System and methods for diagnosis and performance degradation cause identification for artificial intelligence / machine learning models for channel state information compression and decompression

Abstract

Systems and methods for diagnosis and performance degradation cause identification for artificial intelligence (AI) / machine learning (ML) models for channel state information (CSI) compression and decompression are discussed herein. A base station receives, from a user equipment (UE), ground-truth CSI for a reference signal and compressed CSI feedback as generated at a UE encoder using the ground-truth CSI; applies the compressed CSI feedback at a base station decoder to generate first recovered CSI; processes the ground-truth CSI using a base station encoder and the base station decoder to generate second recovered CSI; calculates key performance indicators (KPIs) by applying the ground-truth CSI with each of the first recovered CSI and the second recovered CSI; and identifies a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the KPIs. In other embodiments, reference CSI is used. Various UE behaviors are also discussed.

Description

SYSTEM AND METHODS FOR DIAGNOSIS AND PERFORMANCE DEGRADATION CAUSE IDENTIFICATION FOR ARTIFICIAL INTELLIGENCE / MACHINE LEARNING MODELS FOR CHANNEL STATE INFORMATION COMPRESSION AND DECOMPRESSIONTECHNICAL FIELD

[0001] This application relates generally to wireless communication systems, including wireless communication systems utilizing AI / ML models for CSI 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 industry7groups 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). In1P70697WO1 4898-9140-0001\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 diagram for a case of UE-based AI / ML model monitoring.

[0011] FIG. 3 illustrates a diagram for an inter-vendor training collaboration procedure for training an AI / ML model for CSI compression and decompression that relies on standardized data / a standardized dataset format and a dataset exchange between a base station and a UE.2P70697WO1 4898-9140-0001\1

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

[0013] FIG. 4B illustrates a flow diagram for communications between the base station and a UE corresponding to the use of network-side monitoring of an AI / ML model for CSI compression and decompression after the events described in FIG. 4A.

[0014] FIG. 5 illustrates a diagnostics block as may be used at a network / base station identify a source of performance degradation, according to embodiments discussed herein.

[0015] FIG. 6A and FIG. 6B together illustrate a flow diagram for communications between a base station and a UE corresponding to the use of UE-side monitoring of an AI / ML model for CSI compression and decompression.

[0016] FIG. 7 illustrates a diagnostics block as may be used at a UE to identify a source of performance degradation, according to embodiments discussed herein.

[0017] FIG. 8 illustrates a diagram for the use of a UE reference encoder Eref for purposes of monitoring an AI / ML model for CSI compression and decompression.

[0018] FIG. 9 illustrates a method of a base station, according to embodiments discussed herein.

[0019] FIG. 10 illustrates a method of a UE. according to embodiments discussed herein.

[0020] FIG. 11 illustrates a method of a base station, according to embodiments discussed herein.

[0021] FIG. 12 illustrates a method of a UE, according to embodiments discussed herein.

[0022] FIG. 13 illustrates an example architecture of a wireless communication system, according to embodiments disclosed herein.

[0023] FIG. 14 illustrates a system for performing signaling between a wireless device and a network device, according to embodiments disclosed herein.DETAILED DESCRIPTION

[0024] Various embodiments are described with regard to a UE. However, reference to a UE is merely provided for illustrative purposes. The example embodiments may be3P70697WO1 4898-9140-0001\1utilized 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.

[0025] 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.

[0026] 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.

[0027] Accordingly, in various wireless communication systems, artificial intelligence (AI) / machine learning (ML)-based CSI encoding / compression and decoding / decompression mechanisms may be used to reduce signaling overhead associated with the transmission of CSI from the UE to the base station.

[0028] 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.

[0029] 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 IF).

[0030] The bitstream 118 is then transmitted from the UE 106 to the network 110.4P70697WO1 4898-9140-0001\1

[0031] 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 IV). corresponding to the format of the input 116.

[0032] 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.

[0033] 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 the input 116 is itself sent from the UE 106 to the network side 108 without such encoding / compression.

[0034] The output 120 may correspondingly be referred to variously herein as “decoded,” “decompressed,” “recovered,” etc.

[0035] Discussion herein relates to the monitoring of an AI / ML model for CSI compression and decompression. This monitoring may be used to determine that, for example, a given AI / ML model for CSI compression and decompression is no longer performing with acceptable accuracy / precision (in the sense that, for example, in terms of the diagram 100 of FIG. 1, an output 120 of a decoder 114 exhibits an insufficient match to an input 116 of an 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.

[0036] Embodiments herein relate to mechanisms for performing monitoring of AI / ML models for CSI compression and decompression that can identify or narrow down the causes of a performance degradation, such as whether the cause is a UE-side encoder of the AI / ML model, a network-side decoder of the AI / ML model, and / or due to data drift, etc.

[0037] Mechanisms for the post-deployment performance monitoring of an AI / ML model for CSI compression and decompression that identify and / or narrow down the 5P70697WO1 4898-9140-0001\1causes of performance degradation can aid in the maintenance of good performance in the field.

[0038] Options for such performance monitoring fall into various categories. A first such category includes cases of network-sided monitoring of the AI / ML model (e.g.. base station-based monitoring of the AI / ML model). Considerations for such cases for network-sided monitoring may be developed in consideration of overhead, latency, complexity, monitoring accuracy, and / or UE capability' aspects. Such network-sided monitoring cases may be based on a target CSI that is reported by a UE via an enhanced type 2 (eT2) codebook or some eT2-like high resolution codebook. Additionally, or alternatively, such network-sided monitoring cases may use sounding reference signal (SRS)-based monitoring.

[0039] Another such category for the monitoring of an AI / ML model for CSI compression and decompression includes cases of UE-sided monitoring of the AI / ML model. Such cases may be based on an output of a CSI reconstruction model (a UE-sided decoder). In such examples, the CSI reconstruction model can be the same as the “active” reconstruction model being used at the network (a network-sided decoder) that is being used for substantive inferencing, a reference CSI reconstruction model provided to the UE by the network, or a proxy CSI reconstruction model developed at the UE side. In some cases, for UE-sided monitoring, the monitoring occurs via direct estimation of one or more intermediate key performance indicator(s) (KPI(s)) (such as, for example, a squared generalized cosine similarity (SGCS)), without an actual reconstruction of a target CSI being performed. In some cases, for UE-sided monitoring, the monitoring occurs via an estimation of a monitoring output that is other than an intermediate KPI, without reconstructing a target CSI.

[0040] Preliminarily, it is noted that FIG. 1 as described above introduces its AI / ML model 102 for CSI compression and decompression in terms of an encoder 112 that is found at a UE 106 and a decoder 114 that is located at the base station of a network 110. This discussion corresponds to the active use of an already -trained AI / ML model for CSI compression and decompression purposes, where the encoder 112 and the decoder 114 are “active” in the sense of being used / ready for use in an active inferencing deployment where CSI compression and decompression is used as part of a regular CSI feedback procedure between the UE and the base station.6P70697WO1 4898-9140-0001\1

[0041] Within the monitoring contexts described herein, it will be understood that, in addition to the “active” encoder at a UE and the “active” decoder at the network, other entities associated with the AI / ML model could be used / developed in an intermediate fashion (e.g., to facilitate the monitoring of the “active” (portion(s) of the) AI / ML model). For example, there may be one or more decoders at a UE and / or one or more encoders at a base station that are associated with training and / or monitoring the AI / ML model.

[0042] FIG. 2 illustrates a diagram 200 for a case of UE-based AI / ML model monitoring 202. Preliminarily, a set of CSI and other conditions 204 (which may include CSI 206) is provided to an encoder / decoder pair 208 at a UE. The encoder / decoder pair 208 may include a UE-side encoder 210 and a UE-side decoder 212, as illustrated. Note that in some cases, the UE-side encoder 210 as may be used / planned for use during “active” AI / ML model inferencing. Further, the UE-side decoder 212 may be a UE-trained decoder, a network-provided reference decoder, and / or a copy of a network-side decoder that is used by the UE for purposes of the UE-based AI / ML model monitoring 202 but is not used by the UE during active AI / ML model inferencing.

[0043] The CSI 206 is provided to the UE-side encoder 210, which generates the bitstream 218. The bitstream 218 is then passed to the UE-side decoder 212, which generates the recovered CSI 220. The UE then performs an accuracy inference 214 based on the CSI 206 and the recovered CSI 220 (e.g., a loss between the CSI 206 and the recovered CSI 220 is calculated, as shown). The results of the accuracy inference 214 are then passed along as monitoring results 216 that are used to evaluate the accuracy of one or more of the UE-side encoder 210 and / or the UE-side decoder 212.

[0044] Various embodiments herein relate to mechanisms for AI / ML model training with the goal of developing a decoder / encoder pairing corresponding to an AI / ML model for CSI compression and decompression. In such training contexts, it will be understood that, in addition to an encoder at a UE and a decoder at the network, other entities associated with the AI / ML model could be used / developed in an intermediate fashion (e.g.. to facilitate the training of the AI / ML model). For example, there may be one or more decoders at a UE and / or one or more encoders at a base station that are associated with training the AI / ML model (e.g., as will now be discussed in FIG. 3).

[0045] FIG. 3 illustrates a diagram 300 for an inter-vendor training collaboration procedure for training an AI / ML model for CSI compression and decompression that7P70697WO1 4898-9140-0001\1relies on standardized data / a standardized dataset format and a dataset exchange between a base station and a UE.

[0046] As illustrated, at the network 302, a joint training 306 for a network-side encoder Ei 308 and a network-side decoder Di 310 is performed.

[0047] Then, a dataset 312 is sent from the network 302 to the UE 304. The dataset 312 includes target CSI and CSI feedback information. The CSI feedback information may include model parameters for the AI / ML model at the network 302 (as represented by the network-side encoder Ei 308 and / or the network-side decoder Di 310) Note that the dataset 312 may be a partial dataset.

[0048] Training at the UE 304 then proceeds according to one of a first alternative 314 and a second alternative 316.

[0049] In the first alternative 314, the UE 304 performs training in multiple steps. In a first step 324, the UE 304 trains a UE-side decoder D2318 using the dataset 312 (e.g., the target CSI and the CSI feedback information) received from the network 302. Note that as part of this process, the UE 304 may train the UE-side decoder D2318 in such a way that the UE-side decoder D2318 has a different structure than the structure of the network-side decoder Di 310

[0050] Then, in a second step 326, the UE 304 freezes the model for the UE-side decoder D2318 and performs a joint training procedure with the UE-side decoder D2318 to train a UE-side encoder E2320. Note that as part of this process, the UE 304 may train the UE-side encoder E2320 in such a way that the UE-side encoder E2320 has a different structure than the structure of the network-side encoder Ei 308.

[0051] In a second alternative 316, upon receiving the dataset 312 (e.g., the target CSI and the CSI feedback information), the UE 304 uses the dataset 312 to train a UE-side encoder E2322. Note that as part of this process, the UE 304 may train UE-side encoder E2322 in such a way that the UE-side encoder E2322 has a different structure than the structure of the network-side encoder Ei 308.

[0052] Accordingly, it may be understood that in cases corresponding to such examples, a network vendor first trains a pair of (CSI generation model (a network-side encoder Ei 308), CSI reconstruction model (network-side decoder Di 310)) Then, a corresponding dataset 312 is sent to a UE 304 using the standardized dataset format.8P70697WO1 4898-9140-0001\1

[0053] Note that, corresponding to a case of AI / ML model use for active inferencing as was described according to the AI / ML model 102 described in FIG. 1, the UE-side encoder E2320 (in the case the first alternative 314 was used) or the UE-side encoder E2 322 (in the case the second alternative 316 was used) may be understood to correspond to the encoder 112, while the network-side decoder Di 310 may be understood to correspond to the decoder 114. The network-side encoder Ei 308 and the UE-side decoder D2318 (in the case that the first alternative 314 was used) may be understood to have been used for purposes other than active inferencing (e.g., training purposes, as just described) but not ultimately used corresponding to active inferencing using the AI / ML model.

[0054] Note that the notation for a network-side encoder (Ei), a network-side decoder (Di), a UE-side encoder (E2), and a UE-side decoder (D2) will be followed throughout this disclosure.Embodiments for Network Side AI / ML Model Degradation Diagnosis / Monitoring

[0055] FIG. 4A illustrates a flow diagram 400 for communications between a base station 402 and a UE 404 corresponding to the use of network-side 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 first ground-truth CSI 408 (denoted F) based on its measurement of this reference signal.

[0056] The UE 404 then provides the first ground-truth CSI 408 to a UE encoder E2 410 of the AI / ML model for CSI compression and decompression that is sited at the UE 404. The UE encoder E2410 uses the first ground-truth CSI 408 to generate first compressed CSI feedback 412 (e.g., a bitstream) (denoted c).

[0057] The UE 404 then sends 414 the first ground-truth CSI 408 and the first compressed CSI feedback 412 to the base station 402, as illustrated.

[0058] 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 first ground-truth CSI 408 to the base station 402. 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 first ground-truth CSI 408 may turn at least in part on the reporting mode(s) so used. Note that, due to the nature of9P70697WO1 4898-9140-0001\1the first ground-truth CSI 408 as measuring the reference signal as impacted by the channel between the base station 402 and the UE 404, the first ground-truth CSI 408 is highly reliable in view of any presently applicable environmental factors.

[0059] The base station 402 provides the first compressed CSI feedback 412 as received from the UE 404 to a base station decoder Di 416 of the AI / ML model for CSI compression that is sited at the base station 402. Using the first compressed CSI feedback 412, the base station decoder Di 416 generates first recovered CSI 418 (denoted CSI2, with the “21” portion of this notation corresponding to the fact that the first recovered CSI 418 was generated using the UE encoder E2410 and the base station decoder Di 416).

[0060] The base station 402 then calculates a first KPI 422 (denoted KPh) using both the received first ground-truth CSI 408 (which is understood in context as the actual channel estimation performed by the UE 404) and the first recovered CSI 418 (which is the result of encoding / compression of the first ground-truth CSI 408 at the UE encoder E2410 and then decoding / decompression of that result at the base station decoder Di 416).

[0061] In some cases, the KPI calculation 420 for the first KPI 422 takes the form of an SGCS calculation that results in an indication of the amount of similarity between the first recovered CSI 418 and the first ground-truth CSI 408. Other KPI calculation ty pes may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the first ground-truth CSI 408 and the first recovered CSI 418).

[0062] The first ground-truth CSI 408 as provided to the base station 402 from the UE 404 is also provided 424 at a base station encoder Ei 426, as illustrated. The base station encoder Ei 426 may be, for example, a reference base station encoder or an encoder that was part of a joint training for the base station encoder Ei 426 and the base station decoder Di 416 (refer to, e.g., the discussion provided herein in relation to the joint training of the network-side encoder Ei 308 and the network-side decoder Di 310 of FIG.3). The base station encoder Ei 426 uses the first ground-truth CSI 408 to generate second compressed CSI feedback 428 (denoted ci).

[0063] The base station 402 then provides the second compressed CSI feedback 428 to the base station decoder Di 416. Using the second compressed CSI feedback 428, the10P70697WO1 4898-9140-0001\1base station decoder Di 416 generates second recovered CSI 430 (denoted CSI1±, with the “11” portion of this notation corresponding to the fact that the second recovered CSI 430 was generated using the base station encoder Ei 426 and the base station decoder Di 416.

[0064] The base station 402 then calculates a second KPI 432 (denoted KPI ) using both the received first ground-truth CSI 408 (which is understood in context as the actual channel estimation performed by the UE 404) and the second recovered CSI 430 (which is the result of encoding / compression of the first ground-truth CSI 408 at the base station encoder Ei 426 and then decoding / decompression of that result at the base station decoder Di 416).

[0065] In some cases, the KPI calculation 420 for the second KPI 432 takes the form of an SGCS calculation that results in an indication of the amount of similarity between the second recovered CSI 430 and the first ground-truth CSI 408. Other KPI calculation types may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the first ground-truth CSI 408 and the second recovered CSI 430).

[0066] The base station 402 may be configured to use a performance threshold d to evaluate the first KPI 422 and the second KPI 432. For example, each of the first KPI 422 and the second KPI 432 may be compared to the performance threshold d. The result of these comparisons provides an indication / narrowing of possible sources of performance degradation in the AI / ML model.

[0067] It is noted that the first KPI 422 corresponds to a case of the use of the UE encoder E2410, while the second KPI 432 alternatively corresponds to a case of the use of the base station encoder Ei 426. Accordingly, if the first KPI 422 does not meet the threshold while the second KPI 432 meets the threshold, this informs that the UE encoder E2410 of the UE model is likely a source of any performance degradation in the AI / MU model (a UE-side issue). In some cases, a corresponding response to this determination is that the UE encoder E2410 is retrained or replaced, and / or that the use of the UE encoder E2410 is ended.

[0068] In alternative cases where both the first KPI 422 and the second KPI 432 fail to meet the threshold, this indicates that any performance degradation in the AI / ML model11P70697WO1 4898-9140-0001\1is likely more attributable to either data drift or one or more issues with one of the base station encoder Ei 426 and / or the base station decoder Di 416 (a network-side issue).

[0069] In cases where the KPI of relevance is an SGCS KPI, the ideal / best theoretically possible result of SGCS procedure used to calculate the first KPI 422 / the second KPI 432 is 1, while the worst possible case corresponds to a value 0. Accordingly, the performance threshold 3 may be set at some value between 0 and 1 in such cases. Then, an SGCS-based KPI such as the first KPI 422 and / or the second KPI 432 fails to meet the threshold when it is less than (or, alternatively, less than or equal to) the value of the performance threshold <5, and meets the threshold when it is greater than or equal to (or alternatively, greater than) the value of the performance threshold 3.

[0070] Note that the use of a threshold may be somewhat different in the cases where the KPI of relevance is other than an SGCS KPI. For example, in the case that the KPI of interest is an error calculation KPI, the ideal / best theoretically possible result of the KPI calculation procedure is 0, (while the worst possible case is theoretically unbounded). Accordingly, the performance threshold 3 may be set at some value above 0 in such cases. Then, an error-based KPI calculation that calculates an error between the first ground-truth CSI 408 to the first recovered CSI 418 and / or that calculates an error between the first ground-truth CSI 408 and the second recovered CSI 430 fails to meet the threshold when it is greater than (or, alternatively, greater than or equal to) the value of the performance threshold 3, and meets the threshold when it is less than or equal to (or alternatively, less than) the value of the performance threshold 3.

[0071] Assuming that the result of the comparisons of the first KPI 422 and the second KPI 432 result in a conclusion that the UE encoder E2410 is the source of a performance degradation in the AI / ML model, it may be that the UE encoder E2410 is replaced and / or retrained. Then, the system may perform another iteration of the procedure with the replaced / retrained UE encoder.

[0072] FIG. 4B illustrates a flow diagram 434 for communications between the base station 402 and the UE 404 corresponding to the use of network-side monitoring of an AI / MU model for CSI compression and decompression after the events described in FIG.4A. As illustrated, the flow diagram 434 corresponds to an iteration that occurs after the UE encoder E2410 has been replaced with and / or retrained into the UE encoder E'2436.12P70697WO1 4898-9140-0001\1

[0073] The base station 402 sends 438 the UE 404 a reference signal. The UE 404 generates second ground-truth CSI 444 (denoted Vnext) based on its measurement of this reference signal.

[0074] The UE 404 then provides the second ground-truth CSI 444 to the UE encoder E'2436 of the AI / ML model for CSI compression and decompression that is sited at the UE 404. Recall that the UE encoder E'2436 has replaced the UE encoder E2410 and / or represents a retrained version of the UE encoder E2410, as discussed. The UE encoder E'2436 uses the second ground-truth CSI 444 to generate third compressed CSI feedback 440 (e.g., a bitstream) (denoted c').

[0075] The UE 404 then sends 442 the second ground-truth CSI 444 and the third compressed CSI feedback 440 to the base station 402, as illustrated. The reporting of the second ground-truth CSI 444 may be along the lines of the reporting of the first groundtruth CSI 408 as previously discussed.

[0076] The base station 402 provides the third compressed CSI feedback 440 as received from the UE 404 to the base station decoder Di 416. Using the third compressed CSI feedback 440, the base station decoder Di 416 generates third recovered CSI 446 (denoted CSI2'i, with the "‘2'1” portion of this notation corresponding to the fact that the third recovered CSI 446 was generated using the UE encoder E'2436 and the base station decoder Di 416.

[0077] The base station 402 then calculates a third KPI 448 (denoted KPk ) using both the received second ground-truth CSI 444 (which is understood in context as the actual channel estimation performed by the UE 404) and the third recovered CSI 446 (which is the result of encoding / compression of the second ground-truth CSI 444 at the UE encoder E'2436 and then decoding / decompression of that result at the base station decoder Di 416).

[0078] In some cases, the KPI calculation 420 for the third KPI 448 takes the form of an SGCS calculation that results in an indication of the amount of similarity between the third recovered CSI 446 and the second ground-truth CSI 444, consistent with the manner of calculating the first KPI 422 and / or the second KPI 432 as previously described. Again, other KPI calculation types may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the second ground-truth CSI 444 and the third recovered CSI 446).13P70697WO1 4898-9140-0001\1

[0079] Note that according to the embodiment of FIG. 4B, the UE 404 also sends 442 the base station 402 the second ground-truth CSI 444 in addition to the third compressed CSI feedback 440. Accordingly, the base station 402 can use the first ground-truth CSI 408 (which was used as the training dataset at the UE 404 for training the UE encoder E'2 436) with its base station encoder Ei 426 to generate a fourth compressed CSI feedback 450 (denoted c2l- Then, the fourth compressed CSI feedback 450 is applied with the base Astation decoder Di 416 to generate a fourth recovered CSI 452 (denoted CSIll next, with the “11” portion of this notation corresponding to the fact that the fourth recovered CSI 452 was generated using the base station encoder Ei 426 and the base station decoder Di 416).

[0080] The base station 402 then performs a KPI calculation 420 using the second ground-truth CSI 444 and the fourth recovered CSI 452 that results in the fourth KPI 454 (denoted KP \ The form of the fourth KPI 454 may be the same as the form of the third KPI 448. as discussed.

[0081] FIG. 4B corresponds to a case where the third KPI 448 (corresponding to the change from the UE encoder E2410 to the UE encoder E'2436) and the fourth KPI 454 are each compared to the performance threshold d.

[0082] It is noted that the third KPI 448 corresponds to a case of the use of the UE encoder E'2436 at the UE 404, while the fourth KPI 454 alternatively corresponds to a case of the use of the base station encoder Ei 426. Accordingly, if the third KPI 448 does not meet the performance threshold d while the fourth KPI 454 meets the performance threshold <5, this informs that the UE encoder E'2436 of the UE model is likely a source of any performance degradation in the AI / ML model (e.g., there remains a UE-side issue even after the change from the UE encoder E2410 to the UE encoder E'2436).

[0083] In alternative cases where both the third KPI 448 and the fourth KPI 454 fail to meet the performance threshold <5, this indicates that any performance degradation in the AI / ML model is likely now more attributable to either data drift and / or one or more issues with one of the base station encoder Ei 426 and / or the base station decoder Di 416 (e.g., a network-side issue).

[0084] This procedure may be understood as a determination of whether the fourth recovered CSI 452 is a typical sample corresponding to the second ground-truth CSI 444 (or not).14P70697WO1 4898-9140-0001\1

[0085] FIG. 5 illustrates a diagnostics block 502 as may be used at a network / base station identify a source of performance degradation, according to embodiments discussed herein. As shown, the diagnostics block 502 may be capable of receiving KPIs 504, comparing them to a performance threshold 3506, and then analyzing the results (e.g., relatively to each other, as has been described), to arrive at an AI / ML model degradation source prediction 508. The AI / ML model degradation source prediction 508 may be that a UE-side encoder is a likely source of performance degradation of the AI / ML model, that one of a base station encoder Ei and / or the base station decoder Di is a likely source of performance degradation of the AI / ML model, and / or that data drift likely explains any performance degradation of the AI / ML model, etc., as has been discussed herein.Embodiments for UE Side AI / ML Model Degradation Diagnosis / Monitoring

[0086] FIG. 6A and FIG. 6B together illustrate a flow diagram 600 for communications between a base station 602 and a UE 604 corresponding to the use of UE-side monitoring of an AI / ML model for CSI compression and decompression. Preliminarily, the base station 602 sends 606 the UE 604 a reference signal. The UE 604 generates ground-truth CSI 608 (denoted F) based on its measurement of this reference signal.

[0087] The UE 604 then provides the ground-truth CSI 608 to a UE encoder E2610 of the AI / ML model for CSI compression that is sited at the UE 604. The UE encoder E2 610 uses the ground-truth CSI 608 to generate first compressed CSI feedback 612 (e.g., a bitstream) (denoted c).

[0088] The UE 604 then sends 614 the ground-truth CSI 608 and the first compressed CSI feedback 612 to the base station 402, as illustrated.

[0089] The base station 602 provides the first compressed CSI feedback 612 as received from the UE 604 to a base station decoder Di 616 of the AI / ML model for CSI compression and decompression. Using the first compressed CSI feedback 612, the base Astation decoder Di 616 generates first recovered CSI 618 (denoted CSI21. with the “21” portion of this notation corresponding to the fact that the first recovered CSI 618 was generated using the UE encoder E2610 and the base station decoder Di 616).

[0090] The base station 602 then sends 620 the first recovered CSI 618 to the UE 604. 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 first recovered15P70697WO1 4898-9140-0001\1CSI 618 to the base station 402. 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 first recovered CSI 618 may turn at least in part on the reporting mode(s) so used.

[0091] The UE 604 then calculates a first KPI 632 (denoted KPI.) using both the received ground-truth CSI 608 (which is understood in context as the actual channel estimation performed by the UE 604) and the first recovered CSI 618 as received from the base station 602 (which is the result of encoding / compression of the ground-truth CSI 608 at the UE encoder E2610 and then decoding / decompression of the result by the base station decoder Di 616).

[0092] In some cases, this KPI calculation 622 takes the form of an SGCS calculation that results in an indication of the amount of similarity between the first recovered CSI 618 and the ground-truth CSI 608. Other KPI calculation types may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the first ground-truth CSI 608 and the first recovered CSI 618).

[0093] In sum, it will be understood that the procedure for generating the first KPI 632 includes a reconstruction by the base station 602 of first recovered CSI 618 from UE CSI feedback, where this the first recovered CSI 618 is signaled (e.g., through eT2 feedback or the like) back to the UE 604. Then, the UE computes the first KPI 632 based on the first recovered CSI 618 (e.g.. according to KPf = f(V, CSI21))

[0094] As shown, the UE 604 sends 614 the first ground-truth CSI 608 to the base station 602 along with the first compressed CSI feedback 612. The base station 602 is accordingly enabled to provide 634 the ground-truth CSI 608 at a base station encoder Ei 630, as illustrated. The base station encoder Ei 630 may be, for example, a reference base station encoder or an encoder that was part of a joint training for the base station encoder Ei 630 and the base station decoder Di 616 (refer to, e.g.. the discussion provided herein in relation to the joint training of the network-side encoder Ei 308 and the network-side decoder Di 310 of FIG. 3). The base station encoder Ei 630 uses the first ground-truth CSI 608 to generate second compressed CSI feedback 636 (denoted ci).16P70697WO1 4898-9140-0001\1

[0095] The base station 602 then provides the second compressed CSI feedback 636 to the base station decoder Di 616. Using the second compressed CSI feedback 636, the Abase station decoder Di 616 generates second recovered CSI 638 (denoted CSI^, with the “11” portion of this notation corresponding to the fact that the second recovered CSI 638 was generated using the base station encoder Ei 630 and the base station decoder Di 616. The base station 602 then sends 620 the UE 604 the second recovered CSI 638, as shown.

[0096] The UE 604 then calculates a second KPI 640 (denoted KPIi) using both the received first ground-truth CSI 608 (which is understood in context as the actual channel estimation performed by the UE 604) and the second recovered CSI 638 (which is the result of encoding / compression of the first ground-truth CSI 608 at the base station encoder Ei 426 and then decoding / decompression of that result at the base station decoder Di 616).

[0097] In some cases, the KPI calculation 622 for the second KPI 640 takes the form of an SGCS calculation that results in an indication of the amount of similarity between the second recovered CSI 638 and the first ground-truth CSI 608. Other KPI calculation types may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the first ground-truth CSI 608 and the second recovered CSI 638).

[0098] In sum, it will be understood that the procedure for generating the second KPI 640 includes a use of ground-truth CSI 608 received at a base station from a UE that passed through both an encoder at the base station (the base station encoder Ei 630) and a decoder at the base station (the base station decoder Di 616), such that the base station generates a second recovered CSI 638 CSIithat is passed to the UE 604. Then, the UE 604 computes the second KPI 640 based on the second recovered CSI 638 (e.g., according to KPI2= f(V, CSI1:L)).

[0099] The UE 604 may be configured to use a performance threshold 3 to evaluate the first KPI 632 and the second KPI 640. For example, each of the first KPI 632 and the second KPI 640 may be compared to the performance threshold S. The result of these comparisons provides an indication / narrowing of possible sources of performance degradation in the AI / ML model.17P70697WO1 4898-9140-0001\1

[0100] It is noted that the first KPI 632 corresponds to a case of the use of the UE encoder E2610, while the second KPI 640 alternatively corresponds to a case of the use of the base station encoder Ei 630. Accordingly, if the first KPI 632 does not meet the threshold while the second KPI 640 meets the threshold, this informs that the UE encoder E2610 of the UE model is likely a source of any performance degradation in the AI / ML model (a UE-side issue). In some cases, a corresponding response to this determination is that the UE encoder E2610 is retrained or replaced, and / or that the use of the UE encoder E2610 is ended.

[0101] In alternative cases where both the first KPI 632 and the second KPI 640 fail to meet the threshold, this indicates that any performance degradation in the AI / ML model is likely more attributable to either data drift or one or more issues with one of the base station encoder Ei 630 and / or the base station decoder Di 616 (a network-side issue).

[0102] Details with respect to different possible forms (e.g., SGCS, NMSE. etc.) for the first KPI 632 and or the second KPI 640 and how those forms may meet or fail to meet the performance threshold 0 are along the lines discussed elsewhere herein.

[0103] Corresponding to the diagram 600. it is observed that the UE 604 has possession of a UE decoder D2628. This may be, for example, a proxy decoder that was used as part of a training collaboration procedure (refer to, e.g., the first alternative 314 of FIG. 3). In some embodiments, the UE 604 may provide the first compressed CSI feedback 612 as previously described to the UE decoder D2628 to generate a third recovered CSI 642 (denoted CSI22, with theL‘22’?portion of this notation corresponding to the fact that the third recovered CSI 642 was generated using the UE encoder E2610 and the UE decoder D2628).

[0104] Then, the UE 604 can compute a third KPI 644 (denoted KPI3) using both the received ground-truth CSI 608 (which is understood in context as the actual channel estimation performed by the UE 604) and the third recovered CSI 642 (which is the result of encoding / compression of the ground-truth CSI 608 at the UE encoder E2610 and then decoding / decompression of the result by the UE decoder D2628). This third KPI 644 may be used to make one or more determinations with respect to a likely source of any degradation of the AI / ML model for CSI compression and decompression and take a corresponding response.18P70697WO1 4898-9140-0001\1

[0105] Finally, corresponding to the diagram 600, it is observed that in some embodiments, the base station 602 sends 620 the second compressed CSI feedback 636 to the UE 604 (e.g., along with the first recovered CSI 618 and / or the second recovered CSI 638, as shown). In this case, the UE may provide 646 the second compressed CSI feedback 636 to the UE decoder D2628 and then use the UE decoder D2628 to decode this second compressed CSI feedback 636 to generate a fourth recovered CSI 648A(denoted CSI12, with theL‘12” portion of this notation corresponding to the fact that the fourth recovered CSI 648 was generated using the base station encoder Ei 630 and the UE decoder D2628).

[0106] Then, the UE 604 can compute a fourth KPI 650 (denoted KPP) using both the received ground-truth CSI 608 (which is understood in context as the actual channel estimation performed by the UE 604) and the fourth recovered CSI 648 (which is the result of encoding / compression of the ground-truth CSI 608 at the base station encoder Ei 630 and then decoding / decompression of the result by the UE decoder D2628). This fourth KPI 650 may be used to make one or more determinations with respect to a likely source of any degradation of an AI / ML model for CSI compression and decompression and take a corresponding response.

[0107] In some embodiments, the UE 604 sends 624 KPI feedback 626 (one or more KPIs and / or KPI-based results of the KPI calculation 622) to the base station 602, as illustrated. In such cases, it may be the base station 602 that, for example, compares one or more of these reported KPI(s) to a performance threshold 3. In other cases, the UE 604 may perform the comparison(s) of the one or more KPIs to a performance threshold 3 and then provide a result (e.g., a determination of a likely source of performance degradation) to the base station 602 along the KPI feedback 626. Further, the base station 602 can take decisions responsive to these KPIs and / or results, such as activating / instructing the activation of anew AI / ML model for CSI compression and decompression, de-activating / instructing the de-activation of the current AI / ML model, and / or ending / instructing the end of the use of AI / ML model-based behavior for CSI reporting altogether, etc.

[0108] FIG. 7 illustrates a diagnostics block 702 as may be used at a UE to identify a source of performance degradation, according to embodiments discussed herein. As shown, the diagnostics block 702 may be capable of receiving KPIs 704, comparing them to a performance threshold 3706, and then analyzing the results (e.g., relatively to each19P70697WO1 4898-9140-0001\1other, as has been described), in order to arrive at an AI / ML model degradation source prediction 708. The AI / ML model degradation source prediction 708 may be that a UE-side encoder is a likely source of performance degradation of the AI / ML model, that one of a base station encoder Ei and / or the base station decoder Di is a likely source of performance degradation of the AI / ML model, and / or that data drift likely explains any performance degradation of the AI / ML model, etc., as has been discussed herein.Embodiments for Reference Encoder use for AI / ML Model Degradation Diagnosis / Monitoring

[0109] FIG. 8 illustrates a diagram 800 for the use of a UE reference encoder Eref 808 for purposes of monitoring an AI / ML model for CSI compression and decompression. As illustrated, the embodiment of FIG. 8 contemplates that a UE 802 possesses the UE reference encoder Eref 808 and a UE encoder E2806, and that the base station 804 possesses a base station decoder Di 810.

[0110] The UE encoder E2806 and the base station decoder Di 810 may respectively represent the encoder / decoder pair used by an AI / ML model for CSI compression and decompression during active UE inferencing. The UE reference encoder Eref 808 may be a reference encoder that is preconfigured to and / or configured to the UE 802 for purposes of performing monitoring of that AI / ML model.

[0111] In other words, to assess a UE-side CSI generation model (e.g., the UE encoder E2806), a process contemplated by the embodiment shown in FIG. 8 involves using the active UE encoder E2806 and base station decoder Di 810 that are anticipated for use for active for CSI feedback, alongside a UE reference encoder Eref 808 that is compatible with the base station decoder Di 810.

[0112] To facilitate the monitoring, an inferencing stage is conducted using both the combination of the UE reference encoder Eref 808 and the base station decoder Di 810 and the combination of the UE encoder E2 806 and the base station decoder Di 810. First, as shown, a ground-truth CSI 812 (denoted F) is applied to the UE encoder E2806 to generate first compressed CSI feedback 814 (denoted c). The first compressed CSI feedback 814 is then sent to the base station 804, which applies it with the base station decoder Di 810 to generate the first recovered CSI 816 (denoted CSI21, with the “21” portion of this notation corresponding to the fact that the first recovered CSI 816 was generated using the UE encoder E2806 and the base station decoder Di 810). As the first20P70697WO1 4898-9140-0001\1recovered CSI 816 corresponds to the use of the UE encoder E2806 and the base station decoder Di 810 that are associated with active inferencing use of the AI / ML model, the first recovered CSI 816 may be understood as “active recovered CSI.’’

[0113] Further, as shown, a reference CSI 818 (denoted Frey) is applied at the UE reference encoder Eref 808 to generate a second compressed CSI feedback 820 (e.g., a reference compressed CSI feedback) (denoted cref). The second compressed CSI feedback 820 is then sent to the base station 804, which applies it with the base station decoder Di 810 to generate second recovered CSI 822 (e g., a reference reconstructed A CSI) (denoted CSIre^. with the “re / 1” portion of this notation corresponding to the fact that the second recovered CSI 822 was generated using the UE reference encoder Eref 808 and the base station decoder Di 810. As the second recovered CSI 822 corresponds to the use of the UE reference encoder Eref 808, the second recovered CSI 822 may be understood as “reference recovered CSI.”

[0114] In this case, since each of the UE reference encoder Eref 808 and the UE encoder E2806 use the same CSI reconstruction model (the base station decoder Di 810), any performance difference that is determined by comparing a KPI of the first recovered CSI 816 and a KPI of the first recovered CSI 816 may be understood to largely reflects the distinctions between the UE encoder E2806 and the UE reference encoder Eref 808.

[0115] In some cases, the KPI calculation for the first recovered CSI 816 uses the ground-truth CSI 812 and the first recovered CSI 816. This KPI calculation can take the form of an SGCS calculation that results in an indication of the amount of similarity between the first ground-truth CSI 812 and the first recovered CSI 816. Other KPI calculation types may be used in other embodiments (e.g., such as an error calculationbased KPI that calculates an error between the ground-truth CSI 812 and the first recovered CSI 816).

[0116] In some cases, the KPI calculation for the second recovered CSI 822 uses the reference CSI 818 and the second recovered CSI 822. This KPI calculation can take the form of an SGCS calculation that results in an indication of the amount of similarity between the reference CSI 818 and the second recovered CSI 822. Other KPI calculation types may be used in other embodiments (e.g., such as an error calculation-based KPI that calculates an error between the reference CSI 818 and the second recovered CSI 822).21P70697WO1 4898-9140-0001\1

[0117] Corresponding to such cases, when the KPI of the first recovered CSI 816 corresponding to the use of the UE encoder E2806 and the KPI of the second recovered CSI 822 corresponding to the use of the UE reference encoder Eref 808 each show satisfactory results (e.g., cases where each meet a performance threshold 3 as has been discussed herein and / or cases where the first recovered CSI 816 is close to the second recovered CSI 822), it confirms that the UE encoder E2806, is operating as intended.

[0118] Conversely, if the KPI of the first recovered CSI 816 corresponding to the use of the UE encoder E2806 is worse than the KPI for the second recovered CSI 822 corresponding to the use of the UE reference encoder Eref 808 and / or is not meeting a performance threshold <5, this points to a potential problem UE encoder E2806. However, if in such cases the second recovered CSI 822 corresponding to the use of the UE reference encoder Eref 808 also exhibits poor performance (e.g., with respect to a performance threshold <5), the root cause may lie with the base station decoder Di 810 and / or be driven by issues related to data drift, requiring additional analysis.

[0119] In some embodiments, the base station 804 performs the KPI calculations and analyzes the KPIs. In such cases, the UE 802 may provide the base station 804 with the ground-truth CSI 812 and / or the reference CSI 818 (if not already known at the base station 804) to facilitate this process. The base station 804 may be configured to use a performance threshold 3 to evaluate these KPIs. For example, each of the KPIs may be compared to the performance threshold 3, as is discussed herein. The result of these comparisons provides an indication / narrowing of possible sources of performance degradation in the AI / ML model for CSI compression and decompression, in one or more of the ways detailed herein.

[0120] In some embodiments, the UE 802 performs the KPI calculations and analyzes the KPIs. In such cases, the base station 804 may provide the UE 802 with the first recovered CSI 816 and the second recovered CSI 822 to facilitate this process. The UE 802 may be configured to use a performance threshold 3 to evaluate these KPIs. For example, each of the KPIs may be compared to the performance threshold 3, as is discussed herein. The result of these comparisons provides an indication / narrowing of possible sources of performance degradation in the AI / ML model for CSI compression and decompression, in one or more of the ways detailed herein.22P70697WO1 4898-9140-0001\1

[0121] Note that cases corresponding to the discussion of FIG. 8 where the ground-truth CSI 812 and the reference CSI 818 are equal (e.g., where the ground-truth CSI 812 is reused as the reference CSI 818) are contemplated.Embodiments for Staged Root Cause Identification

[0122] Root cause identification with respect to performance degradations for AI / ML models for CSI compression and decompression as discussed herein may benefit from careful consideration of aspects related to monitoring frequency. In some embodiments, a monitoring procedure may be considered in terms of two stages.

[0123] An initial phase / stage focuses on determining whether the AI / ML model for CSI compression and decompression is the source of performance degradation, or if external factors are at play. This involves assessing the effectiveness of the CSI generation model(s) (e.g., encoder(s)) and the CSI reconstruction model(s) (e.g., decoder(s)) processes, particularly through analyzing the KPI of the path through the UE encoder E2 and a base station decoder Di, which is the case that represents nominal CSI feedback when the AI / ML model is under active inferencing use.

[0124] Then, if this first phase / stage identifies that the AI / ML model as a potential cause, a second phase / stage proceeds from there to isolate the exact issue within the model, enabling a more targeted investigation.

[0125] Accordingly, it may be understood that a first stage / phase exhibits relatively low latency and a relatively lightweight design. This first phase / stage is focused on early detection of performance issues. It operates with low latency, ensuring that potential degradations are flagged in near real-time.

[0126] Further, it may be understood that a second phase / stage represents a relatively more in-depth analysis. Offline operations may be used. This second phase / stage may be triggered by the first phase / stage. In cases where the second phase / stage is triggered based on outcome(s) of the first phase / stage, resources are accordingly allowed to be utilized by the second phase / stage only when necessary.

[0127] The second phase / stage may involve activities such as, for example: signaling the target CSI (through eT2 signaling or the like, corresponding to a high accuracy and high overhead information transfer); passing a target CSI and / or compressed CSI feedback through different combinations of encoder / decoder pairs; performing computations of extra / additional KPI(s); signaling KPI(s) to a diagnostic engine at a UE23P70697WO1 4898-9140-0001\1or a base station; and / or taking the appropriate actions with respect to the AI / ML model for CSI compression and decompression (e.g., a deactivation of an encoder and / or an encoder of the AI / ML model, model retraining of an encoder and / or an encoder of the AI / ML model, performing a model switch from the AI / ML model to a new AI / ML model, etc.).

[0128] Advantages corresponding to the use of the staged procedure include: relatively minimized over-the-air signaling overhead, in that the first phase / stage limits the data exchanged over the network unless / until it affirmatively triggers the second phase / stage, conserving bandwidth and reducing communication costs; optimized resource utilization, in that the staggered nature of the procedure avoids unnecessary computational or operational corresponding to the gated / triggered use of the second phase / stage (which is only run on an as-needed basis); and / or improved system responsiveness, in that by separating relatively immediate monitoring tasks (the first phase / stage) from relatively resource-intensive analysis (the second phase / stage), the system can respond faster to critical situations without being burdened by complex computations.

[0129] FIG. 9 illustrates a method 900 of a base station, according to embodiments discussed herein. The method 900 includes sending 902, method 900 sends, to a UE. a reference signal. The method 900 further includes receiving 904, from the UE, groundtruth CSI for the reference signal and first compressed CSI feedback as generated at a UE encoder using the ground-truth CSI. The method 900 further includes applying 906 the first compressed CSI feedback at a base station decoder to generate first recovered CSI. The method 900 further includes processing 908 the ground-truth CSI using a base station encoder and the base station decoder to generate second recovered CSI. The method 900 further includes performing 910 a first KPI calculation using the groundtruth CSI and the first recovered CSI to generate a first KPI for the first recovered CSI. The method 900 further includes performing 912 a second KPI calculation using the ground-truth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI. The method 900 further includes identifying 914 a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI.

[0130] In some embodiments of the method 900, identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold. In some of these24P70697WO1 4898-9140-0001\1embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold. In some of these embodiments, the first KPI is a first SGCS between the ground-truth CSI and the first recovered CSI; the second KPI is a second SGCS between the ground-truth CSI and the second recovered CSI; and the performance threshold comprises an SGCS threshold. In some of these embodiments, the first KPI is a first error between the ground-truth CSI and the first recovered CSI; the second KPI is a second error between the ground-truth CSI and the second recovered CSI; and the performance threshold comprises an error threshold.

[0131] In some embodiments, the method 900 further includes determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

[0132] FIG. 10 illustrates a method 1000 of a UE, according to embodiments discussed herein. The method 1000 includes identifying 1002, based on a measurement of a reference signal that is received from a base station, ground-truth CSI for the reference signal. The method 1000 further includes applying 1004 the ground-truth CSI at a UE encoder to generate first compressed CSI feedback. The method 1000 further includes sending 1006 the ground-truth CSI and the first compressed CSI feedback to the base station. The method 1000 further includes receiving 1008, from the base station, first recovered CSI associated with using the UE encoder and a base station decoder and second recovered CSI associated with using a base station encoder and the base station decoder. The method 1000 further includes performing 1010 a first KPI calculation using the ground-truth CSI and the first recovered CSI to generate a first KPI for the first recovered CSI. The method 1000 further includes performing 1012 a second KPI calculation using the ground-truth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI. The method 1000 further includes identifying25P70697WO1 4898-9140-0001\11014 a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI.

[0133] In some embodiments of the method 1000, identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold. In some of these embodiments, the first KPI is a first SGCS between the ground-truth CSI and the first recovered CSI; the second KPI is a second SGCS between the ground-truth CSI and the second recovered CSI; and the performance threshold comprises an SGCS threshold. In some of these embodiments, the first KPI is a first error between the ground-truth CSI and the first recovered CSI; the second KPI is a second error between the ground-truth CSI and the second recovered CSI; and the performance threshold comprises an error threshold.

[0134] In some embodiments, the method 1000 further includes determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

[0135] FIG. 11 illustrates a method 1100 of a base station, according to embodiments discussed herein. The method 1100 includes sending 1102, to a UE, a reference signal. The method 1100 further includes receiving 1104, from the UE, ground-truth CSI for the reference signal, first compressed CSI feedback as generated at an active UE encoder using the ground-truth CSI. and second compressed CSI feedback as generated at a reference UE encoder using a reference CSI. The method 11 0 further includes applying 1106s the first compressed CSI feedback at a base station decoder to generate active recovered CSI. The method 1100 further includes applying 1108 the second compressed CSI feedback at the base station decoder to generate reference recovered CSI. The26P70697WO1 4898-9140-0001\1method 1100 further includes performing 1110 a first KPI calculation using the groundtruth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI. The method 1100 further includes performing 1112 a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI. The method 1100 further includes identifying 1114 a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder using the first KPI and the second KPI.

[0136] In some embodiments of the method 1100, identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold. In some of these embodiments, the first KPI is a first SGCS between the ground-truth CSI and the active recovered CSE the second KPI is a second SGCS between the reference CSI and the reference recovered CSI; and the performance threshold comprises an SGCS threshold. In some of these embodiments, the first KPI is a first error between the ground-truth CSI and the active recovered CSI; the second KPI is a second error between the reference CSI and the reference recovered CSI; and the performance threshold comprises an error threshold.

[0137] In some embodiments, the method 1100 further includes determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

[0138] In some embodiments, the method 1100 further includes receiving, from the UE, the reference CSI.

[0139] In some embodiments of the method 1100. the reference CSI is equal to the ground-truth CSI.27P70697WO1 4898-9140-0001\1

[0140] FIG. 12 illustrates a method 1200 of a UE, according to embodiments discussed herein. The method 1200 includes identifying 1202, based on a measurement of a reference signal that is received from a base station, ground-truth CSI for the reference signal. The method 1200 further includes applying 1204 the ground-truth CSI at an active UE encoder to generate first compressed CSI feedback. The method 1200 further includes applying 1206 a reference CSI at a reference UE encoder to generate second compressed CSI feedback. The method 1200 further includes sending 1208, to the base station, the first compressed CSI feedback and the second compressed CSI feedback. The method 1200 further includes receiving 1210, from the base station, active recovered CSI associated with using the UE encoder and a base station decoder and reference recovered CSI associated with using the reference UE encoder and the base station decoder. The method 1200 further includes performing 1212 a first KPI calculation using the groundtruth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI. The method 1200 further includes performing 1214 a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI. The method 1200 further includes identifying 1216 a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder sing the first KPI and the second KPI.

[0141] In some embodiments of the method 1200. identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold. In some of these embodiments, identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold. In some of these embodiments, the first KPI is a first SGCS between the ground-truth CSI and the active recovered CSI; the second KPI is a second SGCS between the reference CSI and the reference recovered CSI; and the performance threshold comprises an SGCS threshold. In some of these embodiments, the first KPI is a first error between the ground-truth CSI28P70697WO1 4898-9140-0001\1and the active recovered CSI; the second KPI is a second error between the reference CSI and the reference recovered CSI; and the performance threshold comprises an error threshold.

[0142] In some embodiments, the method 1200 further includes determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

[0143] In some embodiments of the method 1200. the reference CSI is equal to the ground-truth CSI.

[0144] FIG. 13 illustrates an example architecture of a wireless communication system 1300, according to embodiments disclosed herein. The following description is provided for an example wireless communication system 1300 that operates in conjunction with the LTE system standards and / or 5G or NR system standards as provided by 3 GPP technical specifications.

[0145] As shown by FIG. 13, the wireless communication system 1300 includes UE 1302 and UE 1304 (although any number of UEs may be used). In this example, the UE 1302 and the UE 1304 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.

[0146] The UE 1302 and UE 1304 may be configured to communicatively couple with a RAN 1306. In embodiments, the RAN 1306 may be NG-RAN. E-UTRAN, etc. The UE 1302 and UE 1304 utilize connections (or channels) (shown as connection 1308 and connection 1310, respectively) with the RAN 1306, each of which comprises a physical communications interface. The RAN 1306 can include one or more base stations (such as base station 1312 and base station 1314) that enable the connection 1308 and connection 1310.

[0147] In this example, the connection 1308 and connection 1310 are air interfaces to enable such communicative coupling, and may be consistent with RAT(s) used by the RAN 1306, such as. for example, an UTE and / or NR.

[0148] In some embodiments, the UE 1302 and UE 1304 may also directly exchange communication data via a sidelink interface 1316. The UE 1304 is show n to be configured to access an access point (shown as AP 1318) via connection 1320. By way of example, the connection 1320 can comprise a local wireless connection, such as a 29P70697WO1 4898-9140-0001\1connection consistent with any IEEE 802.11 protocol, wherein the AP 1318 may comprise a Wi-Fi® router. In this example, the AP 1318 may be connected to another network (for example, the Internet) without going through a CN 1324.

[0149] In embodiments, the UE 1302 and UE 1304 can be configured to communicate using orthogonal frequency division multiplexing (OFDM) communication signals with each other or with the base station 1312 and / or the base station 1314 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.

[0150] In some embodiments, all or parts of the base station 1312 or base station 1314 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 1312 or base station 1314 may be configured to communicate with one another via interface 1322. In embodiments where the wireless communication system 1300 is an LTE system (e.g., when the CN 1324 is an EPC), the interface 1322 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 1300 is an NR system (e.g.. when CN 1324 is a 5GC), the interface 1322 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 1312 (e.g., a gNB) connecting to 5GC and an eNB. and / or between two eNBs connecting to 5GC (e.g.. CN 1324).

[0151] The RAN 1306 is shown to be communicatively coupled to the CN 1324. The CN 1324 may comprise one or more network elements 1326, which are configured to offer various data and telecommunications services to customers / subscribers (e.g.. users of UE 1302 and UE 1304) who are connected to the CN 1324 via the RAN 1306. The components of the CN 1324 may be implemented in one physical device or separate physical devices including components to read and execute instructions from a machine-30P70697WO1 4898-9140-0001\1readable or computer-readable medium (e g., a non-transitory machine-readable storage medium).

[0152] In embodiments, the CN 1324 may be an EPC, and the RAN 1306 may be connected with the CN 1324 via an SI interface 1328. In embodiments, the SI interface 1328 may be split into two parts, an SI user plane (Sl-U) interface, which carries traffic data between the base station 1312 or base station 1314 and a serving gateway (S-GW), and the SI -MME interface, which is a signaling interface between the base station 1312 or base station 1314 and mobility management entities (MMEs).

[0153] In embodiments, the CN 1324 may be a 5GC, and the RAN 1306 may be connected with the CN 1324 via an NG interface 1328. In embodiments, the NG interface 1328 may be split into two parts, an NG user plane (NG-U) interface, which carries traffic data between the base station 1312 or base station 1314 and a user plane function (UPF), and the SI control plane (NG-C) interface, which is a signaling interface between the base station 1312 or base station 1314 and access and mobility management functions (AMFs).

[0154] Generally, an application server 1330 may be an element offering applications that use internet protocol (IP) bearer resources with the CN 1324 (e g., packet switched data services). The application server 1330 can also be configured to support one or more communication services (e g., VoIP sessions, group communication sessions, etc.) for the UE 1302 and UE 1304 via the CN 1324. The application server 1330 may communicate with the CN 1324 through an IP communications interface 1332.

[0155] FIG. 14 illustrates a system 1400 for performing signaling 1434 between a wireless device 1402 and a network device 1418, according to embodiments disclosed herein. The system 1400 may be a portion of a wireless communications system as herein described. The wireless device 1402 may be, for example, a UE of a wireless communication system. The network device 1418 may be, for example, a base station (e.g.. an eNB or a gNB) of a wireless communication system.

[0156] The wireless device 1402 may include one or more processor(s) 1404. The processor(s) 1404 may execute instructions such that various operations of the wireless device 1402 are performed, as described herein. The processor(s) 1404 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 hardware31P70697WO1 4898-9140-0001\1device, a firmware device, or any combination thereof configured to perform the operations described herein.

[0157] The wireless device 1402 may include a memory 1406. The memory' 1406 may be anon-transitory' computer-readable storage medium that stores instructions 1408 (which may include, for example, the instructions being executed by the processor(s) 1404). The instructions 1408 may also be referred to as program code or a computer program. The memory 1406 may also store data used by, and results computed by, the processor(s) 1404.

[0158] The wireless device 1402 may include one or more transceiver(s) 1410 that may include radio frequency (RF) transmitter circuitry' and / or receiver circuitry that use the antenna(s) 1412 of the wireless device 1402 to facilitate signaling (e.g., the signaling 1434) to and / or from the wireless device 1402 with other devices (e.g., the network device 1418) according to corresponding RATs.

[0159] The wireless device 1402 may include one or more antenna(s) 1412 (e.g., one, two, four, or more). For embodiments with multiple antenna(s) 1412, the wireless device 1402 may leverage the spatial diversity of such multiple antenna(s) 1412 to send and / or receive multiple 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 1402 may be accomplished according to precoding (or digital beamforming) that is applied at the wireless device 1402 that multiplexes the data streams across the antenna(s) 1412 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).

[0160] In certain embodiments having multiple antennas, the wireless device 1402 may implement analog beamforming techniques, whereby phases of the signals sent by the antenna(s) 1412 are relatively adjusted such that the (joint) transmission of the antenna(s) 1412 can be directed (this is sometimes referred to as beam steering).32P70697WO1 4898-9140-0001\1

[0161] The wireless device 1402 may include one or more interface(s) 1414. The interface(s) 1414 may be used to provide input to or output from the wireless device 1402. For example, a wireless device 1402 that is a UE may include interface(s) 1414 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) 1410 / antenna(s) 1412 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).

[0162] The wireless device 1402 may include an AI / ML model for CSI compression and decompression monitoring module 1416. The AI / ML model for CSI compression and decompression monitoring module 1416 may be implemented via hardware, software, or combinations thereof. For example, the AI / ML model for CSI compression and decompression monitoring module 1416 may be implemented as a processor, circuit, and / or instructions 1408 stored in the memory 1406 and executed by the processor(s) 1404. In some examples, the AI / ML model for CSI compression and decompression monitoring module 141 may be integrated within the processor(s) 1404 and / or the transceiver(s) 1410. For example, the AI / ML model for CSI compression and decompression monitoring module 1416 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) 1404 or the transceiver(s) 1410.

[0163] The AI / ML model for CSI compression and decompression monitoring module 1416 may be used for various aspects of the present disclosure, for example, aspects of FIG. 10 and / or FIG. 12. The AI / ML model for CSI compression and decompression monitoring module 1416 may configure the wireless device 1402 to identify, based on a measurement of a reference signal that is received from a base station, ground-truth CSI for the reference signal; apply the ground-truth CSI at a UE encoder to generate first compressed CSI feedback; send the ground-truth CSI and the first compressed CSI feedback to the base station; receive, from the base station, first recovered CSI associated with using the UE encoder and a base station decoder and second recovered CSI associated with using a base station encoder and the base station decoder; perform a first KPI calculation using the ground-truth CSI and the first recovered CSI to generate a33P70697WO1 4898-9140-0001\1first KPI for the first recovered CSI; perform a second KPI calculation using the groundtruth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI; and identify a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI. In another example, the AI / ML model for CSI compression and decompression monitoring module 1416 may configure the wireless device 1402 to identify, based on a measurement of a reference signal that is received from a base station, ground-truth CSI for the reference signal; applying the ground-truth CSI at an active UE encoder to generate first compressed CSI feedback; apply a reference CSI at a reference UE encoder to generate second compressed CSI feedback; send, to the base station, the first compressed CSI feedback and the second compressed CSI feedback; receive, from the base station, active recovered CSI associated with using the UE encoder and a base station decoder and recovered CSI associated with using the reference UE encoder and the base station decoder; perform a first KPI calculation using the ground-truth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI; perform a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI; and identify a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder using the first KPI and the second KPI.

[0164] The network device 1418 may include one or more processor(s) 1420. The processor(s) 1420 may execute instructions such that various operations of the network device 1418 are performed, as described herein. The processor(s) 1420 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.

[0165] The network device 1418 may include a memory 1422. The memory 1422 may be anon-transitory computer-readable storage medium that stores instructions 1424 (which may include, for example, the instructions being executed by the processor(s) 1420). The instructions 1424 may also be referred to as program code or a computer program. The memory’ 1422 may also store data used by. and results computed by. the processor(s) 1420.

[0166] The network device 1418 may include one or more transceiver(s) 1426 that may include RF transmitter circuitry and / or receiver circuitry that use the antenna(s) 1428 of34P70697WO1 4898-9140-0001\1the network device 1418 to facilitate signaling (e.g., the signaling 1434) to and / or from the network device 1418 with other devices (e.g.. the wireless device 1402) according to corresponding RATs.

[0167] The network device 1418 may include one or more antenna(s) 1428 (e.g., one, two, four, or more). In embodiments having multiple antenna(s) 1428, the network device 1418 may perform MIMO, digital beamforming, analog beamforming, beam steering, etc., as has been described.

[0168] The network device 1418 may include one or more interface(s) 1430. The interface(s) 1430 may be used to provide input to or output from the network device 1418. For example, a network device 1418 that is a base station may include interface(s) 1430 made up of transmitters, receivers, and other circuitry' (e.g., other than the transceiver(s) 1426 / antenna(s) 1428 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.

[0169] The network device 1418 may include an AI / ML model for CSI compression and decompression monitoring module 1432. The AI / ML model for CSI compression and decompression monitoring module 1432 may be implemented via hardware, software, or combinations thereof. For example, the AI / ML model for CSI compression and decompression monitoring module 1432 may be implemented as a processor, circuit, and / or instructions 1424 stored in the memory' 1422 and executed by the processor(s) 1420. In some examples, the AI / ML model for CSI compression and decompression monitoring module 1432 may be integrated within the processor(s) 1420 and / or the transceiver(s) 1426. For example, the AI / ML model for CSI compression and decompression monitoring module 1432 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) 1420 or the transceiver(s) 1426.

[0170] The AI / ML model for CSI compression and decompression monitoring module 1432 may be used for various aspects of the present disclosure, for example, aspects of FIG. 9 and / or FIG. 11. For example, the AI / ML model for CSI compression and decompression monitoring module 1432 may configure the network device 1418 to send,35P70697WO1 4898-9140-0001\1to a UE, a reference signal; receive, from the UE, ground-truth channel state information (CSI) for the reference signal and first compressed CSI feedback as generated at a UE encoder using the ground-truth CSI; apply the first compressed CSI feedback at a base station decoder to generate first recovered CSI; process the ground-truth CSI using a base station encoder and the base station decoder to generate second recovered CSI; perform a first KPI calculation using the ground-truth CSI and the first recovered CSI to generate a first KPI for the first recovered CSI; perform a second KPI calculation using the ground-truth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI; and identify a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI. In another example, the AI / ML model for CSI compression and decompression monitoring module 1432 may configure the network device 1418 to send, to a UE, a reference signal; receive, from the UE, ground-truth CSI for the reference signal, first compressed CSI feedback as generated at an active UE encoder using the ground-truth CSI, and second compressed CSI feedback as generated at a reference UE encoder using a reference CSI; apply the first compressed CSI feedback at a base station decoder to generate active recovered CSI; apply the second compressed CSI feedback at the base station decoder to generate reference recovered CSI; perform a first KPI calculation using the ground-truth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI; perform a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI; and identifying a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder using the first KPI and the second KPI.

[0171] Embodiments contemplated herein include an apparatus comprising means to perform one or more elements of any one or more of the method 1000 and / or the method 1200. This apparatus may be. for example, an apparatus of a UE (such as a wireless device 1402 that is a UE, as described herein).

[0172] 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 any one or more of the method 1000 and / or the method 1200. This non-transitory computer-readable media may be, for example, a memory of a UE (such as a memory 1406 of a wireless device 1402 that is a UE, as described herein).36P70697WO1 4898-9140-0001\1

[0173] Embodiments contemplated herein include an apparatus comprising logic, modules, or circuitry to perform one or more elements of any one or more of the method 1000 and / or the method 1200. This apparatus may be, for example, an apparatus of a UE (such as a wireless device 1402 that is a UE, as described herein).

[0174] 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 any one or more of the method 1000 and / or the method 1200. This apparatus may be, for example, an apparatus of a UE (such as a wireless device 1402 that is a UE. as described herein).

[0175] Embodiments contemplated herein include a signal as described in or related to one or more elements of any one or more of the method 1000 and / or the method 1200.

[0176] 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 any one or more of the method 1000 and / or the method 1200. The processor may be a processor of a UE (such as a processor(s) 1404 of a wireless device 1402 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' 1406 of a wireless device 1402 that is a UE, as described herein).

[0177] Embodiments contemplated herein include an apparatus comprising means to perform one or more elements of any one or more of the method 900 and / or the method 1100. This apparatus may be, for example, an apparatus of a base station (such as a network device 1418 that is a base station, as described herein).

[0178] 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 any one or more of the method 900 and / or the method 1100. This non-transitory computer-readable media may be, for example, a memory of a base station (such as a memory 1422 of a network device 1418 that is a base station, as described herein).

[0179] Embodiments contemplated herein include an apparatus comprising logic, modules, or circuitry to perform one or more elements of any one or more of the method 37P70697WO1 4898-9140-0001\1900 and / or the method 1100. This apparatus may be, for example, an apparatus of a base station (such as a network device 1418 that is a base station, as described herein).

[0180] 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 any one or more of the method 900 and / or the method 1100. This apparatus may be, for example, an apparatus of a base station (such as a network device 1418 that is a base station, as described herein).

[0181] Embodiments contemplated herein include a signal as described in or related to one or more elements of any one or more of the method 900 and / or the method 1100.

[0182] 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 any one or more of the method 900 and / or the method 1100. The processor may be a processor of a base station (such as a processor(s) 1420 of a network device 1418 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 1422 of a network device 1418 that is a base station, as described herein).

[0183] 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.

[0184] 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.38P70697WO1 4898-9140-0001\1

[0185] 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.

[0186] 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.

[0187] 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.

[0188] 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 the description is not to be limited to the details given herein, but may be modified within the scope and equivalents of the appended claims.39P70697WO1 4898-9140-0001\1

Claims

CLAIMS1. A method of a base station, comprising:sending, to a user equipment (UE), a reference signal;receiving, from the UE, ground-truth channel state information (CSI) for the reference signal and first compressed CSI feedback as generated at a UE encoder using the ground-truth CSI;applying the first compressed CSI feedback at a base station decoder to generate first recovered CSI;processing the ground-truth CSI using a base station encoder and the base station decoder to generate second recovered CSI;performing a first key performance indicator (KPI) calculation using the groundtruth CSI and the first recovered CSI to generate a first KPI for the first recovered CSI;performing a second KPI calculation using the ground-truth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI; and identifying a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI.

2. The method of claim 1, wherein identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold.

3. The method of claim 2, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold.

4. The method of claim 2, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold.

5. The method of claim 2, wherein:the first KPI is a first squared generalized cosine similarity (SGCS) between the ground-truth CSI and the first recovered CSI;40P70697WO1 4898-9140-0001\1the second KPI is a second SGCS between the ground-truth CSI and the second recovered CSI; andthe performance threshold comprises an SGCS threshold.

6. The method of claim 2, wherein:the first KPI is a first error between the ground-truth CSI and the first recovered CSI;the second KPI is a second error between the ground-truth CSI and the second recovered CSI; andthe performance threshold comprises an error threshold.

7. The method of claim 1, further comprising determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

8. A method of a user equipment (UE), comprising:identifying, based on a measurement of a reference signal that is received from a base station, ground-truth channel state information (CSI) for the reference signal;applying the ground-truth CSI at a UE encoder to generate first compressed CSI feedback;sending the ground-truth CSI and the first compressed CSI feedback to the base station;receiving, from the base station, first recovered CSI associated with using the UE encoder and a base station decoder and second recovered CSI associated with using a base station encoder and the base station decoder;performing a first key performance indicator (KPI) calculation using the groundtruth CSI and the first recovered CSI to generate a first KPI for the first recovered CSI;performing a second KPI calculation using the ground-truth CSI and the second recovered CSI to calculate a second KPI for the second recovered CSI; and identifying a cause for a performance degradation corresponding to using the UE encoder and the base station decoder using the first KPI and the second KPI.

9. The method of claim 8, wherein identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold.41P70697WO1 4898-9140-0001\110. The method of claim 9, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold.

11. The method of claim 9, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold.

12. The method of claim 9, wherein:the first KPI is a first squared generalized cosine similarity (SGCS) between the ground-truth CSI and the first recovered CSI;the second KPI is a second SGCS between the ground-truth CSI and the second recovered CSI; andthe performance threshold comprises an SGCS threshold.

13. The method of claim 9. wherein:the first KPI is a first error between the ground-truth CSI and the first recovered CSI;the second KPI is a second error between the ground-truth CSI and the second recovered CSI; andthe performance threshold comprises an error threshold.

14. The method of claim 8, further comprising determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

15. A method of a base station, comprising:sending, to a user equipment (UE). a reference signal;receiving, from the UE. ground-truth channel state information (CSI) for the reference signal, first compressed CSI feedback as generated at an active UE encoder using the ground-truth CSI, and second compressed CSI feedback as generated at a reference UE encoder using a reference CSI;42P70697WO1 4898-9140-0001\1applying the first compressed CSI feedback at a base station decoder to generate active recovered CSI;applying the second compressed CSI feedback at the base station decoder to generate reference recovered CSI;performing a first key performance indicator (KPI) calculation using the groundtruth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI;performing a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI; and identifying a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder using the first KPI and the second KPI.

16. The method of claim 15, wherein identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold.

17. The method of claim 16, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold.

18. The method of claim 16, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold.

19. The method of claim 16, wherein:the first KPI is a first squared generalized cosine similarity (SGCS) between the ground-truth CSI and the active recovered CSI;the second KPI is a second SGCS between the reference CSI and the reference recovered CSI; andthe performance threshold comprises an SGCS threshold.

20. The method of claim 16, wherein:43P70697WO1 4898-9140-0001\1the first KPI is a first error between the ground-truth CSI and the active recovered CSI;the second KPI is a second error between the reference CSI and the reference recovered CSI; andthe performance threshold comprises an error threshold.

21. The method of claim 15, further comprising determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

22. The method of claim 15, further comprising receiving, from the UE, the reference CSI.

23. The method of claim 15, wherein the reference CSI is equal to the ground-truth CSI.

24. A method of a user equipment (UE), comprising:identifying, based on a measurement of a reference signal that is received from a base station, ground-truth channel state information (CSI) for the reference signal;applying the ground-truth CSI at an active UE encoder to generate first compressed CSI feedback;applying a reference CSI at a reference UE encoder to generate second compressed CSI feedback;sending, to the base station, the first compressed CSI feedback and the second compressed CSI feedback;receiving, from the base station, active recovered CSI associated with using the UE encoder and a base station decoder and reference recovered CSI associated with using the reference UE encoder and the base station decoder;performing a first key performance indicator (KPI) calculation using the groundtruth CSI and the active recovered CSI to generate a first KPI for the active recovered CSI;performing a second KPI calculation using the reference CSI and the reference recovered CSI to calculate a second KPI for the reference recovered CSI; and identifying a cause for a performance degradation corresponding to using the active UE encoder and the base station decoder sing the first KPI and the second KPI.44P70697WO1 4898-9140-0001\125. The method of claim 24, wherein identifying the cause for the performance degradation using the first KPI and the second KPI comprises comparing each of the first KPI and the second KPI to a performance threshold.

26. The method of claim 25, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying the UE encoder as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI meets the performance threshold.

27. The method of claim 25, wherein identifying the cause of the performance degradation using the first KPI and the second KPI further comprises identifying one of the base station decoder and a data drift as the cause of the performance degradation when the first KPI does not meet the performance threshold and the second KPI does not meet the performance threshold.

28. The method of claim 25, wherein:the first KPI is a first squared generalized cosine similarity (SGCS) between the ground-truth CSI and the active recovered CSI;the second KPI is a second SGCS between the reference CSI and the reference recovered CSI; andthe performance threshold comprises an SGCS threshold.

29. The method of claim 25, wherein:the first KPI is a first error between the ground-truth CSI V and the active recovered CSI;the second KPI is a second error between the reference CSI Vreand the reference recovered CSI; andthe performance threshold comprises an error threshold.

30. The method of claim 24, further comprising determining that the first KPI does not meet a performance threshold; and wherein the second KPI calculation is performed in response to the determination that the first KPI does not meet the performance threshold.

31. The method of claim 24, wherein the reference CSI is equal to the ground-truth CSI.

32. An apparatus comprising means to perform the method of any of claim 1 to claim 31.45P70697WO1 4898-9140-0001\133. 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 31.

34. An apparatus comprising logic, modules, or circuitry to perform the method of any of claim 1 to claim 31.

35. 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 8 to claim 14 and claim 24 to claim 31.

36. 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 1 to claim 7 and claim 15 to claim 23.46P70697WO1 4898-9140-0001\1

Images