Beam reporting for AI-enabled time domain prediction

Abstract

Systems and methods are provided for beam reporting using artificial intelligence (AI) or machine learning (ML) time domain prediction. A user equipment (UE) receives a first configuration for a first set of downlink (DL) reference signals (RS) from a base station; the UE measures a first set of DL RSs at a plurality of measurement opportunities, and determines, based on the measurements, one or more selected beams corresponding to the first set of DL RSs for reporting to the base station. The UE then transmits to the base station an indication of the one or more selected beams and feedback data corresponding to at least one selected beam of the one or more selected beams.

Description

TECHNICAL FIELD

[0001] This application relates generally to wireless communication systems, including beam management using artificial intelligence (AI) and / or machine learning (ML). BACKGROUND

[0002] Wireless mobile communication technology uses various standards and protocols to transmit data between base stations and wireless communication devices. For example, wireless communication system standards and protocols can include, for example, 3rd Generation Partnership Project (3GPP) Long-Term Evolution (LTE) (e.g., 4G), 3GPP New Radio (NR) (e.g., 5G), and Institute of Electrical and Electronics Engineers (IEEE) 802.11 standards (commonly referred to as Wi-Fi ® ).

[0003] As contemplated by 3GPP, different wireless communication system standards and protocols can use various radio access networks (RANs) for communication between base stations (which can also be commonly referred to as RAN nodes, network nodes, or simply nodes) of the RAN and wireless communication devices referred to as user equipment (UE). A 3GPP RAN can include, for example, a Global System for Mobile Communications (GSM), Enhanced Data Rates for GSM Evolution (EDGE) RAN (GERAN), Universal Terrestrial Radio Access Network (UTRAN), Evolved Universal Terrestrial Radio Access Network (E-UTRAN), and / or Next Generation Radio Access Network (NG-RAN).

[0004] Each RAN can use one or more radio access technologies (RATs) to perform communication between base stations and UEs. For example, a GERAN implements GSM and / or EDGE RAT, a UTRAN implements Universal Mobile Telecommunications System (UMTS) RAT or other 3GPP RAT, an E-UTRAN implements LTE RAT (sometimes referred to simply as LTE), and an NG-RAN implements NR RAT (which is sometimes referred to herein as 5G RAT, 5G NR RAT, or simply NR). In certain deployments, an E-UTRAN can also implement NR RAT. In certain deployments, an NG-RAN can also implement LTE RAT.

[0005] A base station used by a RAN can correspond to that RAN. One example of an E-UTRAN base station is an Evolved Universal Terrestrial Radio Access Network (E-UTRAN) Node B (also commonly denoted as an Evolved Node B, Enhanced Node B, eNodeB, or eNB). One example of an NG-RAN base station is a Next Generation Node B (sometimes also referred to as a gNodeB or gNB).

[0006] The RAN, through its connection to a core network (CN), also provides connectivity to external entities. For example, an E-UTRAN can utilize an Evolved Packet Core (EPC), and NG-RAN can utilize a 5G Core Network (5GC). BRIEF DESCRIPTION OF DRAWINGS

[0007] To easily identify the discussion of any particular element or act, one or more of the highest three digits in a figure number refer to the figure number in which that element is first introduced.

[0008] Figure 1 An AI-based beam selection utilizing partial measurements that can be used is illustrated in accordance with certain embodiments.

[0009] Figure 2 An example of using an AI model or ML model to predict information of downlink (DL) beams at a future time based on historical measurements is illustrated in accordance with certain embodiments.

[0010] Figure 3 Using an AI model or ML model for time-domain prediction when the measured beam is the same as the predicted beam is illustrated in accordance with certain embodiments.

[0011] Figure 4 Using an AI model or ML model for time-domain prediction when the measured beam is different from the predicted beam is illustrated in accordance with certain embodiments.

[0012] Figure 5 Different aspects that can be combined are illustrated in accordance with certain embodiments.

[0013] Figure 6 A bitmap for indicating strong beams measured at a first occasion Al is illustrated in accordance with one embodiment.

[0014] Figure 7 A common bitmap for indicating strong beams for multiple occasions is illustrated in accordance with certain embodiments.

[0015] Figure 8 is a flowchart illustrating a method for a UE to perform beam management utilizing time-domain prediction in accordance with certain embodiments.

[0016] Figure 9 is a flowchart illustrating a method for a base station to perform time-domain prediction for beam management in accordance with certain embodiments.

[0017] Figure 10 Test data provisioning and performance monitoring for AI beam management is illustrated in accordance with certain embodiments.

[0018] Figure 11Examples of test data provisioning and performance monitoring for AI beam management based on certain implementation schemes are provided.

[0019] Figure 12 This is a flowchart illustrating a method, according to certain implementations, for a base station to provide test data to a UE for performance monitoring of a model used for beam management.

[0020] Figure 13 This is a flowchart illustrating a method for performance monitoring of beam management enabled by AI by a UE, according to certain implementation schemes.

[0021] Figure 14 An example architecture of a wireless communication system according to the implementation scheme disclosed herein is illustrated.

[0022] Figure 15 A system for performing signaling between a wireless device and a network device according to an embodiment disclosed herein is illustrated. Detailed Implementation

[0023] Various implementations are described with respect to the UE. However, references to the UE are provided for illustrative purposes only. The example implementations can be used with any electronic component that can establish a connection to a network and is configured with hardware, software, and / or firmware for exchanging information and data with the network. Therefore, the UE as described herein is used to represent any suitable electronic component.

[0024] The beam management process may include a base station performing beam scanning on multiple transmit beams. The base station can use each transmit beam to transmit, for example, a Channel State Information Reference Signal (CSI-RS) for beam management. To enable the UE to perform receive beam scanning, the base station can use the transmit beam to transmit (e.g., using repetition) each CSI-RS within the same reference signal (RS) resource set multiple times, allowing the UE to scan through the receive beam in multiple transmission instances. For example, if the base station has a set of N_1 transmit beams and the UE has a set of M_1 receive beams, then the CSI-RS can be transmitted M times on each of the N_1 transmit beams, allowing the UE to receive M_1 instances of the CSI-RS per transmit beam. In other words, for each transmit beam of the base station, the UE can perform a beam scan through the UE's receive beam. Therefore, the beam management process allows the UE to use different receive beams to measure CSI-RS on different transmit beams, supporting the selection of base station transmit beam / UE receive beam pairs. The UE can report measurements to the base station, enabling the base station to select one or more beam pairs for communication between the base station and the UE. Although this example has been described in conjunction with CSI-RS, the beam management process can also be performed using synchronization signal blocks (SSBs) in a similar manner to that described above.

[0025] Beamforming uses phased arrays to generate highly directional transmit links to overcome path loss attenuation of millimeter wave links. Directional links use fine alignment of transmit beams and receive beams implemented through a set of operations known as beam management, which includes beam selection, beam reporting, beam switching, beam tracking, etc. However, using a large number of antennas creates high overhead in beam management in terms of power and time consumption, reference signal overhead, and computational complexity. AI techniques can be used to improve performance and reduce overhead by predicting the quality of beam pairs and selecting the optimal beam via measuring only a limited number of beams.

[0026] For example, Figure 1 AI-based beam selection using partial measurements that can be used is illustrated in accordance with certain embodiments. In the illustrated example, a UE can measure and report the reference signal received power (RSRP) of only some of the beams at a first time 102. The base station can use an AI model 104 to predict the RSRP of other beams at a future second time 106. The base station can also use the AI model 104 to predict the optimal or strongest beam with the largest RSRP. The base station can then inform the UE of the optimal or strongest beam. Thus, rather than reporting the RSRP of all beam pairs through an exhaustive beam search, the AI model 104 (or ML model) allows the UE to reduce the overhead by measuring and reporting the RSRP of only some of the beams.

[0027] Figure 2 An example of using an AI model or ML model to predict information of downlink (DL) beams at future times based on historical measurements is illustrated in accordance with certain embodiments. In the illustrated example, a UE measures a plurality of beams (e.g., performs RSRP measurements on eight RS beams) at a first measurement occasion Al and again at a second measurement occasion A2. At a given time (e.g., a current time), the base station or UE uses an AI / ML model to determine the best beam or beam pair to use at a first future time Fl and a second future time F2.

[0028] In Figure 2 In certain embodiments of the illustrated example, beams in a plurality of instances or measurement occasions of set B are used to predict beams in set A for one or more instances of time. Based on measurements at measurement occasion Al and a second measurement occasion A2 (each block of set B beams), beam predictions for set A beams are generated at two instances of time (a first future time Fl and a second future time F2).

[0029] Various options can exist for the Set-A beams and Set-B beams. In a first alternative (Alt. 1), the Set-A beams and Set-B beams are different from each other (i.e., Set-B is not a subset of Set-A). In a second alternative (Alt. 2), the Set-B beams are a subset of the Set-A beams (but Set-A and Set-B are not identical). In a third alternative (Alt. 3), the Set-A beams and Set-B beams are identical. It can be noted that the beam pattern of Set-A and Set-B can be selected based on the particular implementation. It can be noted that Set-B can be defined as the set of beams for which the UE performs measurements, or a subset of beams for which the UE performs measurements, and the RSRP of beams from this set are used as input to the AI / ML model.

[0030] Embodiments of set A and set B .

[0031] The network can configure measurement resources (such as SSB, CSI-RS resources, channel state information (CSI) resource sets, CSI resource configurations, etc.) for the UE to perform beam measurements.

[0032] When using beams in multiple instances of Set-B to predict beams in Set-A for one or more instances of time, the UE can perform beam measurements at multiple occasions M. For example, there are slots and / or orthogonal frequency division multiplexing (OFDM) symbols of at least one measurement resource. To simplify the discussion below, it can be assumed that the measurement occasions are indexed by a starting time t1, t2,..., tMand a corresponding 1,..., Mindex. For each occasion, the UE can perform beam measurements at multiple measurement resources P1, P2,..., PM, where 1≤ m≤ M. In some cases, P1= P2=... = PM= P, i.e., the number of measurements is the same for each occasion. M m In some cases, P1= P2=... = PM= P, i.e., the number of measurements is the same for each occasion. M

[0033] In certain embodiments, Set-A, which is the set of beams for which predictions are generated for the AI model, includes CSI-RS resources, e.g., {NZP-CSI-RS-ResourceId-1, NZP-CSI-RS-ResourceId-2,..., NZP-CSI-RS-ResourceId-32} of non-zero power (NZP) CSI-RS resources.

[0034] In certain embodiments, Set-A, which is the set of beams for which predictions are generated for the AI model, includes SSBs, e.g., {ssb-Index SSB-Index-1, ssb-Index SSB-Index-1,..., ssb-Index SSB-Index-8}.

[0035] ​​In certain embodiments, the set A (which is the set of beams for which the AI model generates predictions) can include SSB and CSI-RS resources, e.g., {ssb-Index ssb-Index-1, ssb-Index ssb-Index-1, …, ssb-Index ssb-Index-8} and {NZP-CSI-RS-ResourceId-1, NZP-CSI-RS-ResourceId-2, …, NZP-CSI-RS-ResourceId-32}, which can be provided by two resource sets, one for SSB and the other for CSI-RS resources.

[0036] In certain embodiments (e.g., for Alt. 3 discussed above), the set A (which is the set of beams for which the AI model generates predictions) can include CSI-RS resources, e.g., {NZP-CSI-RS-ResourceId-1, NZP-CSI-RS-ResourceId-2, …, NZP-CSI-RS-ResourceId-32}.

[0037] In certain embodiments, the measurement set (set B) can include CSI-RS resources. When each CSI-RS resource is associated with a NZP-CSI-RS-ResourceId (as in 3GPP Technical Specification (TS) 38.331), then set B includes CSI resources from {NZP-CSI-RS-ResourceId-1, NZP-CSI-RS-ResourceId-2, …, NZP-CSI-RS-ResourceId-32}.

[0038] In certain embodiments (e.g., for Alt. 2), the set A (which is the set of beams for which the AI model generates predictions) can include CSI-RS resources, e.g., {NZP-CSI-RS-ResourceId-1, NZP-CSI-RS-ResourceId-2, …, NZP-CSI-RS-ResourceId-32}.

[0039] The measurement set (set B) can include CSI-RS resources. When each CSI-RS resource is associated with a NZP-CSI-RS-ResourceId (as in 3GPP TS 38.331), then set B includes CSI resources from {NZP-CSI-RS-ResourceId-2, NZP-CSI-RS-ResourceId-4, …, NZP-CSI-RS-ResourceId-27}.

[0040] When set B is equal to set A (e.g., Alt. 3 above, where both set A and set B are {1, 2, …, 32}), the time-domain beam prediction can not depend on the correlation between spatial beams. For example, Figure 3 Time-domain prediction using an AI or ML model is illustrated when the measured beams are different from the predicted beams, according to certain embodiments. Even though the UE does not know how two different beams are correlated, by two measurements of beam 17 at first measurement occasion Al and second measurement occasion A2, the RSRP of beam 17 at first future time Fl can be predicted by using a long short-term memory (LSTM) model. In this case, time-domain prediction at the UE is feasible, and the analog beam information from the base station (e.g., gNB) is not needed.

[0041] When set B is not equal to set A (e.g., Alt. 1 or Alt. 2 above, where set B = {2, 4, 9, 11, 18, 20, 25, 27} and set A = {1, 2, …, 32}), then the analog beam information of the transmitted beams from the base station is useful for time-domain prediction. However, it can be difficult to obtain the analog beam information at the UE side. For example, the infrastructure provider can not be willing to disclose such information. In such cases, it can be more feasible for the base station to perform inference based on the measurements made by the UE on set B.

[0042] For example, Figure 4 Time-domain prediction using an AI or ML model is illustrated when the measured beams are different from the predicted beams, according to certain embodiments. In this example, the UE measures the RSRP of some of the beams 402 (the illustrated shaded beams) at first measurement occasion Al and second measurement occasion A2. The UE does not measure the other beams 404 (the illustrated unshaded beams). Because the UE does not have the analog beam information indicating how the measured beams are correlated with the unmeasured beams, the UE can not be able to accurately predict the RSRP of unmeasured beam 17 at first future time Fl.

[0043] Conventionally, the network can act on the beam measurements made by the UE, and the number of reported beams in the beam report can be limited. However, as seen from the example Figure 4 It can be seen that the number of reported beams can be large. For illustration clarity, only two instances of set B measurement are shown. In practice, more instances can be used. After the conventional beam report, the feedback overhead can be too much, which can make AI-based beam management less useful.

[0044] Thus, certain implementations herein reduce feedback overhead when the measured beams (i.e., set B) are different from the predicted beams (set A) and AI-based beam management for time-domain inference is performed at the network side (e.g., base station). Certain such implementations reduce feedback overhead by indicating strong beams via one or more bitmaps or a combined index. In addition, or in other implementations, quantization of the strongest beam can be per-occasion or across occasions, and a corresponding differential quantization is provided. Certain implementations also use a two-part feedback for beam reporting to handle varying uplink control information (UCI) payload sizes.

[0045] Other implementations use a similar reporting scheme for test data transmitted from the base station to the UE for performance monitoring of the AI model by the UE.

[0046] Selection and indication of strong beams .

[0047] In one aspect (Aspect A), RSRP reporting for weak beams can be omitted when the UE indicates strong beams to the network.

[0048] In one implementation of Aspect A (Implementation A-l), selection and indication of strong beams can be per-occasion. Thus, beam strength fluctuations between different occasions can be accommodated. In one such implementation, the number of selected strong beams can be fixed (e.g., by standard specification) and / or set by network configuration. For example, the network can configure a percentage of set B beams. In another such implementation, the number of selected strong beams can vary within a range according to the UE’s discretion (e.g., subject to standard specification and / or network configuration). For example, the network can configure a percentage of set B beams (e.g., 25% or 50%), and the UE can select the number of set B beams within that range. The range can be specified by an upper limit, a lower limit, or both an upper and lower limit on the number of reported beams.

[0049] In another implementation of Aspect A (Implementation A-2), selection and indication of strong beams can be across occasion groups. Thus, channel correlation can be exploited to save on feedback overhead for indication of strong beams. In one such implementation, the number of selected strong beams can be fixed (e.g., by standard specification) and / or set by network configuration. In another such implementation, the number of selected strong beams can vary within a range according to the UE’s discretion (e.g., subject to standard specification and / or network configuration). The range can be specified by an upper limit, a lower limit, or both an upper and lower limit on the number of reported beams.

[0050] Quantization scheme for RSRP .

[0051] In another aspect (Aspect B), a different quantization scheme can be selected for reporting RSRP values.

[0052] In one embodiment of aspect B (embodiment B-1), one or more (N) reference beams can be selected as reference beams. The RSRP of each of the remaining reporting beams is quantized with respect to the RSRP of the reference beams using differential quantization. Thus, the overhead is reduced.

[0053] In another embodiment of aspect B (embodiment B-2), the RSRP of the beams is quantized individually (without differential quantization). Thus, the accuracy of the reporting beam strength can be increased (e.g., at the cost of increased overhead).

[0054] Quantizer design .

[0055] In another aspect (aspect C), different quantizer designs can be selected to reduce the overhead of the beam reporting.

[0056] In one embodiment of aspect C (embodiment C-1), uniform quantization (in the log domain) using a rounding up and / or rounding and / or rounding down function can be used. This design can provide simplicity and compatibility with legacy designs.

[0057] In another embodiment of aspect C (embodiment C-2), non-uniform quantization (in the log domain) using a rounding up and / or rounding and / or rounding down function can be used. Non-uniform quantization can provide improved quantizer performance compared to the performance provided by uniform quantization.

[0058] The inventors have observed from simulation evaluations that the inference accuracy is acceptable with a quantization step size of 1 dB for the RSRP of the reference beam (i.e., the strongest beam) and a quantization step size of 2 dB for the differential RSRP, with a 30 dB dynamic range, compared to the case where the RSRP is not quantized. It can be appreciated that the differential RSRP values can be set to low values (e.g., -30 dB) for weak beams without substantially affecting the inference performance. Furthermore, it can be possible to change the quantization step size. For example, in the objective of minimizing quantization error, the quantization boundaries can be identified, e.g., according to the Lloyd algorithm.

[0059] Selection and indication of reference beams .

[0060] In another aspect (aspect D), the selection and indication of the reference beams can be per measurement occasion or per measurement occasion group.

[0061] In one embodiment of aspect D (Embodiment D-1), for each occasion, a reference beam can be selected. Specifically, a fixed number (N) of reference beams can be selected. For example, for N=2, one reference beam n is selected among the groups of measurement resources at a single occasion, 1≤n≤N. This can help bring awareness of channel dynamics to the network. Group 1 can correspond to an analog beam associated with one polarization at the base station antenna array. Group 2 can correspond to an analog beam associated with another polarization at the base station antenna array. For N=1, in certain embodiments, a single reference beam is selected for all measurement resources at a single occasion. In another example, multiple set B patterns (e.g., set B pattern 1 and set B pattern 2 are interleaved, and the number of beams in their union is half the number of beams in set A) are used for beam measurements. Then, occasions with the same set pattern can be grouped together, and N can correspond to the number of set B patterns or a multiple of the number of set B patterns (e.g., there are 2 set B patterns, and N=4, so there can be two groups for each set B pattern).

[0062] In another embodiment of aspect D (Embodiment D-2), for a group of occasions, a reference beam is selected. Specifically, a fixed number (N) of reference beams can be selected. For example, for N=2, one reference beam n is selected among the groups of measurement resources across the group of occasions, 1≤n≤N. For N=1, in certain embodiments, a single reference beam is selected for all measurement resources across the group of occasions or all occasions. This can reduce feedback overhead.

[0063] Figure 5 Different aspects that can be combined are illustrated according to certain embodiments. As shown, when Embodiment B-1 is selected to use N reference beams, the embodiment combinations include {A-1, A-2} x {B-1} x {C-1, C-2} x {D-1, D-2}, which provides eight combinations, such as {A-1} x {B-1} x {C-1} x {D-1}. Alternatively, Embodiment B-2 is selected, and no reference beams are used, the embodiment combinations include {A-1, A-2} x {B-2} x {C-1, C-2}, which provides four combinations, such as {A-1} x {B-2} x {C-1}.

[0064] Example implementation - bitmap on strong beams per occasion .

[0065] In one embodiment, a bitmap is used to indicate strong beams within a set of measured beams. For example, Figure 6A bitmap 602 is illustrated for indicating strong beams measured at a first occasion Al according to one embodiment. Darkly shaded beams correspond to strong beams, which are each represented as "1" in bitmap 602. For beams represented as "0" in bitmap 602 (lightly shaded beams), a differential RSRP can be assumed to have a small value (e.g., -30 dB), which can not need to be reported by the UE. Note that the strong beams can be different for each occasion. For example, beam 604 can be measured to have a strong RSRP value at occasion Al and a weak RSRP value at occasion A2. Thus, while bitmap 602 for occasion Al has a "1" for beam 604, a bitmap (not shown) for occasion A2 would have a "0" for beam 604. Thus, different bitmaps are useful for different measurement occasions. In certain embodiments, for each occasion, the strongest beam is identified and its RSRP is quantized with high resolution. When using a bitmap, it can be useful to order the measurement resources. For example, the measurement resources can be ordered according to CSI-RS resource index or other scheme (e.g., in row-major or column-major fashion, etc.) in order to achieve a common understanding of the indication of strong beams between the UE and gNB.

[0066] Example implementation - combined index on strong beams per occasion .

[0067] In one embodiment, a combination index can be used to indicate a strong beam within a set of measured beams. The construction of the combination index can be referred to in, for example, 3GPP TS 38.214. The strong beam can be different for each occasion. Thus, different combination indices are evoked. The combination index can be designed to indicate a fixed number of selected beams or a number of selected beams within a range. For each occasion, the strongest beam can be identified and its RSRP quantized with high resolution.

[0068] In the case of a combination index, it can be useful to selectively order the measurement resources. For example, the measurement resources can be ordered according to CSI-RS resource index or other scheme (e.g., in row-major or column-major fashion, etc.) in order to achieve a common understanding of the indication of strong beams between the UE and gNB.

[0069] Example implementation - bitmap on strongest beams across occasions .

[0070] In one embodiment, a bitmap can be used to indicate a strong beam within a set of measured beams. As discussed above with reference to Figure 6 The strong beam can be different for each occasion. Thus, different bitmaps can be evoked. The strongest beam among all occasions or groups of occasions can be identified and the differential encoding of the remaining beams can be made according to the identified strongest beam.

[0071] Example implementation - combined index on strongest beams across occasions .

[0072] In one embodiment, a combination index can be used to indicate the strong beams within the set of measured beams. The construction of the combination index can be referred to in, for example, 3GPP TS 38.214. For each occasion, the strong beams can be different. Thus, different combination indices can be evoked. The strongest beam among all occasions or groups of occasions can be identified, and the differential encoding of the remaining beams can be made in accordance with the identified strongest beam.

[0073] Common bitmap or combined index for multiple occasions .

[0074] Considering the time-domain correlation among beams, even though the indices of the strong beams can vary over time (e.g., occasion indices, A1, A2, etc.), using a single bitmap or a single combination index for a group of occasions or all occasions can be more efficient, as the bitmap or combination index does not directly contribute to the RSRP feedback. Thus, in certain embodiments, a single bitmap or combination index is used for multiple occasions.

[0075] For example, Figure 7 A common bitmap 702 for indicating strong beams for occasion A1 and occasion A2 is illustrated in accordance with certain embodiments. For illustrative purposes, different patterns are used to show weak beams and strong beams in the common bitmap 702. However, the skilled person will appreciate from the disclosure herein that the bitmap will use ones and zeros to distinguish between weak beams and strong beams. In the illustrated example, the common bitmap 702 indicates beam 704 as a strong beam, even though it is measured to have a stronger RSRP at the first occasion A1 and a weaker RSRP at the second occasion. If multiple set B patterns are utilized (e.g., set B pattern 1 and set B pattern 2), the occasions can be associated with different set B patterns (e.g., occasions 1 and 3 are associated with set B pattern 1, occasions 2 and 4 are associated with set B pattern 2; or in another example, occasions 1 and 2 are associated with set B pattern 1, occasions 3 and 4 are associated with set B pattern 2). Then, for each group of occasions, there can be a separate common bitmap / combination index, and for each group of occasions, there can be a separate strongest beam indication.

[0076] Two-part feedback for beam reporting .

[0077] It can be noted that if differential RSRP is used, the number of beams not associated with the lowest RSRP value (e.g., -30 dB for differential beams) can vary depending on the channel conditions for all considered schemes. Thus, the beam report size can vary, which can result in many blind decodes at the network side to decode the beam report. To avoid blind decodes at the network side, two-part feedback is provided in accordance with certain embodiments.

[0078] For example, in one embodiment, a beam report indicating one or more strong beams includes two parts. The first part of the beam report has a fixed size and indicates the number of strong beams being reported. The second part of the beam report has a variable size based on the number of strong beams indicated in the first part to report the RSRP values of the strong beams.

[0079] In certain embodiments, when multiple bitmaps are reported for different occasions, the first part of the beam report includes the sum of “1”s in the reported bitmaps. In other embodiments, when a common bitmap is reported for multiple occasions, the first part of the beam report includes the total number of “1”s in the common bitmap. In addition, or in other embodiments, one code state can be assigned to represent the lowest RSRP or the lowest differential RSRP (e.g., -30 dB).

[0080] Example beam reporting implementation .

[0081] Certain embodiments assume there are M occasions and P measurements per occasion (e.g., P = 8). At the UE side, there are M . P measurements. At the baseband of the UE modem, the received signals can be represented by a linear scale or in a dB scale (e.g., in dBm). In the following description, the signal representation with dB scale is used. Similar embodiments can be derived for the case of signal representation with linear scale.

[0082] Let the measured signals be S m,p dBm, 1≤m≤M, 1≤p≤P.

[0083] The network can configure the number of reported beams (“strong beams”) R as a fraction of the set B size, e.g., R = f(β ∙ P) is the number of reported beams, f(x) is a rounding, ceiling or floor function, and β is a fraction, e.g., β = 0.25 or β = 0.5. The fraction can be provided by radio resource control (RRC) signaling or medium access control (MAC) control element (CE), or multiple candidate values of β are configured to the UE by RRC signaling or MAC CE. Then, one candidate value is selected by dynamic signaling (e.g., a codepoint in DCI) for beam reporting.

[0084] The network can directly configure R without using β (e.g., for P = 8, the network can configure R=5). The multiple candidate values of R can be configured to the UE through RRC signaling or MAC CE. Then, one of the candidate values is selected through dynamic signaling (e.g., codepoint in DCI) for beam reporting. β

[0085] From B beams, select R beams. The selection can be through bitmap or combinatorial index. The selection can be for a single occasion, a group of occasions, or all M occasions. When M is large, a common selection for all M occasions can not be suitable, and the M occasions can be divided into groups of occasions through specification and / or network configuration.

[0086] In some embodiments, the UE precisely selects P beams from R beams. If a bit bitmap is used, a “1” is used to indicate a selected beam, and the number of “1”s in the bitmap is R . If a combinatorial index scheme is used, the combinatorial index is given by , where is the number of combinations from a set with y elements. x

[0087] In some embodiments, the UE can select less than R beams for reporting. If a P bit bitmap is used, a “1” is used to indicate a selected beam, and the number of “1”s in the bitmap can be less than R . If a combinatorial index scheme is used, the combinatorial index is given by , where R lowest is the minimum number of reporting beams. For example, from a performance perspective, if there are at least two reporting beams according to specification and / or network configuration, R lowest = 2. In some embodiments, R lowest = 1 can also be used.

[0088] Quantization with per-occasion reference .

[0089] Continuing the above example, certain embodiments provide quantization with reference beams per occasion. At a given , find S m,p (1 ≤​​p ≤ P Let the largest of the following be S. m,p ,and p r It is a set B Beam index within the beam.

[0090] Signaling for the reference beam can be obtained via a bitmap (e.g., at location). p r A string that has a "1" in one place and a "0" in another. P Bitmap) sends signals to the network to notify about the set B The location or index of the strongest beam. Alternatively, it can be determined by combining indices. Signal to the network about the set B The location or index of the strongest beam.

[0091] Signaling for the reference beam can also be communicated with... P Select from the beams R Beam combining. A two-step process can be used, where the first step includes consumption. The strongest beam (reference beam) of each code state is indicated, and the second step is conditional upon the indication of the strongest beam. This is to signal the remaining... P -1 beam selection R -1 beam, the number of code states in the second step is Considering two steps, the total number of code states is... And the signaling overhead is from Given. Under certain conditions (e.g., P =16、 R =1) Under this condition, four bits can be saved through joint signaling of reference beam and beam selection. Alternatively, selection can be performed through a first step and a second step, in which, from P Select from the beams R One beam, in the second step, from R Select one beam from the given beams (the strongest beam): the signaling overhead is... Provided.

[0092] Quantization with per-occasion group reference .

[0093] Continuing with the example above, some implementations provide quantization using the reference beam of each time group. For simplicity of notation, assume all... M A group is formed at certain times. Discovery S m,p (1≤ m ≤ M ,1≤ p ≤ PLet the largest of the following be S. m r, p r ,and m r It is the index of the selected timing. p r It is a set B Beam index within the beam.

[0094] To signal the reference beam, two parts can be used to signal the network about the set. B The location or index of the strongest beam. In one option, both parts have bitmaps. Bitmap-1 can be the location... p r A string that has a "1" in one place and a "0" in another. P Bitmap. Bitmap-2 can be used at the right time. m r A string that has a "1" in one place and a "0" in another. M Bitmap. In other cases, both parts may be related to a composite index, or one part may be through a bitmap and the other through a composite index.

[0095] Composite indexes can be used Alternatively, the network can be signaled with respect to the position or index of the strongest beam in set B at all times by indicating the selected timing within the group and the selected beam at the selected timing.

[0096] For joint coding with cross-time public reporting beam selection, signaling for the reference beam can also be integrated with... P Select from the beams R The beams are combined and can follow a two-step process. Assuming a common choice of reporting beams (“strong beams”) across timing pairs, the first step includes depletion. The strongest beam (reference beam) of each code state is indicated, and the second step is conditional upon the indication of the strongest beam. Then, in order to signal the remaining... P -1 beam selection R -1 beam, the number of code states in the second step is Considering two steps, the total number of code states is... And the signaling overhead is from Given. It can be verified under certain conditions (e.g.) P =16、 R =1), 4 bits can be saved through joint signaling of reference beam and beam selection. Alternatively, selection can be performed through a first step and a second step, in which, from P Select from the beams ROne beam, in the second step, from R Select one beam from the given beams (the strongest beam): the signaling overhead is... or (The latter is a separate signaling instruction for the timing with the strongest beam and the strongest beam among the R selected beams at that timing.)

[0097] For joint coding of separately reported beam selection at specific times, signaling for the reference beam can also be combined with... P Select from the beams R Beam combination, and can follow a two-step process. Assuming separate reporting beam (“strong beam”) selection at the timing point, the first step includes consumption The strongest beam (reference beam) of each code state indicates the timing. m The second step is conditional upon indicating the strongest beam. Then, in order to signal the remaining... P -1 beam selection R -1 beam, for timing m The number of code states in the second step is Logically, regarding timing m Do not select with beam index p r Beamforming is possible, but the chances are slim. If this is a valid consideration, the choice of each timing point could be changed to... Considering two steps, the total number of code states is... And the signaling overhead is from Given that the total number of code states can be very large, some implementations use separate signaling for reference beams and reporting beam selections to facilitate resolution on the network side. Variations for the handover selection steps, as shown above, can be constructed.

[0098] Differential quantization .

[0099] Continuing with the example above, some implementations provide differential quantization, where the quantizer... Q (∙) Applied to S m,p - or S m,p - This depends on the design choices made above. The quantizer can be uniform, characterized by the highest value (e.g., 0), the lowest value (e.g., -40), and the quantization step size. In some implementations, the quantizer can be non-uniform. The design of a non-uniform quantizer can be done by applying the Lloyd algorithm, where the optimal quantization boundary is found and the quantization error is minimized. Quantization error metrics can be in the linear domain, in the dB domain, etc.

[0100] Example beam reporting procedure .

[0101] Figure 8 This is a flowchart illustrating a method 800 for a UE to perform beam management using time-domain prediction, according to certain embodiments. Method 800 includes: receiving 802 a first configuration for a first set of downlink (DL) reference signals (RS) from a base station at the UE. Method 800 also includes: measuring 804 the first set of DL RS at multiple measurement opportunities at the UE. In some embodiments, the DL RS in the first set may be periodic, semi-persistent, or aperiodic reference signals. In some embodiments, all DL RS in the first set are constrained to be periodic, semi-persistent, or aperiodic. For periodic reference signals, their period and offset within the period are configured by RRC; semi-persistent reference signals may also be associated with a period and offset. Aperiodic reference signals may be triggered by dynamic signaling. In some embodiments, measurement opportunities include reference signals from different periods. In some embodiments, measurement opportunities include clusters of reference signals within a period, where the time interval between two adjacent reference signals within a cluster may be uniform or non-uniform. In some embodiments, measurement opportunities include reference signals from clusters of more than one period. Based on this measurement, method 800 includes: determining 806 one or more selected beams corresponding to a first set of DLRS for reporting to a base station. Method 800 also includes: transmitting from the UE to the base station 808 an indication of the one or more selected beams and feedback data corresponding to at least one of the one or more selected beams.

[0102] In some implementations, method 800 may optionally include: receiving, at the UE, a second configuration for a second set of DL RS based on time-domain prediction feedback data from the base station 810.

[0103] In some embodiments of method 800, determining one or more selected beams includes selecting one or more selected beams at each of a plurality of measurement times. The fixed number of one or more selected beams at each time point may be predetermined or configured by the base station.

[0104] Some implementations of method 800 further include: the UE determining the number of one or more selected beams per timing within a predetermined range or a range configured by the base station.

[0105] In some embodiments of method 800, the indication of one or more selected beams includes a first plot indicating each timing of one or more selected beams.

[0106] In some embodiments of method 800, the indication of one or more selected beams includes a first combination index indicating each timing of the one or more selected beams. For the combination function... , R Number of reporting beams and per timing P The quantity measurement, the first combined index can be obtained by Provided.

[0107] In some implementations of method 800, the UE is configured to select up to [number missing]. R The number of reporting beams, and for the combination function and every opportunity P The measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowest This is the minimum number of reporting beams selected by the UE. Feedback data may include reference signal received power (RSRP) data for one or more selected beams at each timing. In some such implementations, the feedback data may not explicitly include RSRP data for unselected beams corresponding to a first set of DL RS measured by the UE.

[0108] In some embodiments, method 800 further includes: indicating from the UE to the base station the strongest beam of each of a plurality of measurement moments; and quantizing the RSRP value of the strongest beam of each moment at a higher resolution compared to a lower resolution of RSRP data used to quantize the remainder of one or more selected beams for each moment. Some such embodiments also include using differential quantization relative to the RSRP value of the strongest beam of each moment to quantize the remaining RSRP data of one or more selected beams for each moment. Differential quantization may include a uniform quantization function based on high values, low values, and quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors. Indicating the strongest beam of each moment may include using a second bitmap or by... The second combined index is given.

[0109] In some embodiments of method 800, determining one or more selected beams includes selecting one or more selected beams across a time group of multiple measurement times. The fixed number of one or more selected beams across the time group may be predetermined or configured by the base station. Some embodiments further include: the UE determining that the number of one or more selected beams across the time group is within a predetermined range or a range configured by the base station. An indication of one or more selected beams may include one or more first-order diagrams indicating one or more selected beams across the time group. An indication of one or more selected beams may include one or more first combination indices indicating one or more selected beams across the time group. For the combination function... , R The number of reported beams and the measurement of the number of cross-time group P, the first combined index can be obtained by Provided. Alternatively, the UE is configured to select up to [number]. R The number of reporting beams, and where for the combination function and number of cross-time groups P The measurement, the first combined index is by Given, among which r It is an index, and R lowest This is the minimum number of reporting beams selected by the UE. Feedback data may include Reference Signal Received Power (RSRP) data for one or more selected beams across time groups. In other embodiments, the feedback data may not explicitly include RSRP data for unselected beams corresponding to a first set of DL RS measured by the UE.

[0110] In some implementations, method 800 further includes: indicating from the UE to the base station the strongest beam in a time group among a plurality of measurement times; and quantizing the RSRP value of the strongest beam in the time group at a higher resolution compared to a lower resolution of RSRP data used to quantize the remainder of one or more selected beams across the time group. The method may further include: using differential quantization relative to the RSRP value of the strongest beam in the time group to quantize the RSRP data of the remainder of one or more selected beams across the time group. Differential quantization may include a uniform quantization function based on high values, low values, and a quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors. Indicating the strongest beam in the time group may include using a method targeting the time group... M The second bitmap of the timing or by The second combined index is given.

[0111] In some embodiments of method 800, transmitting indication and feedback information for one or more selected beams includes generating a report comprising: a first portion of fixed size for indicating the number of one or more selected beams; and a second portion of variable size for reporting corresponding reference signal received power (RSRP) data based on the number of one or more selected beams.

[0112] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 800. This apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0113] The embodiments contemplated herein include one or more non-transitory computer-readable media, the one or more non-transitory computer-readable media including instructions for causing the electronic device to perform one or more elements of method 800 when executed by one or more processors of the electronic device. The non-transitory computer-readable medium may be, for example, the memory of a UE (such as memory 1506 of a wireless device 1502 (UE), as described herein).

[0114] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 800. This apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0115] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 800. The apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0116] The implementation scheme envisioned herein includes a signal as described in or associated with one or more elements of method 800.

[0117] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution by a processor will cause the processor to perform one or more elements of method 800. The processor may be a processor of the UE (such as processor 1504 of wireless device 1502 (UE), as described herein). These instructions may, for example, reside in the processor and / or in the memory of the UE (such as memory 1506 of wireless device 1502 (UE), as described herein).

[0118] Figure 9 This is a flowchart illustrating a method 900 for a base station to perform temporal prediction for beam management according to certain embodiments. Method 900 includes: transmitting 902 a first set of downlink (DL) reference signals (RS) from the base station to a user equipment (UE). Method 900 also includes: receiving at the base station from the UE an indication of one or more selected beams corresponding to the first set of DL RS measured by the UE at multiple measurement points, and feedback data corresponding to at least one of the one or more selected beams. Method 900 further includes: based on the feedback data, using 906 a neural network model to determine temporal predictions of DL beam information for one or more future time periods.

[0119] In some implementations, method 900 further includes: configuring measurement resources corresponding to a second set of DLRS for the UE at one or more future time periods based on time-domain prediction. In some implementations, the DLRS in the first set may be periodic, semi-persistent, or aperiodic reference signals. In some implementations, all DLRS in the first set are constrained to be periodic, semi-persistent, or aperiodic. For periodic reference signals, their period and offset within the period are configured by RRC; semi-persistent reference signals may also be associated with a period and offset. Aperiodic reference signals may be triggered by dynamic signaling. In some implementations, measurement timing includes reference signals from different periods. In some implementations, measurement timing includes clusters of reference signals within a period, where the time interval between two adjacent reference signals within a cluster may be uniform or non-uniform. In some implementations, measurement timing includes reference signals from clusters of more than one period.

[0120] In some embodiments of method 900, one or more selected beams are selected for each of a plurality of measurement opportunities. The fixed number of one or more selected beams per opportunity is predetermined or configured by the base station. Alternatively, the number of one or more selected beams per opportunity is within a predetermined range or a range configured by the base station.

[0121] In some embodiments of method 900, the indication of one or more selected beams includes a first plot indicating each timing of the one or more selected beams or a first combination index indicating each timing of the one or more selected beams. For the combination function... , R Number of reporting beams, per timing P The quantity measurement, the first combined index can be obtained by Provided.

[0122] In some embodiments of method 900, up to a maximum ofR The number of reporting beams, and for the combination function and every opportunity P The measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowest This is the minimum number of reporting beams selected by the UE. Feedback data may include Reference Signal Received Power (RSRP) data for one or more selected beams at each timing. Alternatively, the feedback data may not explicitly include RSRP data for unselected beams corresponding to a first set of DL RS measured by the UE.

[0123] In some embodiments, method 900 further includes receiving at the base station from the UE: an additional indication of the strongest beam for each of a plurality of measurement moments; and a quantized RSRP value of the strongest beam for each moment at a higher resolution compared to a lower resolution of RSRP data used to quantize the remainder of one or more selected beams for each moment. The RSRP data of the remainder of one or more selected beams for each moment may be quantized using differential quantization relative to the RSRP value of the strongest beam for each moment, or the differential quantization may include a uniform quantization function based on high values, low values, and quantization step size. Differential quantization may be a non-uniform quantization function based on quantization boundaries and minimizing quantization errors. In some embodiments, the additional indication of the strongest beam for each moment includes a second bitmap or is derived from... The second combined index is given.

[0124] In some embodiments of method 900, one or more selected beams are selected across a time group of multiple measurement times. The fixed number of one or more selected beams across a time group may be predetermined or configured by the base station. Alternatively, the number of one or more selected beams across a time group may be within a predetermined range or within a range configured by the base station. Indication of one or more selected beams may include one or more first-order diagrams indicating one or more selected beams across a time group or one or more first combination indices indicating one or more selected beams across a time group. For the combination function... , R Number of reporting beams, cross-time groups P The quantity measurement, the first combined index can be obtained by Provided. In some implementations, at most [number] options are selected. R The number of reporting beams, and for the combination function and cross-time groups P The measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowestThis is the minimum number of reporting beams selected by the UE. Feedback data may include Reference Signal Received Power (RSRP) data for one or more selected beams across time groups. In some implementations, the feedback data may not explicitly include RSRP data for unselected beams corresponding to a first set of DL RS measured by the UE.

[0125] In some embodiments, method 900 further includes receiving at a base station from the UE: another indication of the strongest beam in a time group among a plurality of measurement times; and a quantized RSRP value of the strongest beam in the time group at a higher resolution compared to a lower resolution used to quantize the RSRP data of the remainder of one or more selected beams across the time group. The RSRP data of the remainder of one or more selected beams across the time group may be quantized using differential quantization relative to the RSRP value of the strongest beam in the time group. Differential quantization may include a uniform quantization function based on high values, low values, and quantization step size. In other embodiments, differential quantization includes a non-uniform quantization function based on quantization boundaries and minimizing quantization errors. The other indication of the strongest beam in the time group may include a quantization value for the strongest beam in the time group. M The second bitmap of the timing or by The second combined index is given.

[0126] In some embodiments of method 900, the indication and feedback data for one or more selected beams includes a report comprising: a first portion of fixed size for indicating the number of one or more selected beams; and a second portion of variable size for reporting corresponding reference signal received power (RSRP) data based on the number of one or more selected beams.

[0127] In some implementations, AI / ML inference is performed on the UE side and the feedback overhead of one or more outputs of the AI / ML inference model within one or more time epochs can be reduced in a manner similar to that used for measurements from the measurement timing, including beam selection and indication, quantizer scheme, quantizer design, and aspects of reference beam selection and indication.

[0128] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 900. This apparatus may be, for example, a base station (such as network device 1518 (base station), as described herein).

[0129] The embodiments contemplated herein include one or more non-transitory computer-readable media, which include instructions for causing the electronic device to perform one or more elements of method 900 when executed by one or more processors of the electronic device. The non-transitory computer-readable medium may be, for example, the memory of a base station (such as memory 1522 of network device 1518 (base station), as described herein).

[0130] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 900. This apparatus may be, for example, a base station (such as network device 1518 (base station), as described herein).

[0131] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 900. The apparatus may be, for example, an apparatus for a base station (such as network device 1518 (base station), as described herein).

[0132] The implementation scheme envisioned herein includes a signal as described in or associated with one or more elements of method 900.

[0133] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution of the program by a processing element will cause the processing element to perform one or more elements of method 900. The processor may be a processor of a base station (such as processor 1520 of network device 1518 (base station), as described herein). These instructions may, for example, reside in the processor and / or in the memory of the base station (such as memory 1522 of network device 1518 (base station), as described herein).

[0134] Performance monitoring for AI beam management .

[0135] Some implementations disclosed herein consider using test datasets to perform performance monitoring of AI models on the UE side. Various alternative message flows are provided for provisioning test datasets to the UE. Furthermore, or in other implementations, different schemes are provided to reduce the overhead of delivering test datasets from the base station to the UE. For example, as discussed above regarding reducing feedback overhead for AI-based beam management for time-domain inference, reducing test dataset overhead may include: indicating the selected input and / or output beams via one or more bitmaps or combined indices, using differential quantization to quantize the selected beams per time instance or across time instances, and / or using two-part signaling of the test dataset to handle different test dataset sizes.

[0136] Table 1 illustrates three alternatives for model training and model inference. In the first option (Opt.1), AI / ML model training and inference can be performed on the network (NW) side. In the second option (Opt.2), AI / ML model training and inference can be performed on the UE side. In the third option (Opt.3), AI / ML model training is performed on the NW side, and AI / ML model inference is performed on the UE side. For beam management, as discussed in this paper, the third option (Opt.3) is supported when the AI / ML model is transferred from the NW side to the UE side.

[0137] Table 1 For beam management, UE-side model monitoring may include: the UE monitoring performance metrics and making decisions on model selection, activation, deactivation, handover, and / or fallback operations. Alternatively, in network-side model monitoring, the network monitors performance metrics and makes decisions on model selection, activation, deactivation, handover, and / or fallback operations. In hybrid model monitoring, the UE monitors performance metrics and the network makes decisions on model selection, activation, deactivation, handover, and / or fallback operations.

[0138] As shown in Table 2, NW-side model monitoring can be easily supported for the option of performing both AI / ML model training and inference on the NW side (Opt. 1). The network has beam measurements reported by the UE, and the ground truth can be obtained by configuring beam measurements / reporting to the UE (e.g., on SSB and / or CSI-RS resources). In some implementations, the suitability of the AI ​​model can be guaranteed by design, and performance monitoring is merely a safety measure that can be considered secondary.

[0139] Table 2 For the third option (Opt.3) shown in Table 2, the situation is somewhat similar to Opt.1, as the AI ​​model deployed on the network side may have already been thoroughly tested by the network. However, since the model designer (network side) and the model inference entity (UE) are different, model performance monitoring is more useful in Opt.3 than in Opt.1.

[0140] For the second option (Opt.2) shown in Table 2, since the AI ​​model is designed by the UE side without complete information about network deployment / configuration, it can be based on educated guesses. It should be noted that some schemes exist that use area identifiers (IDs) or dataset IDs to identify network deployment / configuration at the data collection phase and at the inference phase. Matching IDs can suggest matching network deployments / configurations, thus affecting the suitability of using the AI ​​model.

[0141] However, these zone IDs / dataset IDs (especially global zone IDs / dataset IDs) can be used for reverse engineering of network deployments / configurations. For example, the deployment and diurnal adjustments of antenna modules (macro, pico, etc.) by infrastructure vendors or network operators can be collected, analyzed, and identified using the provided zone IDs / dataset IDs from networks in various geographic areas. Infrastructure vendors and / or network operators may be unwilling to allow competitors to acquire this hard-won proprietary technology in a legitimate and relatively straightforward manner.

[0142] It's worth noting that Opt.2 tends to give UE vendors more design freedom. For example, regarding product differentiation, more capable UE vendors can design, deploy, and use AI models that are better than their competitors. Therefore, using Opt.2 can be beneficial.

[0143] Test data provisioning and ground truth data provisioning .

[0144] For AI training and / or inference, validation and test data can be used at different stages of using the AI ​​model. For AI-enabled beam management, as described in this paper, test data can be used in one or both of the first case (Case 1) and the second case (Case 2).

[0145] In the first scenario, test data can be provided even before activating AI-enabled beam management inference. If the test results are unsatisfactory, the AI ​​model is not activated, and conventional beam management methods are used. The first scenario utilizes message exchange between the network and the UE. For example, Figure 10An example of test data provisioning and performance monitoring for AI beam management in a first scenario, according to certain implementations, is illustrated. In the illustrated example, UE 1002 transmits capability signaling 1006 to gNB 1004 to indicate the UE's capability for performance monitoring for AI beam management. In response, gNB 1004 provides UE 1002 with test data 1008, which UE 1002 uses to verify 1010 its AI model. The model verification process may generate performance metrics and / or pass / fail indications for the model. UE 1002 reports 1012 performance metrics and / or pass / fail to gNB 1004. Then, gNB 1004 configures 1014 measurement resources, reporting, etc., for AI-enabled beam management (AI-BM) to UE 1002.

[0146] In the second scenario, test data can be provided after activating AI-enabled beam management inference. Test data can be provided periodically, semi-persistently, or non-periodically (e.g., at the network's discretion). In some implementations, the transmission of test data can be event-triggered (e.g., a binding step that provides test data in response to a certain event).

[0147] For example, Figure 11 An example of test data provisioning and performance monitoring for AI beam management in a second scenario, according to certain implementations, is illustrated. In the illustrated example, UE 1102 transmits capability signaling 1106 to gNB 1104 to indicate the UE's capability for performance monitoring for AI beam management. In response, gNB 1104 configures 1108 measurement resources, reporting, etc., for UE 1102 for AI-BM. Then, gNB 1104 provides 1110 test data to UE 1102. UE 1102 uses the provided test data to verify 1112 its AI model and reports 1114 performance metrics and / or pass / fail to gNB 1104.

[0148] Choose the first option ( Figure 10Beam management AI models can be based on smoothness assumptions. Typically, AI models for AI beam management are much simpler than AI models for AI CSI, which may include several fully connected layers. Arbitrary continuous functions can be approximated by such AI models, which can have abrupt changes within small regions of the input, using a general approximation theorem. However, because AI beam management models are much simpler than AI CSI models, the probability of abrupt changes in AI beam management models is much smaller. If the AI ​​model trainer also ensures that there are no abrupt changes in the output within small regions of the input (e.g., to ensure robust performance in the presence of quantization or measurement errors in the AI / ML input), then the AI ​​beam management model performs well and can be characterized by multiple samples (and interpolation of the test samples can be assumed).

[0149] If RSRP measurement accuracy is not a critical issue, then verifying that the AI ​​model (e.g., trained on the UE side) is sufficiently close to the truth may be sufficient. Therefore, the first case can be used to verify the mathematical model.

[0150] Regarding the second situation ( Figure 11 The argument goes further. The second scenario also verifies the accuracy of UE measurements. For example, if the AI ​​model is fed inputs from highly accurate inputs, then mathematically, the AI ​​model can generate outputs that are sufficiently close to the truth. If, due to measurement errors by the UE, the inferred output may not be consistent with the UE's optimal choice in a large proportion of cases, then the AI ​​model may still be unsuitable for use.

[0151] Test data signaling .

[0152] Test data signaling may include, for example, control plane broadcast, multicast, or dedicated signaling.

[0153] Broadcasting (e.g., provided in System Information Block (SIBx) messages) on the control plane signals that test data is permitted for UEs in the cell to acquire test data, and the system overhead can be fixed. Alternatively, the SIBx message provides broadcast scheduling and transmission information for the test data, such as the MCS level, the number of PRBs, and the number of OFDM symbols in the PDSCH carrying the test data. Due to the size of the test data, multiple PDSCHs may be required, and therefore the PDSCHs may be expanded in the time domain. Compared to cases where the UE is constrained to acquire test data starting from the first segment of the test data, broadcast scheduling allows the UE to acquire portions of the test data from any or more locations in the broadcast schedule.

[0154] For multicast, test data can be carried on the Physical Downlink Shared Channel (PDSCH) scheduled by Group Common Downlink Control Information (DCI), or on the Physical Downlink Control Channel (PDCCH) carrying a Group Common DCI associated with a Group Common Radio Network Temporary Identifier (RNTI). Signaling overhead can be shared among multiple UEs, and specific requirements can be tailored, for example, for UEs concentrated in certain areas within a cell. A UE can be configured with such a Group Common RNTI, and another UE can be configured with the same Group Common RNTI.

[0155] For dedicated signaling, test data can be included in the MAC CE or the payload section of the PDSCH via the user plane. This offers maximum flexibility but can also be associated with maximum overhead.

[0156] Test data composition .

[0157] For beam management using spatial domain predictions, the test dataset may include multiple samples. Each sample includes an input portion and an output portion. For example, the input portion includes eight CSI resources or eight RSRP measurements from the SSB, and the output portion includes four beam IDs associated with each CSI-RS resource / TCI state. In another example, the input portion includes eight CSI resources or eight RSRP measurements from the SSB, and the output portion includes four predicted RSRPs associated with each CSI-RS resource. The predicted values ​​may be formulated as relative values ​​(e.g., the strongest RSRP is 0 dB, the weakest RSRP is -40 dB, etc.).

[0158] For beam management using time-domain prediction, the test dataset may include multiple samples. Each sample may include a sequence of input portions and an output portion. In the first example, a sequence of four input portions may be provided, where each input portion is for a time instance. An input may include eight CSI resources or eight RSRP measurements from the SSB, and the output portion may include four beam IDs, each associated with a CSI-RS resource / TCI state for a future time instance.

[0159] In another example, the input section may include eight CSI resources or eight RSRP measurements from the SSB. For example, Table 3 shows four time instances (Instance-1, Instance-2, Instance-3, and Instance-4), where each time instance includes test data for eight inputs (Input-1, Input-2, ..., Input-8). The output section may include four predicted RSRPs, each associated with a CSI-RS resource for a future time instance. The predicted values ​​may be formulated as relative values ​​(e.g., the strongest RSRP is 0 dB, the weakest RSRP is -40 dB, etc.).

[0160] Table 3 Test data overhead reduction .

[0161] In some implementations, the overhead of the test data can be expressed as (N1xM1xB1) + (N2xM2xB2) bits, where N1 is the number of time instances, M1 is the number of inputs per time instance, B1 is the number of bits used to represent the RSRP of each input, N2 is the number of prediction time instances (e.g., for beam management using spatial domain prediction, N2 = 1; for beam management using temporal domain prediction, N2 can be 1 or greater), M2 is the number of prediction beam IDs / RSRPs, and B2 is the number of bits used to represent the RSRP of each output (if the output is not in the form of a beam ID).

[0162] For the output, if beam IDs are used, a bitmap or combined index can be used to indicate the beam ID, which is more advantageous than using "M2xN2" units. However, with N1 being 100–300 time instances, the size of the test dataset cannot be considered small. It should be noted that CSI-RS resource indexes, SSB resource indexes, or TCI status indexes can be beam IDs.

[0163] Several methods can be used to compress test datasets. If carried on the data plane, the source coding algorithm can be applied to the test data. However, according to the implementation disclosed herein, overhead reduction can be provided if the test dataset is provided to the UE via RRC signaling or MAC CE. Even if beam reporting (i.e., from UE to network) and test dataset allocation for performance monitoring (i.e., from network to UE) are different, the overhead reduction schemes for them can be used for each other.

[0164] Therefore, certain implementations of test data provisioning from base station to UE utilize the selection and indication of strong beams, quantizer schemes for RSRP, quantizer design, selection and indication of reference beams, and / or other aspects described above for beam reporting for AI-enabled beam management.

[0165] For example, some implementations for test data provisioning include indications of the selected beam for each time instance or cross-time instance group from the base station to the UE.

[0166] Furthermore, or in other implementations for test data provisioning, different quantization schemes may be selected for the RSRP values ​​of the reported input and / or output data. For example, one or more (N) reference beams may be selected as reference beams, and the RSRP of each of the remaining reporting beams may be quantized using differential quantization relative to the reference beams. Alternatively, the RSRP of a beam may be quantized individually (without differential quantization).

[0167] Furthermore, or in other implementations for test data provisioning, different quantizer designs can be selected to reduce beam reporting overhead. For example, uniform quantization (in the logarithmic domain) utilizing rounding up and rounding down functions can be used, or non-uniform quantization (in the logarithmic domain) utilizing rounding up and rounding down functions can be used.

[0168] Furthermore, or in other implementations for test data provisioning, the indication of the reference beam can be per time instance or per group of time instances. A time instance may correspond to input data and / or output data.

[0169] In addition, or in other implementations for the provisioning of test data, a bitmap or combined index may indicate the strongest beam for each time instance.

[0170] In addition, or in other implementations for adapting test data, bitmaps or combined indexes may indicate the strongest beam across time instances.

[0171] Furthermore, or in other implementations for test data provisioning, a common bitmap or common composite index may be used for multiple time instances.

[0172] Two-part signaling for test data set construction .

[0173] If the differential RSRP is low, the number of beams not associated with the lowest RSRP value (e.g., -30dB for differential beams) can vary depending on channel conditions. Therefore, the test dataset size can vary, potentially leading to numerous blind detections at the UE side to decode the test dataset. To avoid blind detection at the UE side, two-part signaling can be considered.

[0174] For example, in one implementation, the test dataset sample comprises two parts. The first part of the test dataset sample has a fixed size and indicates the number of selected beams in the test data. The second part of the test dataset sample has a variable size based on the number of selected beams indicated in the first part to report the RSRP value of strong beams.

[0175] In some implementations, when multiple bitmaps are reported for different time instances, the first part includes the sum of the "1s" in the reported bitmaps. In other implementations, when a common bitmap is reported for multiple time instances, the first part includes the total number of "1s" in the common bitmap. Additionally, or in other implementations, a code state may be assigned to represent the minimum RSRP.

[0176] Figure 12 This is a flowchart illustrating a method 1200 for a base station to provide test data to a UE for performance monitoring of a model used for beam management, according to certain embodiments. Method 1200 includes: generating 1202 test data at the base station, the test data including input data corresponding to a plurality of first downlink (DL) beams at a first time instance and output data corresponding to one or more second DL beams at one or more second time instances. Method 1200 further includes: determining 1204 a selected beam from the plurality of first DL beams and one or more second DL beams. Method 1200 further includes: transmitting 1206 an indication of the selected beam and test data corresponding to the selected beam from the base station to the UE. Method 1200 further includes: receiving 1208 test results of the model used for beam management based on the test data from the UE at the base station.

[0177] In some embodiments of method 1200, determining the selected beam includes: selecting the selected beam for each first instance.

[0178] In some implementations of method 1200, the fixed number of beams selected for each first-time instance is predetermined or configured by the base station.

[0179] In some implementations, method 1200 further includes: the UE determining the number of selected beams for each first time instance within a predetermined range or within a range configured by the base station.

[0180] In some embodiments of method 1200, the indication of the selected beam includes a first bit diagram indicating each first time instance of the selected beam.

[0181] In some embodiments of method 1200, the indication of the selected beam includes a first combination index indicating each first time instance of the selected beam. For the combination function... , R Number of reporting beams, each first instance P The quantity measurement, the first combined index can be obtained by Provided. Alternatively, the UE can be configured to select up to a maximum of R The number of reporting beams, and for the combination function and each first instance PThe measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowest It is the minimum number of reporting beams selected by the UE.

[0182] In some embodiments of method 1200, the test data includes reference signal received power (RSRP) data for selected beams in each first time instance. In some embodiments, the test data does not explicitly include RSRP data for unselected beams from a plurality of first DL beams or one or more second DL beams.

[0183] In some embodiments, method 1200 further includes: indicating from the UE to the base station the strongest beam of each first time instance; and quantizing the RSRP value of the strongest beam of each first time instance at a higher resolution compared to a lower resolution of RSRP data used to quantize the remaining portion of the selected beam for each first time instance. The method may further include: using differential quantization to quantize the RSRP data of the remaining portion of the selected beam for each first time instance relative to the RSRP value of the strongest beam of each first time instance. Differential quantization may include a uniform quantization function based on high values, low values, and a quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors. In some embodiments, indicating the strongest beam of each first time instance includes using a second bitmap or... The second combined index is given.

[0184] In some embodiments of method 1200, determining the selected beam includes selecting the selected beam across a group of first time instances. The fixed number of beams selected across the group of first time instances may be predetermined or configured by the base station. Some embodiments further include: the UE determining that the number of beams selected across the group of first time instances is within a predetermined range or a range configured by the base station. An indication of the selected beam may include one or more first bit diagrams indicating the selected beam across the group of first time instances or one or more first combination indices indicating the selected beam across the group of first time instances. For the combination function... , R Number of reporting beams, groups across first-time instances P The measurement of quantity, the first combined index is by Provided. In some implementations, the UE is configured to select up to... R The number of reporting beams, and for the combination function and groups across first-time instances P The measurement of quantity, the first combined index is by Given, among which r It is an index, andR lowest This is the minimum number of reporting beams selected by the UE. In some implementations, the test data includes reference signal received power (RSRP) data for the selected beams across a group of first time instances. In some implementations, the test data does not explicitly include RSRP data for unselected beams from multiple first DL beams and one or more second DL beams.

[0185] In some embodiments, the method further includes: indicating from the UE to the base station the strongest beam in a group of first time instances; and quantizing the RSRP value of the strongest beam in the group of first time instances at a higher resolution compared to a lower resolution of RSRP data used to quantize the remaining portion of selected beams across the group of first time instances. The method may further include: using differential quantization to quantize the RSRP data of the remaining portion of selected beams across the group of first time instances relative to the RSRP value of the strongest beam in the group of first time instances. Differential quantization may include a uniform quantization function based on high values, low values, and a quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors. In some embodiments, indicating the strongest beam in a group of first time instances includes using data for the first time instance... M The second bitmap of the first instance or by The second combined index is given.

[0186] In some embodiments of method 1200, transmitting indication and test data for the selected beams includes generating a report comprising: a first portion of fixed size for indicating the number of selected beams; and a second portion of variable size based on the number of selected beams for reporting corresponding reference signal received power (RSRP) data.

[0187] In some embodiments of method 1200, the input data includes reference signal received power (RSRP) data corresponding to reference signals that correspond to a first DL beam at a first time instance. The output data may include one or more beam identifiers or beam indices corresponding to one or more second DL beams at one or more second time instances. Alternatively, the output data may include multiple predicted RSRP values ​​associated with one or more second DL beams at one or more second time instances, wherein the selected beam is selected for each second time instance or a group across selected time instances.

[0188] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 1200. This apparatus may be, for example, a base station (such as network device 1518 (base station), as described herein).

[0189] The embodiments contemplated herein include one or more non-transitory computer-readable media, which include instructions for causing the electronic device to perform one or more elements of method 1200 when executed by one or more processors of the electronic device. The non-transitory computer-readable medium may be, for example, the memory of a base station (such as memory 1522 of network device 1518 (base station), as described herein).

[0190] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 1200. This apparatus may be, for example, a base station (such as network device 1518 (base station), as described herein).

[0191] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 1200. The apparatus may be, for example, an apparatus for a base station (such as network device 1518 (base station), as described herein).

[0192] The implementation scheme envisioned herein includes a signal as described in or associated with one or more elements of method 1200.

[0193] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution of the program by a processing element will cause the processing element to perform one or more elements of method 1200. The processor may be a processor of a base station (such as processor 1520 of network device 1518 (base station), as described herein). These instructions may, for example, reside in the processor and / or in the memory of the base station (such as memory 1522 of network device 1518 (base station), as described herein).

[0194] Figure 13This is a flowchart illustrating a method 1300 for performance monitoring of AI-enabled beam management performed by a UE, according to certain embodiments. Method 1300 includes: receiving at the UE, from a base station 1302 an indication of a selected beam and test data corresponding to the selected beam, wherein the test data includes input data corresponding to multiple first downlink (DL) beams at a first time instance and output data corresponding to one or more second DL beams at one or more second time instances. Method 1300 further includes: providing input data to an AI model for beam management 1304. Method 1300 further includes comparing the output of the AI ​​model with the output data 1306 to determine a performance metric. Method 1300 further includes transmitting the performance metric from the UE to the base station 1308.

[0195] In some embodiments of method 1300, the selected beam is selected per first time instance. The fixed number of beams selected per first time instance may be predetermined or configured by the base station. Alternatively, the number of beams selected per first time instance may be within a predetermined range or a range configured by the base station.

[0196] In some embodiments of method 1300, the indication of the selected beam includes a first bit diagram indicating each first time instance of the selected beam or a first combination index indicating each first time instance of the selected beam. For the combination function... , R Number of reporting beams, each first instance P The measurement of quantity, the first combined index is by Provided. In some implementations, up to [number] can be selected. R The number of reporting beams, and where for the combination function and each first instance P The measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowest This is the minimum number of reporting beams selected by the UE. Test data may include Reference Signal Received Power (RSRP) data for the selected beams in each first-time instance. In some implementations, the test data may not explicitly include RSRP data for the unselected beams corresponding to the first set of DL RS measured by the UE.

[0197] In some embodiments, method 1300 further includes receiving at the base station from the UE: another indication of the strongest beam for each first time instance; and a quantized RSRP value of the strongest beam for each first time instance at a higher resolution compared to a lower resolution of RSRP data used to quantize the remaining portion of the selected beam for each first time instance. The RSRP data of the remaining portion of the selected beam for each first time instance may be quantized relative to the RSRP value of the strongest beam for each first time instance using differential quantization. Differential quantization may include a uniform quantization function based on high values, low values, and quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on quantization boundaries and minimizing quantization errors. In some embodiments, the other indication of the strongest beam for each first time instance includes a second bitmap or is derived from... The second combined index is given.

[0198] In some embodiments of method 1300, the selected beam is selected across a group of first time instances. The fixed number of beams selected across a group of first time instances may be predetermined or configured by the base station. Alternatively, the number of beams selected across a group of first time instances may be within a predetermined range or within a range configured by the base station. An indication of the selected beam may include one or more first bit diagrams indicating the selected beams across a group of first time instances or one or more first combination indices indicating the selected beams across a group of first time instances. For the combination function... , R Number of reporting beams, groups across first-time instances P The quantity measurement, the first combined index can be obtained by Provided. In some implementations, at most [number] options are selected. R The number of reporting beams, and where for the combination function and groups across first-time instances P The measurement of quantity, the first combined index is by Given, among which r It is an index, and R lowestThis is the minimum number of reporting beams selected by the UE. Test data may include reference signal received power (RSRP) data for the selected beams across a group of first time instances. In some embodiments, the test data does not explicitly include RSRP data for unselected beams corresponding to a first set of DL RS measured by the UE. In some embodiments, the method further includes receiving from the UE at the base station: another indication of the strongest beam in the group of first time instances; and a quantized RSRP value of the strongest beam in the group of first time instances at a higher resolution compared to a lower resolution used to quantize the RSRP data of the remaining selected beams across the group of first time instances. The RSRP data of the remaining selected beams across the group of first time instances may be quantized using differential quantization relative to the RSRP value of the strongest beam in the group of first time instances. In some embodiments, differential quantization includes a uniform quantization function based on high values, low values, and quantization step size. Alternatively, differential quantization may include a non-uniform quantization function based on quantization boundaries and minimizing quantization errors. Another indication of the strongest beam in the group of first time instances includes RSRP data for the group of first time instances. M The second bitmap of the first instance or by The second combined index is given.

[0199] In some embodiments of method 1300, the indication and test data for the selected beams include a report comprising: a first portion of fixed size for indicating the number of selected beams; and a second portion of variable size for reporting corresponding reference signal received power (RSRP) data based on the number of selected beams.

[0200] In some embodiments of method 1300, the input data includes reference signal received power (RSRP) data corresponding to reference signals that correspond to a first DL beam at a first time instance. The output data may include one or more beam identifiers or beam indices corresponding to one or more second DL beams at one or more second time instances, or the output data may include multiple predicted RSRP values ​​associated with one or more second DL beams at one or more second time instances. The selected beam may be selected for each second time instance or a group across selected time instances.

[0201] The embodiments contemplated herein include an apparatus comprising components for performing one or more elements of method 1300. This apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0202] The embodiments contemplated herein include one or more non-transitory computer-readable media, the one or more non-transitory computer-readable media including instructions for causing the electronic device to perform one or more elements of method 1300 when executed by one or more processors of the electronic device. The non-transitory computer-readable medium may be, for example, the memory of a UE (such as memory 1506 of a wireless device 1502 (UE), as described herein).

[0203] The embodiments contemplated herein include an apparatus comprising logic components, modules, or circuitry for performing one or more elements of method 1300. This apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0204] The embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media including instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of method 1300. The apparatus may be, for example, a UE (such as wireless device 1502 (UE), as described herein).

[0205] The implementation scheme envisioned herein includes a signal as described in or associated with one or more elements of method 1300.

[0206] The embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution by a processor will cause the processor to perform one or more elements of method 1300. The processor may be a processor of the UE (such as processor 1504 of wireless device 1502 (UE), as described herein). These instructions may, for example, reside in the processor and / or in the memory of the UE (such as memory 1506 of wireless device 1502 (UE), as described herein).

[0207] Figure 14 An example architecture of a wireless communication system 1400 according to an embodiment disclosed herein is illustrated. The following description is provided for an example wireless communication system 1400 operating in conjunction with LTE system standards and / or 5G or NR system standards provided by 3GPP technical specifications.

[0208] like Figure 14As shown, the wireless communication system 1400 includes UE 1402 and UE 1404 (but any number of UEs may be used). In this example, UE 1402 and UE 1404 are exemplified as smartphones (e.g., handheld touchscreen mobile computing devices capable of connecting to one or more cellular networks), but may also include any mobile or non-mobile computing device configured for wireless communication.

[0209] UE 1402 and UE 1404 can be configured to be communicatively coupled to RAN 1406. In an implementation, RAN 1406 can be NG-RAN, E-UTRAN, etc. UE 1402 and UE 1404 utilize connections (or channels) with RAN 1406 (shown as connection 1408 and connection 1410, respectively), where each connection (or channel) includes a physical communication interface. RAN 1406 may include one or more base stations (such as base station 1412 and base station 1414) implementing connection 1408 and connection 1410.

[0210] In this example, Connection 1408 and Connection 1410 are air interfaces that implement this type of communication coupling and can conform to the RAT used by RAN 1406, such as LTE and / or NR, for example.

[0211] In some implementations, UE 1402 and UE 1404 may also exchange communication data directly via sidelink interface 1416. UE 1404 is shown configured to access an access point (shown as AP 1418) via connection 1420. By way of example, connection 1420 may include a local wireless connection, such as a connection conforming to any IEEE 802.11 protocol, while AP 1418 may include Wi-Fi. ® Router. In this example, AP 1418 can connect to another network (e.g., the Internet) without using CN 1424.

[0212] In the implementation, UE 1402 and UE 1404 may be configured to communicate with each other or with base station 1412 and / or base station 1414 on a multi-carrier communication channel using orthogonal frequency division multiplexing (OFDM) communication signals according to various communication technologies, such as, but not limited to, orthogonal frequency division multiple access (OFDMA) communication technology (e.g., for downlink communication) or single-carrier frequency division multiple access (SC-FDMA) communication technology (e.g., for uplink and ProSe or sidelink communication)); however, the scope of the implementation is not limited in this respect. The OFDM signal may include multiple orthogonal subcarriers.

[0213] In some implementations, all or part of base station 1412 or base station 1414 may be implemented as one or more software entities running on a server computer as part of a virtual network. Furthermore, or in other implementations, base station 1412 or base station 1414 may be configured to communicate with each other via interface 1422. In implementations where wireless communication system 1400 is an LTE system (e.g., when CN 1424 is an EPC), interface 1422 may be an X2 interface. This X2 interface may be defined between two or more base stations (e.g., two or more eNBs, etc.) connected to the EPC and / or between two eNBs connected to the EPC. In implementations where wireless communication system 1400 is an NR system (e.g., when CN 1424 is a 5GC), interface 1422 may be an Xn interface. This Xn interface is defined between two or more base stations (e.g., two or more gNBs, etc.) connected to the 5GC, between base station 1412 (e.g., a gNB) connected to the 5GC and an eNB, and / or between two eNBs connected to the 5GC (e.g., CN 1424).

[0214] RAN 1406 is shown communicatively coupled to CN 1424. CN 1424 may include one or more network elements 1426 configured to provide various data and telecommunications services to customers / subscribers (e.g., users of UE 1402 and UE 1404) connected to CN 1424 via RAN 1406. Components of CN 1424 may be implemented in a single physical device or a separate physical device, including components for reading and executing instructions from machine-readable or computer-readable media (e.g., non-transitory machine-readable storage media).

[0215] In this implementation, CN 1424 may be an EPC, and RAN 1406 may be connected to CN 1424 via S1 interface 1428. In this implementation, S1 interface 1428 may be divided into two parts: an S1 user plane (S1-U) interface, which carries service data between base station 1412 or base station 1414 and the service gateway (S-GW); and an S1-MME interface, which is the signaling interface between base station 1412 or base station 1414 and the mobility management entity (MME).

[0216] In the implementation, CN 1424 may be a 5GC, and RAN 1406 may be connected to CN 1424 via NG interface 1428. In the implementation, NG interface 1428 may be divided into two parts: an NG user plane (NG-U) interface, which carries service data between base station 1412 or base station 1414 and the User Plane Function (UPF); and an S1 control plane (NG-C) interface, which is the signaling interface between base station 1412 or base station 1414 and the Access and Mobility Management Function (AMF).

[0217] Generally, application server 1430 may be an element that provides Internet Protocol (IP) bearer resources (e.g., packet-switched data services) for use with CN 1424. Application server 1430 may also be configured to support one or more communication services (e.g., VoIP sessions, group communication sessions, etc.) for UE 1402 and UE 1404 via CN 1424. Application server 1430 can communicate with CN 1424 via IP communication interface 1432.

[0218] Figure 15 A system 1500 for performing signaling 1534 between a wireless device 1502 and a network device 1518 according to an embodiment disclosed herein is illustrated. System 1500 may be part of a wireless communication system as described herein. Wireless device 1502 may be, for example, a UE of a wireless communication system. Network device 1518 may be, for example, a base station (e.g., an eNB or gNB) of a wireless communication system.

[0219] Wireless device 1502 may include one or more processors 1504. Processor 1504 is executable instructions to perform various operations of wireless device 1502 as described herein. Processor 1504 may include one or more baseband processors, which are implemented using, for example, a central processing unit (CPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), controller, field-programmable gate array (FPGA) device, another hardware device, firmware device, or any combination thereof configured to perform the operations described herein.

[0220] Wireless device 1502 may include memory 1506. Memory 1506 may be a non-transitory computer-readable storage medium that stores instructions 1508, which may include, for example, instructions executed by processor 1504. Instructions 1508 may also be referred to as program code or a computer program. Memory 1506 may also store data used by processor 1504 and results calculated by the processor.

[0221] Wireless device 1502 may include one or more transceivers 1510, which may include radio frequency (RF) transmitter circuitry and / or receiver circuitry, which use the antenna 1512 of wireless device 1502 to facilitate signaling (e.g., signaling 1534) to and / or from wireless device 1502 and other devices (e.g., network device 1518) according to a corresponding RAT.

[0222] Wireless device 1502 may include one or more antennas 1512 (e.g., one, two, four, or more). In embodiments with multiple antennas 1512, wireless device 1502 may fully utilize the spatial diversity of these multiple antennas 1512 to transmit and / or receive multiple different data streams on the same time-frequency resource. This behavior may be referred to as, for example, multiple-input multiple-output (MIMO) behavior (referring to multiple antennas used at each of the transmitting and receiving devices to implement this aspect). MIMO transmission by wireless device 1502 may be implemented according to pre-decoding (or digital beamforming) applied to wireless device 1502, which multiplexes the data streams among antennas 1512 based on known or assumed channel characteristics, such that each data stream is received with appropriate signal strength relative to the other streams at a desired location in the spatial domain (e.g., the location of the receiver associated with that data stream). Some implementations may use a single-user MIMO (SU-MIMO) approach (where all data streams are directed to a single receiver) and / or a multi-user MIMO (MU-MIMO) approach (where individual data streams may be directed to individual (different) receivers at different locations in the spatial domain).

[0223] In some implementations with multiple antennas, wireless device 1502 can implement analog beamforming technology, whereby the phase of the signal transmitted by antenna 1512 is relatively adjusted so that the (joint) transmission of antenna 1512 can be directed (this is sometimes referred to as beam control).

[0224] Wireless device 1502 may include one or more interfaces 1514. Interfaces 1514 can be used to provide input to or output to wireless device 1502. For example, wireless device 1502 (UE) may include interfaces 1514 such as microphones, speakers, touchscreens, and buttons to allow users of the UE to input to and / or output to the UE. Other interfaces of such a UE may consist of transmitters, receivers, and other circuitry that allow communication between the UE and other devices (e.g., in addition to the transceiver 1510 / antenna 1512 already described), and may be configured according to known protocols (e.g., Wi-Fi). ® Bluetooth ® (etc.) to perform the operation.

[0225] Wireless device 1502 may include beam management module 1516. Beam management module 1516 may be implemented in hardware, software, or a combination thereof. For example, beam management module 1516 may be implemented as a processor, circuitry, and / or instructions 1508 stored in memory 1506 and executed by processor 1504. In some examples, beam management module 1516 may be integrated within processor 1504 and / or transceiver 1510. For example, beam management module 1516 may be implemented in a combination of software components (e.g., executed by a DSP or general-purpose processor) and hardware components (e.g., logic gates and circuitry) within processor 1504 or transceiver 1510.

[0226] Beam management module 1516 can be used in various aspects of this disclosure, such as Figure 8 , Figure 10 , Figure 11 and Figure 13 All aspects.

[0227] Network device 1518 may include one or more processors 1520. Processor 1520 is executable instructions to perform various operations of network device 1518 as described herein. Processor 1520 may include one or more baseband processors, which are implemented using, for example, a CPU, DSP, ASIC, controller, FPGA device, another hardware device, firmware device, or any combination thereof configured to perform the operations described herein.

[0228] Network device 1518 may include memory 1522. Memory 1522 may be a non-transitory computer-readable storage medium that stores instructions 1524, which may include, for example, instructions executed by processor 1520. Instructions 1524 may also be referred to as program code or a computer program. Memory 1522 may also store data used by processor 1520 and results calculated by the processor.

[0229] Network device 1518 may include one or more transceivers 1526, which may include RF transmitter circuitry and / or receiver circuitry that uses the antenna 1528 of network device 1518 to facilitate signaling (e.g., signaling 1534) to and / or from network device 1518 and other devices (e.g., wireless device 1502) in accordance with the corresponding RAT.

[0230] Network device 1518 may include one or more antennas 1528 (e.g., one, two, four or more). In embodiments having multiple antennas 1528, network device 1518 may perform MIMO, digital beamforming, analog beamforming, beam control, etc., as described.

[0231] Network device 1518 may include one or more interfaces 1530. Interface 1530 can be used to provide input to or output to network device 1518. For example, network device 1518 as a base station may include interface 1530 consisting of transmitters, receivers and other circuitry (e.g., in addition to the transceiver 1526 / antenna 1528 already described), which enables the base station to communicate with other equipment in the core network and / or enables the base station to communicate with external networks, computers and databases, etc., for the purpose of operating, managing and maintaining the base station or other equipment operatively connected to the base station.

[0232] Network device 1518 may include beam management module 1532. Beam management module 1532 may be implemented in hardware, software, or a combination thereof. For example, beam management module 1532 may be implemented as a processor, circuitry, and / or instructions 1524 stored in memory 1522 and executed by processor 1520. In some examples, beam management module 1532 may be integrated within processor 1520 and / or transceiver 1526. For example, beam management module 1532 may be implemented using a combination of software components (e.g., executed by a DSP or general-purpose processor) and hardware components (e.g., logic gates and circuitry) within processor 1520 or transceiver 1526.

[0233] Beam management module 1532 can be used in various aspects of this disclosure, such as Figure 9 , Figure 10 , Figure 11 and Figure 12 All aspects.

[0234] For one or more embodiments, at least one of the components set forth in one or more of the foregoing figures may be configured to perform one or more operations, techniques, processes, and / or methods as described herein. For example, a baseband processor as described herein in conjunction with one or more of the foregoing figures may be configured to operate according to one or more of the examples set forth herein. Similarly, circuitry associated with a UE, base station, network element, etc., as described above in conjunction with one or more of the foregoing figures may be configured to operate according to one or more of the examples set forth herein.

[0235] Unless otherwise explicitly stated, any of the above embodiments may be combined with any other embodiment (or combination of embodiments). The foregoing description of one or more specific embodiments provides illustration and description, but is not intended to be exhaustive or to limit the scope of the embodiments to the precise form disclosed. In view of the teachings above, modifications and variations are possible, or modifications and variations may be derived from practice of various embodiments.

[0236] Implementations and specific embodiments of the systems and methods described herein may include various operations embodied in machine-executable instructions to be executed by a computer system. The computer system may include one or more general-purpose or special-purpose computers (or other electronic devices). The computer system may include hardware components, including specific logical parts for performing the operations; or may include a combination of hardware, software, and / or firmware.

[0237] It should be recognized that the systems described herein include descriptions of specific implementations. These implementations may be combined into a single system, partially integrated into other systems, divided into multiple systems, or otherwise partitioned or combined. Furthermore, it is contemplated that parameters, attributes, aspects, etc., of one implementation may be used in one implementation. For clarity, these parameters, attributes, aspects, etc., are described only in one or more implementations, and it should be recognized that, unless specifically stated herein, these parameters, attributes, aspects, etc., may be combined with or substituted for parameters, attributes, aspects, etc., of another implementation.

[0238] As is widely recognized, the use of personally identifiable information should comply with privacy policies and practices that are generally accepted to meet or exceed industry or governmental requirements for protecting user privacy. Specifically, personally identifiable information data should be managed and processed to minimize the risk of unintentional or unauthorized access or use, and the nature of authorized use should be clearly explained to users.

[0239] Although the foregoing has been described in considerable detail for clarity, it will be apparent that certain changes and modifications can be made without departing from the principles of the invention. It should be noted that there are many alternative ways to implement both the processes and apparatus described herein. Therefore, embodiments of the invention should be considered illustrative rather than restrictive, and this description is not limited to the details given herein, but can be modified within the scope and equivalents of the appended claims.

Claims

1. A method for performing beam management using time-domain prediction for user equipment (UE), the method comprising: The UE receives a first configuration from the base station for a first set of downlink (DL) reference signals (RS); The first set of DL RS is measured at the UE at multiple measurement points; Based on the measurements, one or more selected beams corresponding to the first set of DL RS are determined for reporting to the base station; as well as The UE sends to the base station an indication of one or more selected beams and feedback data corresponding to at least one of the selected beams.

2. The method according to claim 1, further comprising: The UE receives from the base station the feedback data based on the time-domain prediction for a second set of DL RS.

3. The method of claim 1, wherein determining the one or more selected beams comprises: Each of the plurality of measurement opportunities selects one or more selected beams.

4. The method of claim 3, wherein the fixed number of the one or more selected beams at each timing is predetermined or configured by the base station.

5. The method according to claim 3, further comprising: The number of the one or more selected beams at each timing is determined by the UE within a predetermined range or within a range configured by the base station.

6. The method of claim 3, wherein the indication of the one or more selected beams comprises: The first diagram indicating each timing of the one or more selected beams.

7. The method of claim 3, wherein the indication of the one or more selected beams includes a first combination index indicating each timing of the one or more selected beams.

8. The method of claim 7, wherein for the combination function , R Number of reporting beams and per timing P The quantity measurement, the first combined index is by Provided.

9. The method of claim 7, wherein the UE is configured to select at most R The number of reporting beams, and where for the combination function and every opportunity P The quantity measurement, the first combined index is by Given, among which r It is an index, and R lowest It is the minimum number of reporting beams selected by the UE.

10. The method of claim 9, wherein the feedback data includes reference signal received power (RSRP) data of the one or more selected beams at each timing.

11. The method of claim 10, wherein the feedback data does not explicitly include the RSRP data of the unselected beam corresponding to the first set of DL RS measured by the UE.

12. The method of claim 10, further comprising: The UE indicates to the base station the strongest beam of each of the plurality of measurement opportunities; as well as The RSRP value of the strongest beam at each time point is quantized at a higher resolution compared to the lower resolution of the RSRP data used to quantize the remainder of the one or more selected beams at each time point.

13. The method according to claim 12, further comprising: Differential quantization is used to quantize the RSRP data of the remaining portion of the one or more selected beams at each time point relative to the RSRP value of the strongest beam at each time point.

14. The method of claim 13, wherein the differential quantization comprises a uniform quantization function based on high values, low values, and quantization step size.

15. The method of claim 13, wherein the differential quantization comprises a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors.

16. The method of claim 12, wherein indicating the strongest beam at each moment comprises using a second bitmap or by... The second combined index is given.

17. The method of claim 1, wherein determining the one or more selected beams comprises: Select one or more selected beams across the time group of the plurality of measurement times.

18. The method of claim 17, wherein the fixed number of the one or more selected beams across the time group is predetermined or configured by the base station.

19. The method of claim 17, further comprising: The number of one or more selected beams across the time group is determined by the UE within a predetermined range or within a range configured by the base station.

20. The method of claim 17, wherein the indication of the one or more selected beams comprises one or more first-order maps indicating the one or more selected beams across the timing group.

21. The method of claim 17, wherein the indication of the one or more selected beams includes one or more first combination indices indicating the one or more selected beams across the timing group.

22. The method of claim 21, wherein for the combination function , R Number of reporting beams and across the timing group P The quantity measurement, the first combined index is by Provided.

23. The method of claim 21, wherein the UE is configured to select at most R The number of reporting beams, and where for the combination function and across the timing group P The quantity measurement, the first combined index is by Given, among which r It is an index, and R lowest It is the minimum number of reporting beams selected by the UE.

24. The method of claim 23, wherein the feedback data includes reference signal received power (RSRP) data for the one or more selected beams across the timing group.

25. The method of claim 24, wherein the feedback data does not explicitly include the RSRP data of the unselected beam corresponding to the first set of DL RS measured by the UE.

26. The method of claim 24, further comprising: The UE indicates to the base station the strongest beam in the timing group among the plurality of measurement timings; as well as The RSRP value of the strongest beam in the timing group is quantized at a higher resolution compared to the lower resolution of the RSRP data used to quantify the remainder of the one or more selected beams across the timing group.

27. The method according to claim 26, further comprising: Differential quantization is used to quantize the RSRP data of the remaining portion of the one or more selected beams across the timing group relative to the RSRP value of the strongest beam in the timing group.

28. The method of claim 27, wherein the differential quantization comprises a uniform quantization function based on high values, low values, and quantization step size.

29. The method of claim 27, wherein the differential quantization comprises a non-uniform quantization function based on determining quantization boundaries and minimizing quantization errors.

30. The method of claim 26, wherein indicating the strongest beam in the timing group comprises: Use for the timing group M The second bitmap of the timing or by The second combined index is given.

31. The method of claim 1, wherein transmitting the indication and feedback data for the one or more selected beams comprises: Generate a report, which includes: A first portion of fixed size used to indicate the number of the one or more selected beams; and A second portion of variable size, used for reporting corresponding reference signal received power (RSRP) data, based on the number of the one or more selected beams.

32. A method for a base station to perform temporal prediction for beam management, the method comprising: The first set of downlink (DL) reference signals (RS) is sent from the base station to the user equipment (UE); Received from the UE at the base station: Indication of one or more selected beams corresponding to the first set of DL RS measured by the UE at multiple measurement times; Feedback data corresponding to at least one of the selected beams; Based on the feedback data, a neural network model is used to determine the temporal prediction of DL beam information for one or more future time periods.

33. The method according to claim 32, further comprising: Based on the time-domain prediction, measurement resources corresponding to a second set of DL RS are configured for the UE at one or more future time periods.

34. The method of claim 32, wherein the one or more selected beams are time-selected among the plurality of measurement times.

35. The method of claim 34, wherein the fixed number of the one or more selected beams at each timing is predetermined or configured by the base station.

36. The method of claim 34, wherein the number of the one or more selected beams at each timing is within a predetermined range or within a range configured by the base station.

37. The method of claim 34, wherein the indication of the one or more selected beams comprises: The first diagram indicating each timing of the one or more selected beams.

38. The method of claim 34, wherein the indication of the one or more selected beams includes a first combination index indicating each timing of the one or more selected beams.

39. The method of claim 38, wherein for the combination function , R Number of reporting beams, per timing P The quantity measurement, the first combined index is by Provided.

40. The method of claim 38, wherein up to R number of reporting beams can be selected, and wherein for the combination function and every opportunity P The quantity measurement, the first combined index is by Given, among which r It is an index, and R lowest It is the minimum number of reporting beams selected by the UE.

41. The method of claim 40, wherein the feedback data includes reference signal received power (RSRP) data of the one or more selected beams at each timing.

42. The method of claim 41, wherein the feedback data does not explicitly include the RSRP data of the unselected beam corresponding to the first set of DL RS measured by the UE.

43. The method according to claim 41, further comprising: Received from the UE at the base station: Another indication of the strongest beam of each of the plurality of measurement opportunities; and The RSRP value quantized at a higher resolution for the strongest beam at each time point compared to the lower resolution of the RSRP data used to quantize the remaining portion of the one or more selected beams at each time point.

44. The method of claim 43, wherein the RSRP data of the remaining portion of the one or more selected beams at each timing is quantized using differential quantization relative to the RSRP value of the strongest beam at each timing.

45. The method of claim 44, wherein the differential quantization comprises a uniform quantization function based on high values, low values, and quantization step size.

46. ​​The method of claim 44, wherein the differential quantization comprises a non-uniform quantization function based on quantization boundaries and minimizing quantization errors.

47. The method of claim 43, wherein the other indication of the strongest beam for each timing comprises a second bitmap or is derived from... The second combined index is given.

48. The method of claim 32, wherein the one or more selected beams are selected across a time group of the plurality of measurement times.

49. The method of claim 48, wherein a fixed number of the one or more selected beams across the time group is predetermined or configured by the base station.

50. The method of claim 48, wherein the number of the one or more selected beams across the timing group is within a predetermined range or a range configured by the base station.

51. The method of claim 48, wherein the indication of the one or more selected beams comprises one or more first-order maps indicating the one or more selected beams across the timing group.

52. The method of claim 48, wherein the indication of the one or more selected beams includes one or more first combination indices indicating the one or more selected beams across the timing group.

53. The method of claim 52, wherein for the combination function , R Number of reporting beams, across the aforementioned timing groups P The quantity measurement, the first combined index is by Provided.

54. The method of claim 52, wherein at most R number of reporting beams are selected, and wherein for the combination function and across the timing group P The quantity measurement, the first combined index is by Given, among which r It is an index, and R lowest It is the minimum number of reporting beams selected by the UE.

55. The method of claim 54, wherein the feedback data includes reference signal received power (RSRP) data for the one or more selected beams across the timing group.

56. The method of claim 55, wherein the feedback data does not explicitly include the RSRP data of the unselected beam corresponding to the first set of DL RS measured by the UE.

57. The method of claim 55, further comprising: Received from the UE at the base station: Another indication of the strongest beam in the group of the plurality of measurement opportunities; and The RSRP value quantized at a higher resolution for the strongest beam in the timing group compared to the lower resolution of the RSRP data used to quantize the remaining portion of the selected beams across the timing group.

58. The method of claim 57, wherein the RSRP data of the remaining portion of the one or more selected beams across the timing group is quantized using differential quantization relative to the RSRP value of the strongest beam in the timing group.

59. The method of claim 58, wherein the differential quantization comprises a uniform quantization function based on high values, low values, and quantization step size.

60. The method of claim 58, wherein the differential quantization comprises a non-uniform quantization function based on quantization boundaries and minimizing quantization errors.

61. The method of claim 57, wherein the further indication of the strongest beam in the timing group includes targeting the timing group... M The second bitmap of the timing or by The second combined index is given.

62. The method of claim 32, wherein the indication and feedback data for the one or more selected beams include a report, the report comprising: A first portion of fixed size used to indicate the number of the one or more selected beams; and A second portion of variable size, used for reporting corresponding reference signal received power (RSRP) data, based on the number of the one or more selected beams.

Images