Beam determination in deployment with network control repeaters

Abstract

According to one aspect, a terminal device is provided that may be configured to receive a message from a network device, the message including a message of an indication of a downlink reference signal resource associated with a beam allocated to a network control repeater. The terminal device may be configured to determine an optimal beam from among the beams provided by the network device and the beams provided by the network control relay based on a determination between the beams indicated in the message and the beams predicted by the machine learning model.

Description

Technical Field

[0001] Various example embodiments relate generally to the field of telecommunication systems. In particular, some example embodiments relate to a solution for determining an optimal beam in a deployment with network controlled repeaters. Background Art

[0002] In wireless communication networks, such as New Radio (NR), Network Controlled Repeaters (NCRs) can be used to extend the coverage of base stations, such as gNodeBs (gNBs). NCRs are in-band radio frequency (RF) repeaters that relay signals sent from gNBs to improve network coverage. Therefore, NCRs can be considered extensions of gNBs and have beamforming capabilities to provide access beams toward out-of-coverage areas within the gNB area.

[0003] Because NCR is transparent to user equipment (UE), the UE may not know whether it is connected to the parent gNB via NCR or using a direct link to the gNB. Therefore, the UE may use a machine learning (ML) model trained only with gNB beams and without any NCR access beams to predict the best Tx beam or best beam pair (Tx / Rx beam). This means that when the UE is served by NCR, the ML model used for beam prediction may not be able to predict the best beam identifier and / or reference signal received power (RSRP). In addition, the UE may not know the reason for the ML model failure. Summary of the Invention

[0004] This summary is provided to introduce some concepts in a simplified form that are further described below. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

[0005] When the wireless communication network includes a network repeater (e.g., a network-controlled repeater), example embodiments may provide a solution for determining an optimal beam. This benefit may be achieved by the features of the independent claims. Further implementations are provided in the dependent claims, the description, and the drawings.

[0006] According to a first aspect, a terminal device may include at least one processor and at least one memory storing instructions, which instructions, when executed by the at least one processor, cause the terminal device to perform: receiving a message from a network device, the message including: an indication of downlink reference signal resources associated with a beam allocated to a network-controlled repeater; and determining a beam transmitted by the network device and a best beam among the beams transmitted by the network-controlled repeater based on a decision between the beam indicated in the message and the beam predicted by a machine learning model.

[0007] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: running a machine learning model on beams sent by the network device to predict an optimal beam among the beams sent by the network device; measuring a reference signal received power for each beam allocated to a network-controlled repeater; and determining the beam sent by the network device and the optimal beam among the beams sent by the network-controlled repeater based on the predicted reference signal received power of the optimal beam among the beams sent by the network device and the measured reference signal received power of the beams allocated to the network-controlled repeater.

[0008] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: obtaining a network-controlled repeater-compatible machine learning model that is capable of predicting an optimal beam sent by the network-controlled repeater; and running the network-controlled repeater-compatible machine learning model for the beam sent by the network device and the beam sent by the network-controlled repeater to predict the optimal beam among the beam sent by the network device and the beam sent by the network-controlled repeater.

[0009] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: collecting measurement data related to the beam transmitted by the network controlled repeater, the measurement data including: the measured reference signal received power of each beam transmitted by the network controlled repeater; and training a network controlled repeater compatible machine learning model based on the collected measurement data and the machine learning model.

[0010] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: applying transfer learning in training a network-controlled repeater-compatible machine learning model based on the collected measurement data and the parameters of the machine learning model trained for the beam sent by the network device.

[0011] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: training a network-controlled repeater-compatible machine learning model by randomly initializing parameters.

[0012] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: sending a network-controlled repeater-compatible machine learning model to a network device.

[0013] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: receiving a network-controlled repeater-compatible machine learning model from a network device.

[0014] In an example embodiment of the first aspect, the instructions, when executed by at least one processor, cause the terminal device to perform: sending information related to capabilities of the machine learning model for beam prediction to the network device.

[0015] In an example embodiment of the first aspect, the instructions, wherein the beam comprises: a synchronization signal block beam, or a channel state information reference signal beam.

[0016] According to a second aspect, a network device may include at least one processor and at least one memory storing instructions, which, when executed by the at least one processor, cause the network device to perform: sending a message to a terminal device, the message including: an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay.

[0017] In an example embodiment of the second aspect, the instructions, when executed by at least one processor, cause the network device to perform: receiving a network-controlled repeater-compatible machine learning model from a terminal device, the network-controlled repeater-compatible machine learning model being capable of predicting an optimal beam transmitted by the network-controlled repeater.

[0018] In an example embodiment of the second aspect, the instructions, when executed by at least one processor, cause the network device to perform: sending a network-controlled repeater-compatible machine learning model to a second end device served by the network-controlled repeater.

[0019] In an example embodiment of the second aspect, the instructions, wherein the beam comprises: a synchronization signal block beam, or a channel state information reference signal beam.

[0020] According to a third aspect, a method includes: receiving, by a terminal device, a message from a network device, the message including: an indication of downlink reference signal resources associated with a beam allocated to a network-controlled repeater; and determining, by the terminal device, a beam transmitted by the network device and a best beam among the beams transmitted by the network-controlled repeater based on a decision between the beam indicated in the message and the beam predicted by a machine learning model.

[0021] In an example embodiment of the third aspect, the method further comprises: running a machine learning model on beams transmitted by the network device to predict a best beam among the beams transmitted by the network device; measuring a reference signal received power for each beam allocated to the network controlled repeater; and determining the best beam among the beams transmitted by the network device and the beams transmitted by the network controlled repeater based on the predicted reference signal received power of the best beam among the beams transmitted by the network device and the measured reference signal received power of the beams allocated to the network controlled repeater.

[0022] In an example embodiment of the third aspect, the method further includes: obtaining a network-controlled repeater-compatible machine learning model, the network-controlled repeater-compatible machine learning model being capable of predicting an optimal beam sent by the network-controlled repeater; and running the network-controlled repeater-compatible machine learning model for the beam sent by the network device and the beam sent by the network-controlled repeater to predict the optimal beam among the beam sent by the network device and the beam sent by the network-controlled repeater.

[0023] In an example embodiment of the third aspect, the method further includes: collecting measurement data related to the beams transmitted by the network-controlled repeater, the measurement data including: the measured reference signal received power of each beam transmitted by the network-controlled repeater; and training a network-controlled repeater compatible machine learning model based on the collected measurement data and the machine learning model.

[0024] In an example embodiment of the third aspect, the method further comprises applying transfer learning in training a network controlled repeater compatible machine learning model based on the collected measurement data and parameters of the machine learning model trained for the beams transmitted by the network device.

[0025] In an example embodiment of the third aspect, the method further comprises training a network controlled repeater compatible machine learning model by randomly initializing parameters.

[0026] In an example embodiment of the third aspect, the method further comprises sending the network-controlled repeater-compatible machine learning model to the network device.

[0027] In an example embodiment of the third aspect, the method further comprises receiving a network-controlled repeater-compatible machine learning model from a network device.

[0028] In an example embodiment of the third aspect, the method further comprises: sending information related to capabilities of the machine learning model for beam prediction to the network device.

[0029] In an example embodiment of the third aspect, instructions, wherein the beam comprises: a synchronization signal block beam, or a channel state information reference signal beam.

[0030] According to a fourth aspect, a method comprises sending, by a network device to a terminal device, a message comprising an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay.

[0031] In an example embodiment of the fourth aspect, the method further comprises receiving a network-controlled repeater-compatible machine learning model from the terminal device, the network-controlled repeater-compatible machine learning model being capable of predicting an optimal beam transmitted by the network-controlled repeater.

[0032] In an example embodiment of the fourth aspect, the method further comprises sending the network-controlled repeater-compatible machine learning model to a second terminal device served by the network-controlled repeater.

[0033] According to a fifth aspect, a computer program comprises instructions for causing an apparatus to perform the method of the third aspect.

[0034] According to a sixth aspect, a computer program comprises instructions for causing an apparatus to perform the method of the fourth aspect.

[0035] According to a seventh aspect, a computer-readable medium comprises a computer program comprising instructions for causing an apparatus to perform the method of the third aspect.

[0036] According to an eighth aspect, a computer-readable medium comprises a computer program comprising instructions for causing an apparatus to perform the method of the fourth aspect.

[0037] According to the ninth aspect, a device may include components for: receiving, by a terminal device, a message from a network device, the message including: an indication of a downlink reference signal resource associated with a beam allocated to a network-controlled repeater; and determining, by the terminal device, a beam sent by the network device and a best beam among the beams sent by the network-controlled repeater based on a decision between the beam indicated in the message and the beam predicted by a machine learning model.

[0038] According to a tenth aspect, an apparatus may comprise means for sending, by a network apparatus to a terminal apparatus, a message comprising an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay.

[0039] Many of the attendant features will be more readily appreciated as the same become better understood by reference to the following detailed description considered in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which are included to provide a further understanding of example embodiments and constitute a part of this specification, illustrate example embodiments and together with the description assist in understanding the example embodiments. In the drawings:

[0041] Figure 1 A system according to an example embodiment is illustrated.

[0042] Figure 2A An example of a method according to an example embodiment is illustrated.

[0043] Figure 2B An example of a method according to an example embodiment is illustrated.

[0044] Figure 3A wireless communication system according to an example embodiment is illustrated.

[0045] Figure 4 A flow chart according to an example embodiment is illustrated.

[0046] Figure 5 A wireless communication system according to an example embodiment is illustrated.

[0047] Figure 6 A flow chart according to an example embodiment is illustrated.

[0048] Figure 7 A wireless communication system according to an example embodiment is illustrated.

[0049] Figure 8 Illustrated is an example for updating an ML model using transfer learning according to example embodiments.

[0050] Figure 9 An example of an apparatus configured to practice one or more example embodiments is illustrated.

[0051] In the drawings, the same reference numerals are used to refer to the same components. DETAILED DESCRIPTION

[0052] Reference will now be made in detail to example embodiments, examples of which are illustrated in the accompanying drawings. The detailed description provided below in conjunction with the drawings is intended as a description of this example and is not intended to represent the only form in which this example may be constructed or utilized. This description sets forth the functionality of this example and the sequence of steps for constructing and operating this example. However, the same or equivalent functions and sequences may be implemented by different examples.

[0053] Figure 1 A wireless communication system according to an example embodiment is illustrated. The wireless communication system includes a base station 100 (e.g., a gNB) and a network relay (e.g., a network control relay (NCR) 102). NCR 102 is an in-band radio frequency (RF) relay that relays signals sent from gNB 100 to improve network coverage. NCR 102 is transparent to user equipment (UE) 110.

[0054] gNB 100 provides backhaul to NCR 102, which acts as an extension of gNB 100 and has beamforming capabilities to provide access beams 106A-106D that extend beyond the coverage area of ​​gNB 100. This means that in addition to managing its own Tx / Rx beams, gNB 100 will also manage access beams 106A-106D for NCR 102. NCR 102 can forward an additional set of SSBs specifically assigned to NCR 102. gNB 100 can also exchange side control link information 112 with NCR 102 to control access beams 106A-106D.

[0055] Because NCR 102 is transparent to UE 110, UE 110 may not know whether it is connected to parent gNB 100 via NCR 102 or using a direct link to gNB 100. Therefore, UE 110 may use a machine learning (ML) model trained using only gNB beams and without any NCR access beams to predict the best Tx beam or best beam pair (Tx / Rx beam). This means that when the UE is served by the NCR, the ML model used for beam prediction may not be able to predict the best beam identifier and / or reference signal received power (RSRP). In addition, the UE may not know the reason for the failure of the ML model.

[0056] Because NCR 102 is transparent to UE 110, UE 110 may not know whether it is connected to parent gNB 100 via NCR 102 or using a direct link to gNB 100. Therefore, UE 110 may use an ML model trained using only gNB beams 104A-104E and without using any access beams 106A-106D to predict the best Tx beam or best beam pair (Tx / Rx beam). This may mean that when UE 110 is served by NCR 102, the ML model used for beam prediction may not predict the best beam identifier and / or RSRP. Furthermore, UE 110 may not know the reason for the ML model's failure.

[0057] In the following, to address the above shortcomings, the current framework for beam prediction is extended to generalize the ML model operation to network architectures including NCR.

[0058] Figure 2A An example of a method according to an example embodiment is illustrated. The method may be implemented by a terminal device, such as a user equipment wirelessly connected to a network device (e.g., a base station or gNB).

[0059] At 200, a message including an indication of downlink reference signal resources associated with a beam allocated to an NCR may be received from a network device. The beam may include, for example, a synchronization signal block (SSB) beam or a channel state information reference signal (CSI-RS) beam. A synchronization signal block identifier may be associated with the corresponding beam transmitted by the NCR. The gNB may provide backhaul to the NCR, which serves as an extension of the gNB and has beamforming capabilities to provide access beams toward directions beyond coverage in the gNB area.

[0060] At 202 , a best beam among the beams transmitted by the network device and the beams transmitted by the NCR may be determined based on a decision between the beams indicated in the message and the beams predicted by the ML model.

[0061] In an example embodiment, the UE can be configured to run an ML model on beams transmitted by the gNB to predict the best beam among beams transmitted by the network device. Because the ML model was trained in a legacy network without an NCR, it provides acceptable results only for beams transmitted by the gNB. Regarding beams transmitted by the NCR, the UE can be configured to measure the RSRP of each beam allocated to the NCR. The terminal device can then be configured to determine the best beam among the beams transmitted by the network device and the beams transmitted by the network NCR based on the reference signal received power of the predicted best beam among the beams transmitted by the network device and the measured reference signal received power of the beams allocated to the NCR. This solution provides a hybrid beam reporting implementation in which the best beam is determined between the predicted beam and the measured beam. In an example embodiment, the UE can compare the RSRP of the best beam transmitted by the gNB and predicted by the ML model with the RSRP measured on the beam transmitted by the NCR. If the RSRP of the best predicted beam by the ML model is higher than the RSRP measured using the beam sent by the NCR, the UE may report the beam identifier and / or RSRP of the best beam predicted by the ML model. This may occur as the UE approaches the beam sent by the gNB. If the RSRP of the beam sent by the NCR is higher than the RSRP of the best predicted beam by the ML model, the UE may report the beam identifier and / or RSRP of the best beam sent by the NCR. This may occur as the UE approaches the beam sent by the NCR.

[0062] In an example embodiment, the UE may be configured to: obtain an NCR-compatible ML model capable of predicting the best beam transmitted by the NCR, for example, an identifier and / or RSRP corresponding to the best beam; and run the NCR-compatible ML model on the beams transmitted by the network device and the beams transmitted by the NCR to predict the best beam among the beams transmitted by the network device and the beams transmitted by the NCR. For example, the NCR-compatible ML model may be run by inputting measurements corresponding to a subset of the beams transmitted by the network device and the subset of the beams transmitted by the NCR. This may provide a solution in which the NCR-compatible ML model has already been trained elsewhere and the UE can use the ready-made ML model.

[0063] In an example embodiment, the UE may be configured to collect measurement data related to the beams transmitted by the NCR, the measurement data comprising: a measured reference signal received power for each beam transmitted by the NCR. Based on the collected measurement data and the ML model, the UE may be configured to: train the NCR-compatible ML model. For example, the UE may be configured to: train the NCR-compatible ML model using the collected measurement data as input to the ML model and using the corresponding identifier of the best beam from the NCR as output of the ML model. Since the ML model has been updated (i.e., trained) to take into account the beams transmitted by the NCR, there is no need to perform any additional RSRP measurements to determine the best beam. Instead, the best beam may be determined based solely on prediction because the NCR-compatible ML model also takes into account the beams transmitted by the NCR. In an example embodiment, the NCR-compatible ML model may be a supervised NCR-compatible ML model. In an example embodiment, alternatively, the training may be performed using semi-supervised and unsupervised learning.

[0064] In an example embodiment, the UE can be configured to apply transfer learning to train an NCR-compatible ML model based on collected measurement data and the parameters of the ML model trained for the beams transmitted by the network device. The transfer learning solution can use an original ML model that has been previously trained, and then the original ML model can be updated for the new task of predicting the best beam in a network with NCR. In another example embodiment, the NCR ML model can be trained by randomly initializing parameters.

[0065] In an example embodiment, the UE can be configured to send an NCR-compatible ML model to the gNB. The gNB can store the ML model for later use by other UEs. The UE can also attach metadata to the header of the file containing the model. Using the metadata, the gNB can identify other UEs (e.g., with the same vendor identifier) ​​that may need to use the ML model. In an example embodiment, the UE can also encrypt the ML model file to preserve proprietary designs. In this case, the gNB can only read the metadata, but the ML model itself cannot be accessed by the gNB.

[0066] In an example embodiment, the UE can be configured to receive an NCR-compatible ML model from the gNB. The gNB may have already received the NCR-compatible ML model from another UE, and the NCR-compatible ML model can then be provided to the UE. Since the ML model has already been updated (i.e., trained) to account for the beams transmitted by the NCR, no additional RSRP measurements need to be performed to determine the best beam. Instead, the best beam can be determined based solely on predictions, as the NCR-compatible ML model also accounts for the beams transmitted by the NCR.

[0067] In an example embodiment, the UE can be configured to send information related to the capabilities of the ML model used for beam prediction to the gNB. Indications of the capabilities related to the ML model (e.g., the preferred number of beams to be measured and the number of beams to be predicted) can help reach a consensus between the gNB and the UE. The ML model may not be compatible with NCR, but it can still support at least the number of Tx beams sent by the gNB.

[0068] Figure 2B An example of a method according to an example embodiment is illustrated. The method may be implemented by a network device (e.g., a base station or gNB) serving a terminal device (e.g., a UE).

[0069] At 204, the gNB may be configured to send a message to the UE, the message including, for example, an indication of downlink reference signal resources associated with a beam allocated to the NCR. The beam may include, for example, a synchronization signal block (SSB) beam or a channel state information reference signal (CSI-RS) beam. Each synchronization signal block identifier may be associated with a corresponding beam transmitted by the NCR. When providing the indication of the beam to the UE, the gNB enables the UE to correctly consider the beam transmitted by the NCR when determining the best beam among the beams transmitted by the gNB and the NCR.

[0070] In an example embodiment, the gNB may be configured to receive an NCR-compatible ML model from the UE that is capable of predicting the best beam transmitted by the NCR. For example, the NCR-compatible ML model may be capable of predicting an identifier and / or RSRP corresponding to the best beam. The gNB may then be configured to send the NCR-compatible ML model to a second terminal device served by the NCR. This may enable a solution in which the gNB sends an updated ML model to the UE that has been updated (i.e., trained) to account for the beam transmitted by the NCR, without requiring any additional RSRP measurements to be performed with the UE to determine the best beam. Instead, the best beam may be determined based solely on the prediction because the NCR-compatible ML model also accounts for the beam transmitted by the NCR.

[0071] Figure 3 A wireless communication system according to an example embodiment is illustrated. The wireless communication system includes a network node 300 (e.g., a base station or gNB) and an NCR 302. The NCR 302 is an in-band radio frequency (RF) repeater that relays signals sent from the gNB 300 to improve network coverage. The NCR 302 is transparent to user equipment (UE) 108, 110.

[0072] The gNB 300 provides backhaul to the NCR 302, which acts as an extension of the gNB 300 and has beamforming capabilities to provide access beams 306A-306B towards directions beyond coverage in the gNB 300 area. This means that in addition to managing its own Tx / Rx beams 304A-304G, the gNB 300 will also manage the access beams 306A-306B of the NCR 302. The NCR 302 can forward an additional set of SSBs specifically assigned to the NCR 302. The gNB 300 can also exchange side control link information 312 with the NCR 302 to control the access beams 306A-306B. Figure 3 In the example shown, the gNB provides beams 304A-304G, and the NCR 302 provides beams 306A-306B. Figure 3 Also shown are two terminal devices 308 and 310, such as UEs. Instead of using SSBs, CSI-RS can be used. SSBs / CSI-RSs are transmitted from the gNB 300 and / or NCR 302 to the UE 310 in a defined beam direction.

[0073] Figure 4 A flow chart according to an example embodiment is shown. Figure 3 Discuss together Figure 4 . Figure 4The example shown in FIG3 assumes that gNB beam 304B is used to backhaul NCR 302, which in turn serves UE 310 with access beams 306A-306B having a different shape / direction / power than gNB beams 304A-304G. It is also assumed that the ML model used has been trained in a legacy network without NCR 302. It is also assumed that NCR 302 is operated by gNB 300 with side control link information 312 exchanged between gNB 300 and NCR 302 to control NCR access beams 306A-306B.

[0074] At 400, UE 308, 310 may be configured to send a message to gNB 300 that includes an indication of capabilities related to the ML model applied by UE 308, 310 for optimal beam prediction. This may include the number of beams to be measured and the number of beams to be predicted. The indication of capabilities related to the ML model (e.g., the preferred number of beams to be measured and the number of beams to be predicted) may help reach a consensus between gNB 300 and UE 308, 310. The ML model may not be compatible with NCR 302, but it may still support at least the number of Tx beams transmitted by gNB 300. At 402, UE 308, UE 310 is configured to receive a message from gNB 300 that includes an indication of downlink reference signal resources associated with a beam assigned to NCR 302, e.g., SSB identifiers assigned to NCR 302, each SSB identifier associated with a corresponding beam transmitted by NCR 302. In another example embodiment, a channel state information reference signal (CSI-RS) may be used instead of using SSBs. This enables UE 308, UE 310 to measure all NCR access beams 306A-306B one beam at a time to establish an optimal beam with a higher RSRP.

[0075] At 404, in response to receiving the message at 402, UEs 308 and 310 are configured to run an ML model to predict the best beam RSRP for beams from gNB 300 (thus excluding beams 306A-306B from NCR 302). The RSRP of beams 306A-306B allocated to NCR 302 is measured using conventional procedures. In other words, gNB 302 may configure UEs 308 and 310 to perform additional measurements for NCR access beams 306A-306B while changing the NCR access beam direction via the control link. UEs 308 and 310 may store the measured vectors and corresponding beam identifiers (extracted by decoding the SSB) for NCR access beams 306A-306B in memory.

[0076] At 406, UE 308, 310 may be configured to report the best beam to gNB 300. If the RSRP of the best beam predicted by the ML model is higher than the RSRP of the best beam measured from NCR 302, UE 308, 310 reports the RSRP of the best beam. Otherwise, UE 308, 310 reports the best beam from NCR 302. If the RSRP of the best beam predicted by the ML model is higher than the RSRP measured using NCR access beams 306A-306B, the UE reports the beam identifier and / or RSRP of the best beam predicted by the ML model. This may occur for UE 308. If the RSRP of NCR access beams 306A-306B is higher than the RSRP of the best predicted gNB beams 304A-304G predicted by the ML model, the UE reports the beam identifier and / or RSRP of the best beam measured using NCR access beams 306A-306B. This is what might happen to UE 310.

[0077] Figure 5 A wireless communication system according to an example embodiment is illustrated. The wireless communication system includes a network node 500 (e.g., a base station or gNB) and an NCR 502. The NCR 502 is an in-band radio frequency (RF) repeater that relays signals sent from the gNB 500 to improve network coverage. The NCR 502 is transparent to user equipment (UE) 508, 510.

[0078] The gNB 500 provides backhaul to the NCR 502, which acts as an extension of the gNB 500 and has beamforming capabilities to provide access beams 506A-506D towards directions beyond coverage in the gNB 500 area. This means that in addition to managing its own Tx / Rx beams 504A-504G, the gNB 500 will also manage the access beams 506A-506D of the NCR 502. The NCR 502 can forward additional SSB sets specifically assigned to the NCR 502. The gNB 500 can also exchange side control link information 512 with the NCR 502 to control the access beams 506A-506D. Figure 5 In the example shown in , the gNB provides beams 504A-504G, and the NCR 502 provides beams 506A-506D. Figure 5 Also illustrated are two terminal devices 508, 510, eg UEs.

[0079] Figure 6 A flow chart according to an example embodiment is shown. Figure 5 Discuss together Figure 6 . Figure 6 The example shown in FIG5 assumes that gNB beam 504B is used to backhaul NCR 502, which in turn serves UE 510 using access beams 506A-506D, which have different shapes / directions / powers than gNB beams 504A-504G. It is also assumed that the ML model used has been trained in a legacy network without NCR 502. It is also assumed that NCR 502 is operated by gNB 500 using side control link information 512 exchanged between gNB 500 and NCR 502 to control NCR access beams 506A-506D.

[0080] At 600, UE 508, 510 is configured to send a message to gNB 500, the message including an indication of capabilities associated with the ML model applied by UE 508, 510 for optimal beam prediction. The indication of capabilities associated with the ML model (e.g., the preferred number of beams to be measured and the number of beams to be predicted) can help reach a consensus between gNB 500 and UE 508, 510. The ML model may not be compatible with NCR 502, but it may still support at least the number of Tx beams transmitted by gNB 500. At 602, UE 508, 510 is configured to receive a message from gNB 500, the message including an indication of downlink reference signal resources associated with beams allocated to NCR 502, e.g., SSB identifiers allocated to NCR 502, each SSB identifier being associated with a respective beam transmitted by NCR 502. In another example embodiment, a channel state information reference signal (CSI-RS) may be used instead of using SSB.

[0081] At 604, in response to receiving the message at 602, the UEs 508, 510 are configured to run an NCR-compatible ML model for predicting optimal beam RSRP for beams from the gNB 500 and the NCR 502. In an example embodiment, the gNB 500 may be configured to send the NCR-compatible ML model to the UEs 508, 510. In another example embodiment, the UEs 508, 510 may train the NCR-compatible ML model based on a previously configured ML model that does not consider the NCR 502 and the beams 506A-506D transmitted by the NCR 502.

[0082] At 606, UE 508, UE 510 may be configured to report the best beam to gNB 500. With respect to UE 508, UE 508 may accurately predict the best gNB beam (i.e., Figure 54 in the example shown in FIG), because the NCR-compatible ML model has been trained to take into account the NCR 502 and the beams 506A-506D transmitted by the NCR 502. With respect to the UE 510, the UE 510 can accurately predict the best NCR access beam (i.e., the beam ID 506A-506D in FIG). Figure 5 Beam ID N+2 in the example shown in ), because the NCR-compatible ML model has been trained to take into account NCR 502 and beams 506A-506D transmitted by NCR 502.

[0083] Figure 7 A flow chart according to an example embodiment is illustrated. Figure 7 Illustrated is a solution for obtaining NCR-compliant ML models.

[0084] At 700, the UEs 508, 510 may be configured to collect data related to the NCR access beams 506A-506D. The data may include, for example, measurements of the NCR access beams 506A-506D paired with identifiers of the beams.

[0085] At 702, UE 508, UE 510 may be configured to train an NCR compatible ML model based on an existing ML model that has not yet considered NCR 502 and its beams 506A-506D using the data collected at 700. In an example embodiment, a transfer learning solution may be used to avoid retraining the entire ML model from scratch. Figure 8 In another example embodiment, the NCR-compatible ML model can be trained by randomly initializing parameters.

[0086] At 704, UE 508, 510 may optionally be configured to send the NCR-compatible ML model to gNB 500, which may store the model for later use by other UEs. UE 508, 510 may also attach metadata to the header of the file containing the NCR-compatible ML model. Using the metadata, gNB 500 may identify other UEs (e.g., with the same vendor identifier) ​​that may need to use the same ML model. In an example embodiment, UE 508, 510 may be configured to encrypt the ML model file to preserve proprietary designs. Thus, gNB 500 may only read the metadata, but the ML model itself is inaccessible to gNB 500.

[0087] At 706, as in Figure 6 As discussed in more detail in the description thereof, UE 508, UE 510 may be configured to run an NCR-compatible ML model for beam prediction.

[0088] Figure 8 An example of using transfer learning to update the ML model 800 according to an example embodiment is illustrated. The ML model 800 has not yet considered the NCR and its beam.

[0089] The ML model 800 takes as input 804 the RSRP measurements of the beam corresponding to the gNB. After layers 806 to 812 and a softmax layer 814, an output 816 is provided. A selection function 818 then selects the best beam 820 based on the output 816.

[0090] Given that the training parameters of the ML model 800 include knowledge for predicting the optimal Tx beam for a gNB, they can be applied to the closely related task of predicting the optimal beam in networks with NCR.

[0091] Subsequently, when implementing transfer learning, in this example, parameters associated with the first two layers 806 and 808 of ML model 800 are used, and ML model 800 is updated by training the remaining layers (i.e., layers 824 and 826) using data collected by measuring the NCR access beam. Other examples of implementing transfer learning may migrate parameters for any number of layers, and / or freeze, retrain, or fine-tune the parameters of the migrated layers to improve prediction accuracy. ML model 802 takes RSRP measurements corresponding to the gNB beam and the NCR beam as input 822. After layers 806, 808, 824, and 826, and a softmax layer 828, an output 830 is provided. A selection function 818 then selects the best beam 832 based on output 830. The best beam can be either a gNB beam or an NCR access beam.

[0092] The NCR-compatible ML model 800 (ie, ML model 802) is compatible with NCR access beams and can be applied to the task of predicting the best beam in a network that also includes NCR.

[0093] Figure 9 An example of an apparatus 900 configured to practice one or more example embodiments is illustrated. Apparatus 900 may include, for example, an access node, a base station, a gNB, a radio network node, or a separate portion thereof, a terminal device, a user node, a user equipment, or generally a device configured to implement the functionality described herein. Although apparatus 900 is shown as a single device, it should be understood that the functionality of apparatus 900 may be distributed across multiple devices, where applicable.

[0094] The apparatus 900 may include at least one processor 902. The at least one processor 902 may include, for example, one or more of a variety of processing devices or processor circuitry, such as, for example, a coprocessor, a microprocessor, a controller, a digital signal processor (DSP), processing circuitry with or without an accompanying DSP, or various other processing devices including integrated circuits, such as, for example, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a microcontroller unit (MCU), a hardware accelerator, a dedicated computer chip, etc.

[0095] The apparatus 900 may further include at least one memory 904. The at least one memory 904 may be configured to store, for example, computer program code, such as operating system software and application software. The at least one memory 904 may include: one or more volatile memory devices, one or more non-volatile memory devices, and / or a combination thereof. For example, the at least one memory 904 may be embodied as a magnetic storage device (such as a hard disk drive, a floppy disk, a magnetic tape, etc.), an optical magnetic storage device, or a semiconductor memory (such as a mask ROM, a PROM (programmable ROM), an EPROM (erasable PROM), a flash ROM, a RAM (random access memory), etc.).

[0096] The device 900 may also include a communication interface 908, which is configured to enable the device 900 to send information to other devices and / or receive information from other devices. In one example, the device 900 can use the communication interface 908 to send or receive signaling information and data according to at least one data communication or cellular communication protocol. The communication interface 908 can be configured to provide at least one wireless radio connection, such as, for example, a 3GPP mobile broadband connection (e.g., 3G, 4G, 5G, 6G, etc.). The communication interface 908 may include or be configured to be coupled to at least one antenna to send and / or receive radio frequency signals. One or more of the various types of connections may also be implemented as separate communication interfaces, which may be coupled to one or more of a plurality of antennas, or configured to be coupled to one or more of a plurality of antennas. The communication interface 908 may include a receiver, a transmitter, or a transceiver.

[0097] When the apparatus 900 is configured to implement certain functions, some components and / or components of the apparatus 900 (e.g., at least one processor 902 and / or at least one memory 904) may be configured to implement the functions. In addition, when the at least one processor 902 is configured to implement certain functions, the functions may be implemented using, for example, program code 906 included in the at least one memory 904.

[0098] The functions described herein may be performed at least in part by one or more computer program product components (such as software components). According to an embodiment, the device may include a processor or processor circuit system, such as a microcontroller, which is configured by a program code to perform the operations and functions described herein when executed. Program code 906 is provided as an example of an instruction, which, when executed by at least one processor 902, enables the performance of the device. Alternatively or additionally, the functions described herein may be performed at least in part by one or more hardware logic components. Illustrative types of hardware logic components that can be used include, for example, but not limited to: field programmable gate array (FPGA), application specific integrated circuit (ASIC), application specific standard product (ASSP), system-on-chip system (SOC), complex programmable logic device (CPLD), and graphics processing unit (GPU).

[0099] The apparatus 900 may be configured to perform or cause the performance of any aspect of the method(s) described herein, for example, functions performed by a UE or gNB. Furthermore, a computer program may include instructions that, when executed, cause the apparatus to perform any aspect of the method(s) described herein. The computer program may be stored on a computer-readable medium. Furthermore, the apparatus 900 may include components for performing any aspect of the method(s) described herein. In one example, such components may include: at least one processor 902; at least one memory 904 including program code 906 (instructions) configured to, when executed by the at least one processor 902, cause the apparatus 900 to perform the method(s). Generally speaking, the computer program instructions may be executed on components providing general processing functionality. Thus, the method(s) may be computer-implemented, for example, based on algorithm(s) executable by the general processing functionality, an example of which is the at least one processor 902. The component may include transmitting and / or receiving components, such as one or more radio transmitters or receivers, which may be coupled to one or more antennas, or (multiple) transmitters or (multiple) receivers of a wired communication interface, or be configured to be coupled to one or more antennas, or (multiple) transmitters or (multiple) receivers of a wired communication interface.

[0100] Any range or device value given herein may be extended or altered without losing the effect sought. Furthermore, any embodiment may be combined with another embodiment unless expressly prohibited.

[0101] Although the subject matter has been described in language specific to structural features and / or acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as examples of implementing the claims, and other equivalent features and acts are intended to be within the scope of the claims.

[0102] It should be understood that the benefits and advantages described above may relate to one embodiment or may relate to several embodiments. The embodiments are not limited to embodiments that solve any or all of the problems described, or embodiments that have any or all of the benefits and advantages described. It will also be understood that reference to "an" item may refer to one or more of these items.

[0103] The steps or operations of the methods described herein may be performed in any suitable order, or simultaneously where appropriate. Additionally, individual blocks may be deleted from any method without departing from the scope of the subject matter described herein. Aspects of any embodiment described above may be combined with aspects of any other embodiment described to form additional embodiments without losing the effects sought.

[0104] The term "comprising" is used herein to mean including the identified methods, blocks, or elements, but that such blocks or elements do not comprise an exclusive list and the method or apparatus may contain additional blocks or elements.

[0105] As used in this application, the term "circuitry" may refer to one or more or all of the following: (a) hardware circuitry implementations only (such as implementations in analog and / or digital circuitry only) and (b) combinations of hardware circuitry and software, such as, as applicable: (i) a combination of analog and / or digital hardware circuitry with software / firmware, and (ii) any portion of hardware processor(s) with software (including digital signal processor(s), software, and memory(s) that work together to enable a device such as a mobile phone or server to perform various functions, and (c) hardware circuitry and / or processor(s), such as microprocessor(s) or portions of microprocessor(s), that require software (e.g., firmware) to operate, but that software may be absent when not required for operation. This definition of circuitry applies to all uses of the term in this application, including in any claims.

[0106] It should be understood that the above description is given by way of example only and that various modifications may be made by those skilled in the art. The above description, examples, and data provide a complete description of the structure and use of the example embodiments. Although various embodiments have been described above with a certain degree of particularity or with reference to one or more individual embodiments, those skilled in the art may make various changes to the disclosed embodiments without departing from the scope of this specification.

Claims

1. A terminal device (510), comprising: at least one processor; as well as at least one memory storing instructions that, when executed by the at least one processor, cause the terminal device to: receiving (200) a message from a network device (500), the message comprising: an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay (502), wherein the network controlled relay is transparent to the terminal device; as well as Determining (202) a best beam among beams (504A-G) transmitted by the network device (500) and beams (506A-C) transmitted by the network controlled repeater (502) based on a decision between the beam indicated in the message and the beam predicted by the machine learning model.

2. The terminal device (510) according to claim 1, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: running the machine learning model on the beams transmitted by the network device (500) to predict an optimal beam among the beams transmitted by the network device; measuring a reference signal received power for each beam allocated to the network controlled repeater (502); as well as The best beam among the beam transmitted by the network device (500) and the beam transmitted by the network controlled repeater (502) is determined based on the predicted reference signal received power of the best beam of the beam transmitted by the network device and the measured reference signal received power of the beam allocated to the network controlled repeater.

3. The terminal device (510) according to claim 1, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: obtaining a network-controlled repeater-compatible machine learning model capable of predicting the optimal beam transmitted by the network-controlled repeater; and For the beam sent by the network device and the beam sent by the network controlled repeater, run (604) the network controlled repeater compatible machine learning model to predict the best beam among the beam sent by the network device and the beam sent by the network controlled repeater.

4. The terminal device (510) according to claim 3, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: collecting (700) measurement data associated with the beam transmitted by the network controlled repeater, the measurement data comprising: a measured reference signal received power for each beam transmitted by the network controlled repeater; as well as Based on the collected measurement data and the machine learning model, the network controlled repeater compatible machine learning model is trained (702).

5. The terminal device (510) according to claim 4, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: Transfer learning is applied in training the network-controlled repeater-compatible machine learning model based on the collected measurement data and the parameters of the machine learning model trained for the beam sent by the network device.

6. The terminal device (510) according to claim 4, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: The network control relay is trained to be compatible with the machine learning model by randomly initializing parameters.

7. The terminal device (510) according to any one of claims 4 to 6, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: The network controlled repeater compatible machine learning model is sent (704) to the network device (500).

8. The terminal device (510) according to claim 3, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: The network-controlled repeater-compatible machine learning model is received from the network device.

9. The terminal device (510) according to any one of claims 1 to 8, wherein the instructions, when executed by the at least one processor, cause the terminal device to execute: Information related to the capabilities of the machine learning model for beam prediction is sent (600) to the network device (500).

10. The terminal device (510) according to any one of claims 1 to 9, wherein the instructions, wherein the beam comprises: Synchronization signal block beam, or channel state information reference signal beam.

11. A network device (500), comprising: at least one processor; as well as at least one memory storing instructions that, when executed by the at least one processor, cause the network device to: A message is sent (204) to a terminal device (510), the message including an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay.

12. The network device (500) of claim 11, wherein the instructions, when executed by the at least one processor, cause the network device to: A network-controlled repeater-compatible machine learning model is received (704) from the terminal device (510), the network-controlled repeater-compatible machine learning model being capable of predicting an optimal beam transmitted by the network-controlled repeater.

13. The network device (500) of claim 12, wherein the instructions, when executed by the at least one processor, cause the network device to: The network-controlled repeater-compatible machine learning model is sent to a second end device served by the network-controlled repeater.

14. A method comprising: Receiving (200) by a terminal device (510) from a network device (500) a message comprising: an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay, wherein the network controlled relay is transparent to the terminal device; and The terminal device (510) determines (202) a best beam among the beams sent by the network device and the beams sent by the network controlled repeater based on a decision between the beams indicated in the message and the beams predicted by the machine learning model.

15. A method comprising: A message is sent (204) by a network device (500) to a terminal device (510), the message comprising an indication of downlink reference signal resources associated with a beam allocated to a network controlled relay, wherein the network controlled relay is transparent to the terminal device.

16. A computer program comprising instructions for causing an apparatus to perform the method according to claim 14.

17. A computer program comprising instructions for causing an apparatus to perform the method according to claim 15.

Images