UE location method and system

Abstract

The present disclosure relates to methods for determining a UE location within a cellular radio network. The UE (110) is associated with a base station, BS (112), of the network and the network includes at least one reconfigurable intelligent surface, RIS (114), wherein the RIS (114) has a RIS configuration set to serve the UE (110), the RIS configuration including a BS component and a UE component. In an embodiment, the method comprises obtaining, by a location estimation module (120) a current RIS configuration of the RIS (114); determining, by the location estimation module (120) the UE component of the current RIS configuration by using position information about the position of the BS (112) to compute and remove the BS component from the RIS configuration; and using, by the location estimation module (120) the UE component of the RIS configuration to infer an estimation of the location of the UE (110).

Description

[0001] UE LOCATION METHOD AND SYSTEM

[0002] The project leading to this application has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101017011.

[0003] The present invention relates to methods and systems for determining a UE location within a cellular radio network.

[0004] Localization is one of the key enabling elements of 5G and beyond mobile network generations, which have set as target not only to increase localization accuracy, but also to reduce the latency of the localization process, while increasing its range and reliability. At the same time, 5G and beyond networks should deliver higher device communication speeds compared to previous generations. In this context, localization plays a crucial role for improving user experience and for enabling advanced services, e.g., for smartphones, laptops, or tablets, autonomous mobile vehicles, wearable devices, loT deployments, etc., in 5G terminology, generally referred to as UE (User Equipment). Moreover, accurate device localization is required for several network management functions, for instance, spectrum management, interference minimization or predictive handover processes, etc.

[0005] The efforts to improve localization capabilities must deal with a great amount of technical trade-offs, with cost as the global limiting element. The accuracy, already expected to go below the 1 m range requirement in the upcoming 5G revisions, is expected to be further refined down to 1 -10cm (or even less) for 6G, as it needs to cater for highly demanding applications, such as Industry 4.0, self-driving cars, etc.

[0006] The usage of RIS (Reconfigurable Intelligent Surfaces) devices in localization is currently being studied (for reference, see LIU, Yuanwei, et al. Reconfigurable intelligent surfaces: Principles and opportunities. IEEE communications surveys & tutorials, 2021 , 23. Jg., Nr. 3, S. 1546-1577). However, state-of-the-art RIS-aided localization techniques are largely based on those already available for access nodes equipped with planar antenna arrays, e.g., multiple-input multiple-output (MIMO) base stations (BS). While such techniques are viable, they present several problems when applied to RIS. Indeed, prior art solutions often rely on far-field assumptions, which prove unreasonable for RIS devices due to their much larger number of elements if compared with traditional active antennas.

[0007] On the other hand, some techniques exist that are exploiting the near-field conditions. For instance, ABU-SHABAN, Zohair, et al. Near-field localization with a reconfigurable intelligent surface acting as lens. In: ICC 2021 -IEEE International Conference on Communications. IEEE, 2021. S. 1 -6, propose the usage of the wavefront curvature to locate the UE. In particular, they consider a fixed RIS configuration (independent from the user position) and study the channel state information (CSI). However, it requires probing the channel with additional pilot signals, thus introducing additional overhead. Moreover, the accuracy of their method strongly depends on the selected RIS configuration and exhibits a lower localization accuracy when the RIS is configured in a directional manner (focused on the served user), as the focus reduces the amount of useful information for localization in the obtained CSI.

[0008] In GUIDI, Francesco; DARDARI, Davide. Radio positioning with EM processing of the spherical wavefront. IEEE Transactions on Wireless Communications, 2021 , 20. Jg., Nr. 6, S. 3571 -3586, in order to locate the UE, an additional antenna array is deployed to process the received signal after the RIS, with the noticeable disadvantage of requiring for additional hardware and signal processing unit, and significantly interfering with the communications as it requires to divert the UE signal to the measuring device and receive UE pilots.

[0009] HU, Sha; RUSEK, Fredrik; EDFORS, Ove. Beyond massive MIMO: The potential of positioning with large intelligent surfaces. IEEE Transactions on Signal Processing, 2018, 66. Jg., Nr. 7, S. 1761 -1774 theorizes a very large device (covering vast surfaces of the environment) equipped with signal processing capabilities in each element, akin more to a MIMO antenna that surrounds the covered devices more than a low-cost RIS that reflects the signals between a BS and a connected device.

[0010] WO 2023 / 036448 A1 discloses RIS-aided localization methods, where the RIS is used as an additional anchor point for triangulation. The proposed localization method relies on coordination among the base station and RIS to get the beam pattern configuration for the direct path and reflected path, based on this it extracts the angle of arrival and performs triangulation, i.e. the approach requires additional operations to locate the devices.

[0011] US 2022 / 0247480 A1 relates to RIS-aided localization and discloses a localization method that exploits the received signal at the BS, and forms the different reflected paths provided by the RIS by processing the resulting power-delay profile.

[0012] In view of the above, there may be a desire for an improved concept for UE localization in wireless communication networks.

[0013] In accordance with the invention, the aforementioned desire is addressed by a method for determining a UE location within a cellular radio network, wherein the UE is associated with a base station, BS, of the network and wherein the network includes at least one reconfigurable intelligent surface, RIS, wherein the RIS has a RIS configuration set to serve the UE, the RIS configuration including a BS component and a UE component, the method comprising: obtaining, by a location estimation module, a current RIS configuration of the RIS; determining, by the location estimation module, the UE component of the current RIS configuration by using position information about the position of the BS to compute and remove the BS component from the RIS configuration; and using, the a location estimation module, the UE component of the RIS configuration to infer an estimation of the location of the UE.

[0014] According to embodiments, the present disclosure provides an unobtrusive, RIS driven UE location method and system. More specifically, embodiments relate to a system that localizes wireless devices communicating in a RIS-assisted scenario by analysing and processing the configuration of one or more RISs supporting devices’ communication while being able to refine the estimation using additional data sources. Advantageously, the system enables localization given a RIS configuration without interfering with the RIS’ communication related standard operations. In an example, the location estimation module may be configured, given the configuration of a RIS optimized for communication purposes of a communicating device with the base station, to retrieve the location of the communicating devices without interfering with the RIS’ communication related operations. Alongside the localisation estimation, the location estimation module may run an additional estimator to obtain the accuracy of the localisation. According to further embodiments, the location system may leverage data skewness and Al to improve inference performance by leveraging the statistical distribution of users, and / or by combining the estimation with traditional location estimation methods.

[0015] According to an embodiment, the method for estimating the UE position may comprise a one-time preparation step to retrieve the position of the BS. This step can be done in different ways, in the easiest case by leveraging deployment information. If no deployment information are available, the position of the BS may be determined, e.g., by i) leveraging cooperation and processing with the BS, ii) analysing different RIS configurations, iii) train a respective Al agent, or iv) combinations thereof.

[0016] According to an embodiment, once the BS position is known, the UE location method may proceed by execution of the following steps:

[0017] 1 ) Obtaining the current configuration of an RIS optimized to enhance or worsen the channel conditions between BS and UEs;

[0018] 2) Computing the corresponding component of the RIS configuration from the known position of the BS;

[0019] 3) Extracting the BS component from the RIS configuration from step 1 .

[0020] Localizing the UE may then be performed either by a model-based approach or by a data-based approach. In the model-based approach, a RIS partial UE configuration corresponding to an arbitrary UE position may be reconstructed and the UE position of the best match may be taken as the localization result. In addition, the accuracy of the localization may be obtained by comparing the gradient of the RIS configuration with respect to the user position. On the other hand, in the data- based approach, the obtained RIS configuration may be fed to a pre-trained Al agent to obtain the location of the user and its accuracy. According to an embodiment, a re-training scheme for the location estimation module may be created by establishing a feedback loop with estimations from the location estimation module and external, independent data from available sensing sources (as sensing BSs, radio beacons, UE internal inertial sensors, satellite navigation systems, or other localization infrastructures). Both data may be combined to obtain improved-precision results which, in turn, may be used to iteratively improve the localisation precision of the location estimation module.

[0021] There are several ways how to design and further develop the teaching of the present invention in an advantageous way. To this end, it is to be referred to the dependent claims on the one hand and to the following explanation of preferred embodiments of the invention by way of example, illustrated by the figure on the other hand. In connection with the explanation of the preferred embodiments of the invention by the aid of the figure, generally preferred embodiments and further developments of the teaching will be explained. In the drawing

[0022] Fig. 1 is a schematic view illustrating a UE localization system according to an embodiment of the present disclosure,

[0023] Fig. 2 is a schematic view illustrating a communication scenario with a BS in far- field conditions and a UE in near-field conditions,

[0024] Fig. 3 is a schematic view illustrating a communication scenario with a RIS and a BS configured for communication together with a RIS configuration adapted to the communication conditions of BS and UE according to an embodiment of the present disclosure,

[0025] Fig. 4 shows a representation of the spatial patterns of a RIS configuration matched with the spherical and planar wavefronts corresponding to the UE in near-field, and the BS in far-field, respectively, according to an embodiment of the present disclosure, Fig. 5 is a diagram showing example results of localization of a test dataset in a squared area obtained by a method according to an embodiment of the present disclosure,

[0026] Fig. 6 is a diagram showing position errors of the test dataset with respect to the distance of the located UE to the RIS obtained by a method according to an embodiment of the present disclosure, and

[0027] Fig. 7 is a schematic view illustrating a UE localization system with data refining and continuous training according to an embodiment of the present disclosure.

[0028] Reconfigurable intelligent surfaces (RISs) are thin, nearly-passive, low-cost devices able to programmatically alter the propagation properties of the medium within which they are installed. They usually take the shape of a planar array of antennas, which individually receive impinging RF signals and alter them with some degree of freedom (e.g., by controlling the amplitude and phase of the received signals at each array element) before re-emitting (reflecting) them, allowing to reflect the waves in a controllable fashion. The working principle of the vast majority of nowadays available RIS devices and RIS optimization solutions lies in the control of the phase shift introduced by each array element of the surface before the re-emission of the signal. This allows to control how the RIS reflects the signals and to steer, focus, or vanish the received power at the desired locations. The different configurations of the device (i.e., the set of phase shifts each individual element applies to the reflected wave) can be optimized according to a specific purpose, allowing the RIS to be suitable for different use-cases. For example, it can be used for boosting or flooring the gain between two devices, providing enhanced reflected paths sustaining the communication (for reference, see LIU, Yuanwei, et al. Reconfigurable intelligent surfaces: Principles and opportunities. IEEE communications surveys & tutorials, 2021 , 23. Jg., No. 3, S. 1546-15779).

[0029] Prior art RIS-aided localization techniques are largely based on those already available for access nodes equipped with planar antenna arrays, e.g., multiple-input multiple-output (MIMO) base stations (BS). However, when applying these techniques to RIS, they present several problems as these techniques usually rely on far-field assumptions, which prove unreasonable for RIS devices due to their much larger number of elements if compared with traditional active antennas. For example, a small, indoors 1 x 1 m2RIS, operating at = 2.5cm, would have a near field region dFof 160m, an outdoors 2 x 2 m2RIS at the same frequency would

[0030] 2D2have a near field region of 640m, following dp= — , (for reference, see A

[0031] BJORNSON, Emil; DEMIR, Ozlem Tugfe; SANGUINETTI, Luca. A primer on nearfield beamforming for arrays and reconfigurable intelligent surfaces. In: 2021 55th Asilomar Conference on Signals, Systems, and Computers. IEEE, 2021. S. IOS- 112).

[0032] Additionally, existing solutions require RIS operations specifically designed for the localization task, which negatively impacts communication performance if the RIS is serving both localization and communication enhancement, i.e., localization operation would reduce the time that the RIS is serving for communication improvement or would configure it in a way other than the optimal for communication purposes. This is the case of simple localization techniques based on sweeping the environment with focused RIS configurations at different directions to find the most suitable, and also localization techniques based on refined channel state information at the BS (as described, e.g., in CHEN, Hui, et al. A tutorial on terahertz-band localization for 6G communication systems. IEEE Communications Surveys & Tutorials, 2022).

[0033] When operating for communication purposes, and regardless of the type of application (e.g., channel enhancement or gain flooring), the location of communicating devices directly influences the resulting RIS configuration, i.e., the configuration adapts to the direction of arrival / departure of the signal of a device in far-field, and to the direction of arrival / departure and the distance of a device in nearfield. According to embodiments, the present disclosure provides a location method that exploits this characteristic to design a localization strategy that, given solely the configuration of one or multiple RIS, recovers the position of the connected devices. Differently from state-of-the-art solutions, the method according to embodiments of the present disclosure does not require either time of arrival information (Time of Arrival, ToA, and / or Difference Time of Arrival, DToA), power measurements, or running specific operations at the RIS, e.g., select specific configurations to perform a scan of the surrounding area and compare the received signal at the BS to locate the target. Accordingly, embodiments of the present disclosure provide the technical advantage of i) not interfering with communication enhancement operations at the RIS, ii) not requiring cooperation with BS and UEs, e.g., with channel measurements.

[0034] The above-mentioned solutions would require additional equipment / computational capabilities to get affordable localization accuracy performance. Conversely, according to embodiments, the present disclosure provides a solution that requires only the RIS configuration information to perform the localization task. Hence, it may run directly on the RIS board without requiring communication with the BS and without performing dedicated operations to perform localization. Consequently, it does not interfere with communication-related operations performed by the RIS or the BS. Embodiments of the solution disclosed herein are also compatible both with active RIS and with large active antenna arrays (as those used in massive-MIMO) as long as they fashion a large enough number of elements, and hence operate in the near-field region, or the technique is used to obtain angular information only, to be combined with other sensing sources.

[0035] Generally, the RISs used in the context of embodiments of the present disclosure may be designed and operated following to the concepts described in the document “Reconfigurable Intelligent Surfaces (RIS); Use Cases, Deployment Scenarios and Requirements”, Group Report, ETSI GR RIS 001 V1.1.1 (2023-04), which is hereby incorporated herein by reference.

[0036] Embodiments of the present disclosure provide a method for locating a target, e.g. a UE, wherein the method is based on the RIS configuration of a RIS (wherein the RIS is configured, e.g., by an agent unrelated to the solution proposed herein, to serve the target (e.g., to enhance communications with it or to reduce interference or EM exposure), and that, by only using the resulting RIS configuration as input, infers the location of said target. Further, embodiments of the present disclosure provide a method that, in addition to the location information, provide the accuracy of such location information.

[0037] Fig. 1 provides an overall view of the building blocks of a system according to an embodiment of the present disclosure. According to the illustrated scenario, a target device 110, e.g. any kind of user equipment, UE, is connected to a base station, BS 112, of a cellular communication network. Further, a RIS 114 is provided within the coverage area of the BS 112 to serve the target device 110. Referring to that, a high-level description of the operational steps is provided below. Additional details are provided in the use cases.

[0038] According to an embodiment, an operational step performed by the system may include obtaining the configuration of the RIS 114 (expressed as configuration vector ), either from the RIS 114 itself or from its controlling entity 116. The configuration < > of the RIS 114 is then provided to location estimation module 120, as indicated in Fig. 1 by the respective arrow from the RIS controller 116 to the location estimation module 120 carrying the RIS configuration .

[0039] The location estimation module 120 is configured to estimate, merely based on the received configuration < > of the RIS 114, the location r of the target. To do this, the location estimation module 120 may utilize different technologies. Accordingly, the location procedure may include any of the following steps, which may be executed by a position estimation submodule 122 of the location estimation module 120:

[0040] According to one embodiment, the position estimation submodule 122 may be configured to determine the relative position of the BS 112 with respect to the RIS 114, and then decompose the geometrical information of the RIS configuration phase shifts in the wave coming from the BS 112 and the curved, near-field wave front pointed towards the target device 110. This embodiment is described in further detail in connection with Figs. 2-4 below.

[0041] According to another embodiment, the position estimation submodule 122 may be configured to leverage a digital twin to simulate the propagation conditions of the environment to determine the spatial location where the current RIS configuration is focused. For instance, the digital twin may be constructed based on the knowledge of the spatial characteristics of the propagation environment (e.g., in the form of a 3D model). This allows to accurately reproduce in a virtual environment the effect of different RIS configurations on the service area, i.e. the area served by the RIS 114. For instance, the gain map generated by a given RIS configuration can be reconstructed. From that, it is possible to take the configuration of the RIS 114 in the real environment, obtain from this real world RIS configuration a gain map in the digital twin, and estimate the position of the target device / UE 110 based on the gain map in the digital twin. For example, the point with the highest ratio between the signal strength (in watts) and the square of the separation between the target device / UE 110 and the RIS 114 could be taken as the central point originally intended to be covered by the RIS 114. According to an embodiment, the local curvature of such a ratio in the vicinity of the central point would offer an uncertainty figure from which an accuracy of the location estimation can be inferred.

[0042] According to yet another embodiment, the position estimation submodule 122 may be configured to train a machine learning model, using all location information available, to infer the location of the served user by using only the RIS configuration in a continuous precision-refining loop. For instance, the available location information may include information about the location of the RIS 114 and information on the relative position of the BS 112 with respect to the RIS 114.

[0043] As shown in Fig. 1 , the location estimation module 120 may further include an accuracy estimation submodule 124 that may be configured to compute an accuracy arof the location information obtained in accordance with the embodiments describe above. Such accuracy information make the location data combinable with other sources of information.

[0044] As shown in Fig.1 , the accuracy estimation submodule 124 may receive as inputs the current configuration of the RIS 114 together with the location r of the target device 110 as determined by the location estimation module 120. In one example, to calculate the accuracy of the location estimation, the accuracy estimation submodule 124 may be configured to estimate the uncertainty of the location information considering the phase quantization and the similarity of the current RIS configuration with the configurations that would correspond to adjacent locations. The phase quantization must be considered in the formulas that determine the amount of position information by computing the spatial gradient of the configurations (i.e. , the quotient between configuration distance and spatial distance when comparing said adjacent locations). In another example, the accuracy estimation submodule 124 may be configured to train a machine learning model that predicts the uncertainty of the localization output of the position estimation submodule 122 by comparing estimated position values with a higher accuracy reference. This reference value will typically take the shape of a dataset of configurations obtained for known positions, so the difference between the position estimation and the known position can be computed.

[0045] The dual output of the location estimation module 120, comprised of both location and location accuracy, is advantageous in its ability to be combined with other information sources.

[0046] As already mentioned above, an advantage of the solution proposed herein is the ability of the location estimation module 120 to locate the target / user 110 by observing only the current configuration of the RIS 114, without interfering with the RIS communication capabilities. In this regard, it is noted that the current state of the art depends on probing by either trying different RIS configurations, emitting and analysing successive pilot signals, establishing ad-hoc communications with additional hardware, etc., i.e. by interfering with the communications. Another advantage achieved in some embodiments of the present disclosure is the production of a pair of values (measurement and accuracy) that allows to integrate the location estimation module 120 into larger positioning systems, to refine the measurement with other sensors, to combine the information of several such modules in parallel, and / or to work even when the near-field assumption is broken.

[0047] Fig. 2, in which like components denote like or similar components as in Fig. 1 , depicts a RIS-aided communication scenario with a UE 110 (sometimes herein also denoted target or connected device) located in a portion of the service area of a BS 112 that is under the influence of the presence of a RIS 114. In the depicted example, the UE 110 is located in the near-field regime of the RIS 114 and the BS 112 is in the far-field regime of the RIS 114. These conditions result in a spherical wavefront characterizing the signal propagating between the UE 110 and the RIS 114, while the wave propagating between the BS 112 and the RIS 114 is characterized by a planar wave front at the RIS location. As illustrated in Fig. 2, the planar wavefront is oriented according to the direction of signals between the BS 112 and the RIS 114, while the spherical wavefront is dependent both i) on the direction of the signal between the UE 110 and the RIS 114, and ii) the distance between the UE 110 and the RIS 114. The distance between the UE 110 and the RIS 114 defines the curvature of the spherical wavefront at the RIS location.

[0048] Fig. 3 depicts an example of RIS 114 configured to support the communication gain between the UE 110 and the BS 112. Here, coupling effects in the RIS 114 are ignored for the simplicity of the description of the embodiment, but, as will be appreciated by those skilled in the art, the system according to the concepts described herein can be configured to include appropriate means such as to operate to account for those coupling effects. Taking into account the location of the UE 110, the resulting RIS reflection configuration is dependent on the planar and the spherical wavefronts at the RIS location. In other words, the phase shifters of the individual RIS elements 115 of the RIS 114 controlling the reflection properties of the RIS 114 exhibit a spatial pattern which is matched with the interference pattern between the two wavefronts. Hence, the RIS configuration is dependent on the position of both the BS 112 and the UE 110. It is noted that in this case a RIS solution model can be considered as a reference. Such model provides a RIS configuration given the location of the connected device 110 and the BS 112 as a function of the propagation conditions of such devices 110, 112 with respect to the RIS 114. In particular, as described in detail in Albanese, Antonio, et al. "MARISA: A selfconfiguring metasurfaces absorption and reflection solution towards 6G." IEEE INFOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022 (which is hereby incorporated herein by reference), the RIS configuration < > can be described as a function of rUEand rBS. Hence, the RIS configuration can be retrieved from a combination of the RIS configuration matched to the UE location, denoted (pUE, and the RIS configuration matched to the BS location, denoted which are both represented in Fig. 4. Specifically, Fig. 4 shows a representation of the spatial patterns of the RIS configuration matched with the spherical and planar wavefronts corresponding to the UE 110 in near-field and the BS 112 in far-field, respectively.

[0049] Accordingly, in order to obtain the position of the UE110 merely from the RIS configuration, in an embodiment it may be provided to separate the contribution of the planar wavefront and the spherical one. To do so, the relative position of the BS 110 with respect to the RIS 114 must be known. It is noted that if the BS 110 lies in the near-field of the RIS 114, also signals between the BS 110 and RIS 114 would be characterized by a spherical wavefront at the RIS 114. However, this would not require any modification of the working principle of the solution disclosed herein.

[0050] The BS position can be known a priori if information on the deployment is available. Otherwise, in accordance with embodiments of the present disclosure, it can be inferred at the time of installation of the RIS 114, aided with the location estimation module 120 in several manners. The following non exhaustive list provides exemplary embodiments that can be implemented to obtain it:

[0051] According to an example, the BS position can be inferred, without the cooperation of the BS 112, by establishing the best communication with a connected device that is at a known location and that is able to operate in the network operated by the owner of the BS 112 and inferring the BS position from that. For example, it may be provided to exploit a device with a directional antenna that can communicate with the BS 112 only through the RIS 114. Requesting the RIS configuration at that instant, i.e. when the best communication at the given location is established, would be enough to compute the UE component of the RIS configuration. This UE component can then be subtracted from the RIS configuration to obtain the planar, BS 112 related component of the configuration, which can then be used for future location estimation.

[0052] Another method that can be used without BS cooperation is collecting and processing several RIS configurations. Such configurations will exhibit different patterns due to the mobility of the UE, but the RIS-BS component will be unchanged due to static deployment. This approach can be based on the following principle: The BS position can be inferred from a RIS configuration and a UE position. Similarly, given n UE positions and one RIS configuration, n possible BS locations can be obtained. Accordingly, every RIS configuration corresponds to a space of possible BS locations. Hence, given a set of RIS configurations, the BS can be found in the intersection of the corresponding spaces.

[0053] According to another example, the BS position can also be inferred with the cooperation of the BS 112. For instance, the BS 112 can cooperate in the process by searching for the best retroreflector configuration for the RIS 114 (i.e., the configuration in which the RIS 114 mirrors the signal back to the BS 112). From this configuration, it is trivial to obtain the BS position. Such operations can be coordinated through the control channel 126 established between the BS 112 and the RIS 114.

[0054] Once the position of the BS 112 is known, it is possible to subtract that information from future configurations of the RIS. More specifically, knowing the wavefront associated with the relative position of the BS 112 to the RIS 114, and a given configuration, it is possible to obtain the complementary wavefront (i.e., the one corresponding to the UE 110). That wavefront carries the position of the UE 110 encoded in the form of direction, and curvature. In the simplest analytical way, subtracting the phase shifts corresponding to the BS wavefront on every individual RIS antenna element 115 from the configuration would give the phase shifts of the UE wavefront (for reference, see again Albanese, Antonio, et al. "MARISA: A selfconfiguring metasurfaces absorption and reflection solution towards 6G." IEEE INFOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022).

[0055] Given that, the position of the connected device 110 can be inferred by finding which position would produce the RIS configuration which is most similar to the one at hand. To do so, one may proceed by considering a spherical wave front model to describe the phase at each RIS element 115 as UE — f(.rUE N x 1 where <pUEe (0,2TT] is the vector containing the phase shifts of each of the N elements 115 of the RIS 114, and whose elements can be described as where rndenotes the location of the n-th RIS element 115 with respect to the position of a reference element rrefof the RIS 114, rUEdenotes the location of the UE 110 with respect to the reference element, andOdenotes a phase shift applied to all the elements 115 of the RIS 114 depending on the specific application, for example to create constructive(destructive) interference at the location of the UE 110 for channel gain enhancement(minimization). The reference element is an arbitrarily chosen element of the RIS 114: all other elements will be configured to interfere with the reference element as constructively as possible.

[0056] It should be noted that for the sake of simplicity the above equations refer to a linear arrangement of the RIS elements 115. However, as will be appreciated by those skilled in the art, the equations can be easily extended to the more general case of a RIS with a two-dimensional arrangement of the RIS elements 115, e.g. in form of an N x M array of RIS elements 115.

[0057] The above equations describe the dependency of the RIS configuration with the UE position. According to embodiments of the present disclosure, from that the UE position may be obtained as the solution of the following optimization problem: where <f>UEis the UE component of the RIS configuration, and - 4>UE \\ denotes the norm of the difference between <pUEand f(rUE). Here, the norm of the difference between the vectors is used as a metric of the similarity between the vectors, and this can be substituted by other equivalent metrics as the cosine similarity, the Manhattan distance, or others: Different objective functions can be used, for example by maximizing the similarity of the two configuration vectors f ( UE)and UE- According to an embodiment, the accuracy of the UE position can be obtained by computing the gradient of the RIS configuration when the position varies around the central, best-match value,

[0058] In other words, when two similar positions would be expected to yield two very similar RIS configurations, the position information extracted from the configuration will be highly affected by inaccuracies, noise, etc., and hence will show a low accuracy. Conversely, if one can appreciate easily noticeable changes in the configurations yielded by two similar positions, those two positions can be easily told apart. For instance, such changes in configurations can be computed for 3 orthogonal directions in space, which is one of the possible computations of the accuracy of the position estimation itself. According to an embodiment, given the quantized nature of the RIS configuration, the amount of bit changes on RIS configuration per length unit may be taken as a suitable metric.

[0059] According to an embodiment of the present disclosure, the location of the UE 110 may be determined by exploiting the correlation between user location and RIS configuration. To do so, several different approaches can be realized, either with or without cooperation with an existing sensing infrastructure.

[0060] In an embodiment that does not require any cooperation with a sensing infrastructure (and that, hence, is particularly suitable if, for example, no sensing infrastructure is available), it may be provided that the UE location method includes a step of preparing a dataset of pairs ‘position’ - ‘RIS configuration’. Based on this dataset a machine learning algorithm may be trained to produce a UE location by processing the RIS configuration. After obtaining the first agent in this way, a second agent may be trained with the same dataset, adding the first agent’s position estimation to forecast the uncertainty of said estimation. This approach brings a unique benefit: environmental properties (for example in the form of large reflectors) would not produce location errors, as the training of the agents takes those properties into account.

[0061] In another embodiment including a cooperation with an existing sensing infrastructure, current location information retrieved from the sensing infrastructure can be used as a “true location”. This location information may be paired with the corresponding RIS configuration at each time to produce a training dataset in an automated way. The training dataset may then be used for training a machinelearning agent to estimate the respective location. If the agent trained in this way estimates the position with an error distribution independent from the error committed by the existing sensing infrastructure (due to the wildly different nature of the measurement estimation process), this process can be indefinitely carried out in a continuous manner, further refining the results until reaching a desired accuracy.

[0062] This embodiment has been implemented by the inventors in a simulation as a proof of concept. Using a neural network trained with RIS configurations, the trained agent is able to reconstruct the position of users, given only a RIS configuration that the agent has never seen for training, with remarkable accuracy. The respective results obtained for a test dataset are illustrated in Figs. 5 and 6. Specifically, Fig. 5 shows example results of localization in a squared area. The centre of the circles indicates the actual position of the UEs and the diameter indicates the corresponding location error. Results are obtained with a RIS comprising 21 x 21 elements with only 2 phases per element (0 or 180 degrees). The RIS location is marked with an x. Fig.6 shows the position errors of the test dataset with respect to the distance of the located UE to the RIS. Again, results are obtained with a RIS comprising 21 x 21 elements with only 2 phases per element (0 or 180 degrees).

[0063] In the above use case, where there is a collaboration with existing sensing infrastructure and where the distribution of the errors of location information provided by the sensors of the sensing infrastructure and the errors of location information obtained by location estimation module 120 is independent of each other, the results may be further refined. For instance, as schematically illustrated in Fig. 7, for data driven architectures, a continuous training system 130 with a sustained cycle of re-training may be implemented that continuously refines the position estimations. The use of independent devices for the continuous training system 130 on the one hand and the location estimation module 120 on the other hand brings many-fold advantages: The devices require very different amounts of processing power, wherein the training efforts usually tax much more compared to the prediction efforts. Furthermore, compared to the location estimations, the results of the training are much less time-sensitive. Both circumstances encourage the placement of the training module 130 in a separate environment (e.g., cloud infrastructure).

[0064] According to an embodiment, the setup depicted in Fig. 7 may be configured to cyclically execute the following operations:

[0065] The system may include a location refinement module 132 configured to collect the separate and independent position information from both the location estimation module 120 and any external available sensors. A communication channel may be established between the location refinement module 132, the location estimation module 120 and the continuous training system 130. As exemplarily shown in Fig. 7, the position information from external available sensors may include ToA information from the BS 112 and / or inertial information from the UE 110.

[0066] According to an embodiment, the location refinement module 132 may be further configured to also collect the accuracy figures of both estimations. The accuracy readings are a critical part of the process, as they allow to meaningfully combine several sources. Hence, the accuracy values corresponding to each individual location estimation, both from the location estimation module 120 and from the external, cooperating infrastructure, are central for the refinement of the estimations. These accuracy values may be transmitted via the same communication channels as described before.

[0067] According to an embodiment, the location refinement module 132 may include a data combination filter 133 that is configured to combine the position estimations from available sources, weighted by their respective accuracies, to produce a third estimate (denoted ‘refined position information’ in Fig. 7) with higher accuracy. This can be done in different ways depending on the statistical nature of the errors faced. In the general case, assuming normal distribution of inaccuracies, the normal distributions of the values used as input may be multiplied and renormalized in the form of a new Gaussian distribution, from which the new value and the new accuracy may be extracted.

[0068] The location measurements may be executed, e.g., periodically or on demand. With a given periodicity (for instance, every n new measurement), the position information obtained by the location estimation module 120 and the refined position information obtained by the location refinement module may be fed as inputs to the continuous training system 130 to re-train the location estimation module 120 with the improved- accuracy data. The result of the training is then re-loaded in the position and accuracy estimation submodules 122, 124.

[0069] Many modifications and other embodiments of the invention set forth herein will come to mind to the one skilled in the art to which the invention pertains having the benefit of the teachings presented in the foregoing description and the associated drawings. Therefore, it is to be understood that the invention is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.

[0070] L ist of refe re n ce s i g n s

[0071] 110 target device, UE

[0072] 112 base station, BS

[0073] 114 RIS

[0074] 115 RIS element

[0075] 116 RIS controller

[0076] 120 location estimation module

[0077] 122 position estimation submodule

[0078] 124 accuracy estimation submodule

[0079] 126 control channel

[0080] 130 continuous training system

[0081] 132 location refinement module

[0082] 133 data combination filter

Claims

C l a i m s1. A method for determining a UE location within a cellular radio network, wherein the UE (110) is associated with a base station, BS (112), of the network and wherein the network includes at least one reconfigurable intelligent surface, RIS (114), wherein the RIS (114) has a RIS configuration set to serve the UE (110), the RIS configuration including a BS component and a UE component, the method comprising: obtaining, by a location estimation module (120), a current RIS configuration of the RIS (114); determining, by the location estimation module (120), the UE component of the current RIS configuration by using position information about the position of the BS (112) to compute and remove the BS component from the RIS configuration; and using, by the location estimation module (120), the UE component of the RIS configuration to infer an estimation of the location of the UE (110).

2. The method according to claim 1 , further comprising: reconstructing a RIS partial UE configuration corresponding to an arbitrary UE position; matching the obtained RIS configuration without the BS component to the reconstructed RIS partial UE configuration; and providing the UE position of the best match as the UE localization result.

3. The method according to claim 1 or 2, further comprising: determining the relative position of the BS (112) with respect to the RIS (114); and decomposing the geometrical information of the RIS configuration phase shifts in the wave coming from the BS (112) and the curved, near-field wave front pointed towards the UE (110).

4. The method according to any of claims 1 to 3, further comprising: determining the wavefront of the BS component associated with the relative position of the BS (112) with regard to the RIS (114);obtaining the phase shifts of the complementary wavefront of the UE component by subtracting the phase shifts corresponding to the determined wavefront of the BS component on each RIS element (115) from the current RIS configuration.

5. The method according to claim 4, further comprising: using a spherical wave front model describing the phase shifts at each RIS element (115) to determine which position would produce the phase shifts that are most similar to the obtained phase shifts of the complementary wavefront of the UE component.

6. The method according to any of claims 1 to 5, further comprising: constructing a digital twin based on a knowledge of the spatial characteristics of the propagation environment in an area served by the RIS (11 ); obtaining from the configuration of the RIS (114) in the real environment a gain map in the digital twin; and estimating the position of the UE (110) based on the gain map in the digital twin.

7. The method according to any of claims 1 to 6, further comprising: preparing a first training dataset including pairs of ‘position’ - ‘RIS configuration’; and using the first training dataset to train a first machine-learning agent to determine a UE location by processing the RIS configuration.

8. The method according to claim 7, further comprising: preparing a second training dataset by adding the first agent’s position estimation to the first training dataset; and using the second training dataset to train a second machine-learning agent to forecast the uncertainty of the first agent’s position estimation.

9. The method according to any of claims 1 to 6, further comprising: obtaining current location information from a sensing infrastructure;preparing a training dataset by pairing the obtained location information with the corresponding RIS configuration; and using the prepared training dataset to train a machine-learning agent to estimate the respective location.

10. The method according to any of claims 1 to 9, further comprising: generating refined position information by combining the estimation of the location of the UE (110) inferred by the location estimation module (120) with position estimations from available sources, weighted by their respective accuracies.11 . The method according to claim 10, further comprising: providing the estimation of the location of the UE (110) inferred by the location estimation module (120) and the generated refined position information as inputs to a training system (130); and using the training system (130) to re-train the location estimation module (120) with the refined position information.

12. A system for determining a UE location within a cellular radio network, in particular for execution of a method according to any of claims 1 to 11 , wherein the UE (110) is associated with a base station, BS (112), of the network and wherein the network includes at least one reconfigurable intelligent surface, RIS (114), wherein the RIS (114) has a RIS configuration set to serve the UE (110), the RIS configuration including a BS component and a UE component, the system comprising a location estimation module (120) including one or processors which, alone or in combination, are configured to provide for execution of the following steps: obtaining a current RIS configuration of the RIS (114); determining the UE component of the current RIS configuration by using position information about the position of the BS (112) to compute and remove the BS component from the RIS configuration; and using the UE component of the RIS configuration to infer an estimation of the location of the UE (110).

13. The system according to claim 12, further comprising a location refinement module (132) configured to generate refined position information by combining the position estimations from the location estimation module (120) with position information from available sources, weighted by their respective accuracies.

14. The system according to claim 13, further comprising a training system (130) configured to receive the position information obtained by the location estimation module (120) and the refined position information obtained by the location refinement module (132) as inputs, and re-train the location estimation module (120) with the refined position information.

15. The system according to claim 14, wherein the location estimation module(120) and the training system (130) are implemented on different computing platforms.

Images