Performance monitoring of artificial intelligence or machine learning enabled channel state information compression

Abstract

A monitoring engine enabled with artificial intelligence or machine learning is used for monitoring the performance of an autoencoder for compressing and reconstructing channel state information. The inputs of the monitoring engine comprise at least (a) the quantized channel state information, (b) a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder, wherein the actual encoder and the hypothetical encoder are located at a user equipment, and (c) a squared generalized cosine similarity reported at the output of a monitoring module located at the user equipment, wherein the monitoring module estimates the squared generalized cosine similarity from the quantized channel state information or by using a proxy decoder. The outputs of the monitoring engine comprise at least the identified root cause of a degradation in the performance of the autoencoder.

Description

METHOD OF PERFORMANCE MONITORING FOR AI / ML-ENABLED CSI COMPRESSIONTECHNICAL FIELD

[0001] The examples and non-limiting example embodiments relate generally to communications and, more particularly, to a method of performance monitoring for AI / ML- enabled compression.BACKGROUND

[0002] It is known for a communication device to gain access to a communication network via an access network node.BRIEF DESCRIPTION OF THE DRAWINGS

[0003] The foregoing aspects and other features are explained in the following description, taken in connection with the accompanying drawings.

[0004] FIG. 1 is a block diagram of one possible and non-limiting system in which the example embodiments may be practiced.

[0005] FIG. 2 shows UE-side CSI encoder model training for NW-first separate training set up assisted by the relationship between the average MSE at the encoder output and E2E SGCS.

[0006] FIG. 3 shows AI / ML performance monitoring engine assisted monitoring operation for the CSI compression-reconstruction two-sided model case.

[0007] FIG. 4 shows training data generation for the AI-ML-enabled monitoring engine model.

[0008] FIG. 5 shows input / output variables of the AI / ML-enabled monitoring engine model.

[0009] FIG. 6 shows a signaling diagram view of AI / ML performance monitoring engine assisted monitoring operation.

[0010] FIG. 7 is an example apparatus configured to implement the examples described herein.

[0011] FIG. 8 shows a representation of an example of non-volatile memory media used to store instructions that implement the examples described herein.

[0012] FIG. 9 is an example method, based on the examples described herein.

[0013] FIG. 10 is an example method, based on the examples described herein.

[0014] FIG. 11 is an example method, based on the examples described herein.

[0015] FIG. 12 is an example method, based on the examples described herein.DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0016] Turning to FIG. 1, this figure shows a block diagram of one possible and nonlimiting example in which the examples may be practiced. A user equipment (UE) 110, radio access network (RAN) node 170, and network element(s) 190 are illustrated. In the example of FIG. 1, the user equipment (UE) 110 is in wireless communication with a wireless network 100. A UE is a wireless device that can access the wireless network 100. The UE 110 includes one or more processors 120, one or more memories 125, and one or more transceivers 130 interconnected through one or more buses 127. Each of the one or more transceivers 130 includes a receiver, Rx, 132 and a transmitter, Tx, 133. The one or more buses 127 may be address, data, or control buses, and may include any interconnection mechanism, such as a series of lines on a motherboard or integrated circuit, fiber optics or other optical communication equipment, and the like. The one or more transceivers 130 are connected to one or more antennas 128. The one or more memories 125 include computer program code 123. The UE 110 includes a module 140, comprising one of or both parts 140-1 and / or 140- 2, which may be implemented in a number of ways. The module 140 may be implemented in hardware as module 140-1, such as being implemented as part of the one or more processors 120. The module 140-1 may be implemented also as an integrated circuit or through other hardware such as a programmable gate array. In another example, the module 140 may be implemented as module 140-2, which is implemented as computer program code 123 and is executed by the one or more processors 120. For instance, the one or more memories 125 and the computer program code 123 may be configured to, with the one or more processors120, cause the user equipment 110 to perform one or more of the operations as described herein. The UE 110 communicates with RAN node 170 via a wireless link 111.

[0017] The RAN node 170 in this example is a base station that provides access for wireless devices such as the UE 110 to the wireless network 100. The RAN node 170 may be, for example, a base station for 5G, also called New Radio (NR). In 5G, the RAN node 170 may be a NG-RAN node, which is defined as either a gNB or an ng-eNB. A gNB is a node providing NR user plane and control plane protocol terminations towards the UE, and connected via the NG interface (such as connection 131) to a 5GC (such as, for example, the network element(s) 190). The ng-eNB is a node providing E-UTRA user plane and control plane protocol terminations towards the UE, and connected via the NG interface (such as connection 131) to the 5GC. The NG-RAN node may include multiple gNBs, which may also include a central unit (CU) (gNB-CU) 196 and distributed unit(s) (DUs) (gNB-DUs), of which DU 195 is shown. Note that the DU 195 may include or be coupled to and control a radio unit (RU). The gNB-CU 196 is a logical node hosting radio resource control (RRC), SDAP and PDCP protocols of the gNB or RRC and PDCP protocols of the en-gNB that control the operation of one or more gNB-DUs. The gNB-CU 196 terminates the Fl interface connected with the gNB -DU 195. The Fl interface is illustrated as reference 198, although reference 198 also illustrates a link between remote elements of the RAN node 170 and centralized elements of the RAN node 170, such as between the gNB-CU 196 and the gNB- DU 195. The gNB -DU 195 is a logical node hosting RLC, MAC and PHY layers of the gNB or en-gNB, and its operation is partly controlled by gNB-CU 196. One gNB-CU 196 supports one or multiple cells. One cell may be supported with one gNB -DU 195, or one cell may be supported / shared with multiple DUs under RAN sharing. The gNB-DU 195 terminates the Fl interface 198 connected with the gNB-CU 196. Note that the DU 195 is considered to include the transceiver 160, e.g., as part of a RU, but some examples of this may have the transceiver 160 as part of a separate RU, e.g., under control of and connected to the DU 195. The RAN node 170 may also be an eNB (evolved NodeB) base station, for LTE (long term evolution), or any other suitable base station or node.

[0018] The RAN node 170 includes one or more processors 152, one or more memories 155, one or more network interfaces (N / W I / F(s)) 161, and one or more transceivers 160 interconnected through one or more buses 157. Each of the one or more transceivers 160 includes a receiver, Rx, 162 and a transmitter, Tx, 163. The one or more transceivers 160 areconnected to one or more antennas 158. The one or more memories 155 include computer program code 153. The CU 196 may include the processor(s) 152, one or more memories 155, and network interfaces 161. Note that the DU 195 may also contain its own memory / memories and processor(s), and / or other hardware, but these are not shown.

[0019] The RAN node 170 includes a module 150, comprising one of or both parts 150-1 and / or 150-2, which may be implemented in a number of ways. The module 150 may be implemented in hardware as module 150-1, such as being implemented as part of the one or more processors 152. The module 150-1 may be implemented also as an integrated circuit or through other hardware such as a programmable gate array. In another example, the module 150 may be implemented as module 150-2, which is implemented as computer program code 153 and is executed by the one or more processors 152. For instance, the one or more memories 155 and the computer program code 153 are configured to, with the one or more processors 152, cause the RAN node 170 to perform one or more of the operations as described herein. Note that the functionality of the module 150 may be distributed, such as being distributed between the DU 195 and the CU 196, or be implemented solely in the DU 195.

[0020] The one or more network interfaces 161 communicate over a network such as via the links 176 and 131. Two or more gNBs 170 may communicate using, e.g., link 176. The link 176 may be wired or wireless or both and may implement, for example, an Xn interface for 5G, an X2 interface for LTE, or other suitable interface for other standards.

[0021] The one or more buses 157 may be address, data, or control buses, and may include any interconnection mechanism, such as a series of lines on a motherboard or integrated circuit, fiber optics or other optical communication equipment, wireless channels, and the like. For example, the one or more transceivers 160 may be implemented as a remote radio head (RRH) 195 for LTE or a distributed unit (DU) 195 for gNB implementation for 5G, with the other elements of the RAN node 170 possibly being physically in a different location from the RRH / DU 195, and the one or more buses 157 could be implemented in part as, for example, fiber optic cable or other suitable network connection to connect the other elements (e.g., a central unit (CU), gNB-CU 196) of the RAN node 170 to the RRH / DU 195. Reference 198 also indicates those suitable network link(s).

[0022] A RAN node / gNB can comprise one or more TRPs to which the methods describedherein may be applied. FIG. 1 shows that the RAN node 170 comprises TRP 51 and TRP 52, in addition to the TRP represented by transceiver 160. Similar to transceiver 160, TRP 51 and TRP 52 may each include a transmitter and a receiver. The RAN node 170 may host or comprise other TRPs not shown in FIG. 1.

[0023] A relay node in NR is called an integrated access and backhaul node. A mobile termination part of the IAB node facilitates the backhaul (parent link) connection. In other words, the mobile termination part comprises the functionality which carries UE functionalities. The distributed unit part of the IAB node facilitates the so called access link (child link) connections (i.e. for access link UEs, and backhaul for other IAB nodes, in the case of multi-hop IAB). In other words, the distributed unit part is responsible for certain base station functionalities. The IAB scenario may follow the so called split architecture, where the central unit hosts the higher layer protocols to the UE and terminates the control plane and user plane interfaces to the 5G core network.

[0024] It is noted that the description herein indicates that “cells” perform functions, but it should be clear that equipment which forms the cell may perform the functions. The cell makes up part of a base station. That is, there can be multiple cells per base station. For example, there could be three cells for a single carrier frequency and associated bandwidth, each cell covering one-third of a 360 degree area so that the single base station’s coverage area covers an approximate oval or circle. Furthermore, each cell can correspond to a single carrier and a base station may use multiple carriers. So if there are three 120 degree cells per carrier and two carriers, then the base station has a total of 6 cells.

[0025] The wireless network 100 may include a network element or elements 190 that may include core network functionality, and which provides connectivity via a link or links 181 with a further network, such as a telephone network and / or a data communications network (e.g., the Internet). Such core network functionality for 5G may include location management functions (LMF(s)) and / or access and mobility management function(s) (AMF(S)) and / or user plane functions (UPF(s)) and / or session management function(s) (SMF(s)). Such core network functionality for LTE may include MME (mobility management entity) / SGW (serving gateway) functionality. Such core network functionality may include SON (self- organizing / optimizing network) functionality. These are merely example functions that may be supported by the network element(s) 190, and note that both 5G and LTE functions might be supported. The RAN node 170 is coupled via a link 131 to the network element 190. Thelink 131 may be implemented as, e.g., an NG interface for 5G, or an SI interface for LTE, or other suitable interface for other standards. The network element 190 includes one or more processors 175, one or more memories 171, and one or more network interfaces (N / W I / F(s)) 180, interconnected through one or more buses 185. The one or more memories 171 include computer program code 173. Computer program code 173 may include SON and / or MRO functionality 172.

[0026] The wireless network 100 may implement network virtualization, which is the process of combining hardware and software network resources and network functionality into a single, software-based administrative entity, or a virtual network. Network virtualization involves platform virtualization, often combined with resource virtualization. Network virtualization is categorized as either external, combining many networks, or parts of networks, into a virtual unit, or internal, providing network-like functionality to software containers on a single system. Note that the virtualized entities that result from the network virtualization are still implemented, at some level, using hardware such as processors 152 or 175 and memories 155 and 171, and also such virtualized entities create technical effects.

[0027] The computer readable memories 125, 155, and 171 may be of any type suitable to the local technical environment and may be implemented using any suitable data storage technology, such as semiconductor based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, non-transitory memory, transitory memory, fixed memory and removable memory. The computer readable memories 125, 155, and 171 may be means for performing storage functions. The processors 120, 152, and 175 may be of any type suitable to the local technical environment, and may include one or more of general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on a multi-core processor architecture, as nonlimiting examples. The processors 120, 152, and 175 may be means for performing functions, such as controlling the UE 110, RAN node 170, network element(s) 190, and other functions as described herein.

[0028] In general, the various example embodiments of the user equipment 110 can include, but are not limited to, cellular telephones such as smart phones, tablets, personal digital assistants (PDAs) having wireless communication capabilities, portable computers having wireless communication capabilities, image capture devices such as digital cameras having wireless communication capabilities, gaming devices having wireless communicationcapabilities, music storage and playback devices having wireless communication capabilities, internet appliances including those permitting wireless internet access and browsing, tablets with wireless communication capabilities, head mounted displays such as those that implement virtual / augmented / mixed reality, as well as portable units or terminals that incorporate combinations of such functions. The UE 110 can also be a vehicle such as a car, or a UE mounted in a vehicle, a UAV such as e.g. a drone, or a UE mounted in a UAV. The user equipment 110 may be a terminal device, such as mobile phone, mobile device, sensor device etc., where the terminal device is a device used by the user or not used by the user.

[0029] UE 110, RAN node 170, and / or network element(s) 190, (and associated memories, computer program code and modules) may be configured to implement (e.g. in part) the methods described herein. Thus, computer program code 123, module 140-1, module 140-2, and other elements / features shown in FIG. 1 of UE 110 may implement user equipment related aspects of the examples described herein. Similarly, computer program code 153, module 150-1, module 150-2, and other elements / features shown in FIG. 1 of RAN node 170 may implement gNB / TRP related aspects of the examples described herein. Computer program code 173 and other elements / features shown in FIG. 1 of network element(s) 190 may be configured to implement network element related aspects of the examples described herein.

[0030] Having thus introduced a suitable but non-limiting technical context for the practice of the example embodiments, the example embodiments are now described with greater specificity.

[0031] For AI / ML based CSI compression using two-sided model, one of the major topics under discussion is a performance monitoring aspect.

[0032] As regards performance monitoring scheme there are two main flavors, i.e., NW- side monitoring and UE-side monitoring. NW-side monitoring scheme is less favorable since it either requires high uplink overhead at the air interface due to ground truth target CSI reporting from UE to NW (based on the target CSI reported by the UE via legacy eT2 codebook or eT2-like high-resolution codebook) or is restricted to TDD system in case of SRS-based monitoring. The fact that NW-side should simultaneously monitor multiple UEs in a cell and N>1 CSI feedback instances are needed to render statistically reliable monitoring operation would cast more challenges to the NW-side monitoring option.

[0033] In this sense, UE-side monitoring (especially based on the output of the CSI reconstruction model at the UE and / or via direct estimation of intermediate KPI (e.g. SGCS) without reconstructing a target CSI) appear to be more attractive alternative, so long as NW- side is entitled to decide, i.e., model switching, fall back, model re-training, etc., based on monitoring reports from UE-side(s).

[0034] Described herein is a method and procedure to identify the root cause of performance degradation, i.e., NW-side issue, UE-side issue, or data drift, under the framework of the UE- side monitoring assisted to NW-side decision-making. Assumed is NW-first type 3 separate or type 2 joint sequential training when it comes to AI / ML model training collaboration types. As regards inter- vendor training collaboration support aspects, standardized reference model structure and parameter exchange between the NW-side and UE-side, standardized data or data format and dataset exchange between the NW-side and UE-side, and standardized model format and reference model exchange between the NW-side and UE-side have been considered.

[0035] Below is a brief explanation on the techniques of which the knowledge can be useful for understanding of the herein described scheme.

[0036] UE-side monitoring techniques

[0037] UE-side monitoring scheme requires UE-side to derive E2E performance metric, i.e., BLER, SGCS, etc. This can be achieved for UE-side either by incorporating CSI reconstruction model, e.g., proxy decoder model, actual decoder model, etc., or by developing direct intermediate (SGCS) estimator. The former deems more challenging for UE, due to its presumably high computational burden at UE. Hence the latter alternative is more appealing from UE perspective. This topic has been investigated, and a numerical evaluation result has been provided on monitoring accuracy of the direct SGCS estimator. An independent study has validated feasibility of the similar approach (low complexity UE-side E2E performance monitoring scheme via monitoring bits) as well. A multi-layer perceptron (MLP)-based CSI drift detector may be implemented, which is designed to observe quantized latent vector (CSI feedback) for prediction of the reconstructed CSI performance in terms of SGCS to determine channel drift at NW-side. The detection capability performance of the multi-layer perceptron (MLP)-based CSI drift detector may be validated via numerical experiments. The multi-layer perceptron (MLP)-based CSI drift detector can be easily adopted at UE-side for CSI driftdetection. In short, direct estimation of the intermediate KPI looks more promising and viable from technical feasibility and from a monitoring capability perspective compared to US-side monitoring based on the output of the CSI reconstruction model at the UE.

[0038] Novel metric for UE-side model training for two-sided model

[0039] Average MSE, computed at the output of the CSI compression model (encoder) in comparison with the output of the hypothetical CSI encoder, can be used as a good guidance for estimate of the E2E performance metric, i.e., SGCS. This utilizes empirically acquired average MSE-SGCS correlation in preceding CSI reconstruction model training phase at the NW-side (underlining assumption: NW-first training scenario; see FIG. 2 for reference). This scheme may be used to facilitate UE-side CSI encoder model training for NW-first training type (Type 3 separate or Type 2 sequential) by using MSE as a loss function, but it may be possible to find another useful application of this technique for the monitoring use case with slight extension / modification.

[0040] In particular, FIG. 2 shows UE-side CSI encoder model training for a NW-first separate training set up, assisted by the relationship between average MSE at the encoder output and E2E SGCS. At (1) of FIG. 2, the NW (170, 190) trains the decoder 204 with the hypothesis encoder 206 using the E2E SGCS loss function. At (2) of FIG. 2, the NW (170, 190) sends the training dataset (X, Y) to the UE 110. At (3) of FIG. 2, the UE 110 trains the actual encoder 202 using the MSE loss function of the encoder’s output codeword.

[0041] As described previously, there have been efforts to enhance UE-side monitoring scheme by developing a direct intermediate KPI (SGCS) estimation method without having to adopt a proxy / actual CSI decoder at the UE-side. However, there has been no investigation to the best of the authors’ knowledge on how to identify root cause of performance degradation which can be an issue at the UE-side, NW-side or of data drift.

[0042] Note here that the above-mentioned goal shall be achieved with careful consideration of the required overhead in the air interface.

[0043] Main idea of the examples described herein is to introduce multiple checkpoints, i.e., output of CSI encoder, output of the model monitoring module (whether it can be proxy / actual CSI decoder, monitoring bit generator, or direct SGCS estimator), in an effort to monitor two- sided model performance via the dedicated AI / ML monitoring model for identification of rootcause of the performance degradation, i.e., UE-side, NW-side or data drift, and for prediction of the ultimate KPI, i.e., BLER (or consequently, throughput). Especially, monitoring at the output of CSI encoder at UE-side plays critical role, hence it is useful for the UE-side, in parallel with its own actual CSI encoder, to incorporate a hypothetical CSI encoder (which is used at NW-side to mimic an UE-side CSI encoder to develop the actual CSI decoder to take E2E performance target as its primary model optimization goal during NW-side model development & training phase in case of NW-first training) to compute MSE for monitoring purposes. Here, MSE is used as a metric to measure deviation of the actual CSI encoder output with respect to its counterpart of the hypothetical CSI encoder. MSE at CSI encoder output retains strong correlation with SGCS of the E2E encoder-decoder chain, so it can be useful for UE-side CSI encoder model monitoring with consideration of E2E performance impact. Acquisition of the hypothetical CSI encoder at UE-side can be done either by direct model / parameter transfer from NW-vendor(s) or by relevant data set sharing from NW-side to UE-side.

[0044] Another main novel aspect of the examples described herein is to resort to another (dedicated) AI / ML model (denoted as “monitoring engine” to differentiate it from model monitoring module like SGCS estimator) for identification of the root cause of performance degradation. To this end, developed and described herein is an AI / ML-enabled monitoring engine model with controlled deterioration factors being imposed at training data generation phase for supervised learning of root cause identification. The following (but not limited to) variables are used as its input, i.e., MSE at CSI encoder output (with incorporation of hypothetical encoder), SGCS at model monitoring module (i.e., SGCS estimator) output, quantized CSI feedback (latent vector) of the actual CSI encoder model, to name a few.

[0045] Main motivation is to make AI / ML monitoring engine model learn underlining patterns of multi-checkpoint multi-dimensional input variables while a certain performance deterioration factor, i.e., UE-side CSI encoder deviation / failure, NW-side CSI decoder deviation / failure, or CSI data drift, is imposed intentionally at training data generation phase. Moreover, during the training data generation phase, AI / ML monitoring engine can be plugged into the link level or system level simulator / emulator such that CRC check outcome (ACK7NACK) of the precoded PDSCH transmission based on AI / ML-enabled CSI compression-reconstruction and subsequent precoding matrix calculation should be collected and served as label, together with imposed performance deterioration factor. By using trainingdata with MSE, SGCS, CSI feedback as inputs to the monitoring engine model whereas root cause and ACK / NACK as labels, the monitoring engine model can be trained in a supervised learning fashion to be capable of identifying the root cause as well as of predicting a chance of CRC failure (NACK), i.e., the ultimate KPI, rather than intermediate KPI, i.e., SGCS. The rationale here is that deviation in encoder / decoder or data characteristics can be tolerated so long as the resulting precoded PDSCH can be decoded successfully.

[0046] Advantages and technical effects of the herein described scheme include the following (1-4):

[0047] 1. Model-free and data-driven monitoring engine development allows constant update / improvement so long as new data set becomes available and is anticipated to better capture patterns of the multi-checkpoint multi-dimensional input variables.

[0048] 2. Monitoring scheme can be trained to predict ACK / NACK (which has direct impact to ultimate KPI, i.e., throughput) as well as to identify the root cause of performance degradation.

[0049] 3. Low additional overhead in the air interface: MSE and SGCS only (both are scalar value), in case of monitoring engine model inference at NW- side

[0050] 4. Flexibility: monitoring engine model can be placed either at NW-side or at UE- side

[0051] The main objective of the examples described herein is to provide a technically viable performance monitoring scheme which is applicable to AI / ML-enabled CSI compression use case, that is a two-sided model. When it comes to training collaboration scheme, NW-first type 2 joint sequential or NW-first type 3 separate training is the target set up. As regards inter-vendor training collaboration options, the herein described scheme is applicable to one option (standardized reference model structure and parameter exchange between NW-side and UE-side), another option (standardized data and / or dataset format and dataset exchange between NW-side and UE-side), and another option (standardized model format and reference model exchange between NW-side and UE-side) conditioned to the NW-first approach. In this configuration, NW-side develops its own CSI reconstruction (decoder) model first, based on either standard specified reference decoder model structure or its own proprietary design. As a byproduct of this procedure, NW-side develops ahypothetical CSI compression (encoder) model as well. Then NW-side is supposed to share with UE-side either generated training data set or trained decoder and / or encoder model (structure / ) parameters, depending on the agreed inter-operability option. UE-side should train its own encoder model via shared data set or shared encoder or decoder model (structure / ) parameters. UE-side is assumed to have access to model monitoring module, which is an AI / ML model to take CSI feedback (CSI encoder output) as input and to generate either estimate of the intermediate KPI (like SGCS) or monitoring bits for performance monitoring purposes.

[0052] The following main novel and inventive features characterize the examples described herein. See FIG. 3 for corresponding illustrations. FIG. 3 shows AI / ML performance monitoring engine assisted monitoring operation for the CSI compressionreconstruction two-sided model case. The top portion of FIG. 3 shows the monitoring engine at the NW-side (including monitoring inference 302), and the bottom portion of FIG. 3 shows the monitoring engine at the UE-side (including monitoring inference 352).

[0053] 1. Described herein are methods of the performance monitoring framework for UE- sided monitoring by incorporation of AI / ML-enabled monitoring engine for identification of the root cause of performance degradation.

[0054] La. Prerequisite is for UE-side to have access to the hypothetical encoder model having been developed at NW-side.

[0055] l.a.i. In one embodiment, the model structure and trained model parameters of the hypothetical CSI encoder model which has been used to train the actual CSI decoder model at NW-side are to be transferred from NW-side to UE-side. l.a.i.1. Delivery of the hypothetical encoder model may take place via vendor-to-vendor agreement prior to the actual model deployment in cell, i.e., IODT phase for example, l.a.i.2. Delivery of the hypothetical encoder model may also take place over the air (as illustrated in FIG. 3 at 304 and 354).

[0056] La .ii. In another embodiment, NW-side may share with UE-side training data set(s) consisting of {(target CSI, CSI feedback)} for hypothetical encoder model training at UE- side. The model structure of the hypothetical encoder may be shared by NW-side in advance.

[0057] La .iii. As regards l.a.i. and l.a.ii., the following LCM-related aspects can beconsidered, l.a.iii.l. When model transfer or dataset transfer takes place, the NW can assign an ID (proxy-model-ID) for the transferred proxy model (hypothetical encoder model) / dataset. l.a.iii.2. The UE may report capability of supporting certain proxy models, by sending the supported proxy-model-IDs later, l.a.iii.3. Also, the monitoring engine (l.b.) may run per proxy-model-ID and NW / UE can trigger LCM actions on proxy-model based on the monitoring engine outcome.

[0058] l.b. The main monitoring procedure is executed by an AI / ML-enabled “monitoring engine” (302, 352) which can identify the root cause (UE-side, NW-side or data drift) of performance degradation. The UE-side monitoring engine is given with reference 352, and the NW-side monitoring engine is given with reference 302. This monitoring engine can be additionally designed for prediction of successful decoding of the precoded PDSCH. Referring to FIG. 5, input variables of the monitoring engine (302, 352) are MSE 502 at CSI encoder output (with incorporation of hypothetical encoder), SGCS 504 at model monitoring module output, quantized CSI feedback 506 (latent vector) of the actual CSI encoder model in use at UE, etc. Outputs of the monitoring engine (302, 352) are identified root cause 508 and predicted ACK 510 or predicted NACK 512 (or its probability).

[0059] l.b.i. In one embodiment, AI / ML monitoring engine 302 may be placed at NW-side (170, 190) as illustrated in the top subfigure of FIG. 3. l.b.i.l. In one embodiment, AI / ML- assisted model monitoring module (denoted as a hexagon in FIG. 3, such as monitoring module 306) at UE may produce as output 332 (the output 332 denoted as Out mm) monitoring bits instead of intermediate KPI like SGCS, and at 308, these monitoring bits together with MSE 330 are fed back to the NW-side (170, 190) via air interface. In this case, a certain preprocessing unit 310 is required at NW-side for interpretation / conversion of monitoring bits to E2E SGCS estimates 334. l.b.i.2. In another embodiment, AI / ML-assisted model monitoring module 306 at UE may directly calculate E2E SGCS estimates. In this case, at 312, intermediate KPI 332 (like SGCS) with MSE 330 are directly fed back to NW-side via an air interface, l.b.i.3. As regards to l.b.i.l. and l.b.i.2., the NW may trigger monitoring reporting of different quantities jointly (as described above) or separately, such that the UE may report, for example, monitoring bits alone, or MSE metric alone or both together. The input variables of the monitoring engine may need to be supplied at different rate.

[0060] l.b.ii. In another embodiment, AI / ML monitoring engine 352 may be placed at UE- side as illustrated in bottom subfigure of FIG. 3. In this case, AI / ML-assisted modelmonitoring module at UE (such as monitoring module 314 and monitoring module 316) may not produce monitoring bits but calculate E2E SGCS (318, 320) directly, as SGCS (318, 320) should be used as an input variable to the monitoring engine 352, together with other input variables like MSE (322, 324), CSI feedback 326 (latent vector), etc. l.b.ii.l. In one embodiment, with reference to FIG. 5, AI / ML-enabled monitoring engine (302, 352) can take channel-related information 514, e.g., SINR 516, Doppler shift / spread 518, delay spread 520, etc., as additional input variables, l.b.ii.2. In another embodiment, UE may monitor outcomes of the AI / ML-enabled monitoring inference engine 352 for multiple (N>1) feedback occasions to acquire statistically reliable monitoring result. A certain level of procedures regarding determination of “Event” 328 can be defined in a standard, and based on it, UE may be able to detect Event 328 and report it on detection of Event 328. See the bottom subfigure of FIG. 3. l.b.ii.3. As regards “Event” definition and its associated procedures in l.b.ii.2.: The output of the monitoring engine 352 may include one or more sequences of values (monitoring results) whose value range and / or statistical properties can be used to detect one or more causes of failures. Based on the capabilities of the monitoring engine 352 indicated by a UE, the gNB may configure certain events or conditions that trigger a UE report of the monitoring results. For example, the monitoring engine 352 may output a failure probability for each predefined and supported cause of failure. The gNB may configure a reporting condition defined as the probability of that failure exceeding a certain threshold. If the monitoring result exceeds the configured threshold, a UE triggers a report including the failure probability value.

[0061] The herein described scheme may require either only MSE and SGCS (or monitoring bits) or event ID for additional feedback over the air.

[0062] 2. Methods of training data generation for the AI / ML-enabled monitoring engine model.

[0063] The main objective of this procedure is to generate a training data set (440, 450, 460) with intentionally imposed performance deterioration factor, i.e., UE-side CSI encoder model deviation, NW-side CSI decoder model deviation or CSI data drift from the channels being used for the encoder / decoder model development. This configured deterioration factor can be used as a label of the generated / collected training data for the corresponding set up.

[0064] 2. a. In one embodiment, pre-trained CSI encoder model 410, pre-trained CSIdecoder model 414, pre-trained monitoring module model 416 and pre-trained hypothetical CSI encoder model 412 are integrated into LLS or SLS (418) to generate ACK7NACK as a label 420, as illustrated in FIG. 4. Regarding training data elements, input variables can be (MSE 470 between CSI encoder output from the actual CSI encoder at UE and that from the hypothetical CSI encoder, CSI feedback 474 (i.e., CSI encoder output from the actual CSI encoder at UE), SGCS 472 reported at output of the monitoring module (proxy decoder or SGCS estimator)), whereas corresponding labels consist of (root cause 476 (configured imposed performance deterioration factor), decoding result 420 (CRC success / fail; ACK / NACK) of the associated precoded PDSCH). As depicted in FIG. 4, “true SGCS” 478 (E2E SGCS measured at the output of the actual CSI decoder with target CSI being considered; denoted as SGCSt) can be collected as well during training data generation phase, but it cannot be used as one of input variables for monitoring engine model training / inference as neither UE nor NW has access to it in practice. However, this true SGCS 478 may be used as one of outputs, i.e., learning objective.

[0065] 2.a.i. For the “UE-side” issue case 402 (top subfigure of FIG. 4), certain increasing levels of noise are injected 415 to weights of the trained CSI encoder model 410 to the extent that brings about CRC failure, while weights of other trained models, i.e., hypothetical CSI encoder 412, monitoring module 416 and CSI decoder 414, are frozen. Alternatively, the UE- side encoder model 410 can be replaced with one or more known / identified problematic encoder models.

[0066] 2.a.ii. For “NW-side” issue case 404 (middle subfigure of FIG. 4), certain increasing levels of noise are injected 425 to weights of the CSI decoder model 414 to the extent that brings about CRC failure, while weights of other trained models, i.e., hypothetical CSI encoder 412, monitoring module 416 and CSI encoder 410, are frozen. Alternatively, the NW- side decoder model 414 can be replaced with one or more known / identified problematic decoder models.

[0067] 2. a .iii. For “data drift” issue case 406 (bottom subfigure of FIG. 4), certain increasing levels of channel deviation (for example) are induced 435 by increasing mobile speed, delay shift / spread, interferences with respect to the channels having been used for the model training to the extent that brings about CRC failure, while weights of other trained models, i.e., CSI encoder 410, hypothetical CSI encoder 412, monitoring module 416 and CSI decoder 414, are frozen. In FIG. 4, CSI encoder 410 may correspond to actual CSI encoder 305 ofFIG. 3, NW-side decoder model 414 may correspond to decoder model 301 of FIG. 3, hypothetical CSI encoder 412 may correspond to hypothetical CSI encoder 303 of FIG. 3, and monitoring module 416 may correspond to monitoring module 306 of FIG. 3, monitoring module 314 of FIG. 3, or monitoring module 316 of FIG. 3.

[0068] 2.a.iv. As regards the practical aspects of the data collection (irrespective of 2.a.i.- 2.a.iii.), throughput (or average BLER, equivalently) can be used as indicator over the relatively long term, instead of tracking ACK / NACK for every individual precoded PDSCH.

[0069] 2.a.v. Alternatively for 2.a.i. and 2.a.ii., CSI encoder / decoder models which have been trained on the training data set of the different channel configuration / cell characteristics can be used to induce performance degradation owing to model / trained data characteristics mismatch.

[0070] 2.b. In another embodiment, additional input variables may be collected as well in case they are anticipated to be available at monitoring engine bearing side and it would lead to performance enhancement.

[0071] 2.b.i. In one embodiment in which the monitoring engine bearing side is UE-side, channel related information, e.g., SINR, Doppler shift / spread, delay shift / spread, etc., can be included as input variables. 2.b.ii. In another embodiment in which the monitoring engine bearing side is NW-side, a link adaptation scheme related information, i.e., CQI, TBS, etc., can be included as input variables.

[0072] 3. Methods of supervised-learning based training for the A I / ML-cnablcd monitoring engine model.

[0073] This procedure allows the AI / ML monitoring engine model to learn underlining distinctive pattern behind the multi-source multi-dimensional input variables in an effort to identify a root cause of the possible performance degradation, if any, as well as to predict possibility of demodulation success of the precoded PDSCH. As the labeled training data set is to be provided by embodiment 2., supervised learning practice can be used for the model training. An Exemplary model input / output interface is illustrated in FIG. 5. In particular, FIG. 5 shows input / output variables of the AI / ML-enabled monitoring engine model.

[0074] 3. a. In FIG. 5, two separate models are configured, i.e., one model 501 for root causeidentification, the other model 503 for ACK7NACK prediction, which share the identical input variables (MSE 502, SGCS 504, z?506, and channel related information 514).

[0075] 3.a.i. Input variables consist of (but not limited to) the followings: (MSE 502 (such as MSE 330, MSE 322, or MSE 324) between CSI encoder output from the actual CSI encoder (e.g. encoder 305) at UE and that from the hypothetical CSI encoder (e.g. encoder 303), CSI feedback 506 (i.e., CSI encoder output (e.g. CSI FB 307 or CSI FB 326) from the actual CSI encoder 305 at UE), SGCS 504 (e.g. SGCS 334, SGCS 318, SGCS 320) reported at output of the monitoring module, for example monitoring module 306, monitoring module 314, or monitoring module 316 (proxy decoder or SGCS estimator)). 3.a.i.l. In one embodiment in which the monitoring engine 352 bearing side is UE-side, channel related information 514, e.g., SINR 516, Doppler shift / spread 518, delay shift / spread 520, etc., can be included as input variables (shown in FIG. 5), assuming that corresponding input variables have been collected according to the training data generation embodiment 2.b.i. 3.a.i.2. In another embodiment in which the monitoring engine 302 bearing side is NW-side, a link adaptation scheme related information, i.e., CQI, TBS, etc., can be included as input variables (not shown in FIG. 5), assuming that corresponding input variables have been collected according to the training data generation embodiment 2.b.ii.

[0076] 3.a.ii. Output labels consist of: (root cause 508, i.e., NW-side 522, UE-side 520 or data drift 524, decoding result 530 (CRC success / fail; ACK 510 or NACK 512 or its probability) of the associated precoded PDSCH). 3.a.ii.1. Optionally, true SGCS (“SGCSt” 478 in FIG. 4) collected in embodiment 2. a. can be used as one of outputs (not shown in FIG. 5) such that the model should be trained to predict true SGCS, based on the input variables enumerated in embodiment 3.a.i.

[0077] 3.b. Alternatively, one monolithic model (that combines both model 501 and model 503) can be used for both root cause identification 508 and ACK7NACK prediction (510 / 512) simultaneously.

[0078] 3.c. In general (for a. and b.), the model training characteristics can be seen as classification or regression, and the underlining AI / ML model structure can be selected accordingly.

[0079] The signaling diagram between UE 110 and NW (170, 190) can be found in FIG. 6, which is MSC-view of FIG. 3. The steps of FIG. 6 are given as follows:

[0080] 1: Steps 2 to 4 can be considered as a [Preparation Phase], in which

[0081] 2: NW asks UE about its capabilities,

[0082] 3: UE provides its capability related information, e.g., support of AIML-enabled CSI compression feature, performance monitoring capability (incl. availability of hypothetical CSI encoder model from a specific NW vendor / version), etc.

[0083] 4: After examining UE capability report (Step 3), NW transfers the latest hypothetical CSI encoder model to UE for monitoring purposes, if it deems necessary. This step can be skipped if UE is already equipped with the latest version of the model on device.

[0084] 5: (Optional) NW may want to configure performance monitoring operation details, e.g., periodicity and time offset of monitoring occasions, monitoring -related dedicated feedback elements configuration (e.g., MSE, SGCS or monitoring bits in case of model inference at NW-side; Event index / ID in case of model inference at UE-side), number of CSI feedback occasions for monitoring (A>1), etc.

[0085] 6: [CSI encoding / decoding Inference and Monitoring Phase]: Performance monitoring inference can take place either at NW-side [AltO]; Steps 8 to 16 or at UE-side [Altl] ; Steps 18 to 20.

[0086] 8-12: Steps 8 to 12 are ordinary AUML-enabled CSI compression-reconstruction procedures. These procedures are not explicitly illustrated in Step 18 for the sake of brevity, but it is assumed that the same operation (including Step 13) is ongoing as well within Step 18.

[0087] 13: UE performs CSI compression (encoding) by using the hypothetical CSI encoder, after more time-critical CSI encoding process via its own actual CSI encoder has been completed (Step 10). UE can compute MSE between outcome from the hypothetical encoder and that of the actual CSI encoder.

[0088] 14: UE reports model monitoring-related parameters, i.e., MSE, SGCS or monitoring bits, to NW. Note here that overall there are two feedback messages from UE to NW (Step 11 and 14). Step 11 deems more time critical, as its feedback is to be used for precoding matrix calculations at NW for subsequent DL PDSCH. In case that timing requirement of Step 11 allows, Step 11 and 14 may be merged to reduce the number of OTAsignals.

[0089] 15: (Optional) At least one of the monitoring parameters may require pre-processing, e.g., monitoring bits interpretation / conversion to SGCS, in case monitoring bit generator is employed at UE instead of direct SGCS estimator.

[0090] 16: AI / ML-cnablcd monitoring engine inference takes place at NW. UE-reported monitoring parameters like (MSE, SGCS) are used as its input variables, together with dequantized CSI feedback (encoder outcome). Monitoring engine’s outputs consist of a possible root cause and probability of CRC success (or ACK7NACK likelihood). Optionally, link adaptation related parameters can be provided as extra input variables, i.e., CQI, TBS, etc., on top of the previously mentioned parameters, if it deems beneficial.

[0091] In case of monitoring engine inference at UE (Step 17),

[0092] 18: This procedure consists of ordinary AI / ML-cnablcd CSI compressionreconstruction procedures to be followed by CSI encoding via the hypothetical CSI encoder for computation of MSE, i.e., Steps 8 to 13. Note that this procedure can be repeated by preconfigured number (A>1), to acquire statistically reliable monitoring result.

[0093] 19: AI / ML-enabled monitoring engine inference takes place at UE. As regards input / output interface of the model, please refer to Figure 4. Optionally, DL channel related parameters can be provided as extra input variables, i.e., SINR, Doppler shift / spread, delay shift / spread, etc., if it deems beneficial.

[0094] 20: UE reports model monitoring-related Event occasion in case of Event detection. Here, Event can be defined in 3GPP specifications in a similar way with the measurement reporting framework. Event triggering / cancellation conditions can be based on combination of (root cause, predicted CRC success probability) with pre-defined thresholds. If necessary and beneficial, this threshold value may be subject to dynamic configuration from the NW- side (can be performed in Step 5 for instance).

[0095] 21: (Common for [AltO] Step 7 and [Altl] Step 17) Based on monitoring reports from UE, i.e., [AltO] monitoring parameters or [Altl] Event ID, NW makes a monitoring- related decision, i.e., model switching, fall back, model re-training, etc.

[0096] FIG. 7 is an example apparatus 700, which may be implemented in hardware,configured to implement the examples described herein. The apparatus 700 comprises at least one processor 702 (e.g. an FPGA and / or CPU), one or more memories 704 including computer program code 705, the computer program code 705 having instructions to carry out the methods described herein, wherein the at least one memory 704 and the computer program code 705 are configured to, with the at least one processor 702, cause the apparatus 700 to implement circuitry, a process, component, module, or function (implemented with control module 706) to implement the examples described herein. The one or more memories 704 may include a non-transitory memory, a transitory memory, a volatile memory (e.g. RAM), or a non-volatile memory (e.g. ROM).

[0097] Performance monitoring 730 implements the examples described herein related to performance monitoring for AIML-enabled CSI compression.

[0098] The apparatus 700 includes a display and / or I / O interface 708, which includes user interface (UI) circuitry and elements, that may be used to display aspects or a status of the methods described herein (e.g., as one of the methods is being performed or at a subsequent time), or to receive input from a user such as with using a keypad, camera, touchscreen, touch area, microphone, biometric recognition, one or more sensors, etc. The apparatus 700 includes one or more communication e.g. network (N / W) interfaces (I / F(s)) 710. The communication I / F(s) 710 may be wired and / or wireless and communicate over the Internet / other network(s) via any communication technique including via one or more links 724. The link(s) 724 may be the link(s) 131 and / or 176 from FIG. 1. The link(s) 131 and / or 176 from FIG. 1 may also be implemented using transceiver(s) 716 and corresponding wireless link(s) 726. The communication I / F(s) 710 may comprise one or more transmitters or one or more receivers.

[0099] The transceiver 716 comprises one or more transmitters 718 and one or more receivers 720. The transceiver 716 and / or communication I / F(s) 710 may comprise standard well-known components such as an amplifier, filter, frequency-converter, (de)modulator, and encoder / decoder circuitries and one or more antennas, such as antennas 714 used for communication over wireless link 726.

[0100] The control module 706 of the apparatus 700 comprises one of or both parts 706-1 and / or 706-2, which may be implemented in a number of ways. The control module 706 may be implemented in hardware as control module 706-1, such as being implemented as part ofthe one or more processors 702. The control module 706-1 may be implemented also as an integrated circuit or through other hardware such as a programmable gate array. In another example, the control module 706 may be implemented as control module 706-2, which is implemented as computer program code (having corresponding instructions) 705 and is executed by the one or more processors 702. For instance, the one or more memories 704 store instructions that, when executed by the one or more processors 702, cause the apparatus 700 to perform one or more of the operations as described herein. Furthermore, the one or more processors 702, the one or more memories 704, and example algorithms (e.g., as flowcharts and / or signaling diagrams), encoded as instructions, programs, or code, are means for causing performance of the operations described herein.

[0101] The apparatus 700 to implement the functionality of control 706 may be UE 110, RAN node 170 (e.g. gNB), or network element(s) 190 (e.g. LMF 190). Thus, processor 702 may correspond to processor(s) 120, processor(s) 152 and / or processor(s) 175, memory 704 may correspond to one or more memories 125, one or more memories 155 and / or one or more memories 171, computer program code 705 may correspond to computer program code 123, computer program code 153, and / or computer program code 173, control module 706 may correspond to module 140-1, module 140-2, module 150-1, and / or module 150-2, and communication I / F(s) 710 and / or transceiver 716 may correspond to transceiver 130, antenna(s) 128, transceiver 160, antenna(s) 158, N / W I / F(s) 161, and / or N / W I / F(s) 180. Alternatively, apparatus 700 and its elements may not correspond to either of UE 110, RAN node 170, or network element(s) 190 and their respective elements, as apparatus 700 may be part of a self-organizing / optimizing network (SON) node or other node, such as a node in a cloud.

[0102] The apparatus 700 may also be distributed throughout the network (e.g. 100) including within and between apparatus 700 and any network element (such as a network control element (NCE) 190 and / or the RAN node 170 and / or UE 110).

[0103] Interface 712 enables data communication and signaling between the various items of apparatus 700, as shown in FIG. 7. For example, the interface 712 may be one or more buses such as address, data, or control buses, and may include any interconnection mechanism, such as a series of lines on a motherboard or integrated circuit, fiber optics or other optical communication equipment, and the like. Computer program code (e.g. instructions) 705, including control 706 may comprise object-oriented software configured topass data or messages between objects within computer program code 705, or computer program code (e.g. instructions) 705, including control 706 may include functional, scripting, or procedural code. The apparatus 700 need not comprise each of the features mentioned, or may comprise other features as well. The various components of apparatus 700 may at least partially reside in a common housing 728, or a subset of the various components of apparatus 700 may at least partially be located in different housings, which different housings may include housing 728.

[0104] FIG. 8 shows a schematic representation of non-volatile memory media 800a (e.g. computer / compact disc (CD) or digital versatile disc (DVD)) and 800b (e.g. universal serial bus (USB) memory stick) and 800c (e.g. cloud storage for downloading instructions and / or parameters 802 or receiving emailed instructions and / or parameters 802) storing instructions and / or parameters 802 which when executed by a processor allows the processor to perform one or more of the steps of the methods described herein. Instructions and / or parameters 802 may represent a computer readable medium.

[0105] FIG. 9 is an example method 900 based on the examples described herein. At 910, the method includes receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder. At 920, the method includes determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback. At 930, the method includes wherein the monitoring engine is enabled with artificial intelligence or machine learning. Method 900 may be performed with RAN node 170, one or more network elements 190, or apparatus 700.

[0106] FIG. 10 is an example method 1000 based on the examples described herein. At 1010, the method includes determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder. At 1020, the method includes determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback. At 1030, the method includes wherein the monitoring engine is enabled with artificial intelligence or machine learning. Method 1000 may be performed with UE 110 or apparatus 700.

[0107] FIG. 11 is an example method 1100 based on the examples described herein. At 1110, the method includes determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder. At 1120, the method includes transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder. Method 1100 may be performed with UE 110 or apparatus 700.

[0108] FIG. 12 is an example method 1200 based on the examples described herein. At 1210, the method includes receiving, from a user equipment, a monitoring result prediction. At 1220, the method includes implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment. Method 1200 may be performed with RAN node 170, one or more network elements 190, or apparatus 700..

[0109] The following examples are provided and described herein.

[0110] Example 1. An apparatus including: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: receive, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determine a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0111] Example 2. The apparatus of example 1, wherein the monitoring result comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

[0112] Example 3. The apparatus of example 2, wherein the predicted root cause of performance degradation comprises one of: performance degradation due to the user equipment, or performance degradation due to a network, or data drift.

[0113] Example 4. The apparatus of any of examples 1 to 3, wherein the apparatus is further caused to: receive, from the user equipment, monitoring bits; process the monitoring bitsreceived from the user equipment to generate the squared generalized cosine similarity used as input to the monitoring engine.

[0114] Example 5. The apparatus of any of examples 1 to 4, wherein the apparatus is further caused to: receive, from the user equipment, the squared generalized cosine similarity used as input to the monitoring engine.

[0115] Example 6. The apparatus of any of examples 1 to 5, wherein the apparatus is further caused to acquire the squared generalized cosine similarity used as input to the monitoring engine from at least one of: a direct report of an SGCS estimate from the user equipment, or an outcome of processing of monitoring bits which are reported from the user equipment.

[0116] Example 7. The apparatus of any of examples 1 to 6, wherein the apparatus is further caused to: receive, from the user equipment, the quantized channel state information feedback used as input to the monitoring engine.

[0117] Example 8. The apparatus of any of examples 1 to 7, wherein the apparatus is further caused to: execute a model to determine a prediction of acknowledgment or negative acknowledgment; wherein the model is enabled with artificial intelligence or machine learning.

[0118] Example 9. The apparatus of example 8, wherein the monitoring engine comprises the model used to determine the prediction of acknowledgement or negative acknowledgement.

[0119] Example 10. The apparatus of any of examples 1 to 9, wherein the input to the monitoring engine further comprises link adaptation scheme related information.

[0120] Example 11. The apparatus of any of examples 1 to 10, wherein the link adaptation scheme related information comprises at least one of: a channel quality indicator, or a transport block size.

[0121] Example 12. The apparatus of any of examples 1 to 11, wherein the apparatus is caused to generate training data for the monitoring engine with at least one of: altering parameters of the actual channel state information encoder to cause performance degradation due to the user equipment, for the monitoring result to identify performance degradation due to the user equipment, or replacing the actual channel state information encoder with aproblematic actual channel state information encoder to cause performance degradation due to the user equipment, for the monitoring result to identify performance degradation due to the user equipment, or altering parameters of a channel state information decoder to cause performance degradation due a network, for the monitoring result to identify performance degradation due to the network, or replacing a channel state information decoder with a problematic channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or altering channel parameters to cause performance degradation due to data drift, for the monitoring result to identify performance degradation due to data drift.

[0122] Example 13. An apparatus including: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: determine a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determine a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0123] Example 14. The apparatus of example 13, wherein the monitoring result comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

[0124] Example 15. The apparatus of example 14, wherein the predicted root cause of performance degradation comprises one of: performance degradation due to the apparatus, or performance degradation due to a network, or data drift.

[0125] Example 16. The apparatus of any of examples 13 to 15, wherein the apparatus is further caused to acquire the squared generalized cosine similarity used as input to the monitoring engine from at least one of: a channel state information reconstruction model, or a proxy decoder model, or an actual decoder model, or a direct intermediate squared generalized cosine similarity estimator.

[0126] Example 17. The apparatus of example 16, wherein the apparatus is further caused to execute a monitoring module to determine the squared generalized cosine similarity usedas input to the monitoring engine, wherein the monitoring module used to determine the squared generalized cosine similarity comprises at least one of: the channel state information reconstruction model, or the proxy decoder model, or the actual decoder model, or the direct intermediate squared generalized cosine similarity estimator.

[0127] Example 18. The apparatus of any of examples 13 to 17, wherein the apparatus is further caused to: execute the actual channel state information encoder to determine the quantized channel state information feedback used as input to the monitoring engine.

[0128] Example 19. The apparatus of any of examples 13 to 18, wherein the apparatus is further caused to: execute a model to determine a prediction of acknowledgment or negative acknowledgment; wherein the model is enabled with artificial intelligence or machine learning.

[0129] Example 20. The apparatus of example 19, wherein the monitoring engine comprises the model used to determine the prediction of acknowledgement or negative acknowledgement.

[0130] Example 21. The apparatus of any of examples 13 to 20, wherein the input to the monitoring engine further comprises channel related information.

[0131] Example 22. The apparatus of example 21, wherein the channel related information comprises at least one of: a signal to interference plus noise ratio, or a Doppler shift, or a Doppler spread, or a delay shift, or a delay spread.

[0132] Example 23. The apparatus of any of examples 13 to 22, wherein the apparatus is further caused to: execute the monitoring engine multiple times with the input comprising the mean squared error, the squared generalized cosine similarity, and the quantized channel state information feedback to generate a statistically reliable estimate of the monitoring result.

[0133] Example 24. The apparatus of any of examples 13 to 23, wherein the apparatus is further caused to: receive, from the network entity, a reporting configuration that indicates that the apparatus is to report to the network entity the monitoring result when the monitoring result exceeds a threshold; determine whether the monitoring result exceeds the threshold; and transmit, to the network entity, the monitoring result in response to the monitoring result exceeding the threshold.

[0134] Example 25. The apparatus of any of examples 13 to 24, wherein the apparatus is caused to generate training data for the monitoring engine with at least one of: altering parameters of the actual channel state information encoder to cause performance degradation due to the apparatus, for the monitoring result to identify performance degradation due to the apparatus, or replacing the actual channel state information encoder with a problematic actual channel state information encoder to cause performance degradation due to the apparatus, for the monitoring result to identify performance degradation due to the apparatus, or altering parameters of a channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or replacing a channel state information decoder with a problematic channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or altering channel parameters to cause performance degradation due to data drift, for the monitoring result to identify performance degradation due to data drift.

[0135] Example 26. An apparatus including: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: determine a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmit, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

[0136] Example 27. The apparatus of example 26, wherein the apparatus is further caused to: execute a monitoring module to generate monitoring bits; and transmit, to the network entity, the monitoring bits configured to be used to generate a squared generalized cosine similarity.

[0137] Example 28. The apparatus of any of examples 26 to 27, wherein the apparatus is further caused to: execute a monitoring module to generate a squared generalized cosine similarity; and transmit, to the network entity, the squared generalized cosine similarity.

[0138] Example 29. The apparatus of any of examples 26 to 28, wherein the apparatus is further caused to: execute the actual channel state information encoder to determine quantized channel state information feedback; and transmit, to the network entity, the quantized channelstate information feedback.

[0139] Example 30. An apparatus including: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: receive, from a user equipment, a monitoring result prediction; and implement a monitoring-related decision based on the monitoring result prediction received from the user equipment.

[0140] Example 31. The apparatus of example 30, wherein the monitoring result prediction comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

[0141] Example 32. The apparatus of any of examples 30 to 31, wherein the apparatus is further caused to: transmit, to the user equipment, a reporting configuration that indicates that the user equipment is to report to the apparatus the monitoring result prediction when the monitoring result prediction exceeds a threshold; and receive, from the user equipment, the monitoring result prediction when the monitoring result prediction exceeds the threshold.

[0142] Example 33. A method including: receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0143] Example 34. A method including: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0144] Example 35. A method including: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmitting, to a network entity, the mean squared errorbetween the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

[0145] Example 36. A method including: receiving, from a user equipment, a monitoring result prediction; and implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

[0146] Example 37. An apparatus including: means for receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and means for determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0147] Example 38. An apparatus including: means for determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and means for determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0148] Example 39. An apparatus including: means for determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and means for transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

[0149] Example 40. An apparatus including: means for receiving, from a user equipment, a monitoring result prediction; and means for implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

[0150] Example 41. A computer readable medium including instructions stored thereon for performing at least the following: receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result withexecuting a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0151] Example 42. A computer readable medium including instructions stored thereon for performing at least the following: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

[0152] Example 43. A computer readable medium including instructions stored thereon for performing at least the following: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

[0153] Example 44. A computer readable medium including instructions stored thereon for performing at least the following: receiving, from a user equipment, a monitoring result prediction; and implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

[0154] References to a ‘computer’, ‘processor’, etc. should be understood to encompass not only computers having different architectures such as single / multi-processor architectures and sequential or parallel architectures but also specialized circuits such as field- programmable gate arrays (FPGAs), application specific circuits (ASICs), signal processing devices and other processing circuitry. References to computer program, instructions, code etc. should be understood to encompass software for a programmable processor or firmware such as, for example, the programmable content of a hardware device whether instructions for a processor, or configuration settings for a fixed-function device, gate array or programmable logic device etc.

[0155] The memories as described herein may be implemented using any suitable data storage technology, such as semiconductor based memory devices, flash memory, magneticmemory devices and systems, optical memory devices and systems, non-transitory memory, transitory memory, fixed memory and removable memory. The memories may comprise a database for storing data.

[0156] The term “non-transitory,” as used herein, is a limitation of the medium itself (i.e., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM).

[0157] As used herein, the term ‘circuitry’ may refer to the following: (a) hardware circuit implementations, such as implementations in analog and / or digital circuitry, and (b) combinations of circuits and software (and / or firmware), such as (as applicable): (i) a combination of processor(s) or (ii) portions of processor(s) / software including digital signal processor(s), software, and memories that work together to cause an apparatus to perform various functions, and (c) circuits, such as a microprocessor(s) or a portion of a microprocessor(s), that require software or firmware for operation, even if the software or firmware is not physically present. As a further example, as used herein, the term ‘circuitry’ would also cover an implementation of merely a processor (or multiple processors) or a portion of a processor and its (or their) accompanying software and / or firmware. The term ‘circuitry’ would also cover, for example and if applicable to the particular element, a baseband integrated circuit or applications processor integrated circuit for a mobile phone or a similar integrated circuit in a server, a cellular network device, or another network device.

[0158] It should be understood that the foregoing description is only illustrative. Various alternatives and modifications may be devised by those skilled in the art. For example, features recited in the various dependent claims could be combined with each other in any suitable combination(s). In addition, features from different example embodiments described above could be selectively combined into a new example embodiment. Accordingly, this description is intended to embrace all such alternatives, modifications and variances which fall within the scope of the appended claims.

[0159] The following acronyms and abbreviations that may be found in the specification and / or the drawing figures are given as follows (the abbreviations and acronyms may be appended / combined with each other or with other characters using e.g. a dash, hyphen, slash, letter, or number, and may be case insensitive):3GPP third generation partnership project4G fourth generation5G fifth generation5GC 5G core networkAI / ML, AIML artificial intelligence and machine learningAMF access and mobility management functionACK acknowledgementASIC application- specific integrated circuitBLER block error rateCD compact / computer discCPU central processing unitCRC cyclic redundancy checkCQI channel quality indicatorCSI channel state informationCU central unit or centralized unitDC dual connectivityDEC decoderDL downlinkDSP digital signal processorDU distributed unitDVD digital versatile discE2E end-to-end eNB evolved Node B (e.g., an LTE base station)ENC encoderEN-DC E-UTRAN new radio - dual connectivity en-gNB node providing NR user plane and control plane protocol terminations towards the UE, and acting as a secondary node in EN-DC eT2 enhanced Type IIE-UTRA evolved UMTS terrestrial radio access, i.e., the LTE radio access technologyE-UTRAN E-UTRA networkFl interface between the CU and the DUFPGA field-programmable gate array gNB generalized node B, base station for 5G / NR, i.e., a node providingNR user plane and control plane protocol terminations towards theUE, and connected via the NG interface to the 5GCHypo hypotheticalIAB integrated access and backhaulID identifierI / F interfaceI / O input / outputIODT interoperability development testingKPI key performance indicatorLCM lifecycle managementLLS link level simulatorLMF location management functionLTE long term evolution (4G)MAC medium access controlMLP multilayer perceptron mm monitoring module MME mobility management entityMRO mobility robustness optimizationMSC message sequence chartMSE mean squared errorNACK negative acknowledgementNCE network control element ng or NG new generation ng-eNB new generation eNB NG-RAN new generation radio access networkNR new radioNW networkN / W networkOTA over the airPDA personal digital assistantPDCP packet data convergence protocolPDSCH physical data / downlink shared channelPHY physical layerRAM random access memoryRAN radio access networkRE resource elementRLC radio link controlROM read-only memoryRRC radio resource controlRU radio unitRx receive, or receiver, or receptionSDAP service data adaptation protocolSGCS squared generalized cosine similaritySGW serving gatewaySINR signal to interference and noise ratioSMF session management functionSON self-organizing / optimizing networkSLS system level simulatorSRS sounding reference signalTBS transport block sizeTDD time division duplexTRP transmission reception pointTS technical specificationTx transmit, or transmitter, or transmissionUAV unmanned aerial vehicleUE user equipment (e.g., a wireless, typically mobile device)UI user interfaceUMTS Universal Mobile Telecommunications SystemUPF user plane functionUSB universal serial busX2 network interface between RAN nodes and between RAN and the core networkXn network interface between NG-RAN nodes

Claims

CLAIMSWhat is claimed is:

1. An apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: receive, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determine a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

2. The apparatus of claim 1, wherein the monitoring result comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

3. The apparatus of claim 2, wherein the predicted root cause of performance degradation comprises one of: performance degradation due to the user equipment, orperformance degradation due to a network, or data drift.

4. The apparatus of any of claims 1 to 3, wherein the apparatus is further caused to: receive, from the user equipment, monitoring bits; process the monitoring bits received from the user equipment to generate the squared generalized cosine similarity used as input to the monitoring engine.

5. The apparatus of any of claims 1 to 4, wherein the apparatus is further caused to: receive, from the user equipment, the squared generalized cosine similarity used as input to the monitoring engine.

6. The apparatus of any of claims 1 to 5, wherein the apparatus is further caused to acquire the squared generalized cosine similarity used as input to the monitoring engine from at least one of: a direct report of an SGCS estimate from the user equipment, or an outcome of processing of monitoring bits which are reported from the user equipment.

7. The apparatus of any of claims 1 to 6, wherein the apparatus is further caused to: receive, from the user equipment, the quantized channel state information feedback used as input to the monitoring engine.

8. The apparatus of any of claims 1 to 7, wherein the apparatus is further caused to: execute a model to determine a prediction of acknowledgment or negative acknowledgment ; wherein the model is enabled with artificial intelligence or machine learning.

9. The apparatus of claim 8, wherein the monitoring engine comprises the model used to determine the prediction of acknowledgement or negative acknowledgement.

10. The apparatus of any of claims 1 to 9, wherein the input to the monitoring engine further comprises link adaptation scheme related information.

11. The apparatus of any of claims 1 to 10, wherein the link adaptation scheme related information comprises at least one of: a channel quality indicator, or a transport block size.

12. The apparatus of any of claims 1 to 11, wherein the apparatus is caused to generate training data for the monitoring engine with at least one of: altering parameters of the actual channel state information encoder to cause performance degradation due to the user equipment, for the monitoring result to identify performance degradation due to the user equipment, or replacing the actual channel state information encoder with a problematic actual channel state information encoder to cause performance degradation due to the user equipment, for the monitoring result to identify performance degradation due to the user equipment, or altering parameters of a channel state information decoder to cause performance degradation due a network, for the monitoring result to identify performance degradation due to the network, or replacing a channel state information decoder with a problematic channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or altering channel parameters to cause performance degradation due to data drift, for the monitoring result to identify performance degradation due to data drift.

13. An apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least oneprocessor, cause the apparatus at least to: determine a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determine a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

14. The apparatus of claim 13, wherein the monitoring result comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

15. The apparatus of claim 14, wherein the predicted root cause of performance degradation comprises one of: performance degradation due to the apparatus, or performance degradation due to a network, or data drift.

16. The apparatus of any of claims 13 to 15, wherein the apparatus is further caused to acquire the squared generalized cosine similarity used as input to the monitoring engine from at least one of: a channel state information reconstruction model, or a proxy decoder model, oran actual decoder model, or a direct intermediate squared generalized cosine similarity estimator.

17. The apparatus of claim 16, wherein the apparatus is further caused to execute a monitoring module to determine the squared generalized cosine similarity used as input to the monitoring engine, wherein the monitoring module used to determine the squared generalized cosine similarity comprises at least one of: the channel state information reconstruction model, or the proxy decoder model, or the actual decoder model, or the direct intermediate squared generalized cosine similarity estimator.

18. The apparatus of any of claims 13 to 17, wherein the apparatus is further caused to: execute the actual channel state information encoder to determine the quantized channel state information feedback used as input to the monitoring engine.

19. The apparatus of any of claims 13 to 18, wherein the apparatus is further caused to: execute a model to determine a prediction of acknowledgment or negative acknowledgment ; wherein the model is enabled with artificial intelligence or machine learning.

20. The apparatus of claim 19, wherein the monitoring engine comprises the model used to determine the prediction of acknowledgement or negative acknowledgement.

21. The apparatus of any of claims 13 to 20, wherein the input to the monitoring engine further comprises channel related information.

22. The apparatus of claim 21, wherein the channel related information comprises at least one of: a signal to interference plus noise ratio, ora Doppler shift, or a Doppler spread, or a delay shift, or a delay spread.

23. The apparatus of any of claims 13 to 22, wherein the apparatus is further caused to: execute the monitoring engine multiple times with the input comprising the mean squared error, the squared generalized cosine similarity, and the quantized channel state information feedback to generate a statistically reliable estimate of the monitoring result.

24. The apparatus of any of claims 13 to 23, wherein the apparatus is further caused to: receive, from the network entity, a reporting configuration that indicates that the apparatus is to report to the network entity the monitoring result when the monitoring result exceeds a threshold; determine whether the monitoring result exceeds the threshold; and transmit, to the network entity, the monitoring result in response to the monitoring result exceeding the threshold.

25. The apparatus of any of claims 13 to 24, wherein the apparatus is caused to generate training data for the monitoring engine with at least one of: altering parameters of the actual channel state information encoder to cause performance degradation due to the apparatus, for the monitoring result to identify performance degradation due to the apparatus, or replacing the actual channel state information encoder with a problematic actual channel state information encoder to cause performance degradation due to the apparatus, for the monitoring result to identify performance degradation due to the apparatus, oraltering parameters of a channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or replacing a channel state information decoder with a problematic channel state information decoder to cause performance degradation due to a network, for the monitoring result to identify performance degradation due to the network, or altering channel parameters to cause performance degradation due to data drift, for the monitoring result to identify performance degradation due to data drift.

26. An apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: determine a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmit, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

27. The apparatus of claim 26, wherein the apparatus is further caused to: execute a monitoring module to generate monitoring bits; and transmit, to the network entity, the monitoring bits configured to be used to generate a squared generalized cosine similarity.

28. The apparatus of any of claims 26 to 27, wherein the apparatus is further caused to: execute a monitoring module to generate a squared generalized cosine similarity; andtransmit, to the network entity, the squared generalized cosine similarity.

29. The apparatus of any of claims 26 to 28, wherein the apparatus is further caused to: execute the actual channel state information encoder to determine quantized channel state information feedback; and transmit, to the network entity, the quantized channel state information feedback.

30. An apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: receive, from a user equipment, a monitoring result prediction; and implement a monitoring-related decision based on the monitoring result prediction received from the user equipment.

31. The apparatus of claim 30, wherein the monitoring result prediction comprises at least one of: a predicted root cause of performance degradation, or a predicted successful cyclic redundancy check, or a predicted failed cyclic redundancy check, or a predicted squared generalized cosine similarity.

32. The apparatus of any of claims 30 to 31, wherein the apparatus is further caused to: transmit, to the user equipment, a reporting configuration that indicates that the user equipment is to report to the apparatus the monitoring result prediction when the monitoring result prediction exceeds a threshold; andreceive, from the user equipment, the monitoring result prediction when the monitoring result prediction exceeds the threshold.

33. A method comprising: receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

34. A method comprising: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

35. A method comprising: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channelstate information encoder.

36. A method comprising: receiving, from a user equipment, a monitoring result prediction; and implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

37. An apparatus comprising: means for receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and means for determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

38. An apparatus comprising: means for determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and means for determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

39. An apparatus comprising: means for determining a mean squared error between an output of an actualchannel state information encoder and an output of a hypothetical channel state information encoder; and means for transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

40. An apparatus comprising: means for receiving, from a user equipment, a monitoring result prediction; and means for implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

41. A computer readable medium comprising instructions stored thereon for performing at least the following: receiving, from a user equipment, a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, and quantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

42. A computer readable medium comprising instructions stored thereon for performing at least the following: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and determining a monitoring result with executing a monitoring engine with input comprising: the mean squared error, a squared generalized cosine similarity, andquantized channel state information feedback; wherein the monitoring engine is enabled with artificial intelligence or machine learning.

43. A computer readable medium comprising instructions stored thereon for performing at least the following: determining a mean squared error between an output of an actual channel state information encoder and an output of a hypothetical channel state information encoder; and transmitting, to a network entity, the mean squared error between the output of the actual channel state information encoder and the output of the hypothetical channel state information encoder.

44. A computer readable medium comprising instructions stored thereon for performing at least the following: receiving, from a user equipment, a monitoring result prediction; and implementing a monitoring-related decision based on the monitoring result prediction received from the user equipment.

Images