Systems and methods for using capacity loss as a key performance indicator for channel state information artificial intelligence / machine learning compression / decompression reporting

Abstract

Systems and methods for using capacity loss as a key performance indicator (KPI) for channel state information (CSI) artificial intelligence (AI) / machine learning (ML) compression / decompression reporting are discussed herein. A user equipment (UE) identifies, based on a measurement of a first reference signal received from a base station, a precoding matrix V that is ideal for DL precoding; generates CSI feedback indicating V; applies the CSI feedback at an encoder of an AI / ML mode to generate compressed CSI feedback; sends, to the base station, the compressed CSI feedback; identifies, based on a second measurement of a precoded second reference signal received from the base station, a precoding matrix V c that the base station is using for the DL precoding; calculates, using V and V c , a capacity loss S associated with using V c for the DL precoding; and sends, to the base station, a reporting message reporting the capacity loss CLoss.

Description

SYSTEMS AND METHODS FOR USING CAPACITY LOSS AS A KEY PERFORMANCE INDICATOR FOR CHANNEL STATE INFORMATION ARTIFICIAL INTELLIGENCE / MACHINE LEARNING COMPRESSION / DECOMPRESSION REPORTINGTECHNICAL FIELD

[0001] This application relates generally to wireless communication systems, including wireless communication systems using AI / ML models for CSI feedback compression and decompression.BACKGROUND

[0002] Wireless mobile communication technology uses various standards and protocols to transmit data between a base station and a wireless communication device. 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 standard for Wireless Local Area Networks (WLAN) (commonly known to industry groups as Wi-Fi®).

[0003] As contemplated by the 3GPP, different wireless communication systems' standards and protocols can use various radio access networks (RANs) for communicating betw een a base station of the RAN (which may also sometimes be referred to generally as a RAN node, a network node, or simply a node) and a wireless communication device known as a user equipment (UE). 3GPP RANs can include, for example. Global System for Mobile communications (GSM), Enhanced Data Rates for GSM Evolution (EDGE) RAN (GERAN), Universal Terrestrial Radio Access Netw ork (UTRAN), Evolved Universal Terrestrial Radio Access Network (E-UTRAN), and / or Next-Generation Radio Access Network (NG-RAN).

[0004] Each RAN may use one or more radio access technologies (RATs) to perform communication between the base station and the UE. For example, the GERAN implements GSM and / or EDGE RAT, the UTRAN implements Universal Mobile Telecommunication System (UMTS) RAT or other 3GPP RAT, the E-UTRAN implements LTE RAT (sometimes simply referred to as LTE), and NG-RAN implements NR RAT (sometimes referred to herein as 5G RAT, 5GNR RAT, or simply NR). In1P70532WO1 4913-1823-7501\1certain deployments, the E-UTRAN may also implement NR RAT. In certain deployments, NG-RAN may also implement LTE RAT.

[0005] A base station used by a RAN may 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 evolved Node B, enhanced Node B, eNodeB, or eNB). One example of an NG-RAN base station is a next generation Node B (also sometimes referred to as a g Node B or gNB).

[0006] A RAN provides its communication services with external entities through its connection to a core network (CN). For example, E-UTRAN may utilize an Evolved Packet Core (EPC) while NG-RAN may utilize a 5G Core Network (5GC).

[0007] Frequency bands for 5G NR may be separated into two or more different frequency ranges. For example. Frequency Range 1 (FR1) may include frequency bands operating in sub-6 gigahertz (GHz) frequencies, some of which are bands that may be used by previous standards, and may potentially be extended to cover new spectrum offerings from 410 megahertz (MHz) to 7125 MHz. Frequency Range 2 (FR2) may include frequency bands from 24.25 GHz to 52.6 GHz. Note that in some systems. FR2 may also include frequency bands from 52.6 GHz to 71 GHz (or beyond). Bands in the millimeter wave (mmWave) range of FR2 may have smaller coverage but potentially higher available bandwidth than bands in FR1. Skilled persons will recognize these frequency ranges, which are provided by way of example, may change from time to time or from region to region.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0008] To easily identify the discussion of any particular element or act, the most significant digit or digits in a reference number refer to the figure number in which that element is first introduced.

[0009] FIG. 1 illustrates a diagram showing an example of a two-sided AI / ML model for CSI compression and decompression, according to embodiments discussed herein.

[0010] FIG. 2 illustrates a flow diagram for communications between a base station and a UE corresponding to the use of network-side SGCS-based monitoring of an AI / ML model for CSI compression and decompression.2P70532WO1 4913-1823-7501\1

[0011] FIG. 3 illustrates a flow diagram for communications between a base station and a UE corresponding to the use of UE-side SGCS-based monitoring of an AI / ML model for CSI compression and decompression.

[0012] FIG. 4 illustrates a flow diagram for communications between a base station and a UE corresponding to KPI-based monitoring of an AI / ML model for CSI compression and decompression.

[0013] FIG. 5 illustrates a method of a UE, according to embodiments discussed herein.

[0014] FIG. 6 illustrates a method of a base station, according to embodiments discussed herein.

[0015] FIG. 7 illustrates an example architecture of a wireless communication system, according to embodiments disclosed herein.

[0016] FIG. 8 illustrates a system for performing signaling between a wireless device and a network device, according to embodiments disclosed herein.DETAILED DESCRIPTION

[0017] Various embodiments are described with regard to a UE. However, reference to a UE is merely provided for illustrative purposes. The example embodiments may be utilized with any electronic component that may establish a connection to a network and is configured with the hardware, software, and / or firmware to exchange information and data with the network. Therefore, the UE as described herein is used to represent any appropriate electronic component.

[0018] Downlink channel state information (CSI) may be sent from a UE to a base station through feedback channels. The base station may use the CSI feedback to, for example, reduce interference and increase throughput for massive multiple input multiple output (MIMO) communication.

[0019] It may be understood that such feedback represents a relatively large amount of signaling overhead. In various wireless communication systems, vector quantization or codebook-based feedback may be used with the expectation of reducing this overhead. The feedback quantities resulting from these approaches, however, scale linearly with the number of transmit antennas. Accordingly, these approaches may still, for some cases (e.g., when hundreds or thousands of centralized or distributed transmit antennas are used) represent a high level of signaling overhead.3P70532WO1 4913-1823-7501\1

[0020] Accordingly, in various wireless communication systems, artificial intelligence (AI) / machine learning (ML)-based CSI encoding / compression and decoding / decompression mechanisms may be used in order to reduce signaling overhead associated with the transmission of CSI from the UE to the base station.

[0021] FIG. 1 illustrates a diagram 100 showing an example of a two-sided AI / ML model 102 for CSI compression and decompression, according to embodiments discussed herein. The AI / ML model 102 includes the (logical) UE side 104 (a portion of the AI / ML model 102 that exists at a UE 106) and the (logical) network side 108 (a portion of the AI / ML model 102 that exists at, for example, a base station of a network 110). as illustrated.

[0022] The UE side 104 of the AI / ML model 102 that operates at the UE 106 includes an encoder 112. The encoder 112 is configured to accept an input 116 and to provide a bitstream 118 that is based on that input 116 as output. As illustrated, in some cases, the input 116 may include CSI, such as a raw downlink (DL) channel estimate or precoder information (such as a codebook-based indication of a precoder W and / or a set of eigenvectors corresponding to a precoder W).

[0023] The bitstream 118 is then transmitted from the UE 106 to the network 110.

[0024] The network side 108 of the AI / ML model 102 that operates at the network 110 (e.g.. a base station of the network 110) includes a decoder 114. The decoder 114 is configured to accept the bitstream 118 as input and to decode information therein as output 120 to the network 110 for further processing. In this way, the information from the input 116 is made known to the network 110. Accordingly, in some cases, the output 120 may be understood to include CSI, such as a raw DL channel estimate or precoder information (such as a codebook-based indication of a precoder W and / or a set of eigenvectors corresponding to a precoder W), corresponding to the format of the input 116.

[0025] Based on the encoding mechanism used by the encoder 112, the bitstream 118 may be smaller than the raw or data as presented in the input 116. The encoder 112 may thus be understood to “compress” the input 116 into the bitstream, which is correspondingly understood to represent “compressed” information.

[0026] The result is that the transmission of the bitstream 118 to from the UE 106 to the network 110 results in the use of fewer radio resources than an alternative case where the4P70532WO1 4913-1823-7501\1input 116 is itself sent from the UE 106 to the network side 108 without such encoding / compression.

[0027] The output 120 may correspondingly be referred to variously herein as “decoded,"’ “decompressed,” “recovered,” etc.

[0028] It may be beneficial to monitor the operation of an AI / ML model 102 model for CSI compression and decompression. This monitoring may be used to determine that, for example, a given AI / ML model is no longer performing with acceptable accuracy / precision (in the sense that, for example, the output 120 of the decoder 114 exhibits a sufficient or insufficient match to the input 116 of the encoder 112), and thus that the AI / ML model needs to be retrained and / or that use of the AI / ML model should be ended.

[0029] Various aspects with respect to the utility and / or feasibility of various types of monitoring for an AI / ML model for CSI compression and decompression may be considered. For example, when considering aspects for network-side monitoring for such an AI / ML model, overhead, latency, complexity, monitoring accuracy, and / or corresponding UE capabilities may be relevant factors. Such monitoring may be based on a target CSI that is reported by the UE to the network via an enhanced Type 2 (eT2) codebook or some other eT2-like high resolution codebook. Sounding reference signal (SRS) monitoring may be used.

[0030] As another example, when considering aspects for UE-side monitoring for such an AI / ML model, overhead, latency, complexity, monitoring accuracy, and UE capabilities may be relevant factors. In some cases, the monitoring may be based on an output of a CSI reconstruction model at the UE. The CSI reconstruction model used at the UE may be the same as the CSI reconstruction model (e.g., a decoder) used at the network side, may be a reference CSI reconstruction model (e.g., a reference decoder) provided to the UE by the network, or may be a proxy CSI reconstruction model (e.g., that acts as proxy for a decoder at the network) that is developed at the UE side.

[0031] In some cases, the monitoring may leverage direct estimation of a key performance indicator (KPI) (e.g., that is based on a squared generalized cosine similarity (SGCS) calculation), without using a reconstruction of a target CSI at the UE side.5P70532WO1 4913-1823-7501\1

[0032] In some cases, the monitoring may use estimates of monitoring outputs that are other than intermediate KPIs, without using a reconstruction of a target CSI at the UE side.

[0033] In some cases, the monitoring may be based on a precoded reference signal (e.g.. a channel state information reference signal (CSI-RS), a demodulation reference signal (DMRS)) that is transmitted from the network to the UE that is based on an output of the CSI reconstruction model (e.g., the decoder) used at the network.

[0034] In some cases, the monitoring may be based on an output of the CSI reconstruction model (e.g., the decoder) at the network that is indicated to the UE by the network using an eT2 codebook or some other eT2-like high resolution codebook.

[0035] FIG. 2 illustrates a flow diagram 200 for communications between a base station 202 and a UE 204 corresponding to the use of network-side SGCS-based monitoring of an AI / ML model for CSI compression and decompression. Preliminarily, the base station 202 sends 206 the UE 204 a reference signal. The UE 204 generates ground-truth CSI 208 based on its measurement of this reference signal.

[0036] The UE 204 then provides the ground-truth CSI 208 to an encoder 210 of the AI / ML model for CSI compression that is sited at the UE 204. The encoder 210 uses the ground-truth CSI 208 to generate AI / ML compressed CSI feedback 212 (e.g., a bitstream).

[0037] The UE 204 then sends 214 the ground-truth CSI 208 and the AI / ML compressed CSI feedback 212 to the base station 202, as illustrated.

[0038] In some embodiments, an eT2 codebook or an eT2-like high resolution codebook (which may be even more accurate than an eT2 codebook) is used to report the ground-truth CSI 208 to the base station 202. With respect to a reporting mode used, in some cases one or both of per sample reporting and / or reporting of a number of monitored samples may be used. Note that a determination of a type of KPI to calculate (e.g., SGCS or normalized mean square error (NMSE)) using ground-truth CSI 208 may turn at least in part on the reporting mode(s) so used. (Note that FIG. 2 assumes the case of SGCS use, as will be shown). Note that, due to the nature of the ground-truth CSI 208 as measuring the reference signal as impacted by the channel between the base station 202 and the UE 204, the ground-truth CSI 208 is highly reliable in view of any presently applicable environmental factors.6P70532WO1 4913-1823-7501\1

[0039] The base station 202 provides the AI / ML compressed CSI feedback 212 as received from the UE 204 to a decoder 216 of the AI / ML model for CSI compression that is sited at the base station 202. Using the AI / ML compressed CSI feedback 212, the decoder 216 generates recovered CSI 218.

[0040] The base station 202 then calculates a KPI using both the received ground-truth CSI 208 (which is understood in context as a “target’’ CSI that represents the actual channel estimation performed by the UE 204) and the AI / ML compressed CSI feedback 212 (which is the result of encoding / compression and then decoding / decompression of the ground-truth CSI 208). Based on the result of the KPI calculation, the base station 202 can evaluate the accuracy of the AI / ML model that uses the encoder 210 and the decoder 216 for CSI compression and decompression.

[0041] In the case of FIG. 2, this KPI calculation takes the form of an SGCS calculation 220 that results in an indication of the amount of similarity between the recovered CSI 218 and the ground-truth CSI 208. Based on the amount of similarity indicated by the result of the SGCS calculation 220, the base station 202 can evaluate the accuracy of the AI / ML model that uses the encoder 210 and the decoder 216 for CSI compression and decompression.

[0042] FIG. 3 illustrates a flow diagram 300 for communications between a base station 302 and a UE 304 corresponding to the use of UE-side SGCS-based monitoring of an AI / ML model for CSI compression and decompression. Preliminarily, the base station 302 sends 306 the UE 304 a reference signal. The UE 304 generates ground-truth CSI 308 based on its measurement of this reference signal.

[0043] The UE 304 then provides the ground-truth CSI 308 to an encoder 310 of the AI / ML model for CSI compression that is sited at the UE 304. The encoder 310 uses the ground-truth CSI 308 to generate AI / ML compressed CSI feedback 312 (e.g., a bitstream).

[0044] The UE 304 then sends 314 the AI / ML compressed CSI feedback 312 to the base station 202, as illustrated.

[0045] The base station 302 provides the AI / ML compressed CSI feedback 312 as received from the UE 304 to a decoder 316 of the AI / ML model for CSI compression that is sited at the base station 302. Using the AI / ML compressed CSI feedback 312, the decoder 316 generates recovered CSI 318.7P70532WO1 4913-1823-7501\1

[0046] The base station 302 then sends 320 the recovered CSI 318 to the UE 304. In some embodiments, an eT2 codebook or an eT2-like high resolution codebook (which may be even more accurate than an eT2 codebook) is used to report the recovered CSI 318 to the base station 202. With respect to a reporting mode used, in some cases one or both of per sample reporting and / or reporting of a number of monitored samples may be used. Note that a determination of a type of KPI to calculate (e.g., SGCS or NMSE) using the recovered CSI 318 may turn at least in part on the reporting mode(s) so used. (Note that FIG. 3 assumes the case of SGCS use, as will be shown).

[0047] The UE 304 then calculates a KPI using both the received ground-truth CSI 308 (which is understood in context as a "‘target” CSI that represents the actual channel estimation performed by the UE 304) and the AI / ML compressed CSI feedback 312 as received from the base station 302 (which is the result of encoding / compression and then decoding / decompression of the ground-truth CSI 308). The UE then sends the result of the KPI calculation to the base station 302. Based on the result of the KPI calculation, the base station 302 can evaluate the accuracy of the AI / ML model for CSI compression that uses the encoder 310 and the decoder 316.

[0048] In the case of FIG. 3, this KPI calculation takes the form of an SGCS calculation 322 that results in an indication of the amount of similarity between the recovered CSI 318 and the ground-truth CSI 308. The UE 304 then sends 324 SGCS feedback 326 (the result of the SGCS calculation 322) to the base station 302, as illustrated. Based on the amount of similarity’ indicated by the SGCS feedback 326, the base station 302 can evaluate the accuracy of the AI / ML model for CSI compression that uses the encoder 310 and the decoder 316.

[0049] It can be seen that embodiments according to FIG. 2 and FIG. 3 act to provide some indication of relative similarity between unmodified CSI and recovered CSI corresponding to the use of an AI / ML model for CSI compression. While such indications are intermediately useful for evaluating a generalized accuracy corresponding to the case of AI / ML model use as compared to a case of ground-truth CSI reporting from the UE to the base station, the inherently relative nature of these KPIs does not ultimately quantify, in absolute terms, various applicable effects of the use of the AI / ML model. For example, a metric for a relative similarity between unmodified CSI and recovered CSI as determined according to the KPI calculations discussed in FIG. 2 and FIG. 3 does not provide any absolute understanding of an amount of a capacity loss due8P70532WO1 4913-1823-7501\1to the use of the AI / ML model and / or an amount of throughput loss experienced due to the use of the AI / ML model.

[0050] Accordingly, mechanisms for the generation, reporting, and / or use of KPIs denoting an absolute effect of the use of the AI / ML model (e.g., capacity loss and / or throughput loss) are described herein.

[0051] FIG. 4 illustrates a flow diagram 400 for communications between a base station 402 and a UE 404 corresponding to KPI-based monitoring of an AI / ML model for CSI compression and decompression. Preliminarily, the base station 402 sends 406 the UE 404 a reference signal. The UE 404 generates ground-truth CSI 408 based on its measurement of this reference signal.

[0052] The UE 404 then provides the ground-truth CSI 408 to an encoder 410 of the AI / ML model for CSI compression that is sited at the UE 404. The encoder 410 uses the ground-truth CSI 408 to generate AI / ML compressed CSI feedback 412 (e.g., a bitstream).

[0053] The UE 404 then sends 414 the AI / ML compressed CSI feedback 412 to the base station 402, as illustrated.

[0054] The base station 402 provides the AI / ML compressed CSI feedback 412 as received from the UE 404 to a decoder 416 of the AI / ML model for CSI compression that is sited at the base station 402. Using the AI / ML compressed CSI feedback 412, the decoder 416 generates recovered CSI 418.

[0055] The base station 402 then performs a precoding 420 of a reference signal using a precoder indicated by the recovered CSI 418 to generate a precoded reference signal 422. The base station 402 then sends 424 the precoded reference signal 422 to the UE 404.

[0056] The UE 404 then uses a measurement of the precoded reference signal 422 to perform a KPI calculation 426. In some examples, the UE 404 analyzes the precoded reference signal 422 in view of the ground-truth CSI 408 that was previously identified in order to determine a capacity loss 428 that is attributable to differences between the recovered CSI 418 and the ground-truth CSI 408 due to the encoding and decoding procedure used on the ground-truth CSI 408.

[0057] Note that the capacity loss 428 can be used identify a metric that describes the effects of the use of the AI / ML model in absolute terms. For example, a capacity loss 428 can be used to calculate an Ll-reference signal received power (RSRP) and / or a9P70532WO1 4913-1823-7501\1hypothetical block error rate (BLER) associated with the use of the recovered CSI 418 to form the effective beam formed channel for the precoded reference signal 422.

[0058] The UE 404 then sends 430 the base station 402 the result of the KPI calculation 426 as KPI feedback 432. In some cases, the KPI feedback 432 includes an indication of the capacity loss 428. The base station 402 can use the capacity loss 428 to calculate, for example, an Ll-RSRP and / or a hypothetical BLER associated with the use of encoder 410 and the decoder 416 of the AI / ML model as described.

[0059] In other cases, as part of the KPI calculation 426, the UE 404 may use the capacity loss 428 to calculate such an Ll-RSRP and / or a hypothetical BLER, which is / are then sent to the base station 402 in the KPI feedback 432.

[0060] Note that embodiments corresponding to FIG. 4 enable the UE 404 to compare an AI / ML-based KPI value with a KPI value generated by another (e.g., non- AI / ML) solution for precoder matrix indicator (PMI) generation (e.g., as a sanity check on the use of the AI / ML model as described).

[0061] Embodiments corresponding to FIG. 4 are considered overhead friendly, in that delivery of ground-truth CSI between from the UE 404 to the base station 402 (e.g., as in embodiments corresponding to FIG. 2) and / or delivery of recovered CSI from the base station to the UE (e.g., as in embodiments corresponding to FIG. 3) is / are altogether avoided.

[0062] Various details corresponding to the use of capacity loss as a KPI for CSI reporting (e.g., as described in relation to FIG. 4) are now presented. It may be understood that a UE estimates a DL channel HDL = H. To generate H. the UE performs singular value decomposition (SVD) as H = U A K, where 2, = a,2are the eigenvalues of HH.

[0063] Then, assuming that an applicable the rank is L, the ideal precoding matrix is **v**1= [**v**(1), **v**(2), …, **v**(L)]. In this formulation, an effective channel matrix with ideal precoding feedback Heff(as seen at the UE) is therefore Heff= HV1.

[0064] A general precoding matrix P with the effective matrix can be denoted HP.

[0065] Note that with limited feedback available as between a UE and a base station, P = PT is not possible in reality (as this would correspond to an ideal encoding / decoding through the AI / ML model for CSI feedback).10P70532WO1 4913-1823-7501\1

[0066] It is observed that, within the given framework, there are two cases under consideration. A first of these cases is the P = Vi case, corresponding to the ideal feedback case that is not actually achievable through the AI / ML model because of nonidealities in the encoder and / or decoder of the AI / ML model. Note, however, that a UE has knowledge of the P = Vi precoder. For example, in terms of the embodiment of FIG.4, the UE 404 has access to the P = Vi precoding matrix as indicated by the ground-truth CSI 408.

[0067] A second of these cases is P = Vc case, corresponding to the precoder actually- used by the base station as indicated by recovered CSI subsequent to the encoding and then decoding of UE-generated ground truth CSI through an AI / ML model. The UE has access to the P = Vcprecoder through its receipt and analysis of a reference signal from the base station that the base station has precoded with the precoder indicated by the recovered CSI. For example, in terms of the embodiment of FIG. 4, the precoded reference signal 422 that is sent to the UE 404 is analyzed by the UE 404 to identify the nature of the P = Vcprecoding matrix used by the base station 402.

[0068] Because the UE has knowledge of both P = Vi and P = Vc, it is enabled to derive a capacity loss due to the difference between the effective channels: HP\P=V1and HP| =VC.

[0069] Embodiments for determining capacity loss as described herein function with respect to various kinds of receivers that may be used at a UE. For example, a UE may¬ use a zero force (ZF) receiver. In such cases, it may be understood that:WZF= HP(PTHTHP)"1andSINRk= -; wherekZ.|(PTHTHP)-' |fcfcHis the wireless channel;P is the precoder;SINRfc is an effective signal to noise ratio per layer £; andL is a number of layers.

[0070] In another example, a UE may use a minimized mean square error (MMSE) receiver. In such cases, it may be understood that:11P70532WO1 4913-1823-7501\1WMMSE = HP(~I + PTHTHP)1andSINRk= hj(^l + H / H / T)-1hD; whereH is the wireless channel;P is the precoder;SINR / c is an effective signal to noise ratio per layerho denotes the desired channel for the -th stream: £-th column of HP, andHi represents an interference matrix that is the matrix HP with the A-th column removed.

[0071] In another example, a UE may use an SVD receiver. In such cases, it may be understood that:™SVD = U; andSINRfc|v(fc)TP(fc)|22 — —; whereSj=ij#fc|v(k)TPG)|SINR / . is an effective signal to noise ratio per layer kL is a number of layers;v(£) is a -th vector of the precoding matrix F;p(A) is a Ar-th vector of P (where, for example. P = V\ or P = Vcas described herein, depending on the case under consideration);p( / ) is a / -th vector of;£ is a number of layers;p is a signal to noise ratio; andAA is an eigenvalue of the channel that is associated with a -th transmission layer.12P70532WO1 4913-1823-7501\1Note also that in this formulation, X^ij^k Iv(^)TP0) I2 may be understood as an interference term, while may be understood as a noise term.p^k

[0072] Within the given framework, a capacity C per subcarrier for each of these three receiving techniques can be understood according to C =log2(l + SINRk).Corresponding to the optimal / ideal performance case of P = Ki, the assumption of the above condition for C would result in each receiver achieving C = 2fc=i log2(1 + “)•

[0073] However, as has been discussed, because of the limited nature of feedback through the Al / ML model (e.g., because of non-idealities in either / both the encoder and / or the decoder of the AI / ML model), it may be that an AI / ML model does not actually achieve the P = Vi case. Accordingly, an alternative P = Vc case is also considered, where P = Vc represents a reconstructed precoder as understood / used at the base station according to the use of the AI / ML model use (after an encoding of groundtruth CSI at the UE and a decoding of this result at the base station).

[0074] A capacity loss CLOSS is defined as a capacity difference between the capacity associated with the idealized use of the precoder P = Vi and the capacity associated with the actual use instead of the precoder P = Vc. In other words:LOSS=C\p=vt— CP=Vc.

[0075] An approximate expression for capacity loss at a high signal to noise ratio (SNR) for an SVD receiver CLOS, SVD may accordingly be derived as:C ~ VLInn ( PPl(PPk / L+L \LLOSS, SVD ~ Lk=i10§21 |,T u-hprA,,.,2 L wnere\P / (fc) + |v(fc)Tvc(fc)| / L is a rank of the precoding matrix V and the precoding matrix Vcv(A) is the i-th vector of the precoding matrix F;is the -th vector of the precoding matrix Vc.p is a signal to noise ratio;PiW =lv0)Tvc(^)l2and is understood as an inter-stream interference to layer k caused by the mismatch between Vcand Fi; and13P70532WO1 4913-1823-7501\1 / .k is an eigenvalue of the channel that is associated with the A-th transmission layer used to send the compressed CSI feedback.

[0076] Note that a UE may be configured to use the above approximation for CLOSS. SVD as a full and / or precise equality in order to facilitate the use of embodiments discussed herein.

[0077] Note that in cases for rank 1, L = 1 and thus there is no stream interference. The capacity loss of the SVD receiver CLOSS. SVD in such case may be more simply approximated as:CLOSS. SVD ~ ~ l°g2 lv(l)Tvc(l) I2-

[0078] Note that a UE may be configured to use the above approximation for LOSS, SVD when L = 1 as a full and / or precise equality in order to facilitate the use of embodiments discussed herein.

[0079] At high SNR, performance of ZF and MMSE receives is similar. Accordingly, a bounding expression for capacity’ loss at a high SNR for a ZF receiver or an MMSE receiver CLOSS. ZF / MMSE may be derived as:C LOSS ZF / MMSE < NTX 1 Xk=l Xj = l log2NTk whereNT is a number of transmit (Tx) antenna ports;£ is a rank of the precoding matrix V and the precoding matrix U;v(£) is the A’-th vector of the precoding matrix F;Vc( / ) is the. / -th vector of the precoding matrix U; / .k is a first eigenvalue of the channel that is associated with the A-th transmission layer used to send the compressed CSI feedback; andj is a second eigenvalue of the channel that is associated with the / -th transmission layer used to send the compressed CSI feedback.

[0080] Note that a UE may be configured to use the above bounding expression for CLOSS. ZFD. IMSE as a full and / or precise equality in order to facilitate the use of embodiments discussed herein corresponding to ZF and / or MMSE cases.14P70532WO1 4913-1823-7501\1

[0081] These calculations for CLOSS may be used, for example, as is described herein in relation to the flow diagram 400 of FIG. 4.

[0082] The use and reporting of KPIs such as CLOSS (and / or values derived therefrom) as discussed herein may give the base station a closer / better understanding of system performance than non-absolute KPIs such as SGCS indications. Values such as CLOSS and / or values derived therefrom may provide the network with a better understanding of the effects of throughput on communications to mismatches between an ideal effective channel corresponding to ideal precoding (according to precoding vector P = PT) and the actually received effective channel corresponding to the non-ideal precoding actually in use by the base station (according to precoding vector / 5= PT). Such KPIs may be used for AI / ML model monitoring in lifecycle management (LCM) as a reported quantity, as has been discussed. Such KPIs may be used between many layers.

[0083] FIG. 5 illustrates a method 500 of a UE, according to embodiments discussed herein. The method 500 includes identifying 502, based on a first measurement of a first reference signal that is received from a base station, a precoding matrix V that is ideal for DL precoding. The method 500 further includes generating 504 CSI feedback that indicates the precoding matrix V. The method 500 further includes applying 506 the CSI feedback at a first encoder of a first AI / ML model that is at the UE to generate compressed CSI feedback. The method 500 further includes sending 508, to the base station, the compressed CSI feedback. The method 500 further includes identifying 510, based on a second measurement of a precoded second reference signal that is received from the base station in response to the compressed CSI feedback, a precoding matrix Vcthat the base station is using for the DL precoding. The method 500 further includes calculating 512, using the precoding matrix V and the precoding matrix Vc, a capacity loss CLOSSassociated with using the precoding matrix Vc for the DL precoding. The method 500 further includes sending 514. to the base station, a reporting message reporting the capacity loss CLOSS.

[0084] In some embodiments of the method 500, the UE uses a singular value decomposition (SVD) receiver to receive the first reference signal and the precoded second reference signal; and the capacity loss CLOSSis calculated according to CLOSS= Xfc-i log, —\ where £ is a rank of the precoding matrix V and theprecoding matrix PT; v(E) is a A- th vector of the precoding matrix E; Vc(£) is a A- th vector15P70532WO1 4913-1823-7501\1of the preceding matrix Uc; p is a signal to noise ratio; Pi(k) = 2^-1 - Jv0)Tvc(k)|2; and k is an eigenvalue of the channel that is associated with a A -th transmission layer used to send the compressed CSI feedback.

[0085] In some embodiments of the method 500, the UE uses a zero force (ZF) receiver to receive the first reference signal and the precoded second reference signal; and the capacity loss CLOSSis calculated according to CLOSS= NTxlog2 / Vrk where NT is a number of transmit (Tx) antennaports; L is a rank of the precoding matrix V and the precoding matrix U; v(&) is a A- th vector of the precoding matrix E; vc( / ) is a / -th vector of the precoding matrix Vc', Ak is a first eigenvalue of the channel that is associated with a A-th transmission layer used to send the compressed CSI feedback; and Aj is a second eigenvalue of the channel that is associated with a / -th transmission layer used to send the compressed CSI feedback.

[0086] In some embodiments of the method 500, the UE uses a minimized mean square error (MMSE) receiver to receive the first reference signal and the precoded second reference signal; and the capacity loss CLOSSis calculated according to CLOSS=217? VC( / -)TV( / C) — log2ATk where Nr is a number of transmit (Tx)antenna ports; L is a rank of the precoding matrix V and the precoding matrix U; v(Zr) is a £-th vector of the precoding matrix E; vc( / ) is a / -th vector of the precoding matrix VcAk is a first eigenvalue of the channel that is associated with a Uth transmission layer used to send the compressed CSI feedback; and A}is a second eigenvalue of the channel that is associated with a / -th transmission layer used to send the compressed CSI feedback.

[0087] In some embodiments, the method 500 further includes receiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS, an instruction to stop a using the first encoder at the UE; and stopping a use of the first encoder at the UE in response to the instruction.

[0088] In some embodiments, the method 500 further includes receiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS, an AI / ML model identifier (ID) for a second AI / ML model; and switching from using the first encoder to using a second encoder of the second AI / ML model.16P70532WO1 4913-1823-7501\1

[0089] In some embodiments, the method 500 further includes receiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS. an instruction to perform retraining operations for the first encoder; and performing the retraining operations for the first encoder in response to the instruction.

[0090] FIG. 6 illustrates a method 600 of a base station, according to embodiments discussed herein. The method 600 includes sending 602, to a UE, a first reference signal. The method 600 further includes receiving 604, from the UE, compressed CSI feedback. The method 600 further includes applying 606 the compressed CSI feedback at a first decoder of a first AI / ML model that is at the base station to generate recovered CSI feedback indicating a precoding matrix Vc. The method 600 further includes performing 608 DL precoding of a second reference signal with the precoding matrix Vcto generate a precoded second reference signal. The method 600 further includes sending 610, to the UE, the precoded second reference signal. The method 600 further includes receiving 612, from the UE. in response to the precoded second reference signal, a reporting message reporting a capacity loss CLOSSassociated with using the precoding matrix Vc for the DL precoding.

[0091] In some embodiments, the method 600 further includes determining, based on the capacity loss CLOSS. to stop using the first AI / ML model; sending, to the UE, an instruction to stop using an encoder of the first AI / ML model at the UE; and stopping a use of the first decoder at the base station.

[0092] In some embodiments, the method 600 further includes determining, based on the capacity loss CLOSS, to perform an AI / ML model switch; sending, to the UE, an AI / ML model ID for a second AI / ML model; and switching from using the first decoder to using a second decoder of the second AI / ML model.

[0093] In some embodiments, the method 600 further includes determining, based on the capacity loss CLOSS. that the first AI / ML model is to be retrained; sending, to the UE, an instruction to perform first retraining operations for the encoder; and performing second retraining operations for the first decoder.

[0094] FIG. 7 illustrates an example architecture of a wireless communication system 700, according to embodiments disclosed herein. The following description is provided for an example wireless communication system 700 that operates in conjunction with the LTE system standards and / or 5G or NR system standards as provided by 3GPP technical specifications.17P70532WO1 4913-1823-7501\1

[0095] As shown by FIG. 7. the wireless communication system 700 includes UE 702 and UE 704 (although any number of UEs may be used). In this example, the UE 702 and the UE 704 are illustrated as smartphones (e.g., handheld touchscreen mobile computing devices connectable to one or more cellular networks), but may also comprise any mobile or non-mobile computing device configured for wireless communication.

[0096] The UE 702 and UE 704 may be configured to communicatively couple with a RAN 706. In embodiments, the RAN 706 may be NG-RAN, E-UTRAN, etc. The UE 702 and UE 704 utilize connections (or channels) (shown as connection 708 and connection 710, respectively) with the RAN 706, each of which comprises a physical communications interface. The RAN 706 can include one or more base stations (such as base station 712 and base station 714) that enable the connection 708 and connection 710.

[0097] In this example, the connection 708 and connection 710 are air interfaces to enable such communicative coupling, and may be consistent with RAT(s) used by the RAN 706, such as, for example, an LTE and / or NR.

[0098] In some embodiments, the UE 702 and UE 704 may also directly exchange communication data via a sidelink interface 716. The UE 704 is shown to be configured to access an access point (shown as AP 718) via connection 720. By way of example, the connection 720 can comprise a local wireless connection, such as a connection consistent with any IEEE 802.11 protocol, wherein the AP 718 may comprise a Wi-Fi® router. In this example, the AP 718 may be connected to another network (for example, the Internet) without going through a CN 724.

[0099] In embodiments, the UE 702 and UE 704 can be configured to communicate using orthogonal frequency division multiplexing (OFDM) communication signals with each other or with the base station 712 and / or the base station 714 over a multicarrier communication channel in accordance with various communication techniques, such as, but not limited to. an orthogonal frequency division multiple access (OFDMA) communication technique (e.g., for downlink communications) or a single carrier frequency division multiple access (SC-FDMA) communication technique (e.g., for uplink and ProSe or sidelink communications), although the scope of the embodiments is not limited in this respect. The OFDM signals can comprise a plurality of orthogonal subcarriers.18P70532WO1 4913-1823-7501\1

[0100] In some embodiments, all or parts of the base station 712 or base station 714 may be implemented as one or more software entities running on server computers as part of a virtual network. In addition, or in other embodiments, the base station 712 or base station 714 may be configured to communicate with one another via interface 722. In embodiments where the wireless communication system 700 is an LTE system (e.g., when the CN 724 is an EPC), the interface 722 may be an X2 interface. The X2 interface may be defined between two or more base stations (e.g., two or more eNBs and the like) that connect to an EPC, and / or between two eNBs connecting to the EPC. In embodiments where the wireless communication system 700 is an NR system (e.g., when CN 724 is a 5GC), the interface 722 may be an Xn interface. The Xn interface is defined between two or more base stations (e.g.. two or more gNBs and the like) that connect to 5GC, between a base station 712 (e.g., a gNB) connecting to 5GC and an eNB, and / or between two eNBs connecting to 5GC (e.g., CN 724).

[0101] The RAN 706 is shown to be communicatively coupled to the CN 724. The CN 724 may comprise one or more network elements 726, which are configured to offer various data and telecommunications services to customers / subscribers (e.g., users of UE 702 and UE 704) who are connected to the CN 724 via the RAN 706. The components of the CN 724 may be implemented in one physical device or separate physical devices including components to read and execute instructions from a machine-readable or computer-readable medium (e g., a non-transitory machine-readable storage medium).

[0102] In embodiments, the CN 724 may be an EPC, and the RAN 706 may be connected with the CN 724 via an SI interface 728. In embodiments, the SI interface 728 may be split into two parts, an SI user plane (Sl-U) interface, which carries traffic data between the base station 712 or base station 714 and a serving gateway (S-GW), and the SI -MME interface, which is a signaling interface between the base station 712 or base station 714 and mobility management entities (MMEs).

[0103] In embodiments, the CN 724 may be a 5GC, and the RAN 706 may be connected with the CN 724 via an NG interface 728. In embodiments, the NG interface 728 may be split into two parts, an NG user plane (NG-U) interface, which carries traffic data between the base station 712 or base station 714 and a user plane function (UPF), and the SI control plane (NG-C) interface, which is a signaling interface between the base station 712 or base station 714 and access and mobility management functions (AMFs).19P70532WO1 4913-1823-7501\1

[0104] Generally, an application server 730 may be an element offering applications that use internet protocol (IP) bearer resources with the CN 724 (e.g., packet switched data services). The application server 730 can also be configured to support one or more communication services (e.g., VoIP sessions, group communication sessions, etc.) for the UE 702 and UE 704 via the CN 724. The application server 730 may communicate with the CN 724 through an IP communications interface 732.

[0105] FIG. 8 illustrates a system 800 for performing signaling 834 between a wireless device 802 and a network device 818, according to embodiments disclosed herein. The system 800 may be a portion of a wireless communications system as herein described. The wireless device 802 may be. for example, a UE of a wireless communication system. The network device 818 may be, for example, a base station (e.g., an eNB or a gNB) of a wireless communication system.

[0106] The wireless device 802 may include one or more processor(s) 804. The processor(s) 804 may execute instructions such that various operations of the wireless device 802 are performed, as described herein. The processor(s) 804 may include one or more baseband processors implemented using, for example, a central processing unit (CPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a controller, a field programmable gate array (FPGA) device, another hardware device, a firmware device, or any combination thereof configured to perform the operations described herein.

[0107] The wireless device 802 may include a memory 806. The memory 806 may be a non-transitory computer-readable storage medium that stores instructions 808 (which may include, for example, the instructions being executed by the processor(s) 804). The instructions 808 may also be referred to as program code or a computer program. The memory 806 may also store data used by, and results computed by, the processor(s) 804.

[0108] The wireless device 802 may include one or more transceiver(s) 810 that may include radio frequency (RF) transmitter circuitry and / or receiver circuitry that use the antenna(s) 812 of the wireless device 802 to facilitate signaling (e.g., the signaling 834) to and / or from the wireless device 802 with other devices (e.g., the network device 818) according to corresponding RATs.

[0109] The wireless device 802 may include one or more antenna(s) 812 (e.g., one, two, four, or more). For embodiments with multiple antenna(s) 812, the wireless device 802 may leverage the spatial diversity of such multiple antenna(s) 812 to send and / or receive20P70532WO1 4913-1823-7501\1multiple different data streams on the same time and frequency resources. This behavior may be referred to as. for example, multiple input multiple output (MIMO) behavior (referring to the multiple antennas used at each of a transmitting device and a receiving device that enable this aspect). MIMO transmissions by the wireless device 802 may be accomplished according to precoding (or digital beamforming) that is applied at the wireless device 802 that multiplexes the data streams across the antenna(s) 812 according to known or assumed channel characteristics such that each data stream is received with an appropriate signal strength relative to other streams and at a desired location in the spatial domain (e.g., the location of a receiver associated with that data stream). Certain embodiments may use single user MIMO (SU-MIMO) methods (where the data streams are all directed to a single receiver) and / or multi user MIMO (MU-MIMO) methods (where individual data streams may be directed to individual (different) receivers in different locations in the spatial domain).

[0110] In certain embodiments having multiple antennas, the wireless device 802 may implement analog beamforming techniques, whereby phases of the signals sent by the antenna(s) 812 are relatively adjusted such that the (joint) transmission of the antenna(s) 812 can be directed (this is sometimes referred to as beam steering).

[0111] The wireless device 802 may include one or more interface(s) 814. The interface(s) 814 may be used to provide input to or output from the wireless device 802. For example, a wireless device 802 that is a UE may include interface(s) 814 such as microphones, speakers, a touchscreen, buttons, and the like in order to allow for input and / or output to the UE by a user of the UE. Other interfaces of such a UE may be made up of transmitters, receivers, and other circuitry (e.g., other than the transceiver(s) 810 / antenna(s) 812 already described) that allow for communication between the UE and other devices and may operate according to known protocols (e.g., Wi-Fi®, Bluetooth®, and the like).

[0112] The wireless device 802 may include a capacity loss module 816. The capacity loss module 816 may be implemented via hardware, software, or combinations thereof. For example, the capacity loss module 816 may be implemented as a processor, circuit, and / or instructions 808 stored in the memory 806 and executed by the processor(s) 804. In some examples, the capacity loss module 816 may be integrated within the processor(s) 804 and / or the transceiver(s) 810. For example, the capacity loss module 816 may be implemented by a combination of software components (e.g.. executed by a21P70532WO1 4913-1823-7501\1DSP or a general processor) and hardware components (e.g., logic gates and circuitry ) within the processor(s) 804 or the transceiver(s) 810.

[0113] The capacity loss module 816 may be used for various aspects of the present disclosure, for example, aspects of FIG. 5. For example, the capacity loss module 816 may configured the wireless device 802 to identify, based on a first measurement of a first reference signal that is received from a base station, a precoding matrix V for that is ideal for DL precoding; generate CSI feedback that indicates the precoding matrix V apply the CSI feedback at a first encoder of a first AI / ML model that is at the UE to generate compressed CSI feedback; send, to the base station, the compressed CSI feedback; identify, based on a second measurement of a precoded second reference signal that is received from the base station in response to the compressed CSI feedback, a precoding matrix Vcthat the base station is using for the DL precoding; calculate, using the precoding matrix V and the precoding matrix Kc, a capacity' loss CLOSSassociated with using the precoding matrix U for the DL precoding; and send, to the base station, a reporting message reporting the capacity loss CLOSS. as has been discussed herein.

[0114] The network device 818 may include one or more processor(s) 820. The processor(s) 820 may execute instructions such that various operations of the network device 818 are performed, as described herein. The processor(s) 820 may include one or more baseband processors implemented using, for example, a CPU, a DSP, an ASIC, a controller, an FPGA device, another hardware device, a firmware device, or any combination thereof configured to perform the operations described herein.

[0115] The network device 818 may include a memory 822. The memory 822 may be a non-transitory computer-readable storage medium that stores instructions 824 (which may include, for example, the instructions being executed by the processor(s) 820). The instructions 824 may also be referred to as program code or a computer program. The memory 822 may also store data used by, and results computed by, the processor(s) 820.

[0116] The network device 818 may include one or more transceiver(s) 826 that may include RF transmitter circuitry and / or receiver circuitry that use the antenna(s) 828 of the network device 818 to facilitate signaling (e.g., the signaling 834) to and / or from the network device 818 with other devices (e.g., the wireless device 802) according to corresponding RATs.

[0117] The network device 818 may include one or more antenna(s) 828 (e.g., one, two, four, or more). In embodiments having multiple antenna(s) 828, the network device 81822P70532WO1 4913-1823-7501\1may perform MIMO, digital beamforming, analog beamforming, beam steering, etc., as has been described.

[0118] The network device 818 may include one or more interface(s) 830. The interface(s) 830 may be used to provide input to or output from the network device 818. For example, a network device 818 that is a base station may include interface(s) 830 made up of transmitters, receivers, and other circuitry (e.g., other than the transceiver(s) 826 / antenna(s) 828 already described) that enables the base station to communicate with other equipment in a core network, and / or that enables the base station to communicate with external networks, computers, databases, and the like for purposes of operations, administration, and maintenance of the base station or other equipment operably connected thereto.

[0119] The network device 818 may include a capacity loss module 832. The capacity loss module 832 may be implemented via hardware, software, or combinations thereof. For example, the capacity loss module 832 may be implemented as a processor, circuit, and / or instructions 824 stored in the memory 822 and executed by the processor(s) 820. In some examples, the capacity loss module 832 may be integrated within the processor(s) 820 and / or the transceiver(s) 826. For example, the capacity loss module 832 may be implemented by a combination of software components (e.g., executed by a DSP or a general processor) and hardware components (e.g., logic gates and circuitry) within the processor(s) 820 or the transceiver(s) 826.

[0120] The capacity loss module 832 may be used for various aspects of the present disclosure, for example, aspects of FIG. 6. For example, the capacity loss module 832 may configure the network device 818 to send, to a UE, a first reference signal; receive, from the UE, compressed CSI feedback; apply the compressed CSI feedback at a first decoder of a first AI / ML model that is at the base station to generate recovered CSI feedback indicating a precoding matrix JC; perform DL precoding of a second reference signal with the precoding matrix Vc to generate a precoded second reference signal; send, to the UE, the precoded second reference signal; and receive, from the UE, in response to the precoded second reference signal, a reporting message reporting a capacity loss CLOSSassociated with using the precoding matrix Vc for the DL precoding, as has been discussed herein.23P70532WO1 4913-1823-7501\1

[0121] Embodiments contemplated herein include an apparatus comprising means to perform one or more elements of the method 500. This apparatus may be. for example, an apparatus of a UE (such as a wireless device 802 that is a UE, as described herein).

[0122] Embodiments contemplated herein include one or more non -transitory computer-readable media comprising instructions to cause an electronic device, upon execution of the instructions by one or more processors of the electronic device, to perform one or more elements of the method 500. This non-transitory computer-readable media may be, for example, a memory of a UE (such as a memory 806 of a wireless device 802 that is a UE, as described herein).

[0123] Embodiments contemplated herein include an apparatus comprising logic, modules, or circuitry to perform one or more elements of the method 500. This apparatus may be, for example, an apparatus of a UE (such as a wireless device 802 that is a UE, as described herein).

[0124] Embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of the method 500. This apparatus may be, for example, an apparatus of a UE (such as a wireless device 802 that is a UE, as described herein).

[0125] Embodiments contemplated herein include a signal as described in or related to one or more elements of the method 500.

[0126] Embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution of the program by a processor is to cause the processor to carry out one or more elements of the method 500. The processor may be a processor of a UE (such as a processor(s) 804 of a wireless device 802 that is a UE, as described herein). These instructions may be, for example, located in the processor and / or on a memory of the UE (such as a memory 806 of a wireless device 802 that is a UE, as described herein).

[0127] Embodiments contemplated herein include an apparatus comprising means to perform one or more elements of the method 600. This apparatus may be. for example, an apparatus of a base station (such as a network device 818 that is a base station, as described herein).

[0128] Embodiments contemplated herein include one or more non-transitory computer-readable media comprising instructions to cause an electronic device, upon 24P70532WO1 4913-1823-7501\1execution of the instructions by one or more processors of the electronic device, to perform one or more elements of the method 600. This non-transitory computer-readable media may be, for example, a memory of a base station (such as a memory 822 of a network device 818 that is a base station, as described herein).

[0129] Embodiments contemplated herein include an apparatus comprising logic, modules, or circuitry to perform one or more elements of the method 600. This apparatus may be, for example, an apparatus of a base station (such as a network device 818 that is a base station, as described herein).

[0130] Embodiments contemplated herein include an apparatus comprising: one or more processors and one or more computer-readable media comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform one or more elements of the method 600. This apparatus may be. for example, an apparatus of a base station (such as a network device 818 that is a base station, as described herein).

[0131] Embodiments contemplated herein include a signal as described in or related to one or more elements of the method 600.

[0132] Embodiments contemplated herein include a computer program or computer program product comprising instructions, wherein execution of the program by a processing element is to cause the processing element to carry out one or more elements of the method 600. The processor may be a processor of a base station (such as a processor(s) 820 of a network device 818 that is a base station, as described herein). These instructions may be, for example, located in the processor and / or on a memory' of the base station (such as a memory 822 of a network device 818 that is a base station, as described herein).

[0133] For one or more embodiments, at least one of the components set forth in one or more of the preceding figures may be configured to perform one or more operations. techniques, processes, and / or methods as set forth herein. For example, a baseband processor as described herein in connection with one or more of the preceding figures may be configured to operate in accordance with one or more of the examples set forth herein. For another example, circuitry associated with a UE, base station, network element, etc. as described above in connection with one or more of the preceding figures may be configured to operate in accordance with one or more of the examples set forth herein.25P70532WO1 4913-1823-7501\1

[0134] Any of the above-described embodiments may be combined with any other embodiment (or combination of embodiments), unless explicitly stated otherwise. The foregoing description of one or more implementations provides illustration and description, but is not intended to be exhaustive or to limit the scope of embodiments to the precise form disclosed. Modifications and variations are possible in light of the above teachings or may be acquired from practice of various embodiments.

[0135] Embodiments and implementations of the systems and methods described herein may include various operations, which may be embodied in machine-executable instructions to be executed by a computer system. A computer system may include one or more general-purpose or special-purpose computers (or other electronic devices). The computer system may include hardware components that include specific logic for performing the operations or may include a combination of hardware, software, and / or firmware.

[0136] It should be recognized that the systems described herein include descriptions of specific embodiments. These embodiments can be combined into single systems, partially combined into other systems, split into multiple systems or divided or combined in other ways. In addition, it is contemplated that parameters, attributes, aspects, etc. of one embodiment can be used in another embodiment. The parameters, attributes, aspects, etc. are merely described in one or more embodiments for clarity, and it is recognized that the parameters, attributes, aspects, etc. can be combined with or substituted for parameters, attributes, aspects, etc. of another embodiment unless specifically disclaimed herein.

[0137] It is well understood that the use of personally identifiable information should follow privacy policies and practices that are generally recognized as meeting or exceeding industry or governmental requirements for maintaining the privacy of users. In particular, personally identifiable information data should be managed and handled so as to minimize risks of unintentional or unauthorized access or use, and the nature of authorized use should be clearly indicated to users.

[0138] Although the foregoing has been described in some detail for purposes of clarity, it will be apparent that certain changes and modifications may be made without departing from the principles thereof. It should be noted that there are many alternative ways of implementing both the processes and apparatuses described herein. Accordingly, the present embodiments are to be considered illustrative and not restrictive, and the26P70532WO1 4913-1823-7501\1description is not to be limited to the details given herein, but may be modified within the scope and equivalents of the appended claims.27P70532WO1 4913-1823-7501\1

Claims

CLAIMS1. A method of a user equipment (UE), comprising:identifying, based on a first measurement of a first reference signal that is received from a base station, a precoding matrix V that is ideal for downlink (DL) precoding;generating channel state information (CSI) feedback that indicates the precoding matrix F;applying the CSI feedback at a first encoder of a first artificial intelligence (AI) / machine learning (ML) model that is at the UE to generate compressed CSI feedback;sending, to the base station, the compressed CSI feedback;identifying, based on a second measurement of a precoded second reference signal that is received from the base station in response to the compressed CSI feedback, a precoding matrix Vcthat the base station is using for the DL preceding;calculating, using the precoding matrix V and the preceding matrix Uc, a capacity loss CLOSSassociated with using the precoding matrix Vc for the DL preceding; and sending, to the base station, a reporting message reporting the capacity loss CLOSS.

2. The method of claim 1, wherein:the UE uses a singular value decomposition (SVD) receiver to receive the first reference signal and the precoded second reference signal; andthe capacity loss CLOSSis calculated according to:CLOSS - 2 / <=i l°g2 ( |T,2 ), where:\p;(fc)+|v(fc)1vc(k)| / L is a rank of the precoding matrix V and the precoding matrix U;v(£) is a -th vector of the precoding matrix U;vc(A) is a Uth vector of the precoding matrix Vc,p is a signal to noise ratio;PiW = |v( / )Tvc(fc)|2; andk is an eigenvalue of the channel that is associated with a Uth transmission layer used to send the compressed CSI feedback.

3. The method of claim 1, wherein:28P70532WO1 4913-1823-7501\lthe UE uses a zero force (ZF) receiver to receive the first reference signal and the precoded second reference signal; andthe capacity loss CLOSSis calculated according to:CLOSS —x) 2fc=i ^~vcG)Tv(k) log2 / Vrk where:NT is a number of transmit (Tx) antenna ports;L is a rank of the precoding matrix V and the precoding matrix Ec;v(£) is a -th vector of the precoding matrix K;vc( / ) is a / -th vector of the precoding matrix Ec; / .k is a first eigenvalue of the channel that is associated with a Uth transmission layer used to send the compressed CSI feedback; andis a second eigenvalue of the channel that is associated with a / -th transmission layer used to send the compressed CSI feedback.

4. The method of claim 1, wherein:the UE uses a minimized mean square error (MMSE) receiver to receive the first reference signal and the precoded second reference signal; andthe capacity loss CLOSSis calculated according to:CLOSS — x 1 E / =i ^vc( / )Tv(k) log2 / VTk where:NT is a number of transmit (Tx) antenna ports;£ is a rank of the precoding matrix V and the precoding matrix U;v(£) is a £-th vector of the precoding matrix E;vc( / ) is a / -th vector of the precoding matrix Ec;k is a first eigenvalue of the channel that is associated with a Zr-th transmission layer used to send the compressed CSI feedback; and2 / is a second eigenvalue of the channel that is associated with a / -th transmission layer used to send the compressed CSI feedback.

5. The method of claim 1, further comprising:receiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS, an instruction to stop a using the first encoder at the UE; and stopping a use of the first encoder at the UE in response to the instruction.

6. The method of claim 1, further comprising:29P70532WO1 4913-1823-7501\lreceiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS. an AI / ML model identifier (ID) for a second AI / ML model; and switching from using the first encoder to using a second encoder of the second AI / ML model.

7. The method of claim 1, further comprising:receiving, from the base station, in response to the reporting message reporting the capacity loss CLOSS, an instruction to perform retraining operations for the first encoder; andperforming the retraining operations for the first encoder in response to the instruction.

8. A method of a base station, comprising:sending, to a user equipment (UE). a first reference signal;receiving, from the UE. compressed channel state information (CSI) feedback; applying the compressed CSI feedback at a first decoder of a first artificial intelligence (Al)Zmachine learning (ML) model that is at the base station to generate recovered CSI feedback indicating a precoding matrix IT;performing downlink (DL) precoding of a second reference signal with the precoding matrix Vcto generate a precoded second reference signal;sending, to the UE, the precoded second reference signal; andreceiving, from the UE, in response to the precoded second reference signal, a reporting message reporting a capacity loss CLOSSassociated with using the precoding matrix Vcfor the DL precoding.

9. The method of claim 8, further comprising:determining, based on the capacity loss CLOSS. to stop using the first AI / ML model;sending, to the UE, an instruction to stop using an encoder of the first AI / ML model at the UE; andstopping a use of the first decoder at the base station.

10. The method of claim 8, further comprising:determining, based on the capacity loss CLOSS, to perform an AI / ML model switch;30P70532WO1 4913-1823-7501\1sending, to the UE, an AI / ML model identifier (ID) for a second AI / ML model; andswitching from using the first decoder to using a second decoder of the second AI / ML model.

11. The method of claim 8, further comprising:determining, based on the capacity loss CLOSS. that the first AI / ML model is to be retrained;sending, to the UE, an instruction to perform first retraining operations for an encoder; andperforming second retraining operations for the first decoder.

12. An apparatus comprising means to perform the method of any of claim 1 to claim 11.

13. A computer-readable media comprising instructions to cause an electronic device, upon execution of the instructions by one or more processors of the electronic device, to perform the method of any of claim 1 to claim 11.

14. An apparatus comprising logic, modules, or circuitry to perform the method of any of claim 1 to claim 11.

15. A baseband processor for a user equipment (UE) that is configured to cause the UE to perform one or more elements of any one of claim 1 to claim 7.

16. A baseband processor for a base station that is configured to cause the base station to perform one or more elements of any one of claim 8 to claim 11.31P70532WO1 4913-1823-7501\1

Images