WO2019240771A1

WO2019240771A1 - User device positioning

Info

Publication number: WO2019240771A1
Application number: PCT/US2018/037064
Authority: WO
Inventors: Rajeev Agrawal; Ian Dexter GARCIA; Hua Xu; Igor Filipovich
Original assignee: Nokia Technologies Oy; Nokia USA Inc
Current assignee: Nokia Technologies Oy; Nokia USA Inc
Priority date: 2018-06-12
Filing date: 2018-06-12
Publication date: 2019-12-19
Anticipated expiration: 2020-12-12

Abstract

A device, apparatus and computer program are described comprising: receiving a radio frequency fingerprint for a user device at an input of a neural network, wherein the neural network has an output providing a location estimate for the user device; and using the neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the neural network and/or one or more previous location estimates for the user device generated by the neural network.

Description

USER DEVICE POSITIONING

Field

The present specification relates to user device positioning, in particular to user device positioning using radio frequency (RF) fingerprinting.

Background

Users of mobile communication devices and other user devices are used to using global positioning system (GPS) and similar systems for positioning purposes, for example for navigation. However, such systems are not always effective (e.g. when used indoors). Other uses of positioning data for user devices include location-based advertising and the positioning of devices for use by emergency services.

Radio Frequency (RF) signal characteristics vary in space. Radio frequency fingerprinting refers to radio frequency signal characteristics at different locations. If enough information is known about radio frequency data, then knowledge of a radio frequency fingerprint at a location of a user device can provide information regarding the location of that device.

There remains a need for improvement related to user device positioning.

Summary

In a first aspect, this specification describes an apparatus comprising: means for receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and means for using the recurrent neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network. The radio frequency fingerprint may comprise radio frequency signals that are detectable by said user device. The radio frequency signals detectable by said user device may include radio frequency signals that are not used by the user device (but that are nonetheless detectable).

Some embodiments may comprise means for determining the radio frequency fingerprint for the user device.

In some embodiments, there may be provided means for providing a time- sequence of determined radio frequency fingerprints to the recurrent neural network. Alternatively, or in addition, there may be provided means for providing a time-sequence of location estimates for the user device to the recurrent neural network.

The said radio frequency fingerprint comprises one or more of: Reference Signal Received Power (RSRP) of a serving cell and interfering cells; Reference Signal Received Quality (RSRQ) of a serving cell and interfering cells; Reference Signal Signal-to-noise ratio (RS-SJNR) of a serving cell and interfering cells; Received Signal Strength Indicator (RSSI); signal delays; time of arrival of radio frequency signals; angle of arrival of radio frequency signals; pathloss of serving cell and interfering cells (obtained, for example, through knowledge of received power and power headroom or transmit power); physical cell ID of a serving cell and interfering cells; strongest beam ID; beam signal-to-noise ratio(s) (SINRs); bean pathloss(es); massive MIMO channel snapshot (e.g. magnitude vs. angle, delay and antenna ID) of serving and interfering cells; and massive MIMO channel snapshot (e.g. magnitude vs. angle, delay and antenna ID) of beam(s).

In some embodiments, a means for training the recurrent neural network may be provided. The recurrent neural network may be trained using real-world measurement data and/or modelling data obtained from a cell planning tool. For example, the recurrent neural network may be trained using said modelling data and refined using real-world measurement data. In some embodiments, the recurrent neural network may comprise a mapping between the radio frequency fingerprint measurements and their corresponding location coordinates, wherein the mapping may be updated periodically.

The said means may comprise: at least one processor; and at least one memory including computer program code, the at least one memory and the computer program code configured, with the at least one processor, to cause the performance of the apparatus.

In a second aspect, this specification describes a method comprising: receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and using the recurrent neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network.

The method may comprise determining the radio frequency fingerprint for the user device.

The method may comprise providing a time-sequence of determined radio frequency fingerprints to the recurrent neural network. Alternatively, or in addition, the method may comprise providing a time-sequence of location estimates for the user device to the recurrent neural network.

The method may comprise training the recurrent neural network. The recurrent neural network may be trained using real-world measurement data. The recurrent neural network may be trained using modelling data obtained from a cell planning tool. The recurrent neural network may be trained using modelling data and then refined using real-world measurement data.

In a third aspect, this specification describes an apparatus configured to perform any method as described with reference to the second aspect. In a fourth aspect, this specification describes computer readable instructions which, when executed by computing apparatus, cause the apparatus to perform a method as described with reference to the second aspect.

In a fifth aspect, this specification describes a computer program comprising instructions stored thereon for performing at least the following: receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and using the recurrent neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network.

In a sixth aspect, this specification describes a computer readable medium (such as a non-transitory computer readable medium) comprising program instructions stored thereon for performing at least the following: receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and using the recurrent neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network.

In a seventh aspect, this specification describes an apparatus comprising: at least one processor; and at least one memory including computer program code which, when executed by the at least one processor, causes the apparatus to: receive a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and use the recurrent neural network to generate said location estimate for the user device based on: the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network.

Brief description of the drawings

Example embodiments will now be described, by way of non-limiting examples, with reference to the following schematic drawings, in which:

FIG. 1 is a block diagram of a radio frequency environment in which a user device is provided;

FIG. 2 is a flow chart showing an example algorithm;

FIG. 3 is a block diagram of a system in accordance with an example embodiment;

FIG. 4 is a block diagram of a system in accordance with an example embodiment;

FIG. 5 is a flow chart showing an algorithm in accordance with an example embodiment;

FIG. 6 shows a neural network used in an example embodiment;

FIG. 7 shows a neural network used in an example embodiment;

FIG. 8 is a block diagram of components of a system in accordance with an exemplary embodiment; and

FIGS. 9A and 9B show tangible media, respectively a removable non-volatile memory unit and a compact disc (CD) storing computer-readable code which when run by a computer perform operations according to embodiments.

Detailed description

Accurate user positioning using cellular systems is challenging. The radio environment typically assumes non-line-of-sight (NLOS) signals from/to the cells, and significant multipath conditions. Position accuracy can be relatively low, even with the use of GPS, GNSS and similar systems. Moreover, in indoor environments, GPS/GNSS signals may not be available.

FIG. 1 is a block diagram of a radio frequency environment, indicated generally by the reference numeral 10, in which a user device 12 is provided. As shown in FIG.l , the user device 12 is exposed to multiple different radio frequency sources (indicated schematically in FIG. 1 by the arrows 14 to 17).

In general, every location has a near-unique radio frequency (RF) fingerprint as a result of the radio frequency sources that impinge on that location. For example, the radio frequency sources 14 to 17 in the example environment 10 can be measured to provide a radio frequency fingerprint at the location of the user device 12.

A radio frequency fingerprint can take many forms. For example, a radio frequency fingerprint may include a set of Reference Signal Received Power (RSRP) for all cells that are accessible by a user device. Alternatively, or in addition, Received Signal Strength Indicator (RSSI) and/or pathless measurements may be incorporated within a fingerprint A radio frequency fingerprint can consist of measurements from a single carrier frequency layer or from multiple different carrier frequency layers. Radio frequency fingerprints can also include other non-cellular measurements, such as WiFi. Other radio frequency fingerprint data points may include one or more 5G signals, such as the strength of individual beams in a mMIMO environment. A radio frequency fingerprint may comprise radio frequency signals that are detectable by said user device.

The following is a non-exhaustive list of radio frequency quantities that may be incorporated within a radio frequency fingerprint:

• Reference Signal Received Power (RSRP) of a serving cell and interfering ceils; • Reference Signal Received Quality (RSRQ) of a serving cell and interfering cells;

• Reference Signal Signal-to-noise ratio (RS-SINR) of a serving cell and interfering cells;

• Received Signal Strength Indicator (RSSI);

• Signal delays;

• Time of arrival of radio frequency signals;

• Angle of arrival (AoA) of radio frequency signals;

• Pathless of serving cell and interfering cells (obtained, for example, through knowledge of received power and power headroom or transmit power);

• Physical cell ID of a serving cell and interfering cells;

• Strongest beam ID;

• Beam signal-to-noise ratio(s) (SINRs);

• Beam pathloss(es);

• Massive MIMO channel snapshot (e.g. magnitude vs. angle, delay and antenna ID) of serving and interfering cells; and

• Massive MIMO channel snapshot (e.g. magnitude vs. angle, delay and antenna ID) of beam(s). The quantities may be from downlink or uplink signal measurements. The quantities for uplink or downlink may be jointly used in a radio frequency fingerprint.

FIG. 2 is a flow chart showing an example algorithm, indicated generally by the reference numeral 20.

The algorithm 20 starts at operation 22 where radio frequency fingerprint and location data are collected. That data may be collected, for example, by a user device such as the user device 12. At operation 24, data collected in the step 22 can be added to a database such that a radio frequency fingerprint database can be generated. Since, as noted above, every location has a near-unique radio frequency fingerprint, a detected radio frequency fingerprint can be used to provide a location estimate. However, radio frequency fingerprints are not static. Such fingerprints change over time, due, for example, to changes in radio frequency signals within a given location, changes in user devices within a given location and changes in the local environment. Thus, a database of RF fingerprint data is not sufficient to provide accurate location estimates from RF fingerprint measurements.

FIG. 3 is a block diagram of a system, indicated generally by the reference numeral 30, in accordance with an example embodiment. The system 30 comprises a neural network 32. The neural network 32 is configured to provide a location estimate (for example an estimate of the location of the user device 12 described above).

As shown in FIG. 3, the location estimate is based on RF fingerprint data (such as RF fingerprint data obtained by the user device 12), past RF fingerprint data (for example recent RF fingerprint data obtained by the user device 12) and past location estimated) (for example, recent outputs of the neural network 32). Thus, as described in detail below, the neural network 32 may provide a location estimate based not only on the current RF fingerprint measurement, but also on a time-sequence of recent RF fingerprint measurements and location estimates.

FIG. 4 is a block diagram of a system, indicated generally by the reference numeral 40, in accordance with an example embodiment The system 40 is an example implementation of a recurrent neural network. The system 40 comprises a neural network 42 (e.g. a feedforward neural network), a first time delay line 44 and a second time delay line 46.

As shown in FIG. 4, RF fingerprint data (for example as obtained by the user device 12) is provided at the input of the first time delay line 44. The first time delay line 44 provides one or more outputs to the neural network 42. Thus, the neural network 42 can be provided with RF fingerprint data at different points in time (e.g. at time t, t-1, t-2 etc.).

Similarly, location estimates as output by the neural network 42 are provided at the input of the second time delay line 46. The second time delay line 46 provides one or more outputs to the neural network 42. Thus, the neural network 42 can be provided with location estimate data at different points in time (e.g. at time t-1, t-2 etc.).

FIG. 5 is a flow chart showing an algorithm, indicated generally by the reference numeral 50, in accordance with an example embodiment. The algorithm 50 may be implemented by the systems 30 or 40 described above.

The algorithm 50 starts at operation 52, where RF fingerprint data is provided. For example, RF fingerprint data obtained by the user device 12 may be provided to the neural network 32 or the first time delay line 44.

At operation 54, a time sequence of RF fingerprint data is provided to the relevant neural network. For example, the first time delay line 44 may provide RF fingerprint data measurements from different time points to the neural network 42.

At operation 56, a time sequence of location estimate data is provided to the relevant neural network. For example, the second time delay line 46 may provide location estimate data from different time points to the neural network 42.

At operation 58, a location estimate is generated (for example for the location of the user device 12). The location estimate is based on the radio frequency fingerprint data received in operation 52, the time sequence of radio frequency fingerprint data received in operation 54 and the time sequence of location estimate data received in operation 58. The location estimate may, for example, be a two-dimensional (x,y) or three-dimensional (x,y,z) co-ordinate.

A number of variants of the algorithm 50 are possible. For example, some of the operations 52 to 58 may be carried out in a different order. Furthermore, it is not essential to all embodiments to provide both the operation 54 and the operation 56. Thus, the location estimate may be based on the time sequence of radio frequency fingerprint data received in operation 54 or the time sequence of location estimate data received in operation 58 (or both).

In order for a neural network (such as the neural networks 32 and 42 described above) to be used to provide location estimates, training of the neural networks is required. The training phase provides a mapping between RF fingerprint data and the location co-ordinates for that data. In a mobile communication system, training may, for example, be performed at a base station, at a remote unit, or at a mobile terminal (or at a combination thereof). As radio frequency fingerprint data is continuously changing, training may be performed periodically or continuously to ensure that a neural network being used to provide position estimates is trained on up-to-date data. Indeed, a neural network may be incrementally re-trained each time a new set of (potentially more accurate and/or more up-to-date) RF fingerprint and location data becomes available.

Radio frequency fingerprint data can be obtained in a number of different ways:

• Real-world measurement data can be obtained in which radio frequency signals are measured at multiple locations. Although conceptually simple, such an approach is labour intensive.

· Cell planning tools may be used to provide radio frequency data on the basis of modelling. In principle, such tools enable almost unlimited radio frequency fingerprint maps to be generated for use in training neural networks.

• A hybrid approach that makes use of both cell planning tools and real-world measurement data may be used. For example, initial neural network training may be carried out on the basis of data obtained from cell planning tools. The neural network may then be refined on the basis of real-world data.

FIG. 6 shows a neural network, indicated generally by the reference numeral 60, used in an example embodiment. The neural network 60 is a feedforward neural network including an input layer 62, a plurality of hidden layers 64 and an output layer 66.

The neural network 60 was used in the modelling of a single floor of an indoor scenario (an office building in this case). RF fingerprint maps were obtained using a (commercially available) Winprop cell planning tool assuming RSRP measurements over every point in a square grid which covered the building floor (100x100 metres), where the sides of the squares were 1 metre. The RSRP measurements were estimated from 11 small cells placed throughout the floor. The neural network 60 was trained using the obtained RF fingerprint maps.

As shown in FIG. 6, the neural network 60 Ms a sequential feed forward network architecture including N inputs labelled Ui(t) where i ranges from 1 to N. The inputs Uj(t) are modelled RSRP signals for cell i at time t. The outputs x(t) and y(t) are the estimated two-dimensional (x,y) co-ordinates.

The trained neural network was used to estimate positions. The following data was obtained using a neural network Mving a single hidden layer having 60 hidden nodes. The activation function in the neurons was hyperbolic tangent.

The modelling was repeated on the basis of the neural network 60 Mving two hidden layers (each Mving 60 hidden nodes). The following data was obtained:

Thus, the feedforward neural network of FIG. 6 provides a position accuracy of a few metres.

FIG. 7 shows a neural network, indicated generally by the reference numeral 70, used in an example embodiment. The neural network 70 is a recurrent neural network including an input layer 72, a plurality of hidden layers 74 and an output layer 76, wherein the outputs of the output layer are fed back into the input layer.

The neural network 70 was used in the modelling of multiple floors of a shopping mall on the basis of real-world RSRP measurements. For comparison, the feedforward neural network 60 described above was also tested using the same real-world RSRP measurements for the shopping mall. The RSRP data was obtained by walking through each of 5 floors of the shopping centre, with approximately 5000 data points being collected per floor. Each floor was about 120x120 metres.

The neural network 70 includes a plurality of N inputs Ui(t), each provided at times t to t-K. The inputs Ui(t) are modelled RSRP signals for cell i at time t. The neural network 70 also includes past outputs x and y at times t-l,...t-L. Thus, in total, the neural network 70 includes (N*(K+1) +2L) inputs.

The outputs x(t) and y(t) of the neural network 70 are the estimated two- dimensional (x,y) co-ordinates. Hidden layers consist of fully connected neurons. The activation function was hyperbolic tangent. Z ¹ represents a single tap delay block.

The trained neural network was used to estimate position. The following data was obtained using a feedforward neural network 60:

Thus, as before, a position error in the range of a few metres was observed.

The following data was obtained using a recurrent neural network 70:

In this case, position error was of the order of 1 metre. Thus, the results show that the recurrent neural network 70 outperformed the feed forward neural network.

In the neural network 70, an input i of the neural network is a measured signal from a cell i. If a cell i is not measured or reported, then that input line is set to 0. In implementations of the invention, a mapping may exist between the input line i and the respective real world cell, to enable appropriate formatting of the input of the neural network. The cell descriptor of the real world cell may, for example, be the cell's unique LTE identifier, the cell's PCI or some other identifier use to uniquely identify that cell in the relevant local area. For completeness. Figure 8 is a schematic diagram of components of one or more of the example embodiments described previously, which hereafter are referred to genetically as processing systems 300. A processing system 300 may have a processor 302, a memory 304 closely coupled to the processor and comprised of a RAM 314 and ROM 312, and, optionally, user input 310 and a display 318. The processing system 300 may comprise one or more network/apparatus interfaces 308 for connection to a network/apparatus, e.g. a modem which may be wired or wireless. Interface 308 may also operate as a connection to other apparatus such as device/apparatus which is not network side apparatus. Thus direct connection between devices/apparatus without network participation is possible.

The processor 302 is connected to each of the other components in order to control operation thereof.

The memory 304 may comprise a non-volatile memory, such as a hard disk drive (HDD) or a solid state drive (SSD). The ROM 312 of the memory 314 stores, amongst other things, an operating system 315 and may store software applications 316. The RAM 314 of the memory 304 is used by the processor 302 for the temporary storage of data. The operating system 315 may contain code which, when executed by the processor implements aspects of the algorithms 20 or 50 described above. Note that in the case of small device/apparatus the memory can be most suitable for small size usage i.e. not always hard disk drive (HDD) or solid state drive (SSD) is used.

The processor 302 may take any suitable form. For instance, it may be a microcontroller, a plurality of microcontrollers, a processor, or a plurality of processors.

The processing system 300 may be a standalone computer, a server, a console, or a network thereof. The processing system 300 and needed structural parts may be all inside device/apparatus such as IoT device/apparatus i.e. embedded to very small size In some example embodiment, the processing system 300 may also be associated with external software applications. These may be applications stored on a remote server device/apparatus and may run partly or exclusively on the remote server device/apparatus. These applications may be termed cloud-hosted applications. The processing system 300 may be in communication with the remote server device/apparatus in order to utilize the software application stored there.

Figures 9A and 9B show tangible media, respectively a removable non-volatile memory unit 365 and a compact disc (CD) 368, storing computer-readable code which when run by a computer may perform methods according to example embodiments described above. The removable memory unit 365 may be a memory stick, e.g. a USB memory stick, having internal memory 366 storing the computer-readable code. The memory 366 may be accessed by a computer system via a connector 367. The CD 368 may be a CD-ROM or a DVD or similar. Other forms of tangible storage media may be used. Tangible media can be any device/apparatus capable of storing data/information which data/information can be exchanged between devices/apparatus/network.

Embodiments of the present invention may be implemented in software, hardware, application logic or a combination of software, hardware and application logic. The software, application logic and/or hardware may reside on memory, or any computer media. In an example embodiment, the application logic, software or an instruction set is maintained on any one of various conventional computer-readable media. In the context of this document, a “memory” or“computer-readable medium” may be any non-transitory media or means that can contain, store, communicate, propagate or transport the instructions for use by or in connection with an instruction execution system, apparatus, or device, such as a computer.

Reference to, where relevant,“computer-readable storage medium”,“computer program product”,“tangibly embodied computer program” etc., or a“processor” or“processing circuitry” etc. should be understood to encompass not only computers having differing architectures such as single/multi -processor architectures and sequencers/parallel architectures, but also specialised circuits such as field programmable gate arrays FPGA, application specify circuits ASIC, signal processing devices/apparatus and other devices/apparatus. References to computer program, instructions, code etc. should be understood to express software for a programmable processor firmware such as the programmable content of a hardware device/apparatus as instructions for a processor or configured or configuration settings for a fixed function device/apparatus, gate array, programmable logic device/apparatus, etc.

As used in this application, the term“circuitry” refers to all of the following: (a) hardware-only circuit implementations (such as implementations in only analogue and/or digital circuitry) and (b) to combinations of circuits and software (and/or firmware), such as (as applicable): (i) to a combination of processor(s) or (ii) to portions of processor(s)/software (including digital signal processors)), software, and memory(ies) that work together to cause an apparatus, such as a server, to perform various functions) and (c) to circuits, such as a microprocessors) or a portion of a microprocessors), that require software or firmware for operation, even if the software or firmware is not physically present If desired, the different functions discussed herein may be performed in a different order and/or concurrently with each other. Furthermore, if desired, one or more of the above-described functions may be optional or may be combined. Similarly, it will also be appreciated that the flow diagrams of Figure 2 and 5 are examples only and that various operations depicted therein may be omitted, reordered and/or combined.

It will be appreciated that the above described example embodiments are purely illustrative and are not limiting on the scope of the invention. Other variations and modifications will be apparent to persons skilled in the art upon reading the present specification. Moreover, the disclosure of the present application should be understood to include any novel features or any novel combination of features either explicitly or implicitly disclosed herein or any generalization thereof and during the prosecution of the present application or of any application derived therefrom, new claims may be formulated to cover any such features and/or combination of such features.

Although various aspects of the invention are set out in the independent claims, other aspects of the invention comprise other combinations of features from the described example embodiments and/or the dependent claims with the features of the independent claims, and not solely the combinations explicitly set out in the claims.

It is also noted herein that while the above describes various examples, these descriptions should not be viewed in a limiting sense. Rather, there are several variations and modifications which may be made without departing from the scope of the present invention as defined in the appended claims.

Claims

Claims:

1. An apparatus comprising:

means for receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and

means for using the recurrent neural network to generate said location estimate for the user device based on:

the received radio frequency fingerprint for the user device; and one or more previous radio frequency fingerprints received at the recurrent neural network and/or one or more previous location estimates for the user device generated by the recurrent neural network.

2. An apparatus as claimed in claim 1, wherein said radio frequency fingerprint comprises radio frequency signals that are detectable by said user device.

3. An apparatus as claimed in claim 1 or claim 2, further comprising means for determining the radio frequency fingerprint for the user device.

4. An apparatus as claimed in any one of the preceding claims, further comprising means for providing a time-sequence of determined radio frequency fingerprints to the recurrent neural network.

5. An apparatus as claimed in any one of the preceding claims, further comprising means for providing a time-sequence of location estimates for the user device to the recurrent neural network.

6. An apparatus as claimed in any one of the preceding claims, wherein the radio frequency fingerprint comprises one or more of: Reference Signal Received Power (RSRP) of a serving cell and interfering cells; Reference Signal Received Quality (RSRQ) of a serving cell and interfering cells; Reference Signal Signal-to-noise ratio (RS-SINR) of a serving cell and interfering cells; Received Signal Strength Indicator (RSSI); signal delays; time of arrival of radio frequency signals; angle of arrival of radio frequency signals; pathloss of serving cell and interfering cells; physical cell ID of a serving cell and interfering cells; strongest beam ID; beam signal-to-noise ratio(s) (SINRs); bean pathloss(es); massive MIMO channel snapshot of serving and interfering ceils; and massive MIMO channel snapshot of beam(s).

7. An apparatus as claimed in any one of the preceding claims, further comprising means for training the recurrent neural network.

8. An apparatus as claimed in claim 7, wherein the recurrent neural netwoik is trained using real-world measurement data.

9. An apparatus as claimed in claim 7 or claim 8, wherein the recurrent neural network is trained using modelling data obtained from a cell planning tool.

10. An apparatus as claimed in claim 9, wherein the recurrent neural network as trained using said modelling data is refined using real-world measurement data.

11. An apparatus as claimed in any one of the preceding claims, wherein the recurrent neural network comprises a mapping between the radio frequency fingerprint measurements and their corresponding location coordinates, wherein the mapping is updated periodically.

12. An apparatus as claimed in any one of the preceding claims, wherein the means comprise:

at least one processor; and

at least one memory including computer program code, the at least one memory and the computer program code configured, with the at least one processor, to cause the performance of the apparatus.

13. A method comprising:

receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and

using the recurrent neural network to generate said location estimate for the user device based on:

14. A method as claimed in claim 13, further comprising at least one of:

providing a time-sequence of determined radio frequency fingerprints to the recurrent neural network; and

providing a time-sequence of location estimates for the user device to the recurrent neural network.

15. A computer program comprising instmctions for causing an apparatus to perform at least the following:

receiving a radio frequency fingerprint for a user device at an input of a recurrent neural network, wherein the recurrent neural network has an output providing a location estimate for the user device; and using the recurrent neural network to generate said location estimate for the user device based on: