Apparatus, method and computer program
The latent space prediction scheme using AI/ML models and spatial encoders improves CSI compression in wireless networks by leveraging temporal correlations, enhancing performance and reducing overhead.
Patent Information
- Application Number
- PCT/EP2026/061795
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-09
- Filing Date
- 2026-04-24
- Publication Date
- 2026-11-12
AI Technical Summary
Existing communication systems face challenges in efficiently utilizing temporal domain aspects for channel state information (CSI) compression, leading to suboptimal performance in wireless networks.
Implementing a latent space prediction scheme using AI/ML models, spatial and frequency encoders, and quantizers to predict and compress CSI, leveraging temporal correlations in channel state information.
Enhances overall network performance by improving CSI compression efficiency and reducing feedback overhead, while maintaining low computational complexity and power consumption.
Smart Images

Figure EP2026061795_12112026_PF_FP_ABST
Abstract
Description
TITLEApparatus, Method and Computer ProgramTECHNICAL FIELD
[0001] Various embodiments of this disclosure relate generally to methods, apparatus and computer programs, and in particular, but not exclusively, to a latent space prediction scheme for AIML-enabled two-sided CSI compression.BACKGROUND
[0002] A communication system can be seen as a facility that enables communication sessions between two or more communication devices, or provides communication devices access to a network. A mobile or wireless communication network is one example of a communication network. A communication device may be provided with a service by an application server.
[0003] A mobile or wireless communication network may operate in accordance with standard(s), such as those provided by 3GPP (Third Generation Partnership Project) or ETSI (European Telecommunications Standards Institute). Examples of mobile or wireless communication network that operate in accordance with 3GPP standards are generally referred to as 4G (4th Generation) networks, 5G (5th Generation) network, 5G-Advanced networks and 6G networks.SUMMARY
[0004] Some embodiments of this disclosure will be described with respect to certain aspects. These aspects are not intended to indicate key or essential features of the various example embodiments of this disclosure, nor are they intended to be used to limit the scope of thereof. Other features, aspects, and elements will be readily apparent to a person skilled in the art in view of this disclosure. For example, it should be appreciated that further aspects may be provided by the combination of any two or more of the various aspects described herein.
[0005] In a first aspect there is provided a method comprising providing M sequential channel state informations as input to a spatial and frequency encoder, wherein the output of the spatial and frequency encoder is M latent vectors, providing the M latent vectors as input to a predictor, wherein the output of the predictor comprises N predicted latent vectors for N future time instances in sequence and providing an indication of the N predicted latent vectors to a network entity.
[0006] The method may comprise providing the N predicted latent vectors as input to a quantizer, wherein the output of the quantizer comprises a bit stream comprising one or more bits per dimension of the latent vector; and
[0007] Providing an indication of the N predicted latent vectors to a network entity may comprise providing the bit stream to the network entity for decoding by a spatial and frequency decoder.
[0008] The method may comprise providing the indication of the N predicted latent vectors to the network entity in uplink control information, UCI.
[0009] Each of the N predicted latent vectors may be associated with a UCI field.
[0010] The predictor may comprise a scalar predictor and the output of the predictor comprises N scalar predictions per dimension of the latent vector.
[0011] The predictor may comprise a vector predictor and the output of the predictor comprises N predicted vectors per segment, where the D-dimensional latent vector consists of multiple equally sized segments of size S dimensions.
[0012] The predictor may uses at least one of the following to provide the output an artificial intelligence / machine learning, AI / ML, model, a non-AI / ML based rule based approach, an autoregressive model; or Wiener filtering.
[0013] The M sequential channel state information may be obtained from M channel state information reference signals received over an observation window, wherein the observation window is incremented for each subsequent M sequential channel state information.
[0014] The observation window may be incremented every channel state information reference signal received by the UE.
[0015] The observation window may be incremented every L channel state information reference signals received by the UE, where L > 1.
[0016] In a second aspect there is provided a method comprising providing M sequential channel state informations as input to a spatial and frequency encoder, wherein the output of the spatial and frequency encoder is M latent vectors, providing the latent vectors as input to a predictor, wherein the output of the predictor comprises N predicted latent vectors for N future time instances in sequence, providing the N predicted latent vectors as input to a recurrent quantizer, wherein the output of the recurrent quantizer comprises a bit stream comprising one or more bits per dimension of the latent vector prediction and providing the bit stream to a network entity for decoding by a spatial and frequency decoder.
[0017] The method may comprise providing a first K of the N predicted latent vectors to a quantizer, wherein the output of the quantizer comprises a bit stream comprising one or more bits per dimension of the latent vector.
[0018] For the first K of the N predicted latent vectors, the bit stream provided to the network may comprise the output of the quantizer and for N-K predicted latent vectors, the bit stream provided to the network comprises the output of the recurrent quantizer.
[0019] In a third aspect there is provided a method comprising determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following:a spatial frequency encoder and a quantizers spatial frequency encoder and a recurrent quantizers spatial frequency encoder and a predictor; a spatial frequency encoder, predictor and a quantizer; and a spatial frequency encoder, predictor and recurrent quantizer;and providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.
[0020] The determining may comprise determining to switch based on an indication received from a network or an event.
[0021] The indication may be received in radio resource control signalling, a media access control (MAC) control element (CE), or downlink control information.
[0022] The event may comprise a detected change in temporal correlation of a downlink channel or a detected change in speed of change of a downlink channel.
[0023] Detecting the speed of change of a downlink channel may be based on detecting the speed of the UE.
[0024] Detecting a change in temporal correlation of a downlink channel may be based on a detected Doppler spread of the channel.
[0025] The other one of the plurality of configurations may be indicated to the network.
[0026] The other one of the plurality of configurations may be indicated together with channel feedback information.
[0027] In a fourth aspect there is provided a method comprising providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of: a spatial frequency encoder and a quantizer; a spatial frequency encoderand a recurrent quantizer; a spatial frequency encoder and a predictor; a spatial frequency encoder, predictor and a quantizer; and a spatial frequency encoder, predictor and recurrent quantizer.
[0028] Providing the indication may comprise determining to provide the indication, wherein the determining is based on an event.
[0029] The event may comprise a detected change in temporal correlation of a downlink channel or a detected change in speed of change of a downlink channel.
[0030] Detecting the speed of change of a downlink channel may be based on detecting the speed of the UE.
[0031] Detecting a change in temporal correlation of a downlink channel may be based on a detected Doppler spread of the channel.
[0032] In a fifth aspect there is provided a method comprising performing training of a model for a predictor, wherein the output of the predictor comprises N predicted latent vectors of a spatial and frequency encoder, wherein the input of the predictor comprises M sequential latent vectors, wherein the input training dataset for the training is based on the M latent vectors and wherein the output training dataset for the training is based on N latent vectors.
[0033] The model for the predictor comprises least one of the following: an artificial intelligence / machine learning, AI / ML, model, a non-AI / ML based rule based approach, an autoregressive model; or Wiener filtering.
[0034] Performing training of the model for the predictor may comprise performing prediction training on a latent vector of D dimensions.
[0035] Prediction training may be performed per segment, where the D-dimen-sional latent vector consists of multiple equally sized segments of size S dimensions.
[0036] The training may feed back for each segment a delayed output for every one or more input.
[0037] In a sixth aspect there is provided an apparatus comprising means for performing the method according to the first, second, third, fourth or fifth aspect.
[0038] In a seventh aspect there is provided an apparatus comprising at least one processor, and at least one memory storing instructions which, when executed by the at least one processor, cause the apparatus at least to perform a method according to the first, second, third, fourth or fifth aspect.
[0039] In an eighth aspect there is provided a non-transitory computer readable medium comprising instructions wherein the instructions when executed by at least one processor of an apparatus cause the apparatus to perform the method according to thefirst, second, third, fourth or fifth aspect.
[0040] In a ninth aspect there is provided a computer program comprising instructions which, when executed by an apparatus, cause the apparatus to perform at least the method according to the first, second, third, fourth or fifth aspect.
[0041] Some embodiments of the invention are defined in the dependent claims.
[0042] In the above, many different aspects have been described. As previously noted, it should be appreciated that further aspects may be provided by the combination of any two or more of the aspects described above (or otherwise in this disclosure).
[0043] Various other aspects are also described in the following detailed description and in the claims.BRIEF DESCRIPTION OF THE FIGURES
[0044] Some embodiments will be described, by way of non-limiting and illustrative example only, with reference to the figures, in which:
[0045] Fig. 1 shows an example of a communication network to which examples disclosed herein may be applied;
[0046] Fig. 2 shows a schematic diagram of architecture for two-sided CSI compression with SF encoder / decoder and a quantizer;
[0047] Fig. 3 shows a schematic diagram of architecture for two-sided CSI compression with recurrent quantization;
[0048] Fig. 4 shows a schematic diagram of architecture for two-sided CSI compression with Adaptive Delta Modulation / Demodulation;
[0049] Fig. 5 shows an example of a latent vector element distribution;
[0050] Fig. 6a shows a schematic diagram of a predictor architecture;
[0051] Fig. 6b shows a schematic diagram of a predictor architecture;
[0052] Fig. 7 shows a schematic diagram of a predictor architecture;
[0053] Fig. 8 shows a flowchart of a method according to an example;;
[0054] Fig. 9 shows a flowchart of a method according to an example;
[0055] Fig. 10a shows a schematic diagram of an example architecture for two-sided CSI compression with recurrent quantization and a predictor;
[0056] Fig. 10b shows a schematic diagram of an example architecture for two-sided CSI compression with quantization and a predictor;
[0057] Fig. 11a shows an example CSI observation and prediction operation scenario;
[0058] Fig. 11b shows an example CSI observation and prediction operation scenario;
[0059] Fig. 12 shows a schematic diagram of an example predictor architecture;
[0060] Fig. 13 shows a flowchart of a method according to an example;
[0061] Fig. 14 shows a flowchart of a method according to an example;
[0062] Fig. 15 shows an example call flow between a NW and a UE;
[0063] Fig. 16 shows a flowchart of a method according to an example;
[0064] Fig. 17 shows a block diagram of E2E model training and inference (deployment) procedures for latent space prediction for AI / ML-enabled CSI compression;
[0065] Fig. 18 shows a flowchart of a method according to an example;
[0066] Fig. 19 shows an example of an apparatus.DETAILED DESCRIPTION
[0067] The following embodiments are provided by way of non-limiting and illustrative example. Although the specification may refer to “an”, “one”, or “some” embodiments) in several locations of the text, this does not necessarily mean that each reference is made to the same embodiment(s), or that a particular feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embodiments. Further, when a particular feature, structure, or characteristic is described in connection of an embodiment, it intended such feature, structure, or characteristic may be applied in connection with other embodiments (whether or not explicitly described).
[0068] It shall be understood that although the terms “first,” “second” and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another.
[0069] For the purposes of this disclosure, the phrases “at least one of A or B”, “at least one of A and B”, and “A and / or B” means (A), (B), or (A and B). For the purposes of this disclosure, the phrase “A, B, and / or C” means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C).
[0070] As used herein, the term “or” refers to a non-exclusive “or” unless otherwise indicated (e.g., use of “or else” or “or in the alternative”).
[0071] As used herein, unless stated explicitly, performing a respective feature, step, or functionality “in response to A” does not indicate that the respective feature, step, or functionality is performed immediately after “A” occurs as one or more intervening features, steps, or functionalities may be performed (at least in part) between anoccurrence of the respective feature, step, or function and “A”. Analogously, performing a respective feature, step, or functionality “based on A” does not indicate that the respective feature, step, or functionality is performed solely based on “A” as the respective feature, step, or functionality may be further based on one or more other features, steps, or functionalities in addition to “A”.
[0072] Embodiments described herein may be implemented in a communication network, such as any of the following radio access technologies (RATs): Worldwide Interoperability for Micro-wave Access (WiMAX), Global System for Mobile communications (GSM, 2G), GSM EDGE radio access Network (GERAN), General Packet Radio Service (GRPS), Universal Mobile Telecommunication System (UMTS, 3G) based on basic wideband-code division multiple access (W-CDMA), high-speed packet access (HSPA), Long Term Evolution (LTE), LTE-Advanced, and enhanced LTE (eLTE), 5G (also called NR), or any future RAT such as 6G. Moreover, communication within the communication network may utilize any proper wireless communication technology, comprising but not limited to: Code Division Multiple Access (CDMA), Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), Frequency Division Duplex (FDD), Time Division Duplex (TDD), Multiple-Input Multiple-Output (MIMO), Orthogonal Frequency Division Multiple (OFDM), and / or Discrete Fourier Transform spread OFDM (DFT-s-OFDM).
[0073] As used herein, the term “network device” or “network node” refers to a node in a communication network via which user equipment may access the network and / or which is configured to control radio communication and managing radio resources within a cell. The network node or network device may be referred to as a base station (BS), an access point (AP) or an access node. The network device may be, depending on the applied technology, for example, a node B (NodeB or NB), an evolved NodeB (eNodeB or eNB), an NR NB (also referred to as a gNB), a Remote Radio Unit (RRU), a radio head (RH), a remote radio head (RRH), a relay, an Integrated Access and Backhaul (IAB) node, a low power node, a non-terrestrial network (NTN) or nonground network device, such as a satellite network device, a low earth orbit (LEO) satellite and a geosynchronous earth orbit (GEO) satellite, or an aircraft network device.
[0074] Moreover, in connection of split radio access network (RAN), the network device may refer to a centralised unit (CU) of a base station and / or a distributed unit (DU) of a base station. An interface between CU and DU may be referred to as an F1 interface in NR. In the split RAN architecture, node operations may be carried out, at least partly, in the central / centralized unit, CU, (e.g. server, host or node) operationallycoupled to the DU, (e.g. a radio head / node). One CU may control one or more DUs, acting at least as transmit / receive (Tx / Rx) nodes. In some embodiments, the DUs may comprise e.g. a radio link control (RLC), medium access control (MAC) layer and a physical (PHY) layer, whereas the CU may comprise the layers above RLC layer, such as a packet data convergence protocol (PDCP) layer, a radio resource control (RRC) and an internet protocol (IP) layers. Other functional splits are possible too. In practice, any processing task may be performed in either the CU or the DU and the boundary where the responsibility is shifted between the CU and the DU may depend on the applied implementation.
[0075] The term “terminal device” refers to any end device that may be configured to perform wireless communication. By way of example, a terminal device may be referred to as a communication device, user equipment (UE), a Subscriber Station (SS), or a Mobile Station (MS). The terminal device may include a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones a tablet, a wearable terminal device, a personal digital assistant (PDA), portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and playback appliances, vehicle-mounted wireless terminal devices, USB dongles, an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts), a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like.
[0076] A term “resource”, as used herein, may refer to radio resources in time domain, in frequency domain, in space domain, and / or in code domain. Some examples of resources may include, e.g., a physical resource block (PRB), a radio frame, a subframe, a time slot, a subband, a frequency region, a sub-carrier, a beam, etc. The term “transmission” and / or “reception” may refer to wirelessly transmitting and / or receiving via a wireless propagation channel on radio resources.
[0077] Fig. 1 illustrates an example of a communication network to which examples disclosed herein may be applied. The communication network or a cellular communication network may comprise a network node 110 configured to provide one or more cells, such as cell 100, and a network node 112 configured to provide one or more other cells, such as cell 102. Each cell may, for example, be a macro cell, a micro cell, femto, or a pico cell. The cell may define a coverage area or a service area of thecorresponding access node.
[0078] The network node (110, 112) may be configured to provide a user equipment (UE) 120 (one or more UEs) with wireless access to the communication network. The wireless access may comprise downlink (DL) communication from the network node (110, 112) to the UE 120 and uplink (UL) communication from the UE 120 to the network node (110, 112). Examples of uplink channels may comprise physical uplink control channel (PUCCH) for transmitting control information and physical uplink shared channel (PUSCH) for transmitting data towards the network. Examples of downlink channels may comprise physical downlink control channel (PDCCH) for transmitting control information and physical downlink shared channel (PDSCH) for transmitting data towards the user equipment.
[0079] There may be a plurality of UEs (120, 122) in the system. Each of the plurality of UEs may be served by the same or by different network nodes (110, 112). UE may be configured with dual connectivity (DC), wherein the UE, for example UE 120, may be connected to multiple network nodes (110, 112). The UEs (120, 122) may communicate with each other, in case device-to-device (D2D) communication interface is established between them via a so-called sidelink (SL). Such D2D communications may be referred to as machine-to-machine, peer-to-peer (P2P) communications, or ve-hicle-to-vehicle (V2V), for example.
[0080] In the case of multiple network nodes in the communication network, the network nodes may be connected to each other via an interface. LTE specifications, for example, refer to such an interface as an X2 interface. An interface between an LTE node and a 5G node, or between two 5G nodes may be called an Xn interface.
[0081] The network nodes 110 and 112 may be further connected via another interface to a core network 116 of the communication network. The LTE specifications specify the core network as an evolved packet core (EPC), and the core network may comprise a plurality of entities (e.g. a mobility management entity (MME) and a gateway node). The MME may handle mobility of terminal devices in a tracking area encompassing a plurality of cells and handle signalling connections between the terminal devices and the core network. The gateway node may handle data routing in the core network and to / from the terminal devices. The 5G specifications specify the core network as a 5G core (5GC). The 5GC may, for example, comprise an access and mobility management function (AMF) and a user plane function / gateway (UPF) and other functions. The AMF may handle termination of non-access stratum (NAS) signalling, NAS ciphering & integrity protection, registration management, connection management,mobility management, access authentication and authorization, security context management. The UPF node may, for example, support packet routing and forwarding, packet inspection and quality of service (QoS) handling.
[0082] AI / ML based channel state information (CSI) compression using two-sided model (two sides of an over the air (OTA) interface, e.g., UE and network entity) is being considered. One topic under discussion is temporal domain aspects on top of spatial and frequency domain compression.
[0083] Consideration of temporal domain aspect may lead to overall performance enhancement.
[0084] Fig. 2 shows an example architecture for two-sided CSI compression (using SF encoder-decoder) with history independent elementwise scalar quantization. This may be known as Case 0.
[0085] One example architecture for exploiting the time correlation in CSI compression is shown in Fig. 3 and adds recurrence to a quantizer / inverse quantizer (a.k.a. dequantizer) stage. In the approach shown in Fig. 3, mapping to and from the latent space is done using an SF encoder decoder pair. The quantizer and inverse quantizer are augmented by feeding back state variables s and st' respectively, with delay. This structure may allow the quantizer to achieve lower mean squared error with a given number of bits in btwhen the CSI matrices Vtare correlated in time. This is due to the reason that the latent vectors ztinherit time-dependence properties like those of the CSI matrices. The example shown in Fig. 3 may be known as Case 2. The Case 2 use case may be supported by directly making use of SF encoder / decoder pair which have been developed for Case 0 with adaptation at the latent domain (replacement of historyindependent quantizer with recurrent quantizer).
[0086] Recurrent quantizer-inverse quantizer operation may be enabled either by AI / ML based scheme or by an explicitly designed scheme. According to numerical evaluation results, a SF encoder / decoder with an explicitly designed recurrent quantizer may outperform both SF encoder / decoder with AI / ML based recurrent quantizer and recurrent SFT encoder / decoder with reduced complexity. The latent vector elements which, in an example, have been collected per 5 ms interval generally vary smoothly in time, which renders element-wise tracking appealing.
[0087] Fig. 4 shows an alternative architecture two-sided CSI compression with recurrent quantization where an Adaptive Delta Modulator (AD Mod) and an AdaptiveData Demodulator are positioned between the SF encoder and SF Decoderand the AD mode produces an adaptive delta modulation encoded value based on the latent vector.
[0088] The examples shown in Figs. 3 and 4 do not require retraining / update of SF encoder / decoder which have been acquired for Case 0 via a quantization-aware training manner. The examples of Figs. 3 and 4 exploit temporal correlation of the latent vector elements depicted in Fig. 5 by taking an elementwise (scalar) value tracking.This means that SF encoder / decoder pair at LIE / NW, respectively, developed for Case0 use case may be used for Case 2 use case by updating quantization operation from history independent mode to history dependent (i.e. , recurrent) mode.
[0089] Fig. 5 shows selected trajectories of a subset of (pre-quantization) latent vector sequences obtained by an SF encoder to a sequence of 40 CSI matrices observed per every 5 ms. On the left is shown selected 4 dimensions and on the right another 4 dimensions from 128 dimension.
[0090] The categorization shown in Table 1 has been adopted for the evaluation of temporal domain aspects of AI / ML-based CSI compression using two-sided model.Case Target CSI Whether the UE uses past CSI Whether the network uses past CSI slot(s) information information 0 Present slot No No1 Present slot Yes No2 Present slot Yes Yes3 Future Yes Noslot(s)4 Future Yes Yesslot(s)5 Present slot No YesTable 1
[0091] For the UE, the past CSI information may include past model inputs and / or any information derived from them. For the network, the past CSI information may include past CSI feedback instances and / or any information derived from them. For case3 and case 4, the UE may perform prediction as a separate step or jointly with compression. Similarly, the network may perform prediction as a separate step or jointly with reconstruction. “Target CSI slot(s)” refers to the slot(s) to which the CSI feedback in the report corresponds. “Present slot” refers to the slot of the most recent CSI-RS measurement used to generate the CSI report. “Future slot(s)” includes at least one slot after the present slot and may include the present slot as well. Down-selection is not precluded.
[0092] Two options have been identified to support temporal domain Case 2 and 3. Option 1 requires specifying a new model structure extending the model structure for temporal domain Case 0, with additional layers to account for past CSIs. Option 2 is based on latent space adaptation, i.e., no new model structure is needed, but instead a new quantisation structure is adopted that can process past CSIs. In option 2, as discussed with reference to Figs. 3 and 4, no new model structures need to be specified for inter-vendor training collaboration with exchange of model parameters.
[0093] For Case 3, there has been a focus on separate predictions based on the full channel matrices (or processed channel information like channel eigenvectors) investigations (see Figs. 6a and 6b). This approach might provide some advantages in terms of prediction performance, as the predictor model takes the raw full channel matrices. However, it may incur high model complexity in terms of the number of trainable parameters and FLOPs. The approach may also pose challenges for the required memory size as well since it uses the multiple (M-1 in Fig. 6a) previous full channel matrices.
[0094] Another approach under investigation for Case 3 takes a joint channel prediction and compression scheme (see, e.g. Fig. 7). In addition to the challenges of the separate prediction and compression schemes, this approach necessitates a separate model development for Case 3 besides the model for Case 0. This approach may render re-use of Case 0 model for Case 3 use case challenging. The above-mentioned factors may also contribute to an increase in UE / modem power consumption.
[0095] Case 3 refers to the case in which UE uses M past CSI(s) for prediction of the target CSIs at N future slot(s). Here, M and N may be considered as CSI observation window length and CSI prediction window length, respectively.
[0096] In an example, a CSI predictor takes M (> 2) sequential instances of the measured (observed) CSIs or their pre-processed or encoded versions as its inputs for computation of the future N (> 1) CSI(s). CSI predictor provides N future CSI(s) as its output(s). The objective of CSI prediction is to overcome channel aging. In a typical CSI operation scenario, around the time at which UE receives a payload (PDSCH) which has been precoded by a CSI having been reported by a UE, the CSI having been used for precoding is already outdated due to a channel temporal evolution between CSI-RS transmission for UE-side channel measurement and precoded PDSCH transmission. To prevent this effect from affecting E2E performance in terms of DL throughput, UE can compute forecasts of the future CSI(s) which correspond to CSI(s) at the points of time of upcoming DL transmission, based on previous past CSI observations.
[0097] Fig. 8 shows a flowchart of a method according to an example. The method may be performed at an apparatus. The apparatus may be, comprise, or be comprised in a user equipment.
[0098] At 801, the method comprises providing M sequential channel state informations as input to a spatial and frequency encoder, wherein the output of the spatial and frequency encoder is M latent vectors.
[0099] At 802, the method comprises providing the M latent vectors as input to a predictor, wherein the output of the predictor comprises N predicted latent vectors for N future time instances in sequence.
[0100] At 803, the method comrpises providing an indication of the N predicted latent vectors to a network entity.
[0101] Fig. 9 shows a flowchart of a method according to an example where output of the predictor may be provided as input to a recurrent quantizer. The method may be performed at an apparatus. The apparatus may be, comprise, or be comprised in a user equipment.
[0102] At 901, the method comprises providing M sequential channel state informations as input to a spatial and frequency encoder, wherein the output of the spatial and frequency encoder is M latent vectors.
[0103] At 902, the method comprises providing the latent vectors as input to a predictor, wherein the output of the predictor comprises N predicted latent vectors for N future time instances in sequence.
[0104] At 903, the method comprises providing the N predicted latent vectors as input to a recurrent quantizer, wherein the output of the recurrent quantizer comprises a bit stream comprising one or more bit(s) per dimension of the latent vector prediction.
[0105] At 904, the method comprises providing the bit stream to a network entity for decoding by a spatial and frequency decoder.
[0106] Methods as described with reference to Figs. 8 and 9 may provide a method for CSI prediction for the two-sided AI / ML-enabled CSI compression use case by employing a prediction on the latent space, e.g., output of CSI compression (SF encoding), rather than on the full channel matrix or its pre-processed outcome like channel eigenvectors. The main advantage of this proposal comes from the compactness of the latent space. As CSI compression / encoding performs compression of the channel state information, its output, i.e., the latent vector, shows a highly condensed texture. Moreover, the latent space retains correlation over temporal sequence as observed in Fig. 5.
[0107] As depicted in Figs 10a, 10b and 12, a predictor on the latent space takes M historical (previous) latent vectors ({ztj 1 ■> ’ * ' >z*-Af+i})asits input and provides / V future predictions of the latent vectors (or their individual elements;W+i,zt+2, * ■ • 5Zt+A?i)asitsoutput. Here, it is assumed that the current time stamp is t, at which prediction takes place at UE. Input (M historical latent vectors) of the predictor is acquired by running an SF encoder repeatedly by taking M historical target CSIs, i.e., {Vt, Vt-i, • • • , Vt- +i}whentarget CSI is channel eigenvector, in sequence. Likewise at NW side, N dequantized future latent vector predictions ( zt+i,zt+2,are fec| .nto anF decoder, which is running repeatedly N times, taking one predicted latent vector at a time. The eventual outcome of the SF decoder comprises N reconstructed predicted target CSIs ( I+ J) NOteherethe SF encoder / decoder pair in use are acquired for Case 0 model development.
[0108] Multiple instances of the predictors can be wrapped under one container, as depicted in Figs. 10a and 10b. The decomposition of the predictor container depicted in Figs. 10a and 10b into individual scalar predictors is illustrated in Fig 12.
[0109] The method may comprise providing the N predicted latent vectors as input to a quantizer, wherein the output of the quantizer comprises a bit stream comprising one or more bits per dimension of the latent vector and wherein providing an indication of the N predicted latent vectors to a network entity comprises providing the bit stream to the network entity for decoding by a spatial and frequency decoder.
[0110] In an example, illustrated in Fig. 10b, the output of the predictor is fed into a non-recurrent (history independent) scalar or vector quantizer (referred to as a quantizer). This may provide high-fidelity quantization at the cost of possibly increased feedback overhead.10111] In one embodiment, each individual scalar quantizer may be trained per dimension. This leads to a dedicated scalar quantizer weights per dimension. Scalar quantizers can share the same backbone architecture and structure.
[0112] In another embodiment, one common scalar quantizer can be trained across dimensions. The identical trained weights can be shared with scalar quantizers which work in parallel.
[0113] In an example, output of the predictor may be provided as input to a recurrent quantizer. This example is depicted in Fig. 10a and may reduce feedback overhead over the air.
[0114] The method may comprise providing a first K of the N predicted latent vectors to a quantizer, wherein the output of the quantizer comprises a bit streamcomprising one or more bits per dimension of the latent vector. For the first K of the N predicted latent vectors, the bit stream provided to the network may comprise the output of the quantizer and for N-K predicted latent vectors, the bit stream provided to the network may comprise the output of the recurrent quantizer. This may be referred to as a hybrid approach.
[0115] For example with K=1, a hybrid approach may be adopted wherein the 1st predicted value is to be quantized with normal scalar quantization with full resolution, whereas the subsequent remaining (N-1) predictions of the prediction instance are to be quantized via a recurrent quantizer with reduced number of bits.
[0116] In another embodiment of the hybrid approach in which future predictions are ensued continuously after prediction lead time, the 1st predicted value of the first prediction instance is to be quantized with normal scalar quantization with full resolution, whereas all the subsequent predictions including the future prediction instances are to be quantized via a recurrent quantizer with a reduced number of bits.
[0117] In the above examples described with reference to Figs. 8 and 9, an SF encoder / decoder pair, which has been developed / trained for Case 0 in which neither UE nor NW uses past CSI(s) for encoding / decoding of the target CSI at the present slot, may be directly used for Case 3 in which UE uses multiple past CSI(s) for prediction of the target CSIs at possibly multiple future slot(s).
[0118] CSI observation and prediction operation can be performed in various scenarios. The M sequential channel state information described with reference to Figs. 8 and 9 may be obtained from M channel state information reference signals received over an observation window, wherein the observation window is incremented for each subsequent M sequential channel state information. Figs. 11a and 11b depict two major possible scenarios for better understanding. Note here that there may be other scenarios to be proposed / supported in future cellular communications standards.
[0119] The observation window may be incremented every channel state information reference signal received by the UE. For example, in a first scenario as illustrated in Fig. 11a, a CSI observation window of length M(M=4 in the illustrated example) proceeds in a sliding window fashion as time progresses. Whenever UE acquires a new CSI from the transmitted CSI-RS, it computes and provides N (A / =1 in the example of Fig. 11a) future forecast(s) of CSI (i.e. , N predicted latent vectors) to NW, based on the most recent past M CSIs (including the newly acquired CSI). This scenario assumes continuous CSI-RS transmissions from NW, and it provides straightforward means for performance monitoring of the predictions - as can be observed in Fig. 11a, UE cancompare its (previously computed) predictions with the actual CSI(s) thanks to continuously transmitting CSI-RS. Not to mention, CSI-RS provides a new latest input to CSI predictor as well.
[0120] The observation window may be incremented every L channel state information reference signals received by the UE, where L > 1. For example, in a second scenario illustrated in Fig. 11b, CSI-RS transmission takes a burst-followed by-pause pattern. For the illustrated example in Figure 11b, UE awaits until it acquires M (M=4 in the illustrated example) CSI(s), then it computes and provides N ( / \ / =2 in the example of Fi. 11b) future forecast(s) of CSI (i.e., N predicted latent vectors) to NW. During the time in which NW makes use of these reported CSI forecasts for precoding of DL payloads, NW may hold transmission of CSI-RS(s) during this period to save CSI-RS overhead in the air interface.
[0121] It should be noted that for both scenarios, a predictor can be designed to provide future forecast(s) as soon as it acquires minimum required number of past CSI(s), say 2 instances, as its input for the case at which it is desirable to provide predictions as early as possible at the cost of potential degradation of prediction reliability.
[0122] In our case, we focus on scalar prediction, since it provides scalable architecture. It is scalable in the sense that for various dimension sizes any required number of scalar predictors can be added (or activated) or removed (or deactivated) as necessary (see bottom subfigure of Fig. 12 for better understanding). It is also well aligned with the (per-dimension) recurrent quantizer scheme, as can be observed in Fig. 10a, predictor outputs (forecasted future CSI(s)) can be directly fed into a recurrent quantizer per-dimension, as there may most likely exist temporal correlation between predicted CSI(s). It should be noted that this does not preclude usage of normal (history independent) scalar / vector quantizer. As the latent space predictor operates at the output of CSI encoder prior to quantization, it should be quantization scheme agnostic.
[0123] The predictor may comprise a vector predictor and the output of the predictor comprises N predicted vectors per segment, where the D-dimensional latent vector consists of multiple equally sized segments of size S dimensions. Alternatively, or in addition, the predictor may comprise a scalar predictor, and the output of the predictor may comprise N scalar predictions per dimension of the scalar vector (segment of size S=1 dimension).
[0124] In an example, as illustrated in Figs. 10a and 10b, a scalar predictor is placed at the CSI encoder output per each dimension of the latent vector, which takes M historical latent vector elements per dimension and computes and provides N future latent vector element predictions per dimension.
[0125] For example, prediction may be performed on the latent vectors as a whole ({zbzt-li ' ' * >;€ R where D is dimension of the latent vector).
[0126] In an example, prediction is performed on the latent vectors based on their individual sequential (scalar) elements per dimension ({^M’ * ’ * ’ ■ z E R, i jSdimension index, i.e., « G {0, 2, • • • , -D — 1})asdepicted in Fig. 12.
[0127] The predictor may use an AI / ML model to provide the output. For example, a scalar predictor may be based on AIML-enabled sequence models which take a sequence of data to forecast the next upcoming value(s), e.g., RNN, LSTM, GRU, etc.
[0128] Fig. 12 illustrates an example of the elementwise (a.k.a per-dimension or scalar) latent space predictor. Input data, i.e., M historical (previous) latent vectors (z G R are decomposed and rearranged per-dimen-sion such that M sequence of scalar values can be fed into individual scalar predictor in a per-dimension basis. Each predictor computes N scalar predictions per-dimension, i.e.,^t+2,i5 ’ ’ ’ > where is dimension index, i.e., i E {0, 1, 2, • • • , D — 1}. individual scalar predictors in the predictor container may share the same backbone architecture, e.g., RNN, LSTM, GRU, etc., and its layer struc-ture details. Predictor model training may develop either dedicated per-dimension predictor models or one common predictor model. From a scalability perspective, it is more desirable, if feasible, to develop a common predictor model which can be used irrespective of the dimension index. Depending on implementation practices of the predictor model, a recurrent loop (which is indicated with z-1sign to denote one step delay in Fig. 12) may be connected from the output of the predictor model to the input.
[0129] This elementwise scalar prediction concept makes sense as there exists strong correlation among temporally adjacent latent vector elements when traced per-dimension (seeFig. 5). As illustrated at the southeast corner of Fig. 12, sequential deep neural network models like RNN, LSTM, GRU, etc. are proven to be effective in forecasting future likely values based on monitoring of the past samples. As the task of scalar prediction is relatively simple, it is expected that well-performing sequence models would not command high complexity in terms of the number of trainable model parameters and / or the number of FLOPs.
[0130] Although the above examples describe AIML-based prediction, CSI prediction on a latent space may be implemented by non-AIML schemes, i.e., conventional signal processing algorithm-based schemes, as well.
[0131] The predictor may use a non-AI / ML rules-based approach, such as but not limited to an autoregressive (AR) model or Wiener filtering to provide the output. Forexample, a predictor may be based on non-AI / ML algorithms, e.g., AR model, Wiener filtering, etc.
[0132] In legacy non-AIML-based CSI prediction specified in 3GPP Rel-18 as well as in the AIML-based CSI prediction being specified in 3GPP Rel-19, the CSI report makes use of the same Type-ll PMI codebook, namely Rel-16 eType-ll for a single predicted sample (N = 1) or Rel-18 Type-ll-Doppler for multiple predicted CSIs (N > 1).
[0133] Methods as described with reference to Fig.s 8 and 9 may comprise providing the indication of the N predicted latent vectors to the network entity in uplink control information, UCI. Each of the N predicted latent vectors may be associated with a UCI field.
[0134] For example, in the proposed solution for temporal domain Case 3, the use of PMI codebooks may not be needed, because prediction is done on the latent space vectors rather than the channel, so no PMI calculation is needed after prediction. In an example, the UCI format for the CSI report may be simplified by associating separate UCI fields with the predicted latent space vectors, as illustrated in Figure Table 2, where each predicted CSI is associated with a predicted time slot configured by the NW in a corresponding CSI reporting setting.Latent space Latent space Latent space vector for 1stvector for 2ndvector for Nthpredicted CSI predicted CSI predicted CSI Table 2
[0135] Fig. 13 shows a flowchart of a method according to an example. The method may be performed at an apparatus. The apparatus may be, comprise, or be comprised in a UE.
[0136] In 1301, the method comprises determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following: a spatial frequency encoder and a quantizer, a spatial frequency encoder and a recurrent quantizer, a spatial frequency encoder and a predictor; a spatial frequency encoder, predictor and a quantizer; and a spatial frequency encoder, predictor and recurrent quantizer.
[0137] In 1302, the method comprises providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.
[0138] Determining to switch may comprise determining to switch based on an indication received from a network (e.g., a network entity).
[0139] Fig. 14 shows a flowchart of a method according to an example. The method may be performed at an apparatus. The apparatus may be, comprise or be comprised in a network entity.
[0140] At 1401, the method comprises providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of a spatial frequency encoder and a quantizer, a spatial frequency encoder and a recurrent quantizer, a spatial frequency encoder and a predictor; a spatial frequency encoder, predictor and a quantizer and a spatial frequency encoder, predictor and recurrent quantizer.
[0141] The indication may be provided in radio resource control signalling, a media access control (MAC) control element (CE), or downlink control information (DCI).
[0142] Determining to switch may be based on an event. The event may comprise a detected change in temporal correlation of a downlink channel or a detected change in speed of change of a downlink channel. Detecting the speed of change of a downlink channel may be based on detecting the speed of the UE. Detecting a change in temporal correlation of a downlink channel may be based on a detected Doppler spread of the channel.
[0143] Methods as described with reference to Figs. 13 and 14 may configure / ac-tivate / trigger a CSI report with space-frequency-time compression and prediction, and associated parameters to facilitate configuration of Cases 0 / 2 / 3 and low latency switching between Cases 0 / 2 / 3.
[0144] The other one of the plurality of configurations may be indicated to the network.
[0145] The other one of the plurality of configurations may be indicated together with channel feedback information.
[0146] Fig. 15 depicts an example NW-LIE interaction scenario to facilitate configuration of Cases 0 / 2 / 3 and low latency switching between Cases 0 / 2 / 3.
[0147] In step 1, as a part of UE capability reporting, UE may report AIM L-enabled CSI compression feature related to UE capabilities, including whether the UE supports for (latent space) CSI prediction (and recurrent quantization, if applicable).
[0148] At step 2, configurations of possibly multiple predictor configurations (and recurrent quantizer configuration, if applicable) are notified by NW to UE via higher level signaling like RRC message or MAC CE.
[0149] At step 3, the NW is determined to operate in Case 0 as regards 3GPP-defined temporal domain aspects related use case categorization, i.e., UE to provide an encoded target CSI of the current slot without using the past CSI information. For Case 0, NW is not supposed to use the past CSI information for decoding of target CSI delivered from UE.
[0150] At step 4, the NW indicates UE of the selected Case, i.e., Case 0, and the associated configurations, i.e., (latent space) CSI prediction: disable, (recurrent quantization: disable, if applicable), configurations for selected history independent (non-re-current) quantization, via lower level signaling, i.e., DCI.
[0151] At steps 5 to 6, NW and UE prepare for Case 0 operation (5.) by loading Case 0 CSI decoder model parameter weights with history independent de-quantizer being configured at NW by loading Case 0 CSI encoder model weights with history independent quantizer being configured at UE, respectively.
[0152] At steps 7 to 9, Case 0 operation runs, i.e., CSI-RS transmission for channel measurements / reporting, channel estimation / pre-processing followed by AIML-as-sisted target CSI encoding (CSI compression) at UE and CSI feedback reporting from UE (not shown in Fig. 15). NW reconstructs the target CSI by AIML-assisted CSI decoding (not shown in Fig.15) NW uses this reconstructed target CSI for precoding of upcoming PDSCH transmission.
[0153] At step 10, performance monitoring runs from time to time with both UE and NW being involved. Based on performance monitoring related reports from UE and its analysis at NW, NW may be able to make an informed decision about Case switching.
[0154] In the meantime, temporal correlation of the downlink channel is observed. This is an example of a detected change in temporal correlation of a downlink channel.
[0155] At step 11 , the NW determines to switch from Case 0 to Case 2, based on observations during Case 0 operation so far. For Case 2, both UE and NW are allowed to use past CSI information for encoding / decoding of target CSI at the current slot.
[0156] At step 12, NW indicates UE of the selected Case, i.e., Case 2, and the associated configurations, i.e., (latent space) CSI prediction: disable, recurrent quantization: enable, configurations for recurrent quantization, via lower level signaling, i.e., DCI. This is an example of providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, whereinthe plurality of configurations comprise a spatial frequency encoder and a quantizer and a spatial frequency encoder and a recurrent quantizer.
[0157] At steps 13 to 14. NW and UE prepare for Case 2 operation by configuring recurrent de-quantizer at NW and by configuring recurrent quantizer at UE, respectively. Note here that unloading / loading of encoder / decoder parameter weights is not required, as the Case 0 encoder / decoder models can be directly used for Case 2 when recurrent quantization is adopted.
[0158] At steps 15 to 20, Case 2 operation runs. The main difference from Case 0 operation is that UE and NW exploit temporal correlation at the latent space by using recurrent quantization / dequantization which leads to CSI feedback overhead reduction.
[0159] At step 21, NW determines to switch from Case 2 to Case 3, based on observations during Case 2 operation so far. For Case 3, UE is allowed to use past CSI information for prediction of target CSI(s) at the future slot(s). NW may determine CSI observation and prediction operation scenario for Case 3. The determination may be based on a detected change in speed of change of a downlink channel
[0160] At step 22, NW indicates UE of the selected Case, i.e., Case 3, and the associated configurations, i.e., (latent space) CSI prediction: enable, (recurrent quantization: enable or disable, if applicable), configurations for quantization (which can be either history independent or recurrent), and configurations for selected latent space predictor (e.g., M, N values) and for associated CSI observation and prediction operation scenario, via lower level signaling, i.e., DCI. This is an example of providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprise a spatial frequency encoder and a recurrent quantizer and a spatial frequency encoder, predictor and optionally, a quantizer or a recurrent quantizer. This is an example of indicating the other one of the plurality of configurations to the network, wherein the other one of the plurality of configurations is indicated together with channel feedback information.
[0161] At steps, 23 to 24, the NW and UE prepare for Case 3 operation by configuring CSI decoder setup at NW and by configuring latent space predictor as well as CSI encoder setup at UE, respectively. Here, CSI encoder / decoder setup refers to coordination required for providing M inputs to acquire N predictions at UE and appropriate interpretation of N predictions at NW for operation of Case 3. Note that the Case 0 encoder / decoder models can be directly used for Case 3 when latent space prediction is adopted.
[0162] At steps 25 to 35, Case 3 operation runs. Depending on the configured scenario, CSI-RS may not be transmitted continuously. Note here that latent space predictors are agnostic to quantizer in operation. This means enabling recurrent quantizer is optional, and any kind of quantizer (scalar / vector) can be used together with the proposed latent space predictor.
[0163] Fig. 16 shows a flowchart of a method according to an example. The method may be performed at an apparatus. The apparatus may be, comprise or be comprised in a UE.
[0164] At 1601, the method comrpises providing M+N sequential channel state informations as input to a spatial and frequency encoder, wherein the output of the spatial and frequency encoder is M+N latent vectors.
[0165] At 1602, the method comprises performing training of a model for a predictor, wherein the output of the predictor comprises N predicted latent vectors of a spatial and frequency encoder, wherein the input of the predictor comprises M sequential latent vectors, wherein the input training dataset for the training is based on the M latent vectors and wherein the output training dataset for the training is based on N latent vectors.
[0166] As described with reference to Figs. 8 and 9, the model for the predictor may comprise least one of the following an artificial intelligence / machine learning, AI / ML, model, a non-AI / ML based rule-based approach, an autoregressive model or Wiener filtering.
[0167] Performing training of the model for the predictor may comprise performing prediction training on a latent vector of D dimensions. Prediction training may be performed per segment, where the D-dimensional latent vector consists of multiple equally sized segments of size S dimensions.
[0168] An example of overall model training and deployment procedures is depicted in Fig. 17. At step A, there is E2E model training of an SF encoder / decoder pair with a quantizer-dequantizer in the middle. At step B, the trained weights for the encoder are used to generate a dataset for training the predictor (i.e., M latent vectors which are historical). The N latent vectors in the method of Fig. 17 may serve as ground-truth for prediction.
[0169] In an example, the training data comrpises a pair of (historical target CSI, historical latent vector), i.e., (input to SF encoder, output of SF encoder), sampled at time instances of length (M+N).
[0170] Fig. 18 shows a flowchart of a method of E2E model training and inference.
[0171] Step A in Fig. 17 is an example E2E (Case 0) Model Training Stage.
[0172] At step 1, E2E model training is performed with SF Encoder and SF decoder being directly connected with scalar quantizer - dequantizer in-between (quantization-aware training). Any training collaboration type may be supported, e.g., joint training, joint sequential training, separate sequential training, etc., and any inter-ven-dor collaboration direction I option which has been down selected in 3GPP may be supported.
[0173] At step 2, once SF encoder I decoder model training is completed, the trained model parameters are saved. With trained model parameters (weights) being loaded, input target CSI data in sequence can be fed into the SF encoder model to generate corresponding sequential latent vectors at the output of CSI encoder prior to quantizer, which are to be collected to form a dataset of the latent vectors.
[0174] At step 3, which is optional when a recurrent quantizer needs to be configured) recurrent quantizer configuration parameters are determined based on the collected latent vectors in step 2.
[0175] Step B in Fig. 17 comprises an example predictor model training stage. Based on the collected latent vectors in step 2, predictor model training is performed in a supervised learning manner. The training may be performed as described with reference to Fig. 16 and / or as illustrated in Fig. 17.
[0176] Step C in Fig. 17 comprises an example E2E Model Inference / Deploy-ment Stage (using Case 3)
[0177] At step 5, the trained model parameters / weights are loaded to an SF encoder and SF decoder. The predictor is positioned at the output of SF encoder at UE side. A recurrent or scalar quantizer may be placed at the output of predictor at UE side and place recurrent de-quantizer at NW side, with the configuration parameters for the recurrent quantizer being set as acquired at step 3. Then the entire chain with SF encoder - predictor - Quantizer-deQuantizer - SF decoder is run in an inference mode for E2E performance validation.
[0178] At step 6, upon successful E2E performance validation, the E2E model chain is deployed with latent space predictor being plugged in.
[0179] Note here that in Fig. 17, an example with recurrent quantizer option has been presented (Fig. 10a), whereas there exists an alternative using a history independent scalar / vector quantizer as depicted in Fig. 10b). Note also that the above-described procedures can be extended to vector quantization case as well.
[0180] Methods as described above provide modality, e.g., a straightforward extension from Case 0 to Case 3. The design and specification of the SF encoder / de-coder pair may be independent of the design of the latent space predictor. Previous specified SF encoder / decoder for Case 0 may be maintained and used for Case 3 simply by adopting the proposed predictor scheme at the output of SF encoder prior to quantizer at UE.
[0181] Methods as described above may be compatible with any quantization scheme. Since the prediction is done on the latent space prior to quantization, prediction operation itself is agnostic to quantization scheme in use. This feature renders the proposed scheme compatible with any quantization scheme, i.e. , SQ, VQ and recurrent quantization.
[0182] Methods as described above may provide scalability. Elementwise perdimension scalar predictor architecture allows scalability over the number of dimensions to support. Methods as described above may provide compactness of the predictor model. Elementwise per-dimension scalar predictor architecture may be realized by a model with small complexity. This may lead to low model complexity and memory requirements at UE and may come from the fact that the predictor itself works on compact domain, i.e., the latent space.
[0183] Methods may provide low power consumption at UE / modem due to the low predictor model complexity and memory requirements.
[0184] Methods may provide short lead time for Case switching since case 0 / 2 / 3 use cases are based on Case 0 SF encoder / decoder pair and switching of Cases can be done via enabling / disabling of predictor and / or recurrent quantizer, fast switching between Cases can be supported.
[0185] Fig. 19 shows, by way of example, a block diagram of an apparatus 10. The apparatus 10 comprises, for example, at least one processor 12 and at least one memory 14 storing instructions 15 that, when executed by the at least one processor, cause the apparatus 10 at least to perform the method or methods (or portion(s) thereof) as disclosed herein, and any of the embodiments (or respective portion(s) thereof). In an example, the at least one memory and the instructions (e.g. a computer program code, software), are configured, with the at least one processor, to cause the apparatus 10 to perform the method or methods (or portion(s) thereof) as disclosed herein, and any of the embodiments (or respective portion(s) thereof).
[0186] A processor 12 may comprise circuitry, or be constituted as circuitry or circuitries, the circuitry or circuitries being configured to perform phases of methods inaccordance with embodiments described herein.
[0187] As used herein, the term “circuitry” may refer to one or more or all of the following: (a) hardware-only circuit implementations, such as implementations in only analog and / or digital circuitry, and (b) combinations of hardware circuits and software, such as, as applicable: (i) a combination of analog and / or digital hardware circuit(s) with software / fi rmware and (ii) any portions of hardware processor(s) with software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a user equipment, to perform various functions) and (c) hardware circuit(s) and or processor(s), such as a microprocessor(s) or a portion of a microprocessors), that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation. This definition of circuitry applies to all uses of this term herein, including in any claims. As a further example, as used herein, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.
[0188] The memory 14 may be implemented using any suitable data storage technology. The memory may comprise a database for storing data. The memory 14 may, for example, be at least in part external to apparatus 10 but accessible to apparatus 10.
[0189] The instructions 15 may be comprised in a computer readable medium or a non-transitory computer readable medium. A 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. random access memory, RAM, vs. read only memory, ROM).
[0190] For example, the apparatus 10 is a terminal device, such as a UE. As another example, the apparatus is comprised in such a terminal device, e.g. as a chipset configured to control the terminal device. The apparatus 10 may be caused or configured or comprise means to perform at least the method of Figs. 8, 9, 13 or 16 and / or any one or more of the embodiments described herein.
[0191] As another example, the apparatus 10 is a network entity. In another embodiment, the apparatus is comprised in such a network entity, e.g. as a chipset configured to control the network entity. The apparatus 10 may be caused or configuredor comprise means to perform at least the method of Fig. 14 and / or any one or more of the embodiments described herein.
[0192] The apparatus may comprise one or more entities of any of protocol layers, such as a MAC entity, an RRC entity, an RLC entity, a PDCP entity or a PHY entity. In some embodiments, the entity is configured to perform at least the method of Figs. 5, and / or any one or more of the embodiments described.
[0193] The apparatus 10 comprises a radio interface 16. The radio interface 16 may provide the apparatus 10 with communication capabilities. The radio interface 16 may comprise a receiver configured to receive information in accordance with at least one cellular or non-cellular standard. The radio interface 16 may comprise a transmitter configured to transmit information in accordance with at least one cellular or non-cellular standard. The receiver may comprise more than one receiver. The transmitter may comprise more than one transmitter. The radio interface 16 may comprise a transceiver configured to receive and transmit information in accordance with at least one cellular or non-cellular standard. The transceiver may comprise more than one transceiver.
[0194] The apparatus 10 may comprise a user interface 18 comprising, for example, at least one of a keypad, a microphone, a touch display, a display, a speaker, etc. The user interface 18 may be used to control the apparatus by the user. The user interface 18 may be external to the apparatus 10. For example, the apparatus 10 may be connected to another device, such as a computer, either via wireless or wired connection, and the apparatus 10 is controlled by the user via the computer.
[0195] In an embodiment, at least some of the processes described herein may be carried out by an apparatus comprising means for carrying out at least some of the described processes. Means for performing method steps as disclosed herein may include software and / or hardware components of the apparatus 10. For example, the at least one processor 12, the memory 14, and the computer program code form means for carrying out the method or methods (or portion(s) thereof) as disclosed herein, and any of the embodiments (or respective portion(s) thereof). As used herein the term “means” is to be construed in singular form, i.e. referring to a single element, or in plural form, i.e. referring to a combination of single elements. Therefore, terminology “means for [performing A, B, C]”, is to be interpreted to cover an apparatus in which there is only one means for performing A, B and C, or where there are separate means for performing A, B and C, or partially or fully overlapping means for performing A, B, C. Further, terminology “means for performing A, means for performing B, means forperforming C” is to be interpreted to cover an apparatus in which there is only one means for performing A, B and C, or where there are separate means for performing A, B and C, or partially or fully overlapping means for performing A, B, C.
[0196] Even though this disclosure has been described above with reference to non-limiting and illustrative examples according to the accompanying figures, it is clear that the scope of this disclosure is not restricted thereto - but can be modified in many different ways. As technology advances, it will become apparent to a person skilled in art as to how the disclosure can be further implemented and / or modified in various ways. Further, it is clear to a person skilled in the art that the embodiments described herein may, but are not required to, be combined in various ways with other embodiments described herein.
Claims
28CLAIMS1. An apparatus comprising at least one processor, and at least one memory storing instructions which, when executed by the at least one processor, cause the apparatus at least to perform:determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer;and providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.
2. The apparatus according to claim 1 , wherein the determining comprises determining to switch based on an indication received from a network or an event.
3. The apparatus according to claim 2, wherein the indication is received in radio resource control signalling, a media access control (MAC) control element (CE), or downlink control information.
4. The apparatus according to claim 2, wherein the event comprises a detected change in temporal correlation of a downlink channel or a detected change in speed of change of a downlink channel.
5. The apparatus according to claim 4, wherein detecting the speed of change of a downlink channel may be based on detecting the speed of the UE.
6. The apparatus according to claim 4, wherein detecting a change in temporal correlation of a downlink channel may be based on a detected Doppler spread of the channel.
7. The apparatus according to any one of claims 1 to 6, wherein the other one of the plurality of configurations is indicated to the network.
8. The apparatus according to any one of claims 1 to 7, wherein the other one of the plurality of configurations is indicated together with channel feedback information.
9. An apparatus comprising at least one processor, and at least one memory storing instructions which, when executed by the at least one processor, cause the apparatus at least to perform: providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer.
10. The apparatus according to claim 9, wherein providing the indication comprises determining to provide the indication, wherein the determining is based on an event.
11. The apparatus according to claim 10, wherein the event comprises a detected change in temporal correlation of a downlink channel or a detected change in speed of change of a downlink channel.
12. The apparatus according to claim 11, wherein detecting the speed of change of a downlink channel is based on detecting the speed of the UE.
13. The apparatus according to claim 11, wherein detecting a change in temporal correlation of a downlink channel is based on a detected Doppler spread of the channel.
14. A method comprising:determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer;and providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.
15. A method comprising:providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer.
16. An apparatus comprising means for:determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer;and providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.
17. An apparatus comprising means for:providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer.
18. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to perform:determining to switch between one of a plurality of configurations for determination of channel state information feedback at the apparatus to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of the following:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer;and providing channel state information feedback to the network, wherein the channel state information is determined using the other one of the plurality of configurations.3219. A computer program comprising instructions which, when executed by an apparatus, cause the apparatus to perform:providing an indication to a user equipment to switch between one of a plurality of configurations for determination of channel state information feedback at a user equipment to an other one of the plurality of configurations, wherein the plurality of configurations comprises at least two of:a spatial frequency encoder and a quantizer;a spatial frequency encoder and a recurrent quantizer;a spatial frequency encoder and a predictor;a spatial frequency encoder, predictor and a quantizer; anda spatial frequency encoder, predictor and recurrent quantizer.
