Selecting Positioning Beams for Downlink Positioning

Abstract

A terminal device including circuitry configured to perform: measuring, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein each candidate positioning beam pair represents a combination of a respective receive beam of the terminal device and a respective transmit beam of a respective network node of one or more network nodes; estimating, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on the measured positioning reference signal metrics; selecting one or more candidate positioning beam pairs as positioning beam pairs for downlink positioning.

Claims

1. A terminal device comprising: at least one processor; and at least one non-transitory memory storing instructions that, when executed with the at least one processor, cause the apparatus to perform: measuring, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein said first subset of candidate positioning beam pairs is a subset of a plurality of candidate positioning beam pairs, wherein each candidate positioning beam pair of said plurality of candidate positioning beam pairs represents a combination of a respective receive beam of said terminal device and a respective transmit beam of a respective network node of one or more network nodes; estimating, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on said measured positioning reference signal metrics, wherein said second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of said plurality of candidate positioning beam pairs that is/are not part of said first subset of candidate positioning beam pairs; selecting, at least partially based on said measured positioning reference signal metrics and said estimated positioning reference signal metrics, one or more candidate positioning beam pairs from said plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

2. A terminal device according to claim 1, wherein said plurality of candidate positioning beam pairs represents available combinations of receive beams of said terminal device and transmit beams of said one or more network nodes.

3. A terminal device according to claim 1, where the instructions that, when executed with the at least one processor, cause the apparatus to perform determining said first subset of candidate positioning beam pairs.

4. A terminal device according to claim 1, wherein said first subset of candidate positioning beam pairs is determined with at least one of: (i) selecting candidate positioning beam pairs from said plurality of candidate positioning beam pairs; or (ii) randomly selecting candidate positioning beam pairs from said plurality of candidate positioning beam pairs; or (iii) selecting, for each network node of said one or more network nodes, candidate positioning beam pairs from said plurality of candidate positioning beam pairs representing at least substantially equally spaced receive beams of said terminal device and/or at least substantially equally spaced transmit beams of said respective network node; or (iv) selecting candidate positioning beam pairs from said plurality of candidate positioning beam pairs based on at least one of (i) beam widths of the beams, or (ii) spatial diversity of the beams, or (iii) time.

5. A terminal device according to claim 1, wherein said at least one respective positioning reference signal metric is or represents at least one of: (i) a Received Signal Strength Indicator, RSSI; or (ii) a Reference Signal Received Power, RSRP; or (iii) a Signal-to-Noise Ratio, SNR; or (iv) a power of the strongest channel tap; or (v) a Carrier-to-Interference Ratio, CIR.

6. A terminal device according to claim 1, wherein said at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of said second subset of candidate positioning beam pairs at least partially based on said measured positioning reference signal metrics with: calculating, for each candidate positioning beam pair of said second subset of candidate positioning beam pairs, a respective weighted average of said measured positioning reference signal metrics.

7. A terminal device according to claim 6, wherein: at least two positioning reference signal metrics measured for respective candidate positioning beam pairs of said first subset of candidate positioning beam pairs are used for calculating said respective weighted average of measured positioning reference signal metrics, and/or for each candidate positioning beam pair of said second subset of candidate positioning beam pairs, the positioning reference signal metrics measured for respective candidate positioning beam pairs of said first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node like said respective candidate positioning beam pair of said second subset of candidate positioning beam pairs are used for calculating said respective weighted average of measured positioning reference signal metrics, and/or for each candidate positioning beam pair of said second subset of candidate positioning beam pairs, positioning reference signal metrics that were (i) most recently measured or (ii) measured within a measurement time window are used for calculating said respective weighted average of measured positioning reference signal metrics

8. A terminal device according to claim 6: wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of said second subset of candidate positioning beam pairs, said respective weighted average of measured positioning reference signal metrics is determined at least partially based on at least one of (i) a spatial distance between, or (ii) beam widths of, or (iii) antenna correlation information associated with at least one beam represented with said respective candidate positioning beam pair of said second subset of candidate positioning beam pairs and at least one beam represented with said respective candidate positioning beam pairs of said first subset of candidate positioning beam pairs, and/or wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of said second subset of candidate positioning beam pairs, said respective weighted average of measured positioning reference signal metrics is determined at least partially based on a time of measurement of at least one measured positioning reference signal metrics of said measured positioning reference signal metrics.

9. A terminal device according to claim 1, wherein said at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of said second subset of candidate positioning beam pairs at least partially based on said measured positioning reference signal metrics with using a machine learning algorithm.

10. A terminal device according to claim 9, wherein said machine learning algorithm is implemented as one of: (i) a deep neural network, or (ii) a convolutional neural network, or (iii) a generative adversial network, or (iv) a long-short term memory, or (v) a time delay neural network.

11. A terminal device according to claim 9, wherein said machine learning algorithm receives one or more positioning reference signal metric matrices as input, wherein said one or more positioning reference signal metric matrices comprise(s) said measured positioning reference signal metrics as matrix elements.

12. A terminal device according to claim 11, wherein each of said positioning reference signal metric matrices comprises respective positioning reference signal metrics measured for respective candidate positioning beam pairs of said first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node.

13. A terminal device according to claim 11, wherein said measured positioning reference signal metrics are positioned in said one or more positioning reference signal metric matrices at least partially based on spatial distances between beams represented with said candidate positioning beam pairs of said first subset of candidate positioning beam pairs.

14. A method, performed at least with a terminal device, the method comprising: measuring, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein said first subset of candidate positioning beam pairs is a subset of a plurality of candidate positioning beam pairs, wherein each candidate positioning beam pair of said plurality of candidate positioning beam pairs represents a combination of a respective receive beam of said terminal device and a respective transmit beam of a respective network node of one or more network nodes; estimating, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on said measured positioning reference signal metrics, wherein said second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of said plurality of candidate positioning beam pairs that is/are not part of said first subset of candidate positioning beam pairs; selecting, at least partially based on said measured positioning reference signal metrics and said estimated positioning reference signal metrics, one or more candidate positioning beam pairs from said plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

15. A non-transitory computer readable medium comprising computer program code, the computer program code when executed by a processor of an apparatus causing said apparatus to: measure, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein said first subset of candidate positioning beam pairs is a subset of a plurality of candidate positioning beam pairs, wherein each candidate positioning beam pair of said plurality of candidate positioning beam pairs represents a combination of a respective receive beam of said terminal device and a respective transmit beam of a respective network node of one or more network nodes; estimate, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on said measured positioning reference signal metrics, wherein said second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of said plurality of candidate positioning beam pairs that is/are not part of said first subset of candidate positioning beam pairs; select, at least partially based on said measured positioning reference signal metrics and said estimated positioning reference signal metrics, one or more candidate positioning beam pairs from said plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

Description

BRIEF DESCRIPTION OF THE FIGURES

[0139] Some example embodiments will now be described with reference to the accompanying drawings.

[0140] FIG. 1 is a block diagram of an exemplary embodiment of a terminal device according to the present disclosure;

[0141] FIG. 2 is a block diagram of an exemplary embodiment of a network node according to the present disclosure;

[0142] FIG. 3 is a schematic illustration of an exemplary embodiment of a system according to the present disclosure;

[0143] FIG. 4 is a flow chart illustrating an exemplary embodiment of a method according to the present disclosure;

[0144] FIG. 5 is a flow chart illustrating an exemplary embodiment of determining the first subset of candidate positioning beam pairs according to the present disclosure;

[0145] FIG. 6 is a flow chart illustrating an exemplary embodiment of estimating PRS metrics according to the present disclosure;

[0146] FIG. 7 is a flow chart illustrating a exemplary embodiment of determining weights according to the present disclosure;

[0147] FIG. 8 is a flow chart illustrating another exemplary embodiment of estimating PRS metrics;

[0148] FIG. 9 is a schematic illustration of an PRS metric matrix received as input by a machine learning algorithm according to the present disclosure;

[0149] FIG. 10 is a schematic illustration of an PRS metric matrix provided as output by a machine learning algorithm according to the present disclosure; and

[0150] FIG. 11 is a schematic illustration of examples of tangible and non-transitory computer-readable storage media.

DETAILED DESCRIPTION OF THE FIGURES

[0151] The following description serves to deepen the understanding of the present disclosure and shall be understood to complement and be read together with the description of example embodiments of the present disclosure as provided in the above summary section of this specification.

[0152] While the specific radio system in the examples below is 5G, this is only to be considered a non-limiting example.

[0153] FIG. 1 shows a block diagram of an exemplary embodiment of a terminal device in form of user equipment (UE) 100 according to the present disclosure. For example, UE 100 may be one of a smartphone, a tablet computer, a notebook computer, a smart watch, a smart band, an IoT device or a vehicle.

[0154] UE 100 comprises a processor 101. Processor 101 may represent a single processor or two or more processors, which are for example at least partially coupled, for example via a bus. Processor 101 executes a program code stored in program memory 102 (for example program code causing UE 100 to perform one or more of the embodiments of a method according to the present disclosure or parts thereof, when executed on processor 101), and interfaces with a main memory 103. Program memory 102 may also contain an operating system for processor 101. Some or all of memories 102 and 103 may also be included into processor 101.

[0155] One of or both of a main memory and a program memory of a processor (e.g. program memory 102 and main memory 103) could be fixedly connected to the processor (e.g. processor 101) or at least partially removable from the processor, for example in the form of a memory card or stick.

[0156] A program memory (e.g. program memory 102) may for example be a non-volatile memory. It may for example be a FLASH memory (or a part thereof), any of a ROM, PROM, EPROM, MRAM or a FeRAM (or a part thereof) or a hard disc (or a part thereof), to name but a few examples. For example, a program memory may for example comprise a first memory section that is fixedly installed, and a second memory section that is removable from, for example in the form of a removable SD memory card.

[0157] A main memory (e.g. main memory 103) may for example be a volatile memory. It may for example be a DRAM memory, to give non-limiting example. It may for example be used as a working memory for processor 101 when executing an operating system, an application, a program, and/or the like.

[0158] Processor 101 further controls a communication interface 104 (e.g. radio interface) configured to receive and/or transmit data and/or information. In particular, communication interface 104 is configured to receive radio signals like PRS from a network node, such as a base station (e.g. network node 200 depicted in FIG. 2), by use of more than one receive beam. Moreover, communication interface 104 may be configured to transmit radio signals to a network node, such as a base station, by use of more than one transmit beam. To this end, communication interface 104 may comprise at least one or multiple antennas, for example an antenna array. It is to be understood that any computer program code based processing required for receiving and/or measuring (e.g. evaluating) radio signals may be stored in an own memory of communication interface 104 and executed by an own processor of communication interface 104 and/or it may be stored for example in memory 103 and executed for example by processor 101.

[0159] Communication interface 104 may in particular be configured to communicate according to a cellular communication system like a 2G/3G/4G/5G NR or future generation cellular communication system. UE 100 may use radio interface 104 to communicate with a base station (e.g. network node 200 depicted in FIG. 2). The 2G/3G/4G/5G NR specifications are developed by and retrievable from the third generation partnership project (3GPP, https://www.3gpp.org/).

[0160] For example, the communication interface 104 may further comprise a Bluetooth Low Energy (BLE) and/or Bluetooth radio interface including a BLE and/or Bluetooth transmitter, receiver or transceiver. The Bluetooth specifications are developed by the Bluetooth Special Interest Group and are presently available under https://www.bluetooth.com/. For example, radio interface 104 may additionally or alternatively comprise a Wireless Local Area Network (WLAN) radio interface including at least a WLAN transmitter, receiver or transceiver. WLAN is for example specified by the standards of the IEEE 802.11 family (http://www.ieee.org/).

[0161] The components 102 to 104 of UE 100 may for example be connected with processor 101 by means of one or more serial and/or parallel busses.

[0162] It is to be understood that UE 100 may comprise various other components. For example, UE 100 may optionally comprise a user interface (e.g. a touch-sensitive display, a keyboard, a touchpad, a display, etc.).

[0163] FIG. 2 is a block diagram of an exemplary embodiment of a network node, such as base station (BS) (e.g. a next generation NodeB (gNB) of a 5G NR communication network, or an evolved NodeB (eNodeB) of a LTE communication network, BS, or the like).

[0164] Network node 200 comprises a processor 201. Processor 201 may represent a single processor or two or more processors, which are for example at least partially coupled, for example via a bus. Processor 201 executes a program code stored in program memory 202 (for example program code causing network node 200 to perform alone or together with UE 100 embodiments according to the present disclosure or parts thereof), and interfaces with a main memory 203.

[0165] Program memory 202 may also comprise an operating system for processor 201. Some or all of memories 202 and 203 may also be included into processor 201.

[0166] Moreover, processor 201 controls a communication interface 204 configured to receive and/or transmit data and/or information. In particular, communication interface 204 is configured to transmit radio signals like PRS by use of more than one transmit beam. Moreover, communication interface 104 may be configured to receive radio signals from a terminal device (e.g. UE 100 depicted in FIG. 1) by use of more than one receive beam. To this end, communication interface 204 may comprise at least one or multiple antennas, for example an antenna array.

[0167] Communication interface 204 may in particular be configured to communicate according to a cellular communication system like a 2G/3G/4G/5G NR or future generation cellular communication system. Network node 200 may use radio interface 204 to communicate with a terminal device (e.g. UE 100 depicted in FIG. 1).

[0168] The components 202 to 204 of network node 200 may for example be connected with processor 201 by means of one or more serial and/or parallel busses.

[0169] It is to be understood that network node 200 may comprise various other components.

[0170] FIG. 3 is a schematic illustration of an exemplary embodiment of a system 300 according to the present disclosure.

[0171] System 300 comprises UE 100 and three (i.e. X=3) network nodes 200-1 (denoted by index x=1), 200-2 (denoted by index x=2) and 200-3 (denoted by index x=3). UE 100 of FIG. 3 may be considered to correspond to UE 100 of FIG. 1, and each of network nodes 200-1 to 200-3 may be considered to correspond to network node 200 of FIG. 2. System 300 may be a communication system like a 2G/3G/4G/5G NR or future generation cellular communication system.

[0172] As shown in FIG. 3, each of network nodes 200-1 to 200-3 supports eight (i.e. L=8) transmit beams for transmitting PRSs which are denoted by the respective transmit beam index l=1, 2, 3, 4, 5, 6, 7 or 8 and UE 100 supports four (R=4) receive beams for receiving PRSs which are denoted by the respective receive beam index r=1, 2, 3 or 4. According to this example embodiment, receive beams and transmit beams which are adjacent to each other in the spatial domain have subsequent indexes in receive beam set/grid Gr and the transmit beam set/grid Gl, respectively.

[0173] Each combination of a respective receive beam of UE 100 and a respective transmit beam of a respective network node of network nodes 200-1 and 200-3 is to be understood to be represented by (e.g. to form or define) a candidate positioning beam pair. For each network node of the network nodes 200-1 to 200-3, there are thus R*L=32 respective candidate beam pairs representing all available combinations of receive beams of UE 100 and the respective network node of the network nodes 200-1 to 200-3. In total there are thus X*R*L=96 candidate positioning beam pairs representing all available combinations of receive beams of UE 100 and transmit beams of the network nodes 200-1 to 200-3. These 96 candidate positioning beams pairs may thus be understood to be a plurality of candidate positioning beam pairs representing all available combinations of receive beams of the terminal device and transmit beams of the one or more network nodes.

[0174] It is to be understood that the present disclosure is not limited to the system according to the exemplary embodiment of FIG. 3. In particular, a system according to the present disclosure may comprise further UEs and/or network nodes, for example supporting a different number of receive beams (e.g. more than 4 receive beams, e.g. 8 receive beams) and a different number of transmit beams (e.g. more than 4 transmit beams, e.g. 8, 16 or 32 transmit beams), respectively. A system according to the present disclosure may comprise at least one UE and at least one network node.

[0175] Considering system 300, UE 100 and each network node of network nodes 200-1 to 200-3 could perform an exhaustive beam sweeping to determine the respective positioning beam pair which is to be used for downlink positioning. According to this example, this however means that UE 100 may receive and measure 96 PRS to determine the positioning beam pairs which are to be used for downlink positioning. Such an exhaustive beam sweeping may result in a high latency and may be inefficient if for example about 3% of these PRS (i.e. 3 out of 96, i.e. one per network node) are of interest for downlink positioning. FIG. 4 is a flow chart 400 illustrating an exemplary embodiment of a method according to the present disclosure. The method may serve for beam selection for downlink positioning without performing an exhaustive beam sweeping. Without limiting the scope of the present disclosure, it is assumed in the following that UE 100 which is part of system 300 as depicted in FIG. 3 performs the actions of flow chart 400.

[0176] In optional action 401, UE 100 determines a first subset of candidate positioning beam pairs. The first subset of candidate positioning beam pairs is a subset of the plurality of candidate positioning beam pairs. In particular, the first subset of candidate positioning beam pairs may be a subset of the example of 96 candidate positioning beam pairs representing all available combinations of receive beams of UE 100 and transmit beams of the network nodes 200-1 to 200-3.

[0177] The UE 100 may thus determine the first subset of candidate positioning beam pairs by selecting the first subset of candidate positioning beam pairs from the plurality of candidate positioning beam pairs, i.e the 96 candidate beam pairs representing all available combinations of receive beams of UE 100 and transmit beams of the network nodes 200-1 to 200-3. To this end, UE 100 may receive assistance information from a serving network node (e.g. network node 200-1) representing the plurality of candidate positioning beam pairs or enabling UE 100 to determine the plurality of candidate positioning beam pairs.

[0178] Action 401 is only optional. Alternatively, UE 100 may for example receive positioning assistance information from serving network node (e.g. network node 200-1) representing the first subset of candidate positioning beam pairs.

[0179] An exemplary embodiment of action 401 is flow chart 500 of FIG. 5 as described below in more detail.

[0180] In action 402, UE 100 measures, for each candidate positioning beam pair of the first subset of candidate positioning beam pairs, at least one respective PRS metric.

[0181] As disclosed above, a respective PRS metric measured for a respective candidate positioning beam pair may be understood to be a metric which qualitatively or quantitatively indicates a perceived reception quality of the corresponding PRS transmitted and received by use of the respective combination of transmit beam and receive beam represented by this respective candidate positioning beam pair. Accordingly, measuring a respective PRS metric for a respective candidate positioning beam pair may be understood to mean determining the respective PRS metric based on a respective PRS received by UE 100 such that the measured respective PRS metric qualitatively or quantitatively indicates a perceived reception quality of the respective PRS. In other words, each PRS metric measured in action 402 may indicate a respective reception quality that was perceived by UE 100 when it received the respective PRS.

[0182] Without limiting the present disclosure, it is assumed in the following that UE 100 measures, for each candidate positioning beam pair of the first subset of candidate positioning beam pairs, a respective Reference Signal Received Power (RSRP) in action 402. The RSRP measured for a respective candidate positioning beam pair representing the respective receive beam r of UE 100 and the respective transmit beam l of network node x of network nodes 200-1 to 200-3 at measurement time t is denoted by mRSRP (r, l, x) or mRSRP (r, l, x, t) in the following.

[0183] In action 403, UE 100 estimates, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on the measured positioning reference signal metrics.

[0184] The second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of the plurality of candidate positioning beam pairs (i.e. the 96 candidate positioning beam pairs representing all available combinations of receive beams of UE 100 and transmit beams of the network nodes 200-1 to 200-3) that is/are not part of the first subset of candidate positioning beam pails. In the following, it is assumed that the first subset of candidate positioning beam pairs and the second subset of candidate positioning beam pairs are complements and that the union of the first subset of candidate positioning beam pairs and the second subset of candidate positioning beam pairs equals the plurality of candidate positioning beam pairs. In other words, the second subset of candidate positioning beam pairs comprises all candidate positioning beam pairs of the plurality of candidate positioning beam pairs (i.e. the 96 candidate positioning beam pairs representing all available combinations of receive beams of UE 100 and transmit beams of the network nodes 200-1 to 200-3) that are not part of the first subset of candidate positioning beam pairs.

[0185] As disclosed above, a respective PRS metric estimated for a respective candidate positioning beam pair may be understood to be a metric which qualitatively or quantitatively indicates an expected reception quality of the corresponding PRS transmitted and received by use of the respective combination of transmit beam and receive beam represented by this respective candidate positioning beam pair. Accordingly, estimating a respective PRS metric for a respective candidate positioning beam pair may be understood to mean determining the respective PRS metric based on PRS metrics measured for other candidate positioning beam pairs such that the estimated respective PRS metric qualitatively or quantitatively indicates an expected reception quality of the respective PRS. In other words, each PRS metric estimated in action 403 may indicate a respective reception quality that is expected to be perceived by UE 100 when it receives the respective PRS.

[0186] Without limiting the present disclosure, it is assumed in the following that UE 100 estimates, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, a respective Reference Signal Received Power (RSRP) in action 403. The RSRP estimated for a respective candidate positioning beam pair representing the respective receive beam r of UE 100 and the respective transmit beam l of network node x of network nodes 200-1 to 200-3 at time t is denoted by eRSRP (r, l, x) or eRSRP (r, l, x, t) in the following.

[0187] Exemplary embodiments of action 403 are flow charts 600 to XXX of FIGS. 6 to X as described below in more detail.

[0188] In action 404, UE 100 selects, at least partially based on the measured positioning reference signal metrics and the estimated positioning reference signal metrics, one or more candidate positioning beam pairs from a plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

[0189] A positioning beam pair for downlink positioning may be understood to represent a combination of a transmit beam and a receive beam that is to be used for transmitting and receiving a PRS for downlink positioning.

[0190] As disclosed above, selecting, at least partially based on the measured PRS metrics and the estimated PRS metrics, one or more candidate positioning beam pairs from the plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning may be understood to mean that, for each network node of the one or more network nodes, one respective candidate positioning beam pair from the plurality of candidate positioning beam pairs is selected as respective positioning beam pair for downlink positioning. For example, for each network node of the one or more network nodes, the respective candidate positioning beam pair from the plurality of candidate positioning beam pairs for which the PRS metric indicating the best reception quality of a PRS transmitted by the respective network node is estimated or measured.

[0191] In particular, UE 100 may select, for each network node x of the network nodes 200-1 to 200-3, the one respective candidate positioning beam pair (r, l, x) from the plurality of candidate positioning beam pairs representing receive beam r and transmit beam l of network node x for which the maximum RSRP has been measured in action 402 and estimated in action 403, respectively (e.g. positioning beam pair (r, l, x)=argmax. (mRSPR, eRSRP)).

[0192] FIG. 5 is a flow chart 500 illustrating an exemplary embodiment of action 401 of flow chart 400 of FIG. 4 relating to determining the first subset of candidate positioning beam pairs by UE 100.

[0193] Actions of flow chart 500 are performed iteratively for each network node of network nodes 200-1 to 200-3.

[0194] In action 501, xis set to 1 such that actions 502 and 503 of the first iteration are performed for network node x=1, i.e. network node 200-1.

[0195] In action 502,

[00023]$\frac{R}{a}$

receive beams of UE 100 and

[00024]$\frac{L}{b}$

transmit beams of the respective network node x (i.e. network node 200-1 in the first iteration) are selected, where R is the total number of receive beams of UE 100 and L is the total number of transmit beams of the respective network node x. The selected

[00025]$\frac{R}{a}$

receive beams of UE 100 form the set S.sub.UE; and the selected

[00026]$\frac{L}{b}$

transmit beams of the respective network node x form the set S.sub.x.

[0196] Parameters a and b may be predetermined or selected as a tradeoff between accuracy and latency. Selecting a=1 and b=1 results in an exhaustive beam sweeping as known from the prior art such that parameters a and b are greater than 1 according to the present disclosure, for example a=2 or a=4, b=4 or b=8. For example, UE 100 may for example receive positioning assistance information from the serving network node (e.g. network node 200-1) representing parameters a and b.

[0197] As disclosed above in more detail, the

[00027]$\frac{R}{a}$

receive beams of UE 100 (i.e. forming S.sub.UE) and the respective

[00028]$\frac{L}{b}$

transmit beams of the respective network node (i.e. forming S.sub.x) may be selected by at least one of: [0198] (i) by randomly selecting their beam indexes from the receive beam set/grid Gr={1, . . . R} and the transmit beam set/grid Gl={1, . . . , L}; and/or [0199] (ii) by selecting equally spaced receive beam indexes from the receive beam set/grid Gr={1, . . . , R}

[00029]$\left(i.e.S_{\mathrm{UE}}=1:\frac{R}{a}:R,\right)$  and/or selecting equally spaced transmit beam indexes from the transmit beam set/grid Gl={1, . . . , L}

[00030]$\left(i.e.S_x=1:\frac{L}{b}:L\right);$  and/or [0200] (iii) based on spatial distances between beams such that the spatial distance (i) between any of the selected

[00031]$\frac{R}{a}$  receive beams and between any of the selected

[00032]$\frac{L}{b}$  transmit beams does not exceed a (e.g. predetermined) spatial distance threshold or is at least substantially equal (e.g. within a (e.g. predetermined) spatial distance range).

[0201] It is to be understood that the present disclosure is not limited to these selection schemes.

[0202] In action 503, all available combinations of the

[00033]$\frac{R}{a}$

receive beams of UE 100 (i.e. forming S.sub.UE) and the respective

[00034]$\frac{L}{b}$

transmit beams of the respective network node (i.e. forming S.sub.x) as selected in action 502 added as candidate positioning beam pairs to the first subset of candidate positioning beam pairs.

[0203] In action 504, it is checked whether x is equal to X (i.e. X=3). If x is equal to x, this means that actions 502 and 503 have been performed for each network node of network nodes 200-1 to 200-3 and flow chart 500 is terminated. Otherwise, x is incremented by one (action 505) and actions 502 and 503 of the next iteration are performed for the next network node, for example actions 502 and 503 of the second iteration are performed for network node 200-2 (i.e. x=2).

[0204] FIG. 6 is a flow chart 600 illustrating an exemplary embodiment of action 403 of flow chart 400 of FIG. 4 relating to estimating the PRS metrics.

[0205] In action 601, UE 100 calculates a respective weighted average of measured PRS metrics calculating for each candidate positioning beam pair of the second subset of candidate positioning beam pairs.

[0206] As disclosed above, it is assumed that UE 100 measures, for each candidate positioning beam pair of the first subset of candidate positioning beam pairs, a respective Reference Signal Received Power (RSRP) in action 402. In the following, it is thus assumed that a respective weighted average of mRSRPs is calculated for each candidate positioning beam pair of the second subset of candidate positioning beam pairs in action 601.

[0207] The respective weighted average of mRSRPs calculated for a respective candidate positioning beam pair of the second subset of candidate positioning beam pairs may be determined or considered as respective estimated RSRP of the respective candidate positioning beam pair. The RSRP estimated for a respective candidate positioning beam pair representing the respective receive beam r of UE 100 and the respective transmit beam l of network node x of network nodes 200-1 to 200-3 at measurement time t is denoted by eRSRP (r, l, x) or eRSRP (r, l, x, t) in the following.

[0208] As disclosed above, the weighted average(s) could be calculated “memoryless”. To estimate eRSRP (r, l, x, t), only mRSRPs that were most recently measured in action 402 (i.e. measured at measurement t) are for example used for calculating the respective weighted average according to the “memoryless” approach.

[0209] Considering the “memoryless” approach, four mRSRPs measured for respective four neighboring candidate positioning beams of the first subset of candidate positioning beam pairs at measurement time t could for example be used for calculating the respective weighted average of mRSRPs for a respective candidate positioning beam pair of the second subset of candidate positioning beam pairs. The respective four neighboring candidate positioning beams of the respective candidate positioning beam pair of the second subset of candidate positioning beam pairs representing a respective receive beam r of the terminal device and a respective transmit beam l of a respective network node x of network nodes 200-1 to 200-3 comprises a first candidate positioning beam representing the receive beam having the closest lower index and a transmit beam of network node x (e.g. r−n, l, x), a second respective candidate positioning beam pair representing the receive beam having the closest higher index and a transmit beam of the same respective network node (e.g. r+m, l, x), a third respective candidate positioning beam pair representing the transmit beam of the same respective network node having the closest lower index (e.g. r, l−i, x) and a fourth respective candidate positioning beam pair representing the transmit beam of the same respective network node having the closest higher index (e.g. r, 1+j, x). In the following, it is thus assumed that mRSRP(r−n, l, x, t), mRSRP(r+m, l, x, t), mRSRP(r, l−i, x, t) and mRSRP(r, l+j, x, t) are used for calculating the weighted average to estimate eRSRP(r, l, x, t) according to the memoryless approach.

[0210] As disclosed above, the weights used for calculating the weighted averages may be determined at least partially based in spatial distances between beams. According to the embodiment of FIG. 3, receive beams and transmit beams which are adjacent to each other in the spatial domain have subsequent indexes in receive beam set/grid Gr and the transmit beam set/grid Gl, respectively. Thus, the amount of the differences between beam indexes may be used to determine the weights. Considering the above example, the weights w may for example be determined as follows:

[00035]$w=\frac{1}{n}\mathrm{for}\mathrm{mRSRP}\left(r-n,l,x,t\right),w=\frac{1}{m}\mathrm{for}\mathrm{mRSRP}\left(r+m,l,x,t\right),w=\frac{1}{i}\mathrm{for}\mathrm{mRSRP}\left(r,l-i,x,t\right)\mathrm{and}w=\frac{1}{j}\mathrm{for}\mathrm{mRSRP}\left(r,l+j,x,t\right).$

The weighted average to estimate eRSRP(r, l, x, t) may then be calculated as follows: x, t).

eRSRP(r,l,x,t)=Σ.sub.r′≠r,l′≠l,t′<t.sup.R,L,t-Dw(r′,l′,t′)*mRSRP2(r′,l′,t′)

where D is the length of the memory, selected by implementation.

[0211] Alternatively, the weights could be determined for the “memoryless” approach as shown in FIG. 7 illustrating a exemplary embodiment of determining weights.

[0212] Actions of flow chart 700 are performed iteratively for each network node of network nodes 200-1 to 200-3.

[0213] In action 701, x is set to 1 such that actions 702 to 705 of the first iteration are performed for network node x=1, i.e. network node 200-1.

[0214] In action 702 a first subset mS1 comprising mRSRPs and a second subset mS2 comprising mRSRPs is selected. The first subset mS1 and the second subset mS2 only comprises mRSRPs which were measured for candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a transmit beam of network node x at measurement time t. Moreover, the first subset mS1 may comprise different mRSRPs than the second subset mS2. For example the first subset mS1 and the second subset mS2 are compelements, and the union of the first subset mS1 and the second subset mS2 equals all mRSRPs which were measured for candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a transmit beam of network node x at measurement time t.

[0215] In action 703, weights are calculated based on a relation between mS1 and mS2.

[0216] To this end, the relation between mS1 and mS2 may be defined as follows:

(r,l,x,t)=Σ.sub.r′≠r,l′≠l.sup.R,Lw(r′,l′,x,t)*mS2(r′,l′,x,t).

[0217] Based on this definition, the weights may then be calculated as follows:

ŵ(r′,l′,x,t)=arg min(|mS1−).

[0218] In action 704, the model error E is calculated based on the weights obtained in action 703.

[0219] Without limiting the present disclosure, the error model E may be calculated as follows:

E=∥mS1−Σ.sub.r′≠r,l′≠lŵ(r′,l′,x,t)*mS2(r′,l′,x,t)∥.sup.2.)

[0220] If the calculated model error E is greater than a predetermined error threshold, the flow chart returns to action 702 where different subsets mS1 and mS2 as in the previous iteration(s) are selected. Otherwise, the weights obtained in action 703 are used for calculating the weighted average(s).

[0221] In action 706, it is checked whether x is equal to X (i.e. X=3). If x is equal to x, this means that actions 702 to 705 have been performed for each network node of network nodes 200-1 to 200-3 and flow chart 700 is terminated. Otherwise, x is incremented by one (action 707) and actions 702 to 705 of the next iteration are performed for the next network node.

[0222] Alternatively, the weighted average(s) could be calculated with a “variable memory size” as disclosed above in more detail. To this end, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, only mRSRPs that were measured within a measurement time window (i.e. “variable memory size” approach) may be available (e.g. can be used) for calculating the respective weighted average of mRSRPs. The length of the measurement time window is denoted by T in the following. It may be determined (e.g. set) to correspond to or to be less than the coherence time, i.e. ˜4/fD, where fD is the maximum Doppler shift. To estimate eRSRP (r, l, x, t), only mRSRPs that were measured within the time window t-T are for example used for calculating the respective weighted average according to the “variable memory size” approach.

[0223] FIG. 8 is a flow chart 600 illustrating another exemplary embodiment of action 403 of flow chart 400 of FIG. 4 relating to estimating the PRS metrics.

[0224] In action 801, UE 100 generates, for each network node of network nodes 200-1 to 200-3, a respective measured PRS metric matrix.

[0225] As disclosed above, it is assumed that UE 100 measures, for each candidate positioning beam pair of the first subset of candidate positioning beam pairs, a respective Reference Signal Received Power (RSRP) in action 402. In the following, it is thus assumed that the PRS metrics generated in action 801 are generated based on the mRSRPs measured in action 403. In particular, for each network node x of network nodes 200-1 to 200-3, a respective RSRP matrix M.sub.X E R.sup.R×L may be generated based on the mRSRPs measured for candidate positioning beam pairs representing a transmit beam of the network node x as follows:

[00036]$\begin{array}{cc}{M}_x\left(r,l,t\right)=\{\begin{array}{c}\mathrm{mRSRP}\left(r,l,x,t\right),r\in S_{\mathrm{UE}},l\in S_x\\ -\infty ,\mathrm{otherwise}\end{array},x=1:X.& \left(1\right)\end{array}$

[0226] For example, if there are the following mRSRPs for network node x at measurement time t: mRSRP (1, 3, x, t)=−61 dBm, mRSRP (1, 7, x, t)=−62 dBm, mRSRP (2, 4, x, t)=−63 dBm, mRSRP (2, 8, x, t)=−89 dBm, mRSRP (3, 1, x, t)=−91 dBm, mRSRP (3, 5, x, t)=−78 dBm, mRSRP (4, 2, x, t)=−91 dBm, the RSRP matrix M.sub.x(r, l, t) 900 as illustrated in FIG. 9 may be obtained in action 801. It is to be understood that the receive beam indexes r and the transmit beam indexes 1 are only included for illustrative purposes in FIG. 9 and that they are not part of the RSRP matrix M.sub.x(r, l, t) 900.

[0227] As a result of action 801, RSRP matrices M.sub.1, M.sub.2 and M.sub.3 are obtained.

[0228] In action 802, the RSRP matrices M.sub.1, M.sub.2 and M.sub.3 generated in action 801 are received by a machine learning algorithm as input. The machine learning algorithm estimates the missing matrix elements (i.e. the matrix elements having the value—∞) of the RSRP matrices M.sub.1, M.sub.2 and M.sub.3 (action 803) and provides, for each of the RSRP matrices M.sub.1, M.sub.2 and M.sub.3, a respective densified RSRP matrix O.sub.x where all matrix elements of the respective input matrix M.sub.x having the value—∞ (i.e. the missing matrix elements) are replaced with an estimated RSRP as output (action 804).

[0229] An example of a densified RSRP matrix O.sub.x(r, l, t)1000 where all matrix elements of the respective input matrix M.sub.x(r, l, t) 900 of FIG. 9 having the value—∞ (i.e. the missing matrix elements) are replaced with an estimated RSRP is shown in FIG. 10. Densified RSRP matrix O.sub.x(r, l, t) 1000 may be obtained in action 803 as output of the machine learning algorithm. It is to be understood that the receive beam indexes r and the transmit beam indexes 1 are only included for illustrative purposes in FIG. 9 and that they are not part of the densified RSRP matrix O.sub.x(r, l, t)1000.

[0230] As disclosed above in more details, the machine learning algorithm may be implemented as one of: [0231] (i) a deep neural network (DNN), or [0232] (ii) a convolutional neural network (CNN), or [0233] (iii) a generative adversial network (GAN), or [0234] (iv) a long-short term memory (LSTM), or [0235] (v) a time delay neural network (TDNN).

[0236] In an exemplary DNN implementation, the RSRP matrices M.sub.1, M.sub.2 and M.sub.3 received as input may be processed in parallel to estimate the RSRPs at the position of the missing matrix elements (i.e., the matrix elements having the value—∞). Accordingly, each network node of the network nodes 200-1 to 200-3 may be treated independently and no information may be shared between these different processing channels. In an exemplary CNN implementation, the RSRP matrices M.sub.1, M.sub.2 and M.sub.3 received as input may be processed jointly to estimate the RSRPs at the position of the missing matrix elements (i.e. the matrix elements having the value—∞). For example, the CNN will learn to exploit the distance between any pair of network nodes or their degree of collinearity with the UE 100.

[0237] The exemplary DNN and CNN implementations are examples of implementations where the machine learning algorithm does (e.g. only) receive one respective PRS metric matrix M.sub.x or M.sub.x(t) for each network node x of the network nodes 200-1 to 200-3 (x=1:X). They may employ a Rectified Linear Unit (ReLu) output activation function or a Swish output activation function.

[0238] In an exemplary LSTM or TDNN implementation, the LSTM or TDNN receives, for each network node x of the network nodes 200-1 to 200-3 (x=1:X), a respective set of at least two PRS metric matrices M.sub.x(t1) and M.sub.x(t2) and respective associated UE 100 specific orientation matrices A.sub.x(t1) and A.sub.x(t2) as input. The respective UE 100 specific orientation matrices A.sub.x(t1) and A.sub.x(t2) could be obtained from or determined based on one or more sensors (e.g. gyroscope) of UE 100. Additionally receiving the respective associated UE 100 specific orientation matrices A.sub.x(t1) and A.sub.x(t2) may enable (e.g. the machine learning algorithm) to adapt RSRP matrices M.sub.x(t1) and M.sub.x(t2) which comprise RSRP measured with different UE 100 orientations at different measurement times t1 and t2.

[0239] All RSRP matrices M.sub.x(t) having a time stamp t within a measurement time window are part of the respective set. The length of the measurement time window may be variable, for example it may be determined to be proportional to the speed of UE 100. To give a non limiting example, the length of the measurement time window may be determined (e.g. set) to correspond to or to be less than the coherence time, i.e. ˜4/fD, where fD is the maximum Doppler shift.

[0240] In the following, the complexity for an CNN implementation is exemplary analyzed based on computational costs. Obtaining a respective densified RSRP matrix O.sub.x for each network node x of the network nodes 200-1 to 200-3, consists of performing N transformations f.sub.n(W.sub.nx.sub.n+b.sub.n), where N is the number of layers of the DNN, f.sub.n is the activation function at the output of layer n, x.sub.n, W.sub.n and b.sub.n are the inputs, weight matrix and biases of the n-th layer. For convolutional kernels, the complexity is 2HB(C.sub.in J.sup.2+1)C.sub.out, where H, B and C.sub.in are the input tensor dimensions, J is the core width and C.sub.out is the number of output channels. For an implementation with two layers, the operational cost is

[00037]$\frac{R}{a}\frac{L}{b}\left({\mathrm{XJ}}_1^2+1\right)X_i+\frac{R}{a}\frac{L}{b}\left(X_i{J}_2^2+1\right)X\approx \frac{R}{a}\frac{L}{b}\left(J_2^2+J_1^2\right){\mathrm{XX}}_i\mathrm{FLOPs}.$

If we select J.sub.1=J.sub.2=X.sub.i=X, the maximum cost of densifying the RSRP matrix is

[00038]$\stackrel{\_}{\stackrel{\_}{C_1}}=\frac{2R}{a}\frac{L}{b}{X}^4\mathrm{FLOPs}.$

[0241] The approximate cost for performing an exhaustive beam sweep are C.sub.1=2R.Math.X.Math.P.Math.L.Math.F.Math.log.sub.2 F FLOPs, where F is the FFT size and P is the number of consecutive PRS symbols e.g P∈{1, 2, 4, 6, 8, 12}.

[0242] The CNN implementation comes at a cost

[00039]$C_t=\stackrel{\_}{\stackrel{\_}{C_1}}+\stackrel{\_}{C_1}=\frac{2R}{a}\frac{L}{b}{X}^4+2\frac{C_1}{\mathrm{ab}}=\frac{C_1}{\mathrm{ab}}\left(2+\frac{X^3}{\mathrm{PF}{\log }_{2}F}\right)={\mathrm{pC}}_1\mathrm{FLOPs},$$p=\frac{\left(2+\frac{X^3}{\mathrm{PF}{\log }_{2}F}\right)}{\mathrm{ab}},$

where C.sub.1 is the approximate cost of actions 401 to 402 and C.sub.1 is the approximate cost of action 403. In general, for more hidden layers N>2, the fraction p can be approximated by

[00040]$p=\frac{\left(2+\frac{{\mathrm{NX}}^3}{2{\mathrm{PF}}_{2}F}\right)}{\mathrm{ab}}.$

[0243] FIG. 11 is a schematic illustration of examples of tangible and non-transitory computer-readable storage media according to the present disclosure that may for example be used to implement memory 102 of FIG. 1 or memory 202 of FIG. 2. To this end, FIG. 10 displays a flash memory 1100, which may for example be soldered or bonded to a printed circuit board, a solid-state drive 1101 comprising a plurality of memory chips (e.g. Flash memory chips), a magnetic hard drive 1102, a Secure Digital (SD) card 1103, a Universal Serial Bus (USB) memory stick 1104, an optical storage medium 1105 (such as for example a CD-ROM or DVD) and a magnetic storage medium 1106.

[0244] Moreover, the following exemplary embodiments are to be understood to be disclosed:

Exemplary Embodiment 1

[0245] A terminal device comprising at least one processor and at least one memory including computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the terminal device at least to perform: [0246] measuring, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein the first subset of candidate positioning beam pairs is a subset of a plurality of candidate positioning beam pairs, wherein each candidate positioning beam pair of the plurality of candidate positioning beam pairs represents a combination of a respective receive beam of the terminal device and a respective transmit beam of a respective network node of one or more network nodes; [0247] estimating, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on the measured positioning reference signal metrics, wherein the second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of the plurality of candidate positioning beam pairs that is/are not part of the first subset of candidate positioning beam pairs; [0248] selecting, at least partially based on the measured positioning reference signal metrics and the estimated positioning reference signal metrics, one or more candidate positioning beam pairs from the plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

Exemplary Embodiment 2

[0249] A terminal device according to embodiment 1, wherein the plurality of candidate positioning beam pairs represents (e.g. all) available combinations of receive beams of the terminal device and transmit beams of the one or more network nodes.

Exemplary Embodiment 3

[0250] A terminal device according to any of embodiments 1 and 2, the at least one memory and the computer program code further configured to, with the at least one processor, cause the terminal device to perform: [0251] determining the first subset of candidate positioning beam pairs.

Exemplary Embodiment 4

[0252] A terminal device according to any of embodiments 1 to 3, wherein the first subset of candidate positioning beam pairs is determined by at least one of: [0253] (i) selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs; or [0254] (ii) randomly selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs; or [0255] (iii) selecting, for each network node of the one or more network nodes, candidate positioning beam pairs from the plurality of candidate positioning beam pairs representing at least substantially equally spaced receive beams of the terminal device and/or at least substantially equally spaced transmit beams of the respective network node; or [0256] (iv) selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs based on at least one of (i) beam widths of the beams, or (ii) spatial diversity of the beams, or (iii) time.

Exemplary Embodiment 5

[0257] A terminal device according to any of embodiments 1 to 4, wherein the at least one respective positioning reference signal metric is or represents at least one of: (i) a Received Signal Strength Indicator, RSSI; or [0258] (ii) a Reference Signal Received Power, RSRP; or [0259] (iii) a Signal-to-Noise Ratio, SNR; or [0260] (iv) a power of the strongest channel tap; or [0261] (v) a Carrier-to-Interference Ratio, CIR.

Exemplary Embodiment 6

[0262] A terminal device according to any of embodiments 1 to 5, wherein the at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of the second subset of candidate positioning beam pairs at least partially based on the measured positioning reference signal metrics by: [0263] calculating, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, a respective weighted average of the measured positioning reference signal metrics.

Exemplary Embodiment 7

[0264] A terminal device according to embodiment 6, wherein at least two positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs are used for calculating the respective weighted average of measured positioning reference signal metrics, and/or wherein, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, (e.g. only) positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node like the respective candidate positioning beam pair of the second subset of candidate positioning beam pairs are used for calculating the respective weighted average of measured positioning reference signal metrics, and/or

[0265] wherein, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, (e.g. only) positioning reference signal metrics that were (i) most recently measured or (ii) measured within a measurement time window are used for calculating the respective weighted average of measured positioning reference signal metrics

Exemplary Embodiment 8

[0266] A terminal device according to any of embodiments 6 and 7, wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of the second subset of candidate positioning beam pairs, the respective weighted average of measured positioning reference signal metrics is determined at least partially based on at least one of [0267] (i) a spatial distance between, or [0268] (ii) beam widths of, or [0269] (iii) antenna correlation information associated with

[0270] at least one beam represented by the respective candidate positioning beam pair of the second subset of candidate positioning beam pairs and at least one beam represented by the respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs, and/or

[0271] wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of the second subset of candidate positioning beam pairs, the respective weighted average of measured positioning reference signal metrics is determined at least partially based on a time of measurement of at least one measured positioning reference signal metrics of the measured positioning reference signal metrics.

Exemplary Embodiment 9

[0272] A terminal device according to any of embodiments 1 to 5, wherein the at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of the second subset of candidate positioning beam pairs at least partially based on the measured positioning reference signal metrics by using a machine learning algorithm.

Exemplary Embodiment 10

[0273] A terminal device according to embodiment 9, wherein the machine learning algorithm is implemented as one of: [0274] (i) a deep neural network, or [0275] (ii) a convolutional neural network, or [0276] (iii) a generative adversial network, or [0277] (iv) a long-short term memory, or [0278] (v) a time delay neural network.

Exemplary Embodiment 11

[0279] A terminal device according to any of embodiments 9 and 10, wherein the machine learning algorithm receives one or more positioning reference signal metric matrices as input, wherein the one or more positioning reference signal metric matrices comprise(s) the measured positioning reference signal metrics as matrix elements.

Exemplary Embodiment 12

[0280] A terminal device according to embodiment 11, wherein each of the positioning reference signal metric matrices (e.g. only) comprises respective positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node.

Exemplary Embodiment 13

[0281] A terminal device according to any of embodiments 11 and 12, wherein the measured positioning reference signal metrics are positioned in the one or more positioning reference signal metric matrices at least partially based on spatial distances between beams represented by the candidate positioning beam pairs of the first subset of candidate positioning beam pairs.

Exemplary Embodiment 14

[0282] A method, performed at least by a terminal device, the method comprising: [0283] measuring, for each candidate positioning beam pair of a first subset of candidate positioning beam pairs, at least one respective positioning reference signal metric, wherein the first subset of candidate positioning beam pairs is a subset of a plurality of candidate positioning beam pairs, wherein each candidate positioning beam pair of the plurality of candidate positioning beam pairs represents a combination of a respective receive beam of the terminal device and a respective transmit beam of a respective network node of one or more network nodes; [0284] estimating, for each candidate positioning beam pair of a second subset of candidate positioning beam pairs, at least one respective positioning reference signal metric at least partially based on the measured positioning reference signal metrics, wherein the second subset of candidate positioning beam pairs comprises one or more candidate positioning beam pairs of the plurality of candidate positioning beam pairs that is/are not part of the first subset of candidate positioning beam pairs; [0285] selecting, at least partially based on the measured positioning reference signal metrics and the estimated positioning reference signal metrics, one or more candidate positioning beam pairs from the plurality of candidate positioning beam pairs as positioning beam pairs for downlink positioning.

Exemplary Embodiment 15

[0286] A method according to embodiment 14, wherein the plurality of candidate positioning beam pairs represents (e.g. all) available combinations of receive beams of the terminal device and transmit beams of the one or more network nodes.

Exemplary Embodiment 16

[0287] A method according to any of embodiments 14 and 15 further comprising: [0288] determining the first subset of candidate positioning beam pairs.

Exemplary Embodiment 17

[0289] A method according to any of embodiments 14 to 16, wherein the first subset of candidate positioning beam pairs is determined by at least one of: [0290] (i) selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs; or [0291] (ii) randomly selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs; or [0292] (iii) selecting, for each network node of the one or more network nodes, candidate positioning beam pairs from the plurality of candidate positioning beam pairs representing at least substantially equally spaced receive beams of the terminal device and/or at least substantially equally spaced transmit beams of the respective network node; or [0293] (iv) selecting candidate positioning beam pairs from the plurality of candidate positioning beam pairs based on at least one of (i) beam widths of the beams, or (ii) spatial diversity of the beams, or (iii) time.

Exemplary Embodiment 18

[0294] A method according to any of embodiments 14 to 17, wherein the at least one respective positioning reference signal metric is or represents at least one of: [0295] (i) a Received Signal Strength Indicator, RSSI; or [0296] (ii) a Reference Signal Received Power, RSRP; or [0297] (iii) a Signal-to-Noise Ratio, SNR; or [0298] (iv) a power of the strongest channel tap; or [0299] (v) a Carrier-to-Interference Ratio, CIR.

Exemplary Embodiment 19

[0300] A method according to any of embodiments 14 to 18, wherein the at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of the second subset of candidate positioning beam pairs at least partially based on the measured positioning reference signal metrics by: [0301] calculating, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, a respective weighted average of the measured positioning reference signal metrics.

Exemplary Embodiment 20

[0302] A method according to embodiment 19, wherein at least two positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs are used for calculating the respective weighted average of measured positioning reference signal metrics, and/or wherein, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, (e.g. only) positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node like the respective candidate positioning beam pair of the second subset of candidate positioning beam pairs are used for calculating the respective weighted average of measured positioning reference signal metrics, and/or

[0303] wherein, for each candidate positioning beam pair of the second subset of candidate positioning beam pairs, (e.g. only) positioning reference signal metrics that were (i) most recently measured or (ii) measured within a measurement time window are used for calculating the respective weighted average of measured positioning reference signal metrics

Exemplary Embodiment 21

[0304] A method according to any of embodiments 19 and 20, wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of the second subset of candidate positioning beam pairs, the respective weighted average of measured positioning reference signal metrics is determined at least partially based on at least one of [0305] (i) a spatial distance between, or [0306] (ii) beam widths of, or [0307] (iii) antenna correlation information associated with

[0308] at least one beam represented by the respective candidate positioning beam pair of the second subset of candidate positioning beam pairs and at least one beam represented by the respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs, and/or wherein at least one weight used for calculating, for at least one respective candidate positioning beam pair of the second subset of candidate positioning beam pairs, the respective weighted average of measured positioning reference signal metrics is determined at least partially based on a time of measurement of at least one measured positioning reference signal metrics of the measured positioning reference signal metrics.

Exemplary Embodiment 22

[0309] A method according to any of embodiments 14 to 18, wherein the at least one respective positioning reference signal metric is estimated for each candidate positioning beam pair of the second subset of candidate positioning beam pairs at least partially based on the measured positioning reference signal metrics by using a machine learning algorithm.

Exemplary Embodiment 23

[0310] A method according to embodiment 22, wherein the machine learning algorithm is implemented as one of: [0311] (i) a deep neural network, or [0312] (ii) a convolutional neural network, or [0313] (iii) a generative adversial network, or [0314] (iv) a long-short term memory, or [0315] (v) a time delay neural network.

Exemplary Embodiment 24

[0316] A method according to any of embodiments 22 and 23, wherein the machine learning algorithm receives one or more positioning reference signal metric matrices as input, wherein the one or more positioning reference signal metric matrices comprise(s) the measured positioning reference signal metrics as matrix elements.

Exemplary Embodiment 25

[0317] A method according to embodiment 24, wherein each of the positioning reference signal metric matrices (e.g. only) comprises respective positioning reference signal metrics measured for respective candidate positioning beam pairs of the first subset of candidate positioning beam pairs representing a respective transmit beam of the same respective network node.

Exemplary Embodiment 26

[0318] A method according to any of embodiments 24 and 25, wherein the measured positioning reference signal metrics are positioned in the one or more positioning reference signal metric matrices at least partially based on spatial distances between beams represented by the candidate positioning beam pairs of the first subset of candidate positioning beam pairs.

Exemplary Embodiment 27

[0319] A computer-readable storage medium comprising program instructions of a computer program code stored thereon for causing an apparatus to perform a method according to any of embodiments 14 to 26.

Exemplary Embodiment 28

[0320] A computer program code, the computer program code when executed by a processor of an apparatus causing the apparatus to perform a method according to any of embodiments 14 to 26.

[0321] Any presented connection in the described embodiments is to be understood in a way that the involved components are operationally coupled. Thus, the connections can be direct or indirect with any number or combination of intervening elements, and there may be merely a functional relationship between the components.

[0322] Further, as used in this text, the term ‘circuitry’ refers to any of the following: [0323] (a) hardware-only circuit implementations (such as implementations in only analog and/or digital circuitry) [0324] (b) combinations of circuits and software (and/or firmware), such as: (i) to a combination of processor(s) or (ii) to sections of processor(s)/software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone, to perform various functions) and [0325] (c) to circuits, such as a microprocessor(s) or a section of a microprocessor(s), that require software or firmware for operation, even if the software or firmware is not physically present.

[0326] This definition of ‘circuitry’ applies to all uses of this term in this text, including in any claims. As a further example, as used in this text, the term ‘circuitry’ also covers an implementation of merely a processor (or multiple processors) or section of a processor and its (or their) accompanying software and/or firmware. The term ‘circuitry’ also covers, for example, a baseband integrated circuit or applications processor integrated circuit for a mobile phone.

[0327] Any of the processors mentioned in this text, in particular but not limited to processors 101 and 201 of FIGS. 1 and 2, could be a processor of any suitable type. Any processor may comprise but is not limited to one or more microprocessors, one or more processor(s) with accompanying digital signal processor(s), one or more processor(s) without accompanying digital signal processor(s), one or more special-purpose computer chips, one or more field-programmable gate arrays (FPGAS), one or more controllers, one or more application-specific integrated circuits (ASICS), or one or more computer(s). The relevant structure/hardware has been programmed in such a way to carry out the described function.

[0328] Moreover, any of the actions or steps described or illustrated herein may be implemented using executable instructions in a general-purpose or special-purpose processor and stored on a computer-readable storage medium (e.g., disk, memory, or the like) to be executed by such a processor. References to ‘computer-readable storage medium’ should be understood to encompass specialized circuits such as FPGAs, ASICs, signal processing devices, and other devices.

[0329] Moreover, any of the actions described or illustrated herein may be implemented using executable instructions in a general-purpose or special-purpose processor and stored on a computer-readable storage medium (e.g., disk, memory, or the like) to be executed by such a processor. References to ‘computer-readable storage medium’ should be understood to encompass specialized circuits such as FPGAs, ASICs, signal processing devices, and other devices.

[0330] The wording “A, or B, or C, or a combination thereof” or “at least one of A, B and C” or “at least one of A, B and/or C” may be understood to be not exhaustive and to include at least the following: (i) A, or (ii) B, or (iii) C, or (iv) A and B, or (v) A and C, or (vi) B and C, or (vii) A and B and C.

[0331] It will be understood that the embodiments disclosed herein are only exemplary, and that any feature presented for a particular exemplary embodiment may be used with any aspect of the present disclosure on its own or in combination with any feature presented for the same or another particular exemplary embodiment and/or in combination with any other feature not mentioned. It will further be understood that any feature presented for an example embodiment in a particular category may also be used in a corresponding manner in an example embodiment of any other category.