, it relates to data processing, electromagnetic signals, computer software, and computer-readable media associated with these methods and systems. Background Technology

[0002] There are several reasons why it may be necessary to estimate the current geographic location (mobile device positioning) of a mobile device. Possible motivations include providing navigation instructions or remotely tracking the location of a vehicle. If a device needs to know its own location (e.g., for navigation), it can use a Global Navigation Satellite System (GNSS). This can be more challenging if another entity in the network needs to know the location of a mobile device (e.g., for monitoring traffic conditions or tracking a specific device). This is because it is not possible to induce the mobile device to transmit GPS information to other network entities.

[0003] This application relates to mobile devices configured to communicate using telecommunications networks. These may include user-operated mobile devices, such as smartphones, and may also include machine-to-machine (M2M) devices or other connected devices in the Internet of Things (IoT).

[0004] Existing positioning techniques can be performed using information about the serving cell of a mobile device. The basic technique used to locate mobile devices simply reports the centroid (or the physical location of the serving base station) of the cellular network's serving cell. However, this technique only provides a coarse location of the device with limited accuracy. This accuracy may be insufficient for some applications.

[0005] Other techniques use "fingerprinting" to estimate the location of mobile devices. This technique compares the network conditions of the serving cell with a database of network conditions for that cell at various locations. Network conditions can include cell identifier (cell ID), received signal strength (RSS), angle of arrival (AoA), time of arrival (ToA), and time difference of arrival (TDoA). Fingerprints can require significant effort to collect and may only be effective for a limited number of time frames due to changing conditions within the network.

[0006] Radio Positioning Systems (RPS) are a type of radio fingerprinting geolocation solution that uses information from mobile radio networks to determine the real-time location of mobile handheld devices. RPS can be used to predict the location of mobile terminals. However, these techniques rely on the similarity of spotted locations between reported data from the mobile device and training data. Reported and training data are quite complex, and spot similarity can be a difficult and time-consuming operation for the model or machine learning data processing unit. As a result, the accuracy of location estimation may be insufficient for certain purposes.

[0007] Furthermore, RPS is inefficient in estimating the location of certain categories of mobile devices. This is likely because radio conditions differ for specific groups of mobile devices. For example, flying mobile devices such as drones have different radio characteristics, leading to inaccurate estimations when using standard RPS.

[0008] Another existing technique is "data normalization" or "feature scaling," which attempts to normalize the range of features in MR for the model. While these methods may be useful for eliminating scaling mismatch between reported and training data, this alternative fails to adequately address the problems present in existing techniques such as RPS.

[0009] The purpose of this invention is to overcome these problems. Summary of the Invention

[0010] This application provides an improved method for estimating the geographic location of a mobile device. It also provides an improved method for providing training data to a model, a method for generating a sorted list of signal properties, and a method for associating the signal properties in the sorted list with corresponding cell identifiers. A novel method is proposed to significantly improve the quality of input data by reformatting and reconstructing neighboring cell information, allowing machine learning (ML) models to better learn patterns. By reconstructing measurement reports that provide neighboring cell information in an indeterminate order and instead providing signal data (e.g., signal strength data) in a sorted list of signal properties, where the index of the signal property is associated with the corresponding cell identifier, this invention can improve the accuracy and performance of geographic location estimation models.

[0011] Existing methods rely on artificial intelligence to correlate signal strength measurements with the identifiers of each neighboring cell in the measurement records. In contrast, using the method described in this application, the model is able to directly correlate signal measurements with cell identifiers using the indices of signal measurements in a sorted list. As a result, the accuracy of the geographic location estimated by the model can be improved. Furthermore, the model's input includes signal measurements that are more consistently sorted compared to existing methods using measurement records (MR). This model can readily correlate signal measurements with cell identifiers and therefore can operate more efficiently and effectively.

[0012] The method described in this application is particularly relevant to estimating the geographic location of unmanned aerial vehicles (UAVs). By applying the proposed method to UAV RPS data, ML models with performance exceeding that of existing technology solutions can be constructed.

[0013] A method for estimating the geographic location of a mobile device is provided, wherein the mobile device is configured to communicate using a telecommunications network comprising multiple cells. The method includes obtaining one or more data records from the mobile device, wherein each data record includes multiple signal measurements and corresponding cell identifiers. Each signal measurement relates to a signal received by the mobile device from a cell among the multiple cells attributed to the corresponding cell identifier. The method further includes, for each data record, generating an ordered list (e.g., an array or vector) comprising the signal properties of the multiple signal measurements. Based on the index of the signal properties in the ordered list, each signal property in the ordered list is associated with a corresponding cell identifier assigned to the cell from which it receives signals. In other words, each of the multiple signal measurements appears in the ordered list at a position having an index corresponding to the corresponding cell identifier of the signal measurement. The method also includes estimating the geographic location of the mobile device based on the ordered list of signal properties.

[0014] By using a list of signal measurements sorted according to the cell identifier of the received signal, models using the above method (e.g., machine learning models) are able to estimate the geographic location of mobile devices more accurately.

[0015] Mobile devices can receive signals from only a subset of multiple cells. Each signal can be encoded using a corresponding cell identifier assigned to the cell from which it is received. For each of the multiple received signals, the mobile device can determine one or more signal measurements of the signal (in other words, the mobile device can measure one or more attributes of the signal; for example, the mobile device can determine a signal strength measurement by measuring the received power and / or signal quality measurements). The multiple signal measurements obtained from the mobile device can relate to signals received from different cells at the same time the mobile device is in operation.

[0016] The characteristics of a received signal that are directly determined by measuring the received signal can be called a signal "measurement". If a list contains measurements and also includes numbers indicating that no signal was received, the elements of that list can be called signal "properties".

[0017] Advantageously, the order of signal properties within the sorted list is always the same. As a result, the sorted list of signal properties includes signal measurements associated with the corresponding cell identifiers based on the indices of the signal properties in the sorted list. Consequently, the model can operate more efficiently because it is not necessary to associate measurements with cell identifiers. Furthermore, the model may not require knowledge of the sorting. The model does not need to directly associate cell identifiers with signal properties because it is likely sufficient to know that signal properties at the same position in their respective sorted lists are associated with the same cell identifier.

[0018] Existing methods may use a list of signal strength measurements that have undergone inconsistent ordering. As an example, a first record of signal strength properties at a first time may contain signal strength measurements from six neighboring cells. A signal strength measurement received from cell identifier 123 may appear at position 1 in the list of signal strength properties in the first record. Subsequently, a second record of signal strength properties at a second time may contain signal strength measurements from the same six cells, including signals received from the cell assigned cell identifier 123. However, a signal strength measurement from that cell may appear at position 2 in the list of signal strength properties in the second record. This inconsistency can cause inaccuracies or inefficiencies in existing estimation methods. The method described above can address such problems in the prior art.

[0019] One or more data records may include a first data record corresponding to a first time and a second data record corresponding to a second time different from the first time.

[0020] Each data record may further include an identifier linked to the mobile device. The method may also include using the identifier from the first and second data records to determine that the first and second data records have been obtained from the same mobile device.

[0021] This can be achieved by determining that the first and second identifiers are the same.

[0022] For the second data record, the estimated geographic location of the mobile device can be further based on the estimated geographic location of the mobile device for the first data record.

[0023] By identifying signal measurements from different measurement records that involve the same mobile device, the model can use previous estimates of the mobile device's location to determine an updated estimate of the mobile device's location. This model can utilize the maximum a posteriori (MAP) technique. This can improve the accuracy of geolocation estimation.

[0024] A method is provided for providing training data to a model (e.g., a classifier, support vector machine, neural network, or other machine learning component) to estimate the geographic location of a mobile device based on multiple signal measurements. The method includes obtaining one or more data records from a mobile device (the training mobile device may be different from the mobile device used by the model to estimate its location). Each data record includes multiple signal measurements from the mobile device and corresponding cell identifiers, along with corresponding measurements of the mobile device's geographic location. The mobile device is configured to communicate using a telecommunications network comprising multiple cells. Each signal measurement relates to (indicates) a signal (received signal power) received by the mobile device from a cell among the multiple cells attributed to the corresponding cell identifier. The method also includes, for each data record, generating an ordered list (e.g., an array or vector) of signal properties comprising the multiple signal measurements. Based on the indices of the signal properties in the ordered list, each index of the signal properties in the ordered list is associated with a corresponding cell identifier assigned to the cell from which the signal is received. In other words, each of the multiple signal measurements appears in the ordered list at a position having an index corresponding to the corresponding cell identifier of the signal measurement. The method also includes, for each data record, providing the model with a sorted list of signal properties and corresponding measurements of the mobile device's geographic location.

[0025] Advantageously, if the model is trained using signal properties provided in a predefined order, it may be able to more easily identify the relationship between location data and signal measurement data. As a result, the model may be able to estimate geographic locations more accurately based on the signal measurement data. This is especially important when the model is a machine learning model. Structured signal data using these techniques significantly improves the accuracy of geographic location estimates using machine learning models. This improvement in accuracy is not what a technician would expect, as the structure of the data itself would not affect the results of the ML model.

[0026] The mobile device can be a first mobile device. The sorted list can be a first sorted list. The model can be configured to receive a second sorted list of signal properties obtained from a second mobile device configured to communicate using a telecommunications network as input. Based on the indices of the signal properties in the sorted list of signal properties, the index of each signal property from the second device's sorted list can be associated with a corresponding cell identifier assigned to the cell from which it receives signals. The model can be configured to provide an estimate of the geographical location of the second mobile device as output.

[0027] The second sorted list can be received indirectly from the second mobile device. This is likely because data obtained directly from the mobile device may contain signal measurements that are not sorted according to their corresponding cell identifiers. The data processor can receive the signal measurements and corresponding cell identifiers from the mobile device, process the data, and provide the second sorted list to the model.

[0028] The first and second sorted lists of signal properties can be sorted such that signal measurements in each of the corresponding sorted lists with the same index correspond to the same cell identifier.

[0029] Advantageously, the order of signal properties within each of the first and second sorted lists of signal properties is identical. Thus, each of the first and second sorted lists of signal properties includes a signal measurement associated with the corresponding cell identifier based on the index of the signal property in the corresponding sorted list. As a result, the model can operate more efficiently because it is not necessary to correlate measurements with cell identifiers. The model does not need to directly correlate cell identifiers with signal properties because it is likely sufficient to know that signal properties at the same position in their respective sorted lists are associated with the same cell identifier.

[0030] The accuracy of the model can be improved by providing training data that includes a sorted list of signal values, where the properties in the list associated with the same cell identifier are in the same position as the signal properties in the data (test data) provided to the model to estimate the geographic location of the mobile device.

[0031] The generated list of sorted signal properties may include assigning a value to each signal property in the sorted list, for each signal property having an index corresponding to a cell identifier for which no corresponding signal measurement has been received, indicating that the mobile device has not yet received a signal.

[0032] In other words, for each signal property in the sorted list corresponding to a cell identifier of any cell from which a mobile device has received a signal (and therefore has not yet obtained a signal measurement), an "invalid (null)" value is assigned to that property in the sorted list. Thus, the sorted list of signal properties includes signal properties for each unique cell identifier—values ​​that are non-invalid if measurements have been obtained and values ​​that are invalid if no measurements have been obtained.

[0033] An "invalid" value can be 0 or another value indicating that no signal was received (e.g., 999 or -999).

[0034] Preferably, for a mobile device, receiving signals from different cells assigned the same cell identifier should not be possible. However, theoretically, if this were to happen, then the list of signal properties corresponding to the sorting of the cell identifiers could be filled with signal measurements related to the strongest signal or the signal with the highest quality.

[0035] Mobile devices may be able to communicate over mobile networks according to any defined mobile standard, such as 2G, 3G, 4G, 5G, or any other standard. A mobile device can be a UE (User Equipment), or a device that includes a UE to provide connectivity to the mobile network. A mobile device can be an aerial vehicle. A mobile device can be an unmanned aerial vehicle (UAV or "drone"). A mobile device can be a user device on an unmanned or manned aircraft (e.g., a user device in the pilot's / passenger's pocket of an airplane, a glider, paraglider, jetpack, passenger balloon, airship, etc.). A mobile device can be an airborne device (e.g., a UE attached to a flying animal or weather balloon).

[0036] Drones and other aerial devices operate at altitudes higher than standard handheld mobile devices, which tend to be at ground level during normal operation. As a result, signals received at aerial mobile devices may experience different signal properties (e.g., signal strength and quality) than signals received at the same latitude and longitude at ground level. This could be because there are fewer obstructions to the signal at higher altitudes (e.g., buildings). There may also be fewer sources of interference (such as generators). Furthermore, depending on the drone's altitude, signal strength may be lower at ground level due to the distance from the transmitter (especially for small cells such as femtocells and / or picocells, which may use lower-power transceivers).

[0037] Due to different signal strength distributions and other differences in the signal properties experienced by airborne devices, the above methods may be exclusively applied to airborne devices (such as drones), making both training and testing data relevant to the airborne devices.

[0038] Each data record may further include a height measurement of the mobile device.

[0039] The model can be configured for use with signal measurements of a predetermined range of height, and the height measurement of the mobile device can be within the predetermined range of the height measurement.

[0040] Because drones experience different signal strength distributions and other differences in signal properties, models can be trained and tested using data from drones operating around the same altitude. For example, one model can be applied to drones with altitudes ranging from 100m to 500m. Different models can be used for drones with altitudes between 500m and 2km. Different models can exist for each band at 100m, such that the first model can be used for drones below 100m, the second for drones from 100m to 200m, the third for drones from 200m to 300m, and so on. The intervals can be 200m, 250m, 300m, etc., instead of 100m in the examples above.

[0041] Alternatively, the height of the mobile device can be another parameter provided to the model (during training and testing). This allows the model to potentially account for different signal properties at different heights. When the model utilizes neural networks or other machine learning techniques, height can be incorporated as an additional parameter and automatically taken into account.

[0042] Multiple signal measurements and corresponding cell identifiers obtained from mobile devices can be used to determine whether a handover process should be initiated.

[0043] Mobile devices in telecommunications networks are configured to transmit measurement records, enabling the initiation and configuration of handover procedures. The methods described above can utilize data already available in the network to estimate the geographical location of the mobile device. Advantageously, the device does not need to be configured to transmit additional data. These methods can be used with all existing mobile devices operating in the network. Furthermore, the mobile device may need to send the date for the handover. Therefore, there may not be a mechanism for the mobile device to opt out so that its signal measurements are not used for geographical location estimation. By using these methods instead of GPS / cell ID-based positioning methods, mobile terminal battery life can also be extended.

[0044] Multiple signal measurements may include: one or more signal strength measurements; one or more RSSI measurements; one or more RSRP measurements; one or more signal quality measurements; and / or one or more RSRQ measurements. Multiple signal measurements may include one or more location measurements. The mobile device can determine its location, and this can be included in the data log. The data log may also include one or more timing advance (TA) measurements.

[0045] When the sorted list of signal properties includes multiple measurements for each signal, the cell identifier can be associated with more than one position in the sorted list of signal properties. For example, the sorted list could contain signal strength properties of the cell identifiers in a predefined order, followed by signal quality properties of the cell identifiers in the same predefined order. Alternatively, multiple sorted lists can be provided, such that a first list includes signal strength properties in a predefined order and a second list includes signal quality properties in a predefined order.

[0046] The list of ordered signal properties can be only part of the "fingerprint," and may also include other data from the mobile device, such as altitude (as described above), serving cell data, and other data characterizing the network conditions experienced by the mobile device. Furthermore, other data characterizing the mobile device, such as weight and speed, may be included. Advantageously, by providing additional data indicating network conditions at a geographic location, the model may be able to estimate the mobile device's geographic location more accurately.

[0047] The method may further include receiving multiple signals from a subset of multiple cells at a mobile device. Each signal may include coded data. The coded data may include a cell identifier attributed to the signal received therefrom. The method may further include determining one or more signal measurements for each of the multiple received signals. The method may further include decoding the signal for each of the multiple received signals and determining the cell identifier encoded in that signal. The method may further include transmitting data records including signal measurements and corresponding cell identifiers to cells among the multiple cells. Alternatively, the mobile device may transmit the data records to a base station (e.g., a serving base station) or a server in a telecommunications network.

[0048] Obtaining signal measurements and corresponding cell identifiers may include receiving signal measurements and corresponding cell identifiers from mobile devices.

[0049] Determining the signal measurement for each of a plurality of cell identifiers may further include, if the mobile device has received more than one signal encoded with the cell identifier, then the received signal measurement for the cell identifier is equal to the maximum received signal measurement (e.g., maximum signal strength measurement or maximum signal quality measurement). This may be appropriate where the cell identifier is reused for different cells at different geographical locations within the network (e.g., physical cell ID).

[0050] Each cell can have a unique cell identifier (e.g., a cell global identity, CGI).

[0051] A method is provided that associates each signal property in a sorted list of signal properties with a corresponding cell identifier based on an index of the signal properties in the sorted list. The sorted list of signal properties is suitable for inputting into a model for estimating the geographic location of a mobile device based on multiple signal measurements. The method includes reading one or more data records. Each data record includes multiple signal data entries. Each signal data entry corresponds to a signal received by the mobile device. Each signal data entry includes a signal measurement of the signal and a corresponding cell identifier assigned to the cell from which the signal is received. The method also includes reading the cell identifier of the signal data entry for each of the one or more data records and for each of the multiple signal data entries in the data records. The method further includes associating the signal properties in the sorted list of signal properties with the corresponding unique cell identifier based on an index of the signal properties in the sorted list of signal properties, for each unique cell identifier read from the multiple signal data entries and from the one or more data records.

[0052] A one-to-one mapping from index to cell identifier can exist. This does not mean that the index and cell identifier are the same (although this is a possibility). Instead, each index in the sorted list corresponds to a cell identifier, but multiple indices in the list can refer to the same cell identifier. This can be useful when there is more than one measurement for each signal.

[0053] Advantageously, the above method can be used when processing measurement records received from mobile devices, enabling data to be fed to the model (for estimating the geographic location of the mobile device based on multiple signal measurements) in a consistent manner. As a result, for example, the model can operate more accurately to estimate the geographic location of the mobile device.

[0054] A method is provided for generating an ordered list of signal properties. The method includes reading a data record comprising multiple signal data entries. Each signal data entry corresponds to a signal received by a mobile device. Each signal data entry includes a received signal measurement of the signal and a corresponding cell identifier assigned to the cell from which the signal is received. The method further includes, for each of the multiple signal data entries in the data record, populating the ordered list of signal properties with the signal measurement of the signal data entry. Based on the index of the signal properties in the ordered list of signal properties, the updated signal properties are associated with the corresponding cell identifier of the signal data entry.

[0055] This sorted list can be used for training, validation, and testing (estimation) data.

[0056] Advantageously, the methods described above for processing data can be used to manipulate measurement records from one or more mobile devices into a format that allows models (especially machine learning models) to more accurately estimate the geographic location of one or more mobile devices.

[0057] The method may also include initializing each signal property in the sorted list with a value indicating that no signal was received. For example, the property could be 0 or -99999. This value can vary depending on the type of measurement. For instance, if no signal was received, the signal quality value could be 0, while the signal strength value could be -999.

[0058] Each cell identifier can be a Physical Cell ID (PCI).

[0059] Each cell identifier can be a Cell Global Identity (CGI). A cell identifier can include more than one identifier. For example, a cell identifier can include both PCI and CGI.

[0060] To further improve the accuracy of estimated geographic locations, the model can include some knowledge about how cell identifiers are geographically distributed. This can be done using Automatic Neighbor Relationship (ANR) techniques.

[0061] Associating cell identifiers with signal properties in a sorted list based on indices of those properties can be achieved by defining the structure of the sorted list using an index. The index provides a definition of how elements in the sorted list relate to other data. In this case, the indexing scheme can include a sorted list of cell identifiers, where the sorting of cell identifiers in the indexing scheme is correlated with the sorting of corresponding signal measurements in a sorted list of signal properties.

[0062] A method is provided for estimating the geographic location of a mobile device configured to communicate using a telecommunications network comprising multiple cells. The method includes obtaining one or more data records from the mobile device, wherein each data record includes multiple signal measurements and corresponding cell identifiers. Each signal measurement relates to a signal received by the mobile device from a cell among the multiple cells attributed to the corresponding cell identifier. The method also includes, for each data record, generating an ordered list (e.g., an array or vector) of signal properties comprising the multiple signal measurements. The ordered list of signal measurements conforms to an indexing scheme. The indexing scheme associates each signal property in the ordered list with a corresponding cell identifier assigned to the cell from which it receives signals, based on the index of the signal properties in the ordered list. In other words, each of the multiple signal measurements appears in the ordered list at a position having an index corresponding to the corresponding cell identifier of the signal measurement. The method also includes estimating the geographic location of the mobile device based on the ordered list of signal properties.

[0063] The geographical location of the mobile device can be a first geographical location of the mobile device at a first time. Multiple signal measurements and corresponding cell identifiers can be a first plurality of signal measurements and corresponding cell identifiers. The list of ordered signal properties can be a first ordered list of signal properties. The method may also include estimating a second geographical location of the mobile device at a second time different from the first time. Estimating the second geographical location of the mobile device may include obtaining a second plurality of signal measurements and corresponding cell identifiers from the mobile device. Each signal measurement in the second plurality of signal measurements may involve a signal received by the mobile device from a cell among a plurality of cells attributed to a corresponding cell identifier. Estimating the second geographical location of the mobile device may also include generating a second ordered list of signal properties including the second plurality of signal measurements. The second ordered list may also conform to an indexing scheme. Estimating the second geographical location of the mobile device may also include estimating the second geographical location of the mobile device based on the second ordered list of signal properties.

[0064] Advantageously, the order of signal properties within each of the first and second sorted lists of signal properties is identical (because each list conforms to the same indexing scheme). In this way, each of the first and second sorted lists of signal properties includes a signal measurement, which is associated with a corresponding cell identifier based on the index of the signal property in its respective sorted list, according to the indexing scheme. As a result, the model can perform location estimation more efficiently because it does not need to associate the measurement with the cell identifier. Furthermore, the model may not require knowledge of the indexing scheme. The model may not need to directly associate the cell identifier with the signal property, since it may be sufficient to know that signal properties at the same position in their respective sorted lists are associated with the same cell identifier.

[0065] A method is provided for training a model (e.g., a classifier, neural network, or other machine learning component) to estimate the geographic location of a mobile device based on multiple signal measurements. The method includes obtaining multiple signal measurements and corresponding cell identifiers from the mobile device, along with corresponding measurements of the mobile device's geographic location. The mobile device is configured to communicate using a telecommunications network comprising multiple cells. Each signal measurement relates to (indicates) a signal (received signal power) received by the mobile device from a cell among the multiple cells attributed to a corresponding cell identifier. The method also includes generating an ordered list (e.g., an array or vector) of signal properties comprising the multiple signal measurements. The ordered list conforms to an indexing scheme that associates the index of each signal property in the ordered list with a corresponding cell identifier assigned to the cell from which it receives the signal. In other words, each of the multiple signal measurements appears in the ordered list at a position having an index corresponding to the corresponding cell identifier of the signal measurement. The method also includes providing the ordered list of signal properties and the corresponding measurements of the mobile device's geographic location to the model.

[0066] Multiple signal measurements can be a first set of multiple signal measurements. The mobile device can be a first mobile device. The sorted list can be a first sorted list. The model can be configured to receive a second sorted list of signal properties obtained from a second mobile device configured to communicate using a telecommunications network as input. The second sorted list of signal properties can conform to an indexing scheme. The model can be configured to provide an estimate of the geographic location of the second mobile device as output.

[0067] The second sorted list can be received indirectly from the second mobile device. This is likely because data obtained directly from the mobile device may contain signal measurements that are not sorted according to the indexing scheme. The data processor can receive the signal measurements and corresponding cell identifiers from the mobile device, process the data, and provide the second sorted list to the model.

[0068] The first and second sorted lists of signal properties can conform to an indexing scheme such that signal measurements in each of the corresponding sorted lists with the same index correspond to the same cell identifier.

[0069] Advantageously, the order of signal properties within each of the first and second sorted lists of signal properties is identical (because each list conforms to the same indexing scheme). In this way, each of the first and second sorted lists of signal properties includes a signal measurement, which is associated with a corresponding cell identifier based on the index of the signal property in its respective sorted list, according to the indexing scheme. As a result, the model can operate more efficiently because it is not necessary to associate the measurement with the cell identifier. Furthermore, the model may not require knowledge of the indexing scheme. The model may not need to directly associate the cell identifier with the signal property, as it may be sufficient to know that signal properties at the same position in their respective sorted lists are associated with the same cell identifier.

[0070] In other words, for each signal property in the sorted list corresponding to a cell identifier from which a mobile device receives a signal (and therefore from which no signal measurement is obtained), an "invalid" value is assigned to that property in the sorted list. In this way, the sorted list of signal properties includes signal properties for each unique cell identifier—values ​​that are non-invalid if measurements have been obtained and values ​​that are invalid if no measurements have been obtained.

[0071] Obtaining multiple signal measurements and corresponding cell identifiers from a mobile device may include obtaining corresponding received signal strength measurements and signal quality measurements from the mobile device. The method may also include generating an ordered list of signal properties comprising the multiple received signal strength measurements and signal quality measurements. The ordered list of signal properties may conform to an indexing scheme that associates each signal property in the ordered list with a corresponding cell identifier assigned to the cell from which it receives a signal, based on the index of the signal quality properties in the ordered list. Therefore, (for estimation methods) estimating the geographic location of a mobile device based on the ordered list of signal properties will take into account both signal strength properties and signal quality properties. Alternatively (for training methods), providing the model with an ordered list of signal properties associated with the corresponding measurements of the mobile device's geographic location will then include providing the model with the signal strength properties and signal quality properties (ordered such that their indices are associated with the corresponding cell identifiers) associated with the corresponding measurements of the mobile device's geographic location.

[0072] A method for generating an indexing scheme is provided. The indexing scheme is adapted to associate each signal property in the sorted list of signal properties with a corresponding cell identifier. The sorted list of signal properties is suitable for input into a model (for estimating the geographic location of a mobile device based on multiple signal measurements). The method includes generating the indexing scheme. The method also includes reading one or more data records. Each data record includes multiple signal data entries. Each signal data entry corresponds to a signal received by the mobile device. Each signal data entry includes a received signal measurement of the signal and a corresponding cell identifier assigned to the cell from which the signal is received. The method further includes reading the cell identifier of the signal data entry for each of the one or more data records and for each of the multiple signal data entries in the data records. The method also includes updating the indexing scheme based on the index of the signal properties in the sorted list of signal properties to associate the signal properties in the sorted list with the corresponding unique cell identifier for each unique cell identifier read from the multiple signal data entries and from the one or more data records.

[0073] Advantageously, when processing measurement records received from mobile devices, an indexing scheme generated by the method described above can be used, enabling data to be provided to the model in a consistent manner (e.g., for estimating the geographic location of the mobile device based on multiple signal measurements). As a result, for example, the model can operate more accurately to estimate the geographic location of the mobile device.

[0074] A method is provided for generating an ordered list of signal properties. The method includes reading an indexing scheme that associates each signal property in the ordered list with a corresponding cell identifier based on an index of the signal properties in the ordered list. The method also includes generating the ordered list of signal properties. The method further includes reading a data record comprising multiple signal data entries. Each signal data entry corresponds to a signal received by a mobile device. Each signal data entry includes a received signal measurement of the signal and a corresponding cell identifier assigned to the cell from which the signal is received. The method also includes updating the ordered list of signal properties with the signal measurement of the signal data entry for each of the multiple signal data entries in the data record. Based on the index of the signal properties in the ordered list, the updated signal property is associated with the corresponding cell identifier of the signal data entry, as defined by the indexing scheme.

[0075] The sorted list of signal properties can conform to an indexing scheme. The sorted list of signal properties can contain multiple signal properties defined by the indexing scheme.

[0076] Indexing schemes can be generated using the methods described above.

[0077] Generating a sorted list of signal properties that conforms to an indexing scheme can include initializing each signal property in the sorted list with a property indicating that no signal was received. For example, this property could be 0 or -99999. This value can vary depending on the type of measurement. For instance, if no signal was received, the signal quality value could be 0, while the signal strength value could be -999.

[0078] An indexing scheme may include an ordered list of cell identifiers. The indexing scheme may associate signal properties in an ordered list of signal properties with identifiers in an ordered list of cell identifiers. The index of a signal property in the ordered list of signal properties may be the same as the index of the corresponding cell identifier in the ordered list of cell identifiers.

[0079] In other words, the indexing scheme can associate a signal property that is indexed in the sorted list of signal properties with the corresponding cell identifier that has the same index in the sorted list of cell identifiers.

[0080] A system configured to perform any of the methods described above is also provided. This system may be a telecommunications network or a specific component within a telecommunications network.

[0081] The elements of the system can be provided by a computer system. A computer system may include a processor, memory, data inputs, and data outputs. The computer system's memory may store instructions in software form. When executed on the processor, the instructions cause the computer system to perform the method steps required to implement the techniques described in this application.

[0082] Although some operations are described as being performed by specific elements of the system, these operations can be performed in a distributed manner, where some elements of the operation are performed by one system element and other elements are performed by another system element.

[0083] Computer software is also provided. The computer software includes computer-readable instructions that, when executed by a processor of a computer system, cause the computer system to perform any of the methods described above.

[0084] An electromagnetic signal is also provided. This electromagnetic signal carries computer-readable instructions that, when executed by the processor of a computer system, cause the computer system to perform any of the methods described above.

[0085] A computer-readable medium is also provided. This computer-readable medium includes instructions that, when executed by a processor of a computer system, cause the computer system to perform any of the methods described above. Attached Figure Description

[0086] The invention will now be described in more detail with reference to several non-limiting embodiments depicted in the following figures, wherein:

[0087] Figure 1 A mobile device configured to communicate with a telecommunications network is shown.

[0088] Figure 2 A map showing the example geographic locations of mobile devices and three cells is displayed.

[0089] Figure 3 The system and data flow diagrams are shown.

[0090] Figure 4 A flowchart illustrating a method for estimating the geographic location of a mobile device is shown.

[0091] Figure 5 A flowchart is shown illustrating a method for providing training data to a model used to estimate the geolocation of a mobile device based on multiple signal measurements.

[0092] Figure 6 A flowchart illustrates a method for associating each signal property in a sorted list with its corresponding cell identifier.

[0093] Figure 7 A flowchart is shown illustrating a method for generating a sorted list of signal properties.

[0094] Figure 8 A flowchart illustrating a method for restructuring data in a specific implementation is shown. Detailed Implementation

[0095] The existing information representation format from MR (Measurement Reports) uses nominal numbers to identify neighboring cells. The system used to define these numbers can vary depending on the mobile service provider. These measurement reports can be used to determine when to perform a handover and to which neighboring cell. They can also be used to re-establish a connection after signal loss. To support these operations, neighboring cells in the MR can be ordered from the neighbor with the strongest RSRP at position #1 to the neighbor with a weaker RSRP value at higher numbers. Alternatively, reference signal reception quality can be used as a criterion for classifying the list. Measurement reports are defined in the 3GPP specifications TS36.331 for LTE (4G) and TS38.331 for NR (5G), which are incorporated herein by reference.

[0096] It is known to use measurement reports to estimate the geolocation of mobile devices. RPS technology feeds MR data into machine learning (ML) models for training and geolocation estimation. However, the nominal cell numbers of neighboring cells in the measurement report can be misleading for machine learning (ML) models learning patterns.

[0097] For a given sample MR, a mobile device can report multiple neighboring cells with similar Reference Signal Received Power (RSRP). From the measurement reports, it is not always clear which cell is the #1 neighboring cell in the MR, which is #2, and so on. Instead, MR structured systems can sometimes use arbitrary and nominal numbering for the representation of neighboring cell information. These nominal numbering systems in MR can mislead the ML model and prevent the model from learning useful features (e.g., neighbor PCI) and associating each neighboring cell with its corresponding radio signal profile (e.g., RSRP).

[0098] refer to Figure 1The illustration shows a mobile device 100 configured to communicate with a cellular telecommunications network. The telecommunications network includes multiple cell transceivers 140, 142, and 144, and associated cells 150, 152, and 154. In this particular non-limiting example, the mobile device is an unmanned aerial vehicle (UAV). The mobile device is configured to receive signals 160, 162, and 164 from cells 150, 152, and 154. For each received signal, the mobile device measures the properties of the received signal (e.g., by measuring the received signal strength and / or quality) and determines a cell identifier indicating which cell transmitted the signal. The mobile device generates a measurement record (MR) from the signal measurement and cell identifier information, and this measurement record (MR) is transmitted as signal 168 to one or more cells. Alternatively, the MR may be transmitted to another entity in the network.

[0099] In this figure, each cell transceiver is shown as a separate tower. However, in some cases, a single tower may include cell transceivers serving different cells. A cell transceiver may be a base station. A base station may serve more than one cell. Each of the reference signals is associated with a cell, therefore this application generally refers to signals from a cell. However, the invention may be implemented using a base station identifier instead of a cell identifier.

[0100] refer to Figure 2 Based on signal measurements and cell identifiers in measurement records associated with cells 150, 152, and 154 in the geographic proximity area of ​​the mobile device, the measurement records can be used to estimate the geographic location of the mobile device 100.

[0101] In a telecommunications network, multiple cells can exist that are assigned the same cell identifier. In other words, cell identifiers can be reused at different geographical locations around the network. However, reuse schemes should be planned so that neighboring cells do not share cell identifiers. Furthermore, reuse schemes should be planned so that no cell has more than one neighbor with the same cell identifier. The combination of signals received by a mobile device at any geographical location in the network, along with the corresponding cell identifier and signal measurements of those signals, should be sufficient to provide a good estimate of the mobile device's geographical location. This is because the reuse scheme planning, along with the combination of signals received from all neighboring cells in the geographical area, should allow the model to determine which cell each signal pertains to, even if the cell identifier may not be unique.

[0102] The Physical Cell ID (PCI) can be used as a cell identifier. There are 504 available PCIs for use in LTE telecommunications networks. To avoid PCI conflicts, adjacent cells must not share a PCI. To avoid PCI confusion, no cell in the network can have two neighbors sharing the same PCI.

[0103] To address the data issues encountered in existing RPS (Relational Persistent Cell) technologies, an innovative feature engineering method is provided. This method can be used to combine and reformat neighbor cell data representations into a new format. In this new format, arbitrary and nominal identifiers used as features of neighbor cells that cause confusion in ML models and misunderstandings of neighbor cell information are removed from both training and test data. Instead, under this new format, different neighbor cell PCIs (for both training and test data) are identified and set as new columns / features / elements in the data.

[0104] For each row / record / MR, if a neighboring cell identified by its PCI is reported and presented, its row value (signal properties in a sorted list of signal properties) is filled with its corresponding radio information (i.e., RSRP value). Otherwise, its row value is filled with zero. Alternatively, different values ​​can be used to indicate that a cell identified by its PCI has not been reported (e.g., 999 or -999). In this way, a clear and unambiguous association is established between each neighboring cell and its radio signal profile for all MRs / records collected in the geographic area of ​​interest. This method can greatly improve the quality of the input data and also eliminate ambiguity in neighbor information.

[0105] The newly regenerated subframes (lists ordered by signal properties) containing information about neighboring cells are merged back into the original data frame (which has had the "old" neighboring cell representations removed) to form a new, "modified" data frame. In other words, the tuples in the measurement records providing neighboring cell information are replaced with a list ordered by signal properties. This can be performed on both training and testing data.

[0106] The reconstructed data frames are used to train an existing ML model. This can be done without modifying other data in the measurement records or changing the ML model. In doing so, an enhancement in model prediction accuracy can be observed. This accuracy enhancement may surpass the performance of the previous model.

[0107] In summary, the entire dataset was transformed and reconstructed to provide the ML model with complete and more accurate neighbor information, which in turn enables better detection and learning of hidden patterns and radio FPs.

[0108] Table 1 shows an example of ten measurement records transmitted from a mobile device to the network. Table 2 shows how the measurement records in Table 1 can be restructured before being provided to the model.

[0109]

[0110] Table 1

[0111]

[0112] Table 2

[0113] Table 1 shows an example of ten measurement records (MR1 to MR10) transmitted from a mobile device to the network. In this example, the measurement records (MRs) belong to the training dataset. Therefore, each MR includes latitude and longitude information transmitted from the mobile device. Similarly, validation data can also contain location measurement data. If this is test data instead of training data, the same principle applies, but latitude and longitude measurements will not be present. As can be seen in Table 1, each MR also includes a list of neighbor cell PCI and RSRP values, numbered from 1 to 32. In this example, each MR does not have data for neighbor cells 7 to 32. Therefore, for brevity, the rows corresponding to neighbor cells 8 to 31 are omitted. In this example, each MR also includes some additional information such as the drone's altitude, the serving cell's PCI, the identifier of the serving eNodeB (“Servingencode”), the serving cell's RSRP and RSRQ values, and so on.

[0114] Timing advance (TA) can be added to the measurement record by the eNB. Timing advance is based on T. s T is expressed in units. s Timing Advance (TA) is the basic time unit defined in the 3GPP standard. At the UE, TA is the negative offset between the start of a transmitted uplink subframe and the start of a received downlink subframe. This allows the UE to synchronize uplink and downlink transmissions. The TA value can be continuously measured by the eNB and can be dynamically adapted and signaled back to the eNB. TA is a measurement of time and can therefore be expressed in microseconds. However, it is more commonly expressed in the basic time unit (T). s TA is represented by multiples of TA, as defined in 3GPP standard 36.211. For example, T s = 1 / (subcarrier spacing x block-by-block FFT size); T s = 1 / (15000x2048) seconds = 0.0325 microseconds. This can be used in LTE networks where the subcarrier spacing is 15kHz and the FFT size per block is 2048. These numbers may be different for 5G networks, so the basic time unit will be different (e.g., depending on the spectrum frequency, the subcarrier spacing could be 30kHz or 60kHz).

[0115] Table 2 illustrates how the measurement records (MR1 to MR10) in Table 1 can be restructured before being provided to the model. This particular example shows that each restructured measurement record (MR1′ to MR10′) is represented by columns in the table, and the table includes rows for each PCI in the dataset. In this example, rows corresponding to PCIs for which no RSRP value is available for any MR have been omitted. However, each available PCI can have rows in the table. The table can also be configured such that each MR is represented by rows and the table includes columns for each PCI in the dataset. Importantly, the ordering of the signal properties for each measurement record is the same as the ordering of every other measurement record. Similarly, the ordering of the signal properties for the training and testing data follows the same ordering system.

[0116] It should be noted that the signal strength values ​​associated with PCI 323 are all filled with "0". This is important because, in this example, 323 is the PCI for the serving cell used for all measurement records. In the method used to modify measurement records in this particular example, only the neighbor signal strength measurements from the original MR are used to populate the sorted list that forms the portion of the updated measurement records (MR1′ to MR10′). In an alternative approach, it might be beneficial to combine the signal strength measurements of the serving cell as well as those of the neighboring cells into the sorted list.

[0117] Figure 3 The system and data flow diagram is shown. Reference signals transmitted from one or more cells 150, 152, and 154 are received by mobile device 100. Mobile device 100 determines the signal measurement for each signal and the cell identifier associated with the cell from which the signal is received. Mobile device 100 generates a measurement record including the signal properties and the corresponding cell identifier. Mobile device 100 sends the measurement record to data processor 320. The measurement record can be sent to data processor 320 by sending the measurement record to one or different cells of cells 150, 152, and 154 from which the reference signal is received. The measurement record can also be sent to data processor 320 by sending the measurement record to the base station associated with the cell. The cell / base station provides the measurement record to data processor 320 in the network, or includes data processor 320. Data processor 320 reads the signal measurements and cell identifier from the measurement record and generates a sorted list of signal properties. Data processor 320 provides the sorted list of signal properties, along with any other necessary information from the measurement record, to model 310.

[0118] Measurement records may also include the geographic location of mobile device 100. In this case, the model can use a sorted list of location data and signal properties (along with any other necessary information from the measurement records) as training data to improve the accuracy of future estimates.

[0119] The model can use a sorted list of signal properties to estimate the geographic location of mobile device 100. This geographic location estimate can be used elsewhere in the network. For example, the geographic location estimate can be used to predict the arrival time of a drone at a specified location.

[0120] MRs can be collected and transmitted at intervals of approximately 2–10 seconds. However, this method can also be performed using MRs transmitted at any time interval or as a single, one-time record.

[0121] Although model 310 and data processor 320 are described as being on the network side, these technologies can also be used on the mobile device 100 itself. In this case, the step of transmitting measurement records to the cell is eliminated. The mobile device 100 can directly format / structure the signal measurements into a sorted list and use it with the model to estimate its own position. This can be advantageous when the mobile device 100 is not equipped with other devices for determining position (e.g., GNSS). Furthermore, this can save battery power compared to GNSS technology.

[0122] Figure 4 A flowchart illustrating a method for estimating the geolocation of a mobile device is shown. The method includes the following steps:

[0123] • Figure S401: Obtain one or more data records from a mobile device, wherein each data record includes multiple signal measurements from the mobile device and a corresponding cell identifier;

[0124] • S403: For each data record, generate a sorted list including the signal properties of multiple signal measurements; and

[0125] • S405: For each data record, estimate the geographic location of the mobile device based on a sorted list of signal properties.

[0126] Figure 5 A flowchart illustrates a method for providing training data to a model used to estimate the geolocation of a mobile device based on multiple signal measurements. The method includes the following steps:

[0127] S501: Obtain one or more data records from the mobile device, wherein each data record includes: multiple signal measurements and corresponding cell identifiers from the mobile device and corresponding measurements of the geographical location of the mobile device;

[0128] • S503: For each data record, generate a sorted list including the signal properties of multiple signal measurements; and

[0129] • S505: For each data record, provide the model with a sorted list of signal properties and the corresponding measurement of the mobile device's geographic location.

[0130] Figure 6 A flowchart illustrates a method for associating each signal property in a sorted list with its corresponding cell identifier. The method includes the following steps:

[0131] S601: Read one or more data records, wherein each data record includes multiple signal data entries;

[0132] S603: For each data record and for each signal data entry within that data record, read the cell identifier of the data record; and

[0133] • S605: For each unique cell identifier, associate the signal properties in the sorted list with the corresponding unique cell identifier based on the index of the signal properties in the sorted list.

[0134] Figure 7 A flowchart illustrating a method for generating a sorted list of signal properties is shown. The method includes the following steps:

[0135] S701: Reads data records containing multiple signal data entries; and

[0136] S703: For each signal data entry in the data record, fill the sorted list of signal properties with the signal measurement of the signal data entry.

[0137] Figure 8 This illustrates a method for restructuring data, executed in a specific implementation. The method includes the following steps:

[0138] • S801: Collect and obtain all unique neighbor cell PCIs (Physical Cell IDs) in both the original training set and the test set;

[0139] S803: Generates a list / set of tuples (PCI, RSRP, etc.) for all neighboring cells reported for each line / record;

[0140] • S805: Iteratively, generate such a list / set of neighbor tuples for each record, and store them all in a nested list / set;

[0141] • S807: Associate each unique PCI in a data record with an index in a sorted list of signal properties. This can be achieved using an indexing scheme in the form of table headers, where each unique neighboring PCI is set as a new column / row / feature / element in the table;

[0142] S809: Parses and transforms a nested list / set of neighboring tuples into a sorted list, where each row's value is filled with the RSRP value of its associated tuple. In other words, it transforms tuples into a sorted list with signal-like properties.

[0143] S811: Constructs a reconstructed neighbor cell data frame by parsing and transforming all entries in a nested list / set;

[0144] • S813: Remove the original neighbor cell representation from the original data frame and merge the newly reconstructed neighbor cell data frame with the remaining neighbor cell data frames to form an updated training / test dataset; and

[0145] • S815: Use updated datasets to train and test ML models.

[0146] After data transformation and reformatting, arbitrary and nominal numbers that could confuse and mislead the ML model from the data, which are features of neighboring cells, are removed from both data frames (training and testing).

[0147] The old neighbor cell information representation failed to associate each neighbor cell identified by PCI with its radio signal profile. Thanks to this novel approach, clear and explicit associations are constructed, and these associations can be used by ML models to learn by filtering, combining, and reconstructing neighbor cell information.

[0148] A novel approach is provided to accurately address data challenges encountered in radio network measurement reporting and data collection. Instead of normalizing the data or performing subset feature selection, and without altering the data structure, this novel approach is a "break-and-build" approach that fundamentally alters the neighbor cell representation structure and regenerates it in a style that unlocks more and more accurate neighbor information, allowing ML models to learn patterns and better present the data to the problem at hand. No existing techniques achieve the same or similar results.

[0149] Existing techniques, such as OOB ML models, combinatorial methods, optimization, and others, have not achieved sufficient accuracy in geolocation. This new method is significant in part due to the fact that it disrupts and reconstructs the ML model with an improved representation of neighbor data to better learn patterns and thus better solve the problem.

[0150] This method replaces the "nominal number" with the actual neighbor cell PCI as a new feature / column and the associated Radio Signal Profile (RSRP), which can provide the ML model with more relevant and useful information to learn patterns. In other words, the proposed method can provide the ML model with more accurate and complete neighbor information and avoids misleading the ML model with those "nominal numbers" represented by neighbor cell information.

[0151] Original Neighborhood Community Format:

[0152]

[0153] Table 3

[0154] Before the neighbor data is reformatted and transformed, the ML model may have already “thought” that the two samples contain different neighbor radio fingerprints, which may lead to mis-prediction.

[0155] To a human observer, these two MR records may appear identical. This is likely because human observers will perform subconscious "data rearrangement and relation reduction." However, for the ML model, the nominal numbers (_1, _2) allow the model to identify neighbor_1 as having been assigned distinct values ​​across samples from neighbor_2. Therefore, for this ML model, the feature / column names, i.e., pci_neigh_1 / pci_neigh_2, are important here. Most ML models lack the human-level intelligence to "achieve" a truly meaningful representation of neighbor cells, which should be the actual neighbor cell PCIs: A, B, and their associated radio profiles (RSRPs). What this novel approach does is make this information explicitly "help" the ML model pick it up and learn the correct representation of the neighbor radio fingerprint.

[0156] New Neighborhood Community Format:

[0157] A B C D E Sample 1 va vb 0 0 0 Sample 2 va vb 0 0 0

[0158] Table 4

[0159] After reformatting, the data provides a clearer relationship between PCI and RSRP. The ML model can therefore learn that these records contain the same neighbor radio fingerprints, thus making correct predictions.

[0160] The above is a hypothetical example illustrating the situation. In real-world examples, it can become far more complex, and some patterns are implicit and hidden, making them difficult for human users to understand. This highlights situations where machine learning models can be beneficial. A data-driven approach allows us to learn, understand, and solve challenging problems if AI / machine learning techniques can assist in these processes.

[0161] In this application, the term "mobile device" is used to refer to a device configured to communicate with a telecommunications network. However, the device may also be referred to as "user equipment," "subscriber equipment," "mobile handheld device," "cellular connection equipment," "telecommunications equipment," and the like.

[0162] Although the term "cell" is generally used in this application to refer to the source of a telecommunications signal, those skilled in the art will understand that the signal may originate from multiple different elements within a telecommunications network. For example, the signal may originate from a NodeB, eNodeB, microcell, picocell, femtocell, and the like.

[0163] Specific embodiments of the invention are described in this application, wherein the signal measurement is a reference signal received power. However, in some cases, these methods may additionally or alternatively use other measurements of signal strength. For example, these methods may use the Received Signal Strength Indicator (RSSI) or the Received Channel Power Indicator (RCPI). Furthermore, different measurements of signal strength may be used in future iterations of telecommunications standards. Such measurements of signal strength can be used in the methods described in this application.

[0164] Similarly, the specific embodiments of the present invention described above use the Physical Cell Identifier (PCI) as the cell identifier. These methods may additionally or alternatively use other cell identifiers, such as Cell Identifier (CI), Cell ID (CID), Enhanced Cell ID (E-CID), and the like. These cell identifiers may be used in combination with other identifiers, such as Mobile Country Code (MCC), Mobile Network Code (MNC), Location Area Code (LAC), Location Area Identity (LAI), and the like. Furthermore, global base station identifiers such as Base Station Identifier Code (BSIC), Cell Global Identity (CGI), and the like may be used in the methods described herein, in place of (or in conjunction with) the cell identifier. Moreover, different conventions for identifying cells may be used in future iterations of telecommunications standards. Such cell identifiers may be used in the methods described herein.

[0165] While the specific examples above describe LTE (4G) and NR (5G) applications, these technologies can be used for 2G / 3G / 4G / 5G and beyond. Furthermore, Wi-Fi signals can also be used in a similar way to support location estimation technologies.

[0166] Although specific embodiments of the invention have been described, those skilled in the art will understand that various modifications and variations can be made without departing from the scope of the invention.

Claims

1. A method for estimating the geographic location of a mobile device, the mobile device being configured to communicate using a telecommunications network comprising multiple cells, the method comprising: One or more data records are obtained from a mobile device, wherein each data record includes multiple signal measurements and corresponding cell identifiers, wherein each signal measurement relates to a signal received by the mobile device from a cell among multiple cells attributed to the corresponding cell identifier; For each data record, a sorted list of signal properties, comprising multiple signal measurements, is generated. Based on the index of the signal properties in the sorted list, each signal property in the sorted list is associated with a corresponding cell identifier assigned to the cell from which it received the signal; and For each data record, the geographic location of the mobile device is estimated based on a sorted list of signal properties; Signal properties include signal strength and quality; and The list of signal properties is based on the index of the signal properties. Each signal property in the list is associated with a corresponding cell identifier assigned to the cell from which it receives the signal. Each of the multiple signal measurements appears in the list at an index corresponding to the corresponding cell identifier of the signal measurement.

2. The method according to claim 1, wherein, One or more data records include a first data record corresponding to a first time and a second data record corresponding to a second time different from the first time, wherein each data record further includes an identifier linked to a mobile device, wherein the method further includes: Identifiers from the first and second data records are used to determine that the first and second data records were obtained from the same mobile device.

3. The method according to claim 2, wherein, For the second data record, estimating the geographic location of the mobile device involves using a sorted list based on signal properties and the estimated geographic location of the mobile device from the first data record.

4. A method for providing training data to a model for estimating the geographic location of a mobile device based on multiple signal measurements, the method comprising: Obtain one or more data records from a mobile device, wherein each data record includes: Multiple signal measurements and corresponding cell identifiers, and Corresponding measurements of the geographical location of mobile devices. The mobile device is configured to communicate using a telecommunications network comprising multiple cells, wherein each signal measurement involves a signal received by the mobile device from a cell among the multiple cells attributed to a corresponding cell identifier; For each data record, a sorted list of signal properties, including multiple signal measurements, is generated. Based on the indices of the signal properties in the sorted list, each index of a signal property in the sorted list is associated with a corresponding cell identifier assigned to the cell from which it received the signal; and For each data record, the model is provided with a sorted list of signal properties and corresponding measurements of the mobile device's geographic location; Signal properties include signal strength and quality; and The list of signal properties is sorted based on the index of the signal properties. Each index of the sorted signal properties is associated with the corresponding cell identifier assigned to the cell from which the signal was received. Each of the multiple signal measurements appears in the sorted list at a position with an index corresponding to the corresponding cell identifier of the signal measurement.

5. The method according to claim 4, wherein, The mobile device is the first mobile device, where the sorted list is the first sorted list. The model is configured to receive a sorted list of signal properties obtained from a second mobile device configured to communicate using a telecommunications network as input. Based on the indices of the signal properties in the sorted list, each index of a signal property from the second device is associated with a corresponding cell identifier assigned to the cell from which it received the signal. The model is configured to provide an estimate of the geographic location of the second mobile device as output.

6. The method according to claim 4 or 5, wherein, The generated list of sorted signal properties includes assigning a value to each signal property in the sorted list, for each signal property in the sorted list having an index corresponding to a cell identifier for which no corresponding signal measurement has been obtained, the value indicating that no signal has been received by the mobile device.

7. The method according to claim 4 or 5, wherein, Mobile devices are: User equipment; Aircraft; Unmanned aerial vehicles (UAVs); User equipment on manned or unmanned aircraft; or Aerial equipment.

8. The method according to claim 7, wherein, Each data record also includes a height measurement of the mobile device, wherein the model is configured for use with signal measurements involving a predetermined range of height, and wherein the height measurement of the mobile device is within the predetermined range of the height measurement.

9. The method according to claim 4 or 5, wherein, Multiple signal measurements include: One or more signal strength measurements; One or more RSSI measurements; One or more RSRP measurements; One or more signal quality measurements One or more RSRQ measurements; and / or One or more location measurements.

10. The method according to claim 4 or 5, wherein, The method further includes: Receive multiple signals from a subset of multiple cells at a mobile device, each signal including coded data, wherein the coded data includes a cell identifier attributed to the cell from which the signal was received; For each of the multiple received signals, determine one or more signal measurements; For each of the multiple received signals, the signal is decoded and the cell identifier encoded within the signal is determined; and Data records, including signal measurements and corresponding cell identifiers, are transmitted to cells within multiple cells.

11. A method for indexing signal properties in a sorted list of signal properties, associating each signal property in the sorted list with a corresponding cell identifier, wherein, A sorted list of signal properties is suitable as input into a model for estimating the geographic location of a mobile device, the method comprising: Read one or more data records, wherein each data record includes multiple signal data entries, wherein each signal data entry corresponds to a signal received by the mobile device, and wherein each signal data entry includes: Signal measurement, and The corresponding cell identifier assigned to the cell from which it received the signal; For each data record in one or more data records, and for each signal data entry in multiple signal data entries within the data records, read the cell identifier of the signal data entry; and For each unique cell identifier read from multiple signal data entries and one or more data records, the signal properties in the sorted list of signal properties are associated with the corresponding unique cell identifier based on the index of the signal properties in the sorted list. Signal properties include signal strength and quality; and The index of the signal properties in the sorted list associates each signal property with its corresponding cell identifier. This includes each signal measurement in the sorted list appearing at an index corresponding to the corresponding cell identifier of the signal measurement.

12. A method for generating a sorted list of signal properties, the method comprising: Read a data record comprising multiple signal data entries, wherein each signal data entry corresponds to a signal received by the mobile device, and wherein each signal data entry includes: Signal reception and signal measurement, and The corresponding cell identifier assigned to the cell from which it received the signal; For each signal data entry in a data record, the signal properties in a sorted list of signal properties are populated with the signal measurements of the signal data entry. The updated signal properties are associated with the corresponding cell identifier of the signal data entry, based on the index of the signal properties in the sorted list of signal properties. Signal properties include signal strength and quality; and The list of signal properties is sorted based on the index of the signal properties. The updated signal properties are associated with the corresponding cell identifier of the signal data entry. Each signal measurement in the sorted list appears at a position with an index corresponding to the corresponding cell identifier of the signal measurement.

13. The method of claim 12, further comprising initializing each signal property in the sorted list with a value indicating that no signal has been received.

14. The method according to any of the preceding claims, wherein, The identifier for each cell is: Physical Cell ID, PCI; and / or Community Global Identity, CGI.

15. A system configured to perform the method of any one of claims 1-14.

16. A mobile device, comprising: processor; as well as Computer memory, The computer memory includes instructions that, when executed on the processor, cause the mobile device to perform the method according to any one of claims 1 to 14. The mobile device also includes mobile devices from which one or more data records are obtained.

17. Computer software comprising computer-readable instructions that, when executed by a processor of a computer system, cause the computer system to perform the method according to any one of claims 1 to 14.

18. A computer-readable medium comprising instructions that, when executed by a processor of a computer system, cause the computer system to perform the method according to any one of claims 1 to 14.

Images