Methods and systems for predicting channel state information in wireless communication networks

Abstract

A method for predicting CSI in a wireless communication system, the method comprising: receiving, from a base station: a full CSI-RS configuration indicating a first set of antenna ports; and a list of sub CSI-RS configurations indicating a second set of antenna ports that is a smaller subset of the first set; determining whether the full CSI-RS configuration and a sub CSI-RS configuration are nested or non-nested; selecting, based on the determining, an AI / ML-based or non-AI / ML-based prediction technique; selecting a target size for a CSI representation; predicting, using the selected prediction technique, a CSI comprising: (i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI corresponding to the sub CSI-RS configuration, or (ii) a sub CSI corresponding to the sub CSI-RS configuration based on a measured full CSI corresponding to the full CSI-RS configuration; and transmitting a CSI report based on the predicted CSI.

Description

METHODS AND SYSTEMS FOR PREDICTING CHANNEL STATE INFORMATION IN WIRELESS COMMUNICATION NETWORKS

[0001] The present disclosure relates generally to wireless communication systems, and more particularly, to techniques for optimally sharing channel state information (CSI) in next-generation radio access networks.

[0002] Description of Related Art Third Generation Partnership Project (3GPP) has been actively developing standardized Network Energy-Saving (NES) mechanisms for Sixth-Generation (6G) wireless systems with an objective of ensuring interoperability and broad industry adoption. Recent studies on energy-efficient operation of Fifth-Generation (5G) New Radio (NR) networks, particularly within 3GPP across different Releases and associated research efforts in the wider telecommunications community, have identified multiple approaches for reducing energy consumption across Radio Access Networks (RANs).

[0003] Among these approaches, spatial-domain NES techniques have emerged as an important category. In spatial-domain NES, a network may dynamically deactivate selected antenna elements or Transmission Reception Points (TRPs), particularly in multi-TRP and massive Multiple-Input Multiple-Output (MIMO) deployments, to reduce energy usage when traffic conditions permit. Such spatial deactivation allows the network to operate with reduced active antenna resources, thereby lowering power consumption while maintaining baseline connectivity.

[0004] However, these techniques introduce significant challenges for CSI acquisition and reporting. When antenna elements, panels, or TRPs are turned off or reconfigured, the CSI corresponding to a full configuration of antenna ports may no longer be directly measurable, and the CSI corresponding to a reduced configuration (also referred to as sub-configuration) may not accurately represent the channel characteristics of the full configuration.

[0005] This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the present disclosure. This summary is neither intended to identify key or essential inventive concepts of the present disclosure and nor is it intended for determining the scope of the present disclosure.

[0006] One or more embodiments of the present disclosure may provide a method performed by a user equipment (UE) in a wireless communication system for predicting channel state information (CSI) to support network energy saving, the method comprising: receiving, from a base station: a full CSI reference signal (CSI-RS) configuration indicating a first set of antenna ports; and a list of sub CSI-RS configurations indicating a second set of antenna ports, wherein the second set is a smaller size subset of the first set; determining whether a relationship between the full CSI-RS configuration and a sub CSI-RS configuration of the list of sub CSI-RS configurations corresponds to a nested CSI configuration or a non-nested CSI configuration; selecting, based on the relationship, a prediction technique from a plurality of prediction techniques, wherein the plurality of prediction techniques comprises an artificial intelligence / machine learning (AI / ML) based technique and a non-AI / ML based technique; selecting, for a CSI representation to be predicted based on one or more performance or structure metrics, a target size; predicting, using the selected prediction technique, a CSI comprising at least one of: (i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI that corresponds to the sub CSI-RS configuration, or (ii) a sub CSI corresponding to the sub CSI-RS configuration based on a measured full CSI that corresponds to the full CSI-RS configuration; and transmitting, to the base station, a CSI report based on the predicted CSI.

[0007] One or more embodiments of the present disclosure may provide a user equipment (UE) in a wireless communication system, the UE comprising: at least one transceiver; and at least one processor communicatively coupled to the at least one transceiver, the at least one processor configured to: receive: a full channel state information (CSI) reference signal (CSI-RS) configuration indicating a first set of antenna ports; and a list of sub CSI-RS configurations indicating a second set of antenna ports, wherein the second set is a smaller subset of the first set; determine whether a relationship between the full CSI-RS configuration and a sub CSI-RS configuration corresponds to a nested CSI configuration or to a non-nested CSI configuration; select a prediction method from a plurality of prediction methods, the plurality of prediction methods comprising at least one artificial intelligence / machine learning (AI / ML)-based method and at least one non-AI / ML-based method; predict, using the selected prediction techniques, a CSI comprising at least one of: (i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI that corresponds to the sub CSI-RS configuration, or (ii) a sub CSI corresponding to the sub CSI-RS configuration based on a measured full CSI that corresponds to the full CSI-RS configuration; and transmit a CSI report based on the predicted CSI.

[0008] These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

[0009] FIG.1A illustrates an example configuration in which a network operates with a smaller number of active antenna ports or antenna elements at a transmission side (Tx);

[0010] FIG.1B illustrates an example configuration in which the network activates a larger number of antenna ports or antenna elements at the Tx side;

[0011] FIG. 2 illustrates a relationship between a number of active antenna elements and a resulting beam width in a wireless communication system;

[0012] FIG. 3A illustrates an example scenario of a network operating with a large antenna configuration at a transmission side;

[0013] FIG. 3B illustrates an example scenario the network operating with a smaller antenna configuration at the transmission side;

[0014] FIG. 4 illustrates an example system implemented at a base station for enabling intelligent prediction of Channel State Information (CSI) associated with various antenna port configurations, in accordance with one or more embodiments of the present disclosure;

[0015] FIG. 5 illustrates an example flowchart of a method for generating a required sub configuration CSI from a full configuration CSI, in accordance with one or more embodiments of the present disclosure;

[0016] FIG. 6 illustrates a flowchart of a method for generating a full-configuration CSI from a sub-configuration CSI, in accordance with one or more embodiments of the present disclosure;

[0017] FIG. 7 illustrates an example scenario demonstrating channel propagation characteristics experienced by a user equipment (UE) transition between near-field and far-field regions, in accordance with one or more embodiments of the present disclosure;

[0018] FIG. 8 illustrates an example scenario demonstrating a training phase of an artificial intelligence or machine learning model for predicting CSI across different antenna port configurations, in accordance with one or more embodiments of the present disclosure;

[0019] FIG. 9 illustrates an example scenario demonstrating the training of a full configuration CSI matrix for a particular angular direction using multiple sub configuration CSI matrices obtained for that same angular direction, in accordance with one or more embodiments of the present disclosure;

[0020] FIG. 10 illustrates an example scenario demonstrating a correspondence of a sub configuration CSI matrix with multiple full configuration CSI matrices at different angular directions during the sub to full CSI prediction process, in accordance with one or more embodiments of the present disclosure;

[0021] FIG. 11 illustrates a flow of a method for predicting CSI to support network energy saving, in accordance with an embodiment of the present disclosure; and

[0022] FIG. 12 illustrates an example system for predicting CSI in a wireless communication system, in accordance with one or more embodiments of the present disclosure.

[0023] Further, skilled artisans will appreciate that elements in the drawings are illustrated for simplicity and may not have necessarily been drawn to scale. For example, the flow charts illustrate the method in terms of the most prominent steps involved to help to improve understanding of aspects of the present disclosure. Furthermore, in terms of the construction of the device, one or more components of the device may have been represented in the drawings by conventional symbols, and the drawings may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the drawings with details that will be readily apparent to those of ordinary skill in the art having the benefit of the description herein.

[0024] For the purpose of promoting an understanding of the principles of the present disclosure, reference will now be made to the various embodiments and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the present disclosure is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the present disclosure as illustrated therein being contemplated as would normally occur to one skilled in the art to which the present disclosure relates.

[0025] All of the embodiments of the disclosure described herein are example embodiments, and thus, the disclosure is not limited thereto, and may be realized in various other forms. Each of the embodiments provided in the following description is not excluded from being associated with one or more features of another example or another embodiment also provided herein or not provided herein but consistent with the disclosure.

[0026] It will be understood that when an element, component, layer, pattern, structure, region, or so on (hereinafter collectively "element") of a semiconductor device is referred to as being "over," "above," "on," "below," "under," "beneath," "connected to" or "coupled to" another element of the semiconductor device, it can be directly over, above, on, below, under, beneath, connected or coupled to the other element or an intervening element(s) may be present. In contrast, when an element of a semiconductor device is referred to as being "directly over," "directly above," "directly on," "directly below," "directly under," "directly beneath," "directly connected to" or "directly coupled to" another element of the semiconductor device, there are no intervening elements present.

[0027] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items. As used herein, expressions such as "at least one of," when preceding a list of elements, modify the entire list of elements and do not modify the individual elements of the list. For example, the expression, "at least one of a, b and c," and "at least one of a, b, or c" should be understood as including only a, only b, only c, both a and b, both a and c, both b and c, or all of a, b and c.

[0028] It will be also understood that, even if a certain step or operation of manufacturing an apparatus or structure is described later than another step or operation, the step or operation may be performed later than the other step or operation unless the other step or operation is described as being performed after the step or operation.

[0029] Various embodiments of the present document and terms used therein are not intended to limit the technical features described in this document to specific embodiments, and should be understood to include various modifications, equivalents, or substitutes of the corresponding embodiments.

[0030] In connection with the description of the drawings, similar reference numerals may be used for similar or related components.

[0031] The singular form of a noun corresponding to an item may include one or a plurality of the items unless clearly indicated otherwise in a related context.

[0032] The embodiments of the disclosure described in the present specification and the drawings are only presented as specific examples to easily explain the technical content according to the embodiments of the disclosure and help understanding of the embodiments of the disclosure, not intended to limit the scope of the embodiments of the disclosure. Therefore, the scope of one or more embodiments of the disclosure should be construed as encompassing all changes or modifications derived from the technical spirit of one or more embodiments of the disclosure in addition to the embodiments disclosed herein.

[0033] It will be understood that, although the terms "first", "second", "third", "primary", "secondary", "tertiary", etc., may be used herein to describe various elements, but elements are not limited by these terms. These terms are only used to distinguish one element from another element. For example, without departing from the scope of the disclosure, a first element may be termed as a second element, and a second element may be termed as a first element. The term of "and / or" includes a plurality of combinations of relevant items or any one item among a plurality of relevant items.

[0034] When an element (e.g., a first element) is referred to as being "(functionally or communicatively) coupled" or "connected" to another element (e.g., a second element), the first element may be connected to the second element, directly (e.g., wired), wirelessly, or through a third element.

[0035] In this disclosure, the terms "containing", "including", "comprising", "having", and the like are used to specify features, numbers, steps, operations, elements, components, or combinations thereof, but do not preclude the presence or addition of one or more of the features, numbers, steps, operations, elements, components, or combinations thereof.

[0036] In addition, in the present disclosure, the meaning of "identical" includes cases where properties are similar to each other or similar within a certain range. Furthermore, unless clearly indicated, stated, and / or shown otherwise; as used herein the terms "identical", "uniform", "equal", and / or "the same" mean "substantially identical", "substantially uniform", "substantially equal", "about the same", and / or "substantially the same". The meaning of substantially identical should be understood to include numerical values within manufacturing error ranges, machining or processing tolerances, and / or differences within a range that is so insignificant such that neither the structure nor function of the embodiments disclosed herein are materially altered, inhibited, or destroyed.

[0037] Furthermore, although one or more embodiments may comprise the disclosed features as described herein―as well as additional features not specifically described―other embodiments may instead be completely free of non-disclosed elements. For example, non-disclosed elements may be completely omitted from one or more embodiments of the present disclosure.

[0038] As used in connection with the disclosure, the terms "module" or "unit" may include a unit implemented in hardware, software, or firmware, and may interchangeably be used with other terms such as: portion, part, unit, member, logic, logic block, part, or circuitry. A module may be a single integral component, or a minimum unit or part thereof, adapted to perform one or more functions. For example, according to an embodiment, the module may be implemented in a form of an application-specific integrated circuit (ASIC). Depending on the embodiment(s), a plurality of "modules" may be implemented as a single element, or a single "module" may include a plurality of elements.

[0039] One or more embodiments as set forth herein may be implemented as software including one or more instructions that are stored in a storage medium that is readable by a machine. For example, a processor of the machine may invoke at least one of the one or more instructions stored in the storage medium, and execute it, with or without using one or more other components under the control of the processor. This allows the machine to be operated to perform at least one function according to the at least one instruction invoked. The one or more instructions may include a code generated by a complier or a code executable by an interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Wherein, the term "non-transitory" simply means that the storage medium is a tangible device, and does not include a signal (e.g., an electromagnetic wave), but this term does not differentiate between where data is semi-permanently stored in the storage medium and where the data is temporarily stored in the storage medium.

[0040] According to an embodiment, a method according to one or more embodiments of the disclosure may be included and provided in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read only memory (CD-ROM)), or be distributed (e.g., downloaded or uploaded) online via an application store (e.g., PlayStoreTM), or between two user devices (e.g., smart phones) directly. If distributed online, at least part of the computer program product may be temporarily generated or at least temporarily stored in the machine-readable storage medium, such as memory of the manufacturer's server, a server of the application store, or a relay server.

[0041] According to one or more embodiments, each component (e.g., a module or a program) of the above-described components may include a single entity or multiple entities. According to one or more embodiments, one or more of the above-described components may be omitted, or one or more other components may be added. Alternatively or additionally, a plurality of components (e.g., modules or programs) may be integrated into a single component. In such a case, according to one or more embodiments, the integrated component may still perform one or more functions of each of the plurality of components in the same or similar manner as they are performed by a corresponding one of the plurality of components before the integration. According to one or more embodiments, operations performed by the module, the program, or another component may be carried out sequentially, in parallel, repeatedly, or heuristically, or one or more of the operations may be executed in a different order or omitted, or one or more other operations may be added.

[0042] According to one or more embodiments, in a non-volatile storage medium storing instructions, the instructions may be configured to, when executed by at least one processor, cause the at least one processor to perform at least one operation. The at least one operation may include displaying an application screen of a running application on a display, identifying a data input field included in the application screen, identifying a data type corresponding to the data input field, displaying at least one external electronic device, around the electronic device, capable of providing data corresponding to the identified data type, receiving data corresponding to the identified data type from an external electronic device selected from among the at least one external electronic device through a communication module, and entering the received data into the data input field.

[0043] According to an aspect of the present disclosure, techniques are provided for predicting Channel State Information (CSI) associated with different antenna port configurations in a wireless communication system. A base station (gNB) may configure a full Channel State Information Reference Signal (CSI-RS) configuration corresponding to a full set of antenna ports, and one or more sub CSI-RS configurations corresponding to reduced antenna-port subsets. When the network deactivates antenna elements or transmission reception points (TRPs) as part of energy-saving procedures, a user equipment (UE) may only be able to measure CSI for a sub configuration, even though the network may require CSI for the full configuration to perform accurate beamforming, link adaptation, or scheduling. Conversely, there are scenarios in which the UE measures full-configuration CSI, while the network requires sub-configuration CSI to support energy-efficient down-sizing operations and / or to reduce the feedback control information.

[0044] According to one or more embodiments of the present disclosure, methods are provided for predicting a CSI report corresponding to a larger (full) antenna or port configuration based on a measurement corresponding to a smaller (sub) antenna or port configuration. Based on the received full and sub CSI-RS configurations, the UE determines whether the relationship between the configurations corresponds to a nested CSI configuration, wherein the beams and angular characteristics are aligned, or a non-nested CSI configuration, wherein the angular directions or antenna layouts differ. Depending on this determination, the UE selects either a non-Artificial Intelligence / Machine Learning (non-AI / ML) prediction technique, such as principal component analysis (PCA), singular value decomposition (SVD), or autoregressive moving-average (ARMA) modeling for nested cases, or an AI / ML-based technique for non-nested scenarios to infer the full-configuration CSI from the measured sub-configuration CSI.

[0045] One or more embodiments of the present disclosure may provide methods for predicting a CSI report corresponding to a smaller (sub) antenna port configuration based on a measurement corresponding to a larger (full) antenna port configuration. To perform this prediction, the UE evaluates one or more performance or structure metrics, including an eigenvalue distribution of the full-configuration channel matrix, its rank or sparsity, a signal-to-interference-plus-noise ratio (SINR), and expected throughput or spectral efficiency. The UE may also compute similarity or correlation indices, such as cosine similarity, angular similarity, or squared generalized cosine similarity (SCGS), to identify an optimal sub configuration and to predict the corresponding sub-configuration CSI. These techniques enable the network to transition to energy-saving sub-configurations while still obtaining accurate CSI reports.

[0046] One or more embodiments of the present disclosure may provide methods for predicting CSI for either smaller or larger configurations using channel- or beam-domain characteristics derived from an actual measurement performed with the complementary configuration. The prediction process may utilize features derived from a channel matrix, an eigen-beamforming matrix (e.g., a right eigen matrix V obtained via SVD), or quantized precoding matrices, including wide-band precoding matrices (W1) and / or sub-band precoding matrices (W2). Additional auxiliary information―such as varying channel conditions, electrical tilts, beam relations between full and sub configurations, antenna-panel geometry, or UE positional effects (e.g., near-field versus far-field behavior) may also be incorporated to improve reliability and offset inherent correlation degradations.

[0047] One or more embodiments of the present disclosure may provide a method to predict and reduce inter-beam interference when operating with a sub configuration of antenna ports. Based on predicted CSI and interference characteristics, the UE may avoid particular sub configurations that result in excessive inter-beam interference, thereby mitigating performance degradation while still enabling energy-saving operation. The UE may also periodically measure full-configuration CSI to refine model parameters, update correlation mappings, and maintain prediction accuracy under dynamic channel conditions.

[0048] In a still further aspect, the disclosed techniques are applicable to multi-TRP scenarios, wherein the UE predicts a full-configuration CSI matrices corresponding to multiple TRPs based on sub-configuration measurements, or based on a subset of TRPs chosen using spatial-relation criteria. The prediction methods also support transitions between near-field and far-field regimes, taking into account the UE's varying channel experience when antenna elements or TRPs are selectively activated or deactivated.

[0049] FIG. 1A illustrates an example configuration 100A in which a network operates with a smaller number of active antenna ports or antenna elements at a transmission side (Tx), in accordance with related art. Such a configuration is typically employed in scenarios where the number of active UEs in a cell is low and / or where radio channel conditions experienced by those UEs are generally favorable. Under these conditions, the network may utilize broader beams, since interference between beams is less of a concern, and high-resolution spatial separation is not required.

[0050] In this state, the underlying channel corresponding to a larger antenna array can be represented as Eq. 1

[0051] (1)

[0052] Similarly, the channel corresponding to a smaller antenna array can be written as Eq. 2

[0053] (2)

[0054] Here, denotes the number of layers, and regardless of whether a full or sub configuration is used, the number of layers is bounded by the number of receive antennas at the UE and the multipath richness of the channel.When the UE is stationary or near-stationary, the matrices and tend to be aligned in similar angular directions, reflecting similar dominant spatial signatures. In mobility scenarios, this directional relationship may evolve over time and can be updated using a Doppler-dependent mapping (Eq. 3):

[0055] (3)

[0056] where denotes the Doppler and fun1, fun2represent mapping functions between configurations.

[0057] FIG. 1B illustrates an example configuration 100B in which the network activates a larger number of antenna ports or antenna elements at the Tx side. This configuration may be required when the number of UEs increases, and the corresponding beams must be made narrower to mitigate inter-beam interference between UEs. Additionally, a network may activate more antenna elements when a longer coverage range or enhanced beamforming gain is needed, even if the number of UEs is not large. Narrower beams enable better spatial resolution, but require more antenna elements to form them effectively.

[0058] In practical deployments, the relationship between a full and a sub configuration may follow a nested CSI structure, i.e., antenna elements remain substantially uniformly spaced, with both polarizations preserved, and smaller configurations are subsets of the larger configuration. Table 1 exemplifies such nested CSI behavior, where reducing the number of CSI-RS antenna ports still maintains the overall beam orientation and structure.

[0059]

[0060] Table 1

[0061] However, not all antenna-port reductions follow a nested relationship. Non-nested CSI relationships arise when:

[0062] the antenna layout enforces single polarization in the smaller configuration (Table 2),

[0063]

[0064] Table 2

[0065] the antenna elements exhibit non-uniform spacing or a multi-panel architecture (Table 3), or electrical tilts differ between the full and sub configurations.

[0066]

[0067] Table 3

[0068] In these scenarios, the spatial beams corresponding to the full and sub configurations may not be aligned, and the behavior of the reduced-port beams may differ significantly from the full configuration. Improper reduction of antenna elements, especially without preserving polarization, tilt, or spacing, can result in broader beams, thereby increasing inter-beam interference, degrading CSI accuracy, and reducing overall system performance.

[0069] FIG. 2 illustrates a relationship between a number of active antenna elements and a resulting beam width in a wireless communication system, in accordance with related art. FIG. 2 shows representative radiation patterns for arrays composed of one half-wavelength dipole, two half-wavelength dipoles, four half-wavelength dipoles, and eight half-wavelength dipoles, respectively. As the number of antenna elements increases, the main-lobe beam width becomes progressively narrower, and the array gain increases correspondingly, as indicated by the gain values shown alongside each pattern.

[0070] For example, a single half-wavelength dipole exhibits a broad beam pattern with a beam width of approximately 78 degrees and low directivity. Increasing to two dipoles reduces the beam width to approximately 32 degrees, while four dipoles further reduce it to approximately 15 degrees. With eight dipoles, the beam width becomes very narrow, approximately 7 degrees, and the corresponding gain significantly increases. This behavior reflects the fundamental principle that larger antenna arrays enable narrower beams, higher spatial resolution, and better interference management.

[0071] However, when the network reduces the number of activated antenna elements for energy-saving purposes, the beam width becomes broader. A broader beam covers a larger angular region, which can result in inter-beam interference, particularly when multiple beams or multiple UEs must be spatially separated. Furthermore, the beam relationships between a full antenna configuration and a reduced-element configuration may deviate significantly if the electrical tilt applied to the full configuration is not preserved in the smaller configuration. Misalignment of tilts or non-uniform antenna element reduction can degrade beam fidelity, weaken directional mapping between configurations, and increase the level of unintended interference. As a result, reducing antenna elements without respecting the underlying beam structure may lead to noticeable performance loss.

[0072] FIG. 3A illustrates an example scenario 300A of a network operating with a large antenna configuration at a transmission side 302, in accordance with related art. In such a configuration, a greater number of antenna elements are active, which enables the formation of narrow and well-focused beams for transmitting different Channel State Information Reference Signal (CSI-RS) beams, shown as CSI-RS-1 and CSI-RS-32. Because the beams produced by the large antenna array are highly directional, the angular separation between neighboring beams is sufficiently large. As a result, the beams do not overlap in their main-lobe or side-lobe regions, resulting in no inter-beam interference at the UE. This stable angular relationship allows accurate CSI measurement for each beam and maintains consistent downlink performance even when multiple beams are transmitted simultaneously.

[0073] FIG. 3B illustrates an example scenario 300B the network operating with a smaller antenna configuration at the transmission side 302, in accordance with related art. Because fewer antenna elements are active, the resulting beams are broader, and the spatial resolution of the antenna array is reduced. In this configuration, CSI-RS-1 and CSI-RS-32 exhibit significant beam overlap, and the network cannot create the narrow, high-resolution beams that are possible with the larger antenna configuration. This overlap results in inter-beam interference, where signals intended for one CSI-RS direction spill over into adjacent beams. Such interference degrades the UE's ability to distinguish the CSI associated with each beam and increases measurement error, ultimately reducing the performance and reliability of the downlink channel. This behavior also reflects the risk associated with reducing antenna elements for energy-saving purposes without preserving beam directionality or electrical tilt relationships.

[0074] FIG. 4 illustrates an example system 400 implemented at a base station 412 for enabling intelligent prediction of Channel State Information (CSI) associated with various antenna port configurations, in accordance with one or more embodiments of the present disclosure. The base station 412 includes an N port antenna structure 402, which may physically support a large number of antenna ports, such as 256 ports. For energy saving, the base station 412 may activate only a smaller set of ports and may, for example, turn on only 64 ports out of the 256 ports. Accordingly, the base station 412 transmits M port CSI-RS 404 based on the currently active port set.

[0075] To maintain high quality CSI reporting even when operating with a reduced number of active antenna ports, the base station 412 may utilizes an AI based M to N CSI prediction module 406 (hereinafter referred to as the module 406). The module 406 forms part of an intelligent system trained using a set of channel measurements corresponding to the sub CSI-RS configurations and the full CSI-RS configuration for multiple angular resolutions. The training process may incorporate various forms of auxiliary information, including varying channel conditions, electrical tilts applied to antenna elements, beam relationships between full and sub configurations, and the actual antenna layout for both full and sub configurations. By including these inputs during training, the module 406 learns a robust and accurate mapping between the CSI associated with full and sub configurations.

[0076] Once trained, the module 406 receives, from a user equipment (UE) 414, a measured CSI report corresponding to the currently active sub configuration. Using the learned mapping, the module 406 may predict CSI corresponding to a required configuration size, including the full configuration, larger sub configurations, or smaller sub configurations. In selecting the appropriate mapping, the system 400 may identify whether the relationship between the full and sub configurations corresponds to a nested CSI configuration, where angular directions of the beams remain aligned, or a non-nested CSI configuration, where differences in tilt, antenna layout, or multi panel architecture create different directional behaviors. In the nested case, non AI techniques such as principal component analysis or singular value decomposition may be applied. In the non-nested case, AI or machine learning techniques executed by the module 406 provide improved prediction accuracy.

[0077] The base station 412 may also receives N port CSI-RS feedback 408 from the UE 414 based on the measurement on M port CSI-RS with module 406 being embedded in UE 414, where the UE reports CSI corresponding to a selected configuration size based on its measurements and prediction operations. The predicted CSI generated by the module 406 when present in base station 412 or reported CSI based on N ports when the module 406 is present in UE 414 is then used to support physical downlink shared channel (PDSCH) transmission 410, enabling the base station 412 to schedule downlink transmissions, perform beamforming, and manage interference while benefiting from reduced antenna activation for energy saving. In one embodiment, the UE 414 may implement one or more AI-ML modules to generate the N port CSI-RS feedback 408.

[0078] As shown in FIG. 4, the intelligent system implemented at the base station 412 provides a unified framework for reconstructing CSI corresponding to arbitrary antenna port configurations by utilizing a combination of full and sub configuration training data, auxiliary information, and learned mappings. This allows the network to adapt CSI reporting and transmission operations dynamically, even when antenna resources are selectively deactivated for energy efficiency. At the same time, accurate prediction of CSI across configurations ensures reliable downlink performance and supports advanced scheduling and beamforming strategies.

[0079] FIG. 5 illustrates an example flowchart of a method 500 for generating a required sub-configuration CSI from a full-configuration CSI, in accordance with one or more embodiments of the present disclosure. The method 500 may be performed by the UE 414. The method 500 begins at an input block 502, where the UE 414 receives a full CSI-RS configuration, an angular direction , a list of available sub CSI-RS configurations, and the full-configuration channel matrix H_f. The channel matrix H_f may include information such as eigenvectors, eigenvalues, and matrix rank, which are used to determine the nature of the CSI relationship.

[0080] At decision block 504, the UE 414 may evaluate whether the relationship between the full CSI-RS configuration and a selected sub CSI-RS configuration corresponds to a nested CSI configuration. If the full and sub configurations are aligned in terms of beam directionality, antenna spacing, or angular structure, then nested behavior is identified. If these directional or structural properties do not match, the UE 414 determines that the configuration relationship is non nested.

[0081] When the configuration relationship is identified as nested, the method 500 proceeds to block 506, where the module 406 determines an appropriate target size H_Si for the sub-configuration CSI representation. The target size may be selected as a function of one or more performance or structure metrics, including an L0, L1 norm of the eigenvalue matrix, expected throughput or spectral efficiency, signal to interference plus noise ratio (SINR), or an inter-beam interference estimate. Once an appropriate size is selected, the UE 414 may perform principal component analysis (PCA) on the full-configuration channel matrix, as illustrated by block 508. PCA reduces the dimensionality of the input matrix by identifying principal components that capture the dominant variations of the channel, allowing the sub-configuration channel matrix to be derived while preserving essential information.

[0082] If the configuration relationship is determined to be non-nested at decision block 504, the method 500 proceeds to block 510, where a trained neural network model is used to map the full-configuration channel matrix H_f and the angular direction to a predicted sub-configuration CSI H_Si. The neural network may be trained using multiple pairs of full and sub configuration channel matrices and may incorporate additional information such as electrical tilts, beam relationships between full and sub configurations, and antenna layout to improve prediction accuracy.

[0083] Following either technique, the resulting output is passed to block 512, which identifies the best / optimal CSI-RS sub configuration that satisfies the network requirements based on predicted performance, interference behavior, and CSI representation quality.

[0084] The flow illustrated in FIG. 5 therefore provides a unified framework within the present disclosure for deriving sub-configuration CSI from full-configuration measurements using either a trained AI or machine learning technique or a non AI or non ML technique. This improves the accuracy of CSI reporting while supporting network energy-saving strategies that involve changing the number of active antenna ports.

[0085] FIG. 6 illustrates a flowchart of a method 600 for generating a full-configuration CSI from a sub-configuration CSI, in accordance with one or more embodiments of the present disclosure. The flow begins at block 602, where a network configured full CSI-RS configuration is defined. The base station 412 may dynamically deactivate certain antenna ports for energy-saving purposes, as indicated at block 604, which may include turning off selected antenna elements or panels.

[0086] At block 608, the module 406 may estimate a sub-configuration channel matrix H_Si. The estimated channel matrix may include eigenvectors, eigenvalues, rank information, a signal to interference plus noise ratio, and an angular direction associated with the received CSI-RS. This measurement may reflect the CSI available for the reduced set of active antenna ports.

[0087] At decision block 606, the module 406 may evaluate whether the relationship between the sub-configuration and the full configuration corresponds to a nested CSI configuration. A nested configuration may be identified when the sub-configuration maintains alignment of beam directions, antenna spacing, or polarization relative to the full configuration. If the configuration is nested, the method 600 may proceed to block 612, where a non AI or non machine-learning prediction technique, such as an autoregressive moving average (ARMA) technique, is used to derive the full-configuration CSI from the measured sub-configuration CSI.

[0088] If the configuration is determined to be non-nested, the method 600 may proceed to block 610, where a trained neural network model is used. At block 610, the trained neural network may take the sub-configuration channel matrix H_Si and the angular direction as inputs and generate the predicted full-configuration CSI H_f. The neural network model may be trained using multiple sub and full configuration channel matrix pairs at different angular positions, allowing the model to learn the relationship between the configurations even in cases where the beams differ due to tilt, array geometry, or multiple antenna panels.

[0089] Upon completing either the non AI or non machine-learning technique or the artificial-intelligence or machine-learning technique, the system obtains the predicted full-configuration CSI required by the network for scheduling, beamforming, or link-adaptation processes.

[0090] FIG. 7 illustrates an example scenario 700 demonstrating channel propagation characteristics experienced by a user equipment (UE) transition between near-field and far-field regions, in accordance with one or more embodiments of the present disclosure. FIG. 7 illustrates that such transition affects the modeling and prediction of CSI for different antenna configurations. The illustration shows that, when the UE is located within the near-field region, the electromagnetic wavefront incident on the antenna array exhibits spherical-wave behavior. In this region, signals arriving across the antenna elements do not share a uniform phase front, and the angular characteristics observed by different antenna ports vary non-linearly with distance. As a result, the mapping between sub-configuration CSI and full-configuration CSI must take into account position-dependent curvature of the wavefront.

[0091] As the UE moves farther from the antenna array and crosses the Rayleigh distance, indicated by , the propagation transitions into the far-field region, where the wavefront approximates a planar wave. In the far-field region, the spatial characteristics become more stable, and the angular directionality across antenna elements becomes consistent. This produces behavior similar to the uniformly distributed beam directions illustrated in the upper portion of the figure. The planar-wave assumptions allow the use of simplified angular relationships between full and sub CSI-RS configurations.

[0092] In the near-field region, the angular characteristics spanned by the UE may be spread broadly, and individual antenna elements observe distinctly different spatial signatures. In contrast, in the far-field region, the angular characteristics may be tightly aligned, allowing the beams of the antenna array to follow nearly parallel directions.

[0093] The illustrated scenario 700 indicates an importance of UE presence in near-field or far-field for CSI prediction because the selection of sub-configuration and full-configuration CSI mapping techniques depends on whether the UE operates in the near-field or the far-field. When antenna ports are selectively turned off for energy saving, the UE may move effectively into a different field regime depending on its physical distance and the number of active antenna elements. Therefore, the prediction framework of the present disclosure incorporates models that handle both spherical-wave and planar-wave behaviors, ensuring that the mapping between sub-configuration CSI and full-configuration CSI remains accurate across variable deployment conditions.

[0094] FIG. 8 illustrates an example scenario 800 demonstrating a training phase of an artificial intelligence or machine learning model for predicting CSI across different antenna port configurations, in accordance with one or more embodiments of the present disclosure. As shown, a base station transmits CSI-RS beams associated with sub-configuration CSI matrices and , labeled as CSI-RS-1 and CSI-RS-32, respectively. Each of these sub-configuration CSI matrices corresponds to a different angular direction, represented as and .

[0095] During the training process, the model receives sets of sub-configuration CSI matrices { } and { }as input. These sets may include several variations or specific sub-configurations taken under the same angular direction. Corresponding to each input set, the model uses the full-configuration CSI matrices and as training labels, which represent the spatial channel characteristics of the full antenna port configuration at angular directions and . FIG. 8 visually captures this relationship by showing the sub-configuration beams aligned with the larger full-configuration beams for each angular direction.

[0096] The model therefore learns a mapping between each sub-configuration CSI beam pattern and its corresponding full-configuration CSI for a given angular position. By training on multiple combinations of { } and across several angular directions, the model develops the capability to infer the full-configuration CSI from sub-configuration measurements during the operational (testing) phase. This mapping may be strengthened using auxiliary information such as electrical tilt, beam relationships between full and sub configurations, antenna layout, or other deployment-specific parameters, although these details are abstracted from FIG. 8.

[0097] Accordingly, FIG. 8 represents the principle of training a machine learning model where sub-configuration CSI inputs are paired with full-configuration CSI labels across angular directions, enabling the model to derive accurate full-configuration CSI during real-time inference.

[0098] FIG. 9 illustrates an example scenario 900 demonstrating the training of a full-configuration CSI matrix for a particular angular direction using multiple sub-configuration CSI matrices obtained for that same angular direction, in accordance with one or more embodiments of the present disclosure. As shown, the base station (for example, the base station 412) transmits a CSI-RS beam labeled CSI-RS-1, from which the user equipment derives a set of sub-configuration CSI matrices for a given direction . These matrices are shown as and , which are specific sub-configurations belonging to the larger set { }. The corresponding full-configuration CSI matrix for this angular direction is shown as .

[0099] During the training phase, the model receives the sub-configuration set { } as the input space, and the associated full-configuration CSI as the training label for the angular position . FIG. 9 visually captures how multiple sub-configuration CSI beams combine to represent different spatial resolutions or port selections corresponding to the same angular direction. These sub-configuration beams collectively contribute to learning the mapping that reconstructs the full-configuration CSI.

[0100] For nested CSI relationships, the training framework may apply dimensionality-reduction techniques such as principal component analysis. Such techniques reduce the dimensionality of the full-configuration CSI while retaining principal components associated with the strongest eigenvalues, allowing the system to select the sub-configuration with the most representative eigenvalue structure for angular direction . For non-nested CSI relationships, one or more neural network models may be employed to map the full-configuration matrix to the best matching sub-configuration CSI from the set { }, using learned relationships between the channel patterns.

[0101] During the operational or testing phase, the model can predict a specific sub-configuration CSI from the corresponding full-configuration CSI for the same angular direction. The decision to choose the target sub-configuration may be based on eigenvalue distribution, matrix rank, signal to interference plus noise ratio, and expected throughput or spectral efficiency. The minimum allowable size of a sub-configuration may be limited by beam width and inter-beam interference, ensuring that performance is not degraded. The mapping between full and sub CSI may be periodically updated using refreshed full-configuration measurements to maintain prediction accuracy.

[0102] Accordingly, FIG. 9 reflects the principle that multiple sub-configuration CSI matrices measured at a given angular direction are used together to train a model to reconstruct the corresponding full-configuration CSI, enabling accurate prediction of CSI under different antenna port configurations.

[0103] FIG. 10 illustrates an example scenario 1000 demonstrating a correspondence of a sub-configuration CSI matrix with multiple full-configuration CSI matrices at different angular directions during the sub-to-full CSI prediction process, in accordance with one or more embodiments of the present disclosure. As shown, the base station (for example, the base station 414) transmits a CSI-RS beam labeled CSI-RS-1, from which the user equipment derives a sub-configuration CSI matrix denoted as . This sub-configuration is associated with a particular spatial region observed by the antenna array.

[0104] In the illustrated scenario, the same sub-configuration CSI matrix corresponds to different full-configuration CSI matrices, such as at angular direction and at angular direction . This relationship arises because, for a given sub-configuration, the full-configuration CSI may vary depending on the exact angular direction of the incident signal, even though the UE measures the same sub-configuration structure. FIG. 10 visually demonstrates that a single sub-configuration beam pattern may overlap with or map into multiple full-configuration beam patterns depending on the angle at which the UE observes the CSI-RS.

[0105] During the training phase, the model therefore receives the sub-configuration CSI matrix as input for each training sample, and associates it with the corresponding full-configuration CSI matrices and for directions and . The training process enables the model to learn that a particular sub-configuration may map to different full-configuration channel matrices depending on angular context. This relationship forms an essential part of the learning mechanism used for sub-to-full CSI prediction.

[0106] During the testing phase, the UE (for example, the UE 414) uses the measured sub-configuration CSI matrix along with angular direction information , when available, to derive the appropriate full-configuration CSI. The angular direction helps the model select the correct mapping among potentially multiple full-CSI representations corresponding to the same sub-configuration. When angular information is unavailable, the model may use learned statistical relationships between the sub- and full-configuration CSI patterns.

[0107] To maintain accurate mapping over time, the UE may periodically measure the full-configuration CSI and provide updated information to the system, allowing the model to fine tune the mapping between full and sub configurations. This is important when antenna ports are dynamically turned off or when mobility causes changes in angular direction, Doppler conditions, or near-field to far-field transitions.

[0108] Accordingly, FIG. 10 captures the principle that a sub-configuration CSI matrix may correspond to several possible full-configuration CSI matrices, and the model must be trained to correctly associate each sub-configuration with the appropriate full-configuration CSI based on angular direction and other contextual factors.

[0109] FIG. 11 illustrates a flow of a method 1100 for predicting CSI to support network energy saving, in accordance with an embodiment of the present disclosure. The method 1100 may be performed by the UE 414.

[0110] At step 1102, the method 1100 includes receiving, from the base station 412, a full CSI-RS configuration indicating a first set of antenna ports and a list of sub CSI-RS configurations indicating a second set of antenna ports that is a smaller subset of the first set. In some embodiments, the full CSI-RS configuration is delivered in a Radio Resource Control (RRC) reconfiguration message received at the UE 414. The list of sub CSI-RS configurations may enumerate multiple candidate sub configurations available for CSI measurement and reporting.

[0111] At step 1104, the method 1100 includes determining whether a relationship between the full CSI-RS configuration and a selected sub CSI-RS configuration corresponds to a nested CSI configuration or a non-nested CSI configuration. In a nested CSI configuration, channels corresponding to the full configuration and the sub configuration are aligned and have similar angular direction, and inter-antenna element distance remains the same so that beams observed for the full and sub configurations are in similar directions (e.g., the antenna elements may transmit the beams in substantially the same direction, or "codirectionally"). In a non-nested CSI configuration, angular directions of beams for the full and sub configurations may not be oriented in the same direction and may have no well-defined relationship, for example due to differences in electrical tilts between the configurations or a multi-panel architecture with non-uniform distances between antenna elements in the sub configuration.

[0112] At step 1106, the method 1100 includes selecting, based on the relationship determined at step 1104, a prediction technique from a plurality of prediction techniques. The plurality includes at least one artificial intelligence or machine learning (AI / ML) technique and at least one non AI or non machine-learning technique. In some embodiments, the selected prediction technique operates on quantized precoding representations, such as wide-band precoding matrices or sub-band precoding matrices, in lieu of or in addition to raw channel matrices. The prediction technique may further utilize auxiliary information, such as varying channel conditions, electrical tilts, beam relations between full and sub configurations, and actual antenna layout for both configurations.

[0113] At step 1108, the method 1100 includes predicting, using the selected prediction technique, one of:

[0114] (i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI corresponding to a selected sub CSI-RS configuration, or

[0115] (ii) a sub CSI corresponding to a selected sub CSI-RS configuration based on a measured full CSI corresponding to the full CSI-RS configuration.

[0116] In some embodiments, the method 1100 may include selecting a target size for a CSI representation to be predicted. The target size is selected based on one or more performance or structure metrics, including at least one of an eigenvalue distribution of a covariance representation, a sparsity measure such as L0 or L1 norm, a rank of a full configuration matrix, a signal to interference plus noise ratio (SINR), an expected throughput or spectral efficiency, and an estimate of inter-beam interference. In some embodiments, a minimum size of the sub CSI-RS configuration is constrained by beam width and inter-beam interference, ensuring that reduced configurations do not induce unacceptable performance loss. In one embodiment, the method 1100 may include predicting the sub CSI based on the selected target size.

[0117] In some embodiments where the relationship is nested, the non AI or non machine-learning technique includes a dimensionality-reduction technique such as Principal Component Analysis (PCA), Singular Value Decomposition (SVD), or an Autoregressive Moving Average (ARMA) technique to derive the predicted CSI. For example, PCA may transform an input channel matrix of the full configuration into principal components to derive a channel matrix of the sub configuration, where features that are uncorrelated and have variance above a threshold are retained as principal components, and where PCA is carried out by calculating eigenvectors and eigenvalues from a covariance matrix or through SVD.

[0118] In some embodiments where the relationship is non nested, the AI / ML technique includes using a trained neural network model to map a measured CSI corresponding to one configuration to a predicted CSI corresponding to another configuration. Training can include using different sub configuration channel matrices as input and corresponding full configuration channel matrices as labels for different angular positions, then using the trained model to map a sub configuration channel matrix to a full configuration channel matrix for a given angular direction.

[0119] In further embodiments where the method predicts sub CSI from full CSI, the UE 414 calculates a similarity or correlation index between the full configuration matrix and each sub configuration matrix from the list of sub CSI-RS configurations, and selects the sub configuration corresponding to an optimal index for reporting. The similarity or correlation index may include cosine similarity, angular similarity, Block Error Rate (BLER), Spectral Efficiency (SE), or Squared Generalized Cosine Similarity (SCGS).

[0120] In multi-site deployments, the predicting can include generating a full configuration channel matrices for multiple Transmission and Reception Points (TRPs), where each predicted matrix corresponds to a respective TRP. The prediction may be based on a corresponding set of sub configurations for each TRP, or on a subset of sub configurations where sub configurations from a subset of TRPs are considered according to spatial relations.

[0121] At step 1110, the method 1100 includes transmitting, to the base station 412, a CSI report based on the predicted CSI determined at step 1110. The CSI report may be provided for the target size selected at step 1108 and may be used by the base station 412 for beamforming, link adaptation, scheduling, and energy-saving control.

[0122] In some embodiments, the method 1100 further includes periodically performing full configuration measurements at the UE 414 to fine tune a learned mapping between full and sub configurations or to update parameters of a trained model, thereby maintaining prediction accuracy under mobility, tilt changes, and dynamic antenna activation.

[0123] FIG. 12 illustrates an example system 1200 for predicting CSI in a wireless communication system, in accordance with one or more embodiments of the present disclosure. The system 1200 may correspond to the UE 414 or to any other network equipment, such as the base station 412, as discussed throughout this disclosure.

[0124] The system 1200 includes at least one processor 1202, a memory unit 1204, and a communication unit 1206. The communication unit 1206 operates as a transceiver and is configured to perform one or more functions associated with transmitting and receiving signals through a wireless channel, including receipt of CSI-RS configurations, transmission of CSI reports, and exchange of control or scheduling information.

[0125] The processor 1202 may be implemented using a single processing element or a combination of multiple processing elements. The processor 1202 may include one or more microprocessors, microcontrollers, microcomputers, central processing units, digital signal processors, state machines, logic circuitry, graphics processing units (GPUs), visual processing units (VPUs), or artificial intelligence-dedicated processors such as neural processing units (NPUs). The processor 1202 is configured to fetch, decode, and execute computer-readable instructions stored in the memory unit 1204, including instructions that define operating rules or instructions that implement an artificial intelligence or machine learning model used for CSI prediction. The artificial intelligence or machine learning model may be stored in the memory unit 1204 as a trained model obtained through supervised or unsupervised learning.

[0126] The memory unit 1204 may include any suitable non-transitory computer-readable medium, such as static random access memory (SRAM), dynamic random access memory (DRAM), read-only memory (ROM), erasable programmable read-only memory, flash storage, solid-state drives, magnetic storage, or optical storage. The memory unit 1204 may store channel measurements, precoding matrices, neural-network parameters, correlation metrics, and executable instructions necessary for performing CSI prediction under nested or non-nested CSI configurations.

[0127] In operation, the system 1200 enables execution of functions described in method 1100, including receiving full and sub CSI-RS configurations, determining configuration relationships, selecting a prediction technique, selecting a target CSI representation size, predicting full or sub CSI using one of the techniques, and transmitting the predicted CSI through the communication unit 1206. Accordingly, the system 1200 provides the hardware architecture for supporting CSI prediction-based energy-saving operations in a wireless communication system.

[0128] Accordingly, the present disclosure provides methods and systems for predicting full or sub Channel State Information (CSI) based on channel measurements obtained under sub or full CSI-Reference Signal (CSI-RS) configurations. The objective of the present disclosure is to enable accurate reconstruction of CSI across antenna port configurations while supporting network energy-saving operations.

[0129] The present disclosure provides an intelligent system capable of predicting optimal CSI using a set of channel measurements corresponding to sub-configuration CSI-RS and full-configuration CSI-RS, obtained for each angular resolution. The system may be trained using varying channel conditions, electrical tilts, beam relations between full and sub configurations, and actual antenna layout for full and sub configurations. By incorporating these additional inputs, the trained system improves accuracy and robustness when predicting CSI corresponding to any target configuration. The trained model can be used to predict full-configuration or sub-configuration CSI using, as inputs, the measured CSI associated with the complementary configuration and, optionally, the angular direction.

[0130] The present disclosure also provides a similarity- or correlation-based prediction method for deriving full or sub CSI from measured CSI. This method may compute one or more similarity metrics, such as cosine similarity, angular similarity, block error rate, spectral efficiency, or squared generalized cosine similarity, to determine which sub or full configuration provides the optimal match. The similarity-based approach may also utilize additional inputs such as channel condition variation, electrical tilt differences, beam alignment characteristics, and antenna-panel geometry to offset similarity deviations and improve prediction fidelity. Further embodiments may employ the right eigen matrix, wide-band precoding matrices, or sub-band precoding matrices in place of or in addition to raw channel matrices to represent the CSI for prediction.

[0131] In a further aspect, the present disclosure extends these methods to a multi-Transmission Reception Point (multi-TRP) environment, wherein the system predicts a set of full or sub-configuration CSI matrices for multiple TRPs. Each predicted matrix may correspond to a full configuration, based on sub-configuration measurements obtained from each TRP or a subset of TRPs selected using spatial-relation criteria.

[0132] In one or more embodiments, the present disclosure may extend CSI prediction to near-field channel environments, enabling the system to adapt to scenarios in which the wavefront transitions between spherical-wave behavior in the near field and planar-wave behavior in the far field. This allows the prediction model to maintain accuracy even when the UE experiences different propagation regimes due to changes in antenna activation or proximity to the transmitting node.

[0133] Finally, the present disclosure provides that a user equipment (UE) may communicate its capability to support the disclosed CSI-prediction features to the network using any known signaling mechanism. This allows the network to determine whether the UE can perform prediction for full or sub configurations and to configure the UE accordingly for efficient CSI measurement and reporting in support of network energy-saving operations.

[0134] At least one of the plurality of modules may be implemented through an AI model. A function associated with AI may be performed through the non-volatile memory, the volatile memory, and the processor.

[0135] The processor may include one or a plurality of processors. At this time, one or a plurality of processors may be a general purpose processor, such as a central processing unit (CPU), an application processor (AP), or the like, a graphics-only processing unit such as a graphics processing unit (GPU), a visual processing unit (VPU), and / or an AI-dedicated processor such as a neural processing unit (NPU).

[0136] The one or a plurality of processors control the processing of the input data in accordance with a predefined operating rule or artificial intelligence (AI) model stored in the non-volatile memory and the volatile memory. The predefined operating rule or artificial intelligence model is provided through training or learning.

[0137] Here, being provided through learning means that, by applying a learning technique to a plurality of learning data, a predefined operating rule or AI model of a desired characteristic is made. The learning may be performed in a device itself in which AI according to an embodiment is performed, and / o may be implemented through a separate server / system.

[0138] The AI model may consist of a plurality of neural network layers. Each layer has a plurality of weight values, and performs a layer operation through calculation of a previous layer and an operation of a plurality of weights. Examples of neural networks include, but are not limited to, convolutional neural network (CNN), deep neural network (DNN), recurrent neural network (RNN), restricted Boltzmann Machine (RBM), deep belief network (DBN), bidirectional recurrent deep neural network (BRDNN), generative adversarial networks (GAN), and deep Q-networks.

[0139] The learning technique is a method for training a predetermined target device (for example, a robot) using a plurality of learning data to cause, allow, or control the target device to make a determination or prediction. Examples of learning techniques include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0140] According to the disclosure, in a method for predicting the CSI, the processor may perform a pre-processing operation on the data to convert into a form appropriate for use as an input for the artificial intelligence model. The artificial intelligence model may be obtained by training. Here, "obtained by training" means that a predefined operation rule or artificial intelligence model configured to perform a desired feature (or purpose) is obtained by training a basic artificial intelligence model with multiple pieces of training data by a training technique. The artificial intelligence model may include a plurality of neural network layers. Each of the plurality of neural network layers includes a plurality of weight values and performs neural network computation by computation between a result of computation by a previous layer and the plurality of weight values.

[0141] Reasoning prediction is a technique of logically reasoning and predicting by determining information and includes, e.g., knowledge-based reasoning, optimization prediction, preference-based planning, or recommendation.

[0142] One or more embodiments may provide a method, wherein, in the nested CSI configuration, channels corresponding to the full CSI-RS configuration or the list of sub CSI-RS configurations are substantially aligned with each other and have, respectively, substantially identical angular directions, wherein the wireless communication system comprises antenna elements that are substantially uniformly spaced apart from each other, and wherein the antenna elements are configured to transmit, substantially codirectionally, beams associated with the full CSI-RS configuration or the list of sub CSI-RS configurations.

[0143] One or more embodiments may provide a method, wherein, in the non-nested CSI configuration, at least one of: differences in respective electrical tilts of antenna elements between the full CSI-RS configuration and the sub CSI-RS configuration, or a multi-panel architecture with non-uniform distances between the antenna elements in the sub CSI-RS configuration; causes the antenna elements to transmit: first beams associated with the full CSI-RS configuration and having first angular directions; and second beams associated with the list of sub CSI configurations and having second angular directions, wherein the first angular directions are different from the second angular directions.

[0144] One or more embodiments may provide a method, wherein for the nested CSI configuration, the non-AI / ML based technique comprises using a dimensionality-reduction technique to derive the predicted CSI, the dimensionality-reduction technique comprising Principal Component Analysis (PCA), singular value decomposition (SVD), and / or an autoregressive moving average (ARMA) technique.

[0145] One or more embodiments may provide a method, further comprising: transforming, by performing PCA, an input channel matrix of the full CSI-RS configuration into principal components to derive a channel matrix of the sub CSI-RS configuration, wherein features which are uncorrelated and have variance above a threshold are retained as the principal components, and wherein performing the PCA comprises: calculating eigenvectors and eigenvalues from a covariance matrix, and / or performing SVD.

[0146] One or more embodiments may provide a method, wherein for the non-nested CSI configuration, the AI / ML based technique comprises using a trained neural network model to map, to a predicted CSI corresponding to a first configuration, a measured CSI corresponding to a second configuration.

[0147] One or more embodiments may provide a method, wherein predicting the full CSI comprises: training a model using: different sub-configuration channel matrices as input, and corresponding full-configuration channel matrices as labels for different angular positions; and using the trained model to map a sub-configuration channel matrix to the full-configuration channel matrix for a given angular direction.

[0148] One or more embodiments may provide a method, wherein predicting the sub CSI comprises: calculating a similarity or a correlation index between a full-configuration matrix and each sub-configuration matrix of the list of sub CSI-RS configurations; and selecting a sub-configuration corresponding to an optimal index for reporting, wherein the one or more performance or structure metrics comprise at least one of: cosine similarity, angular similarity, block error rate (BLER), spectral efficiency (SE), or squared generalized cosine similarity (SCGS).

[0149] One or more embodiments may provide a method, wherein the selecting the target size is based on at least one of: an eigenvalue distribution of a covariance representation, a sparsity measure including an L0, L1 norm, a rank of a full-configuration matrix, a signal-to-interference-plus-noise ratio (SINR), an expected throughput or expected spectral efficiency, and / or an estimate of inter-beam interference.

[0150] One or more embodiments may provide a method, wherein a minimum size of the sub CSI-RS configuration is limited by a beam width and inter-beam interference.

[0151] One or more embodiments may provide a method, wherein the selected prediction technique further comprises at least one of: quantized wide-band precoding matrices or sub-band precoding matrices.

[0152] One or more embodiments may provide a method, wherein the prediction technique further comprises, at least one of: varying channel conditions, electrical tilts, beam relations between the full CSI-RS configuration and the sub CSI-RS configuration, or actual antenna layout for the full CSI-RS configuration and the sub CSI-RS configuration.

[0153] One or more embodiments may provide a method, further comprising: tuning, by periodically performing full-configuration measurements, a mapping between the sub CSI-RS configuration and the full CSI-RS configuration; or updating, by periodically performing full-configuration measurements, parameters of a trained model.

[0154] One or more embodiments may provide a method, wherein the predicting the CSI comprises: predicting a set of full-configuration channel matrices, each of the set of full-configuration channel matrices corresponding to a full-configuration channel matrix from a respective Transmission and Reception Point (TRP) of a plurality of TRPs.

[0155] One or more embodiments may provide a method, wherein the predicting the set of full-configuration channel matrices: is based on a corresponding set of sub-configurations for each TRP of the plurality of TRPs, or is based on a subset of sub-configurations wherein sub-configurations from only a subset of TRPs of the plurality of TRPs are considered based on spatial relations.

[0156] One or more embodiments may provide a method, wherein the full CSI-RS configuration is received in a Radio Resource Control (RRC) reconfiguration message.

[0157] One or more embodiments may provide a method, further comprising: predicting, based on the selected target size, the sub CSI corresponding to the sub CSI-RS configuration based on the measured full CSI that corresponds to the full CSI-RS configuration.

[0158] While specific language has been used to describe the present subject matter, any limitations arising on account thereto, are not intended. As would be apparent to a person in the art, various working modifications may be made to the method in order to implement the inventive concept as taught herein. The drawings and the foregoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment.

Claims

1.A method performed by a user equipment (UE) in a wireless communication system for predicting channel state information (CSI) to support network energy saving, the method comprising:receiving, from a base station:a full CSI reference signal (CSI-RS) configuration indicating a first set of antenna ports; anda list of sub CSI-RS configurations indicating a second set of antenna ports, wherein the second set is a smaller size subset of the first set;determining whether a relationship between the full CSI-RS configuration and a sub CSI-RS configuration of the list of sub CSI-RS configurations corresponds to a nested CSI configuration or a non-nested CSI configuration;selecting, based on the relationship, a prediction technique from a plurality of prediction techniques, wherein the plurality of prediction techniques comprises an artificial intelligence / machine learning (AI / ML) based technique and a non-AI / ML based technique;selecting, for a CSI representation to be predicted based on one or more performance or structure metrics, a target size;predicting, using the selected prediction technique, a CSI comprising at least one of:(i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI that corresponds to the sub CSI-RS configuration, or(ii) a sub CSI corresponding to the sub CSI-RS configuration based on a measured full CSI that corresponds to the full CSI-RS configuration; andtransmitting, to the base station, a CSI report based on the predicted CSI.2.The method of claim 1, wherein, in the nested CSI configuration, channels corresponding to the full CSI-RS configuration or the list of sub CSI-RS configurations are substantially aligned with each other and have, respectively, substantially identical angular directions,wherein the wireless communication system comprises antenna elements that are substantially uniformly spaced apart from each other, andwherein the antenna elements are configured to transmit, substantially codirectionally, beams associated with the full CSI-RS configuration or the list of sub CSI-RS configurations.3.The method of claim 1, wherein, in the non-nested CSI configuration, at least one of:differences in respective electrical tilts of antenna elements between the full CSI-RS configuration and the sub CSI-RS configuration, ora multi-panel architecture with non-uniform distances between the antenna elements in the sub CSI-RS configuration;causes the antenna elements to transmit:first beams associated with the full CSI-RS configuration and having first angular directions; andsecond beams associated with the list of sub CSI configurations and having second angular directions,wherein the first angular directions are different from the second angular directions.4.The method of claim 1, wherein for the nested CSI configuration, the non-AI / ML based technique comprises using a dimensionality-reduction technique to derive the predicted CSI,the dimensionality-reduction technique comprising Principal Component Analysis (PCA), singular value decomposition (SVD), and / or an autoregressive moving average (ARMA) technique.5.The method of claim 4, further comprising:transforming, by performing PCA, an input channel matrix of the full CSI-RS configuration into principal components to derive a channel matrix of the sub CSI-RS configuration,wherein features which are uncorrelated and have variance above a threshold are retained as the principal components, andwherein performing the PCA comprises:calculating eigenvectors and eigenvalues from a covariance matrix, and / orperforming SVD.6.The method of claim 1, wherein for the non-nested CSI configuration, the AI / ML based technique comprises using a trained neural network model to map, to a predicted CSI corresponding to a first configuration, a measured CSI corresponding to a second configuration.7.The method of claim 6, wherein predicting the full CSI comprises:training a model using:different sub-configuration channel matrices as input, andcorresponding full-configuration channel matrices as labels for different angular positions; andusing the trained model to map a sub-configuration channel matrix to the full-configuration channel matrix for a given angular direction.8.The method of claim 1, wherein predicting the sub CSI comprises:calculating a similarity or a correlation index between a full-configuration matrix and each sub-configuration matrix of the list of sub CSI-RS configurations; andselecting a sub-configuration corresponding to an optimal index for reporting,wherein the one or more performance or structure metrics comprise at least one of: cosine similarity, angular similarity, block error rate (BLER), spectral efficiency (SE), or squared generalized cosine similarity (SCGS).9.The method of claim 1, wherein the selecting the target size is based on at least one of:an eigenvalue distribution of a covariance representation,a sparsity measure including an L0, L1 norm,a rank of a full-configuration matrix,a signal-to-interference-plus-noise ratio (SINR),an expected throughput or expected spectral efficiency,and / or an estimate of inter-beam interference.10.The method of claim 1, wherein a minimum size of the sub CSI-RS configuration is limited by a beam width and inter-beam interference.11.The method of claim 1, wherein the selected prediction technique further comprises at least one of: quantized wide-band precoding matrices or sub-band precoding matrices.12.The method of claim 1, wherein the prediction technique further comprises, at least one of: varying channel conditions, electrical tilts, beam relations between the full CSI-RS configuration and the sub CSI-RS configuration, or actual antenna layout for the full CSI-RS configuration and the sub CSI-RS configuration.13.The method of claim 1, further comprising:tuning, by periodically performing full-configuration measurements, a mapping between the sub CSI-RS configuration and the full CSI-RS configuration; orupdating, by periodically performing full-configuration measurements, parameters of a trained model.14.The method of claim 1, wherein the predicting the CSI comprises:predicting a set of full-configuration channel matrices, each of the set of full-configuration channel matrices corresponding to a full-configuration channel matrix from a respective Transmission and Reception Point (TRP) of a plurality of TRPs.15.The method of claim 14, wherein the predicting the set of full-configuration channel matrices:is based on a corresponding set of sub-configurations for each TRP of the plurality of TRPs, oris based on a subset of sub-configurations wherein sub-configurations from only a subset of TRPs of the plurality of TRPs are considered based on spatial relations.16.The method of claim 1, wherein the full CSI-RS configuration is received in a Radio Resource Control (RRC) reconfiguration message.17.The method of claim 1, further comprising:predicting, based on the selected target size, the sub CSI corresponding to the sub CSI-RS configuration based on the measured full CSI that corresponds to the full CSI-RS configuration.18.A user equipment (UE) in a wireless communication system, the UE comprising:at least one transceiver; andat least one processor communicatively coupled to the at least one transceiver, the at least one processor configured to:receive:a full channel state information (CSI) reference signal (CSI-RS) configuration indicating a first set of antenna ports; anda list of sub CSI-RS configurations indicating a second set of antenna ports, wherein the second set is a smaller subset of the first set;determine whether a relationship between the full CSI-RS configuration and a sub CSI-RS configuration corresponds to a nested CSI configuration or to a non-nested CSI configuration;select a prediction method from a plurality of prediction methods, the plurality of prediction methods comprising at least one artificial intelligence / machine learning (AI / ML)-based method and at least one non-AI / ML-based method;predict, using the selected prediction techniques, a CSI comprising at least one of:(i) a full CSI corresponding to the full CSI-RS configuration based on a measured sub CSI that corresponds to the sub CSI-RS configuration, or(ii) a sub CSI corresponding to the sub CSI-RS configuration based on a measured full CSI that corresponds to the full CSI-RS configuration; andtransmit a CSI report based on the predicted CSI.19.The UE of claim 18, wherein, in the nested CSI configuration:channels corresponding to the full CSI-RS configuration or the list of sub CSI-RS configurations are substantially aligned with each other and have, respectively, substantially identical angular directions; andwherein the UE comprises antenna elements configured to transmit, substantially codirectionally, beams associated with the full CSI-RS configuration or the list of sub CSI-RS configurations.20.The UE of claim 18, wherein, in the non-nested CSI configuration, at least one of differences in respective electrical tilts of antenna elements between the full CSI-RS configuration and the sub CSI-RS configuration, or a multi-panel architecture with non-uniform distances between the antenna elements in the sub CSI-RS configuration;causes the antenna elements to transmit:first beams associated with the full CSI-RS configuration and having first angular directions; andsecond beams associated with the list of sub CSI configurations and having second angular directions,wherein the first angular directions are different from the second angular directions.

Images