Codebook embedding generation and processing

Abstract

Various aspects of this disclosure relate to wireless communication systems, and in particular, to techniques for generating and processing embeddings representing beams used for communication. Certain aspects provide methods for wireless communication by a wireless node. These methods typically include: receiving an embedding representing a representation associated with a beam; providing the embedding to a machine learning (ML) model; generating one or more communication parameters for communicating using the beam based on the embedding via the ML model; and communicating using the one or more communication parameters.

Description

[0001] Cross-references to related applications

[0002] This application claims priority to U.S. Application No. 17 / 366,823, filed July 2, 2021, which has been assigned to the assignee of this application and whose entire contents are expressly incorporated herein by reference, as fully set forth below and for all applicable purposes. Technical Field

[0003] Various aspects of this disclosure relate to wireless communication systems, and in particular, to techniques for generating and processing embedded representations of beams used for communication. Background Technology

[0004] Compared to lower frequency communication, communication using "millimeter wave" ("mmW" or "mmWave") or near-mmWave radio frequency bands (e.g., 3 GHz–300 GHz) can have higher path loss and shorter range. Accordingly, communication between a base station and a user equipment (UE) can use beamforming to improve path loss and range. For this purpose, both the base station and the UE can each include multiple antennas, such as antenna elements, antenna panels, and / or antenna arrays, to facilitate beamforming.

[0005] Beams used for communication can be characterized using various techniques. One technique involves using an array gain sphere to represent the beam. While the array gain sphere provides detailed information about the beam, its use can be challenging when performing beam processing. Therefore, what is needed is a technique for representing beams using low-dimensional representations to facilitate efficient beam processing. Summary of the Invention

[0006] Some aspects provide methods for wireless communication by a wireless node. These methods typically include: receiving an embedding representing a representation associated with a beam; providing the embedding to a machine learning (ML) model; generating one or more communication parameters for communicating using the beam based on the embedding via the ML model; and communicating using the one or more communication parameters.

[0007] Some aspects provide methods for wireless communication. These methods typically include: receiving a representation associated with a beam; generating an embedding based on the representation; and providing the embedding to a wireless node.

[0008] Other aspects include a processing system configured to perform the methods described above and those described herein; a non-transitory computer-readable medium comprising instructions which, when executed by one or more processors of the processing system, cause the processor system to perform the methods described above and those described herein; a computer program product embodied on a computer-readable storage medium, the computer program product comprising code for performing the methods described above and those further described herein; and a processing system comprising units for performing the methods described above and those further described herein.

[0009] The following description and related figures illustrate certain illustrative features of one or more aspects. Attached Figure Description

[0010] The accompanying drawings depict some aspects of this disclosure and are therefore not to be considered as a limitation on the scope of this disclosure.

[0011] Figure 1 An example of a wireless communication system 100 in which the aspects described herein can be implemented is depicted.

[0012] Figure 2 Describe various aspects of the example base station (BS) and user equipment (UE).

[0013] Figures 3A-3D Describe the various aspects of the data structure used in wireless communication networks.

[0014] Figure 4 An example array gain sphere is shown.

[0015] Figure 5 Example operations for embedding generation and processing are shown in accordance with certain aspects of this disclosure.

[0016] Figure 6 , Figure 7 and Figure 8 An example sampling technique is shown.

[0017] Figure 9 Example techniques for training autoencoders using graph convolutional networks (GCNs) are shown, based on certain aspects of this disclosure.

[0018] Figure 10 This illustrates the generation of embeddings from a discretized sphere according to certain aspects of this disclosure.

[0019] Figure 11 A noise reduction autoencoder is shown, representing certain aspects of this disclosure.

[0020] Figure 12This disclosure illustrates techniques for generating embeddings using Fourier transforms, based on certain aspects of the present disclosure.

[0021] Figure 13 Example operations for beam rotation are shown in accordance with certain aspects of this disclosure.

[0022] Figure 14 Example operations for beam comparison are shown in accordance with certain aspects of this disclosure.

[0023] Figure 15 Example operations for signal quality prediction are shown in accordance with certain aspects of this disclosure.

[0024] Figure 16 This is a flowchart illustrating example operations for wireless communication by a target entity in accordance with certain aspects of this disclosure.

[0025] Figure 17 This is a flowchart illustrating example operations for wireless communication by a network entity in accordance with certain aspects of this disclosure.

[0026] Figure 18 and Figure 19 Example electronic devices are shown in accordance with certain aspects of this disclosure.

[0027] To facilitate understanding, the same reference numerals have been used wherever possible to designate the same elements that are common to the drawings. It is contemplated that elements and features of one embodiment may be beneficially incorporated into other embodiments without further description. Detailed Implementation

[0028] Various aspects of this disclosure provide apparatus and techniques for generating and processing embeddings representing beams used for communication. In some aspects of this disclosure, embeddings representing beam representations can be generated by an encoder trained using an autoencoder training process. Once trained, the encoder can generate values ​​for lookup tables indicating the embeddings representing various beams. In some aspects, the beam representation can be in the form of an array gain sphere. The array gain sphere data can be converted into a graph data structure, and the autoencoder training process can involve using a graph convolutional network to process the graph and training the encoder to generate the lookup table values.

[0029] In some respects, lookup tables can be provided to user equipment (UE) for processing. For example, the UE can retrieve an embedding representing the beam of interest from the lookup table and input the embedding into a machine learning model to generate communication parameters to be used. As an example, the machine learning model can perform beam rotation, predict signal quality parameters associated with the beam (e.g., reference signal received power (RSRP)), calculate inter-beam similarity, perform mobility estimation, or perform codebook representation, as further described herein.

[0030] By providing a low-dimensional representation of the beams that can be stored at the UE and used to perform computations, the use of embeddings as described herein results in reduced storage space and computational complexity at the UE. The generation of embeddings enables various evaluations from a single metric (e.g., inter-beam similarity evaluation) and provides richer representations compared to conventional implementations with codebook representations using descriptive statistics.

[0031] Introduction to wireless communication networks

[0032] Figure 1 An example of a wireless communication system 100 in which the aspects described herein can be implemented is depicted.

[0033] Typically, wireless communication network 100 includes base station (BS) 102, user equipment (UE) 104, and one or more core networks (such as evolved packet core (EPC) 160 and 5G core (5GC) network 190) that interoperate to provide wireless communication services.

[0034] Base station 102 can provide user equipment 104 with an access point to EPC 160 and / or 5GC 190, and can perform one or more of the following functions: transmission of user data, radio channel encryption and decryption, integrity protection, header compression, mobility control functions (e.g., handover, dual connectivity), inter-cell interference coordination, connection establishment and release, load balancing, distribution of non-access stratum (NAS) messages, NAS node selection, synchronization, radio access network (RAN) sharing, multimedia broadcast multicast service (MBMS), subscriber and equipment tracking, RAN information management (RIM), paging, location, delivery of warning messages, and other functions. In various contexts, a base station may include and / or be referred to as gNB, NodeB, eNB, ng-eNB (e.g., an eNB that has been enhanced to provide connectivity to both EPC 160 and 5GC 190), access point, base transceiver, radio base station, radio transceiver, or transceiver functional unit, or transmit / receive point.

[0035] Base station 102 communicates wirelessly with UE 104 via communication link 120. Each of base stations 102 can provide communication coverage for a corresponding geographic coverage area 110, which may overlap in some cases. For example, a small cell 102' (e.g., a low-power base station) may have a coverage area 110' that overlaps with the coverage areas 110 of one or more macro cells (e.g., high-power base stations).

[0036] The communication link 120 between base station 102 and UE 104 may include uplink (UL) (also known as reverse link) transmission from UE 104 to base station 102 and / or downlink (DL) (also known as forward link) transmission from base station 102 to UE 104. The communication link 120 may use multiple-input multiple-output (MIMO) antenna technology, which in various aspects includes spatial multiplexing, beamforming, and / or transmit diversity.

[0037] Examples of UE 104 include cellular phones, smartphones, Session Initiation Protocol (SIP) phones, laptops, personal digital assistants (PDAs), satellite radio units, global positioning systems, multimedia devices, video devices, digital audio players, cameras, game consoles, tablets, smart devices, wearable devices, vehicles, electricity meters, air pumps, large or small kitchen appliances, healthcare devices, implants, sensors / actuators, displays, or other similar devices. Some UE 104 devices may be Internet of Things (IoT) devices (e.g., parking timers, air pumps, toasters, vehicles, heart monitors, or other IoT devices), always-on (AON) devices, or edge processing devices. More generally, UE 104 may also be referred to as a station, mobile station, subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handheld device, user agent, mobile client, or client.

[0038] The wireless communication network 100 includes a training component 199, which can be configured to process beam representations to generate embeddings. The wireless network 100 also includes a beam processing component 198, which can be configured to use machine learning to process the embeddings for communication.

[0039] Figure 2 The example base station (BS) 102 and user equipment (UE) 104 are described.

[0040] Typically, base station 102 includes various processors (e.g., 220, 230, 238, and 240), antennas 234a-t (collectively referred to as 234), transceivers 232a-t (collectively referred to as 232) including modulators and demodulators, and other aspects that enable the wireless transmission of data (e.g., data source 212) and the wireless reception of data (e.g., data sink 239). For example, base station 102 can transmit and receive data between itself and user equipment 104.

[0041] Base station 102 includes a controller / processor 240, which can be configured to implement various functions related to wireless communication. In the described example, controller / processor 240 includes a training component 241, which can represent... Figure 1 Beam processing component 199. It is worth noting that, although depicted as one aspect of controller / processor 240, in other implementations, training component 241 may additionally or alternatively be implemented in various other aspects of base station 102.

[0042] Typically, user equipment 104 includes various processors (e.g., 258, 264, 266, and 280), antennas 252a-r (collectively referred to as 252), transceivers 254a-r (collectively referred to as 254) including modulators and demodulators, and other aspects that enable the wireless transmission of data (e.g., data source 262) and the wireless reception of data (e.g., data sink 260).

[0043] User equipment 102 includes a controller / processor 280, which can be configured to implement various functions related to wireless communication. In the depicted example, the controller / processor 280 includes a beam processing component 281, which can represent... Figure 1 Beam processing component 198. It is worth noting that although depicted as one aspect of controller / processor 280, in other implementations, beam processing component 281 may be implemented additionally or alternatively in various other aspects of user equipment 104.

[0044] Figures 3A-3D Describe for use in wireless communication networks (e.g.) Figure 1 The data structure of the wireless communication network 100) covers various aspects. Specifically, Figure 3A This is a schematic diagram 300 illustrating an example of the first subframe within a 5G (e.g., 5G NR) frame structure. Figure 3B This is a schematic diagram 330 illustrating an example of a DL channel within a 5G subframe. Figure 3C This is a schematic diagram 350 illustrating an example of a second subframe within a 5G frame structure, and... Figure 3D This is a schematic diagram 380 illustrating an example of a UL channel within a 5G subframe.

[0045] Information about this disclosure will be provided later in this public disclosure. Figure 1 , Figure 2 and Figures 3A-3D Further discussion is needed.

[0046] Introduction to mmWave wireless communication

[0047] In wireless communication, the electromagnetic spectrum is typically subdivided into various categories, bands, channels, or other characteristics. This subdivision is usually based on wavelength and frequency, where frequency can also be referred to as carrier, subcarrier, frequency channel, tone, or subband.

[0048] 5G networks can utilize several frequency ranges, which in some cases are defined by standards (such as 3GPP standards). For example, although the 3GPP technical standard TS 38.101 currently defines Frequency Range 1 (FR1) as including 600MHz-6 GHz, specific uplink and downlink allocations may fall outside this general range. Therefore, FR1 is often referred to (interchangeably) as the "below 6 GHz" band.

[0049] Similarly, although TS 38.101 currently defines Frequency Range 2 (FR2) as including 26–41 GHz, specific uplink and downlink allocations may fall outside this general range. FR2 is sometimes referred to (interchangeably) as the “millimeter wave” (“mmW” or “mmWave”) band, although it differs from the Extremely High Frequency (EHF) band (30–300 GHz) designated as “millimeter wave” by the International Telecommunication Union (ITU) because wavelengths at these frequencies are between 1 and 10 millimeters.

[0050] Compared to low-frequency communication, communication using mmWave / near-mmWave radio frequency bands (e.g., 3 GHz–300 GHz) may have higher path loss and shorter distances. Therefore, in Figure 1 In this configuration, the mmWave base station 180 can utilize beamforming 182 with the UE 104 to improve path loss and range. To this end, the base station 180 and the UE 104 may each include multiple antennas, such as antenna elements, antenna panels, and / or antenna arrays, to facilitate beamforming.

[0051] In some cases, base station 180 may transmit beamformed signals to UE 104 in one or more transmit directions 182'. UE 104 may receive beamformed signals from base station 180 in one or more receive directions 182''. UE 104 may also transmit beamformed signals to base station 180 in one or more transmit directions 182''. Base station 180 may receive beamformed signals from UE 104 in one or more receive directions 182''. Base station 180 and UE 104 may then perform beam training to determine the optimal receive and transmit directions for each of base station 180 and UE 104. Notably, the transmit and receive directions for base station 180 may be the same or different. Similarly, the transmit and receive directions for UE 104 may be the same or different.

[0052] Example of array gain spherical processing using machine learning (ML)

[0053] Machine learning (ML)-based methods are becoming an increasingly attractive option for complex communication systems, including mmWave-enabled systems. Machine learning is generally the process of generating trained models (e.g., artificial neural networks, trees, or other structures) that represent a generalized fit to a previously known set of training data. Applying the trained model to new data produces inferences that can be used to gain a deeper understanding of the new data. In some cases, applying a model to new data is described as “running inference” on the new data.

[0054] Both ML and non-ML millimeter-wave applications use information describing mmWave beams as input to generate desired outcomes (e.g., beam scheduling based on UE mobility). Codebook representations provide a different set of measurements to describe the individual beams and the relationships between them. This set of information is currently obtained through empirical studies and simulations. However, codebook representations are not ideal or practical for downstream millimeter-wave applications. Therefore, the performance of mmWave applications that rely on codebook beam representations may be limited by the current representation of the beams. Certain aspects of this disclosure relate to techniques for generating low-dimensional beam representations (also referred to as "codebook embedding" or simply "embedding"), which can be used for any mmWave application, codebook representation, and beam analysis, to name just a few.

[0055] Various techniques can be used to characterize beams. For example, a beam can be represented by a codebook-specific beam ID. However, using only the beam ID to characterize a beam may not be sufficient as a general tool that can be used with any codebook.

[0056] In some cases, beams can be characterized by defining parent-child neighbor relationships, providing a coarse representation of the beam. The parent-child relationship between two beams can indicate the similarity of beam characteristics. However, using parent-child neighbor relationships may not provide a concept of the relative proximity of beams.

[0057] Summary statistics of a beam can be used for beam characterization. These statistics can be extracted from the array gain sphere, which can provide complete information about the beam. Beam-related summary statistics (such as peak phi angle) can also be used. ), theta angle ( θ (and gain) can lead to two beams with very different profiles sharing a misleadingly similar representation.

[0058] Figure 4 Example array gain spheres are shown for various beams. Array gain spheres can be used to characterize multiple beams, such as beams 1 through 5, as illustrated. The array gain sphere indicates the gain associated with the beam at each phi and theta angle. The array gain sphere for beam 1 may have a peak centroid at spherical coordinates 403, and the array gain sphere for beam 5 may have a peak centroid at spherical coordinates 407. For example, the peak centroid of beam 1 may be located at theta angle 30°, phi angle 179°, and have a gain of -95 dB.

[0059] While array gain spheres provide detailed beam information, they can be difficult to use in practice for various applications, particularly user equipment and network components that cannot generate gain sphere data. Codebook representation metrics that act directly on the sphere (such as spherical cross-correlation, overlap ratio, and envelope) are approximations that do not capture non-zero hysteresis relationships. Non-zero hysteresis relationships are derived using non-zero hysteresis analysis, which provides an accurate representation of inter-beam relationships by aligning the spheres at different coordinates, considering the different rotations associated with the array gain spheres, and performing a separate analysis for each rotation. Furthermore, projecting the spheres to a one-dimensional (1D) or two-dimensional (2D) representation can lead to excessive features, distortion, and / or discontinuities for efficient processing.

[0060] Some aspects of this disclosure relate to generating low-dimensional beam representations (e.g., capturing detailed information provided by codebook representations), which can be represented as vectors or other tensor data. Low-dimensional beam representations may also be referred to herein as embeddings. Embeddings can be generated based on array gain measurements of each UE beam on a sphere (e.g., an array gain sphere). In some cases, codebook beam representation metrics of each UE beam can be used as auxiliary inputs for generating embeddings. As an example, embeddings can be used to determine relationships between beams. Embeddings can be generated using ML models trained with various types of processing units, including graphics processing units (GPUs), central processing units (CPUs), neural processing units (NPUs), artificial intelligence (AI) accelerators, application-specific integrated circuits (ASICs), or other processing units, any of which can be "off-target" (e.g., at a network entity such as a network server).

[0061] Embeddings can then be generated off-target using a trained ML model (e.g., using the CPU at the network). For example, a lookup table indicating the embedding for each of a plurality of beam IDs can be generated. The lookup table can be sent to the target (e.g., the UE) to perform various target tasks, such as evaluating inter-beam similarity, performing mobility estimation, and performing codebook representation. As a few examples, embeddings can capture beam properties provided by codebook representations, such as beamwidth, intensity, orientation, and peak gain. By providing a low-dimensional representation of the beam that can be stored at the UE and used to perform computations, using embeddings results in reduced storage space and computational complexity at the UE. In some respects, the training of the ML model and the generation of the lookup table can be performed on the target (e.g., at the UE).

[0062] Figure 5 Example operations for embedding generation and processing are shown. In some aspects, autoencoder techniques can be used to implement encoders for generating embeddings.

[0063] An autoencoder is an artificial neural network used to learn efficient data encoding in an unsupervised manner. The goal of an autoencoder is to learn a representation (encoding) for a dataset by training the network to ignore signal "noise," effectively generating a low-dimensional representation of the input. In some aspects, an autoencoder 590 can be trained end-to-end. Autoencoder 590 may include encoder 502, which receives a representation of a beam (e.g., array gain spherical data) and generates a low-dimensional representation of the beam (e.g., an embedding 503). Autoencoder 590 may also include decoder 504 that attempts to reconstruct the representation of the beam.

[0064] In other words, encoder 502 (also referred to as embedding generator) can be used to learn efficient encoding to generate embeddings 503 based on one or more training codebooks. In some aspects, the objective task of embedding can be optimized (or at least improved). For example, the embedding is provided to decoder 504 to reconstruct (e.g., generate) output 505 for training by optimizing (or at least improving) the objective function (e.g., by comparing the output from the decoder with the input of the encoder and adjusting the weights of the autoencoder accordingly).

[0065] Once the autoencoder has been trained, encoder 502 can be used independently for other tasks, such as generating codes, without using the decoding aspect of the autoencoder model. For example, encoder 502 can receive a target codebook to generate data for lookup table 510, including embeddings for each beam ID, as shown in the figure. The autoencoder training operation and lookup table generation can be performed off-target (e.g., on a server at the network). The lookup table can then be saved and provided to target 520 (e.g., UE) for performing the target task.

[0066] For example, the UE can use lookup table 510 to determine the embeddings 514 and 515 associated with the two beams respectively (e.g., in...). Figure 5 The embeddings 514 and 515 can be provided to the ML model 516 to perform a target task, such as a prediction related to the communication connectivity between the target (e.g., the UE) and the network. As described in more detail herein, the target task can be performed based on the data input 530.

[0067] Certain aspects of this disclosure relate to techniques for converting array gain spheres (or array gain sphere data) into embeddings using graph convolutional networks (GCNs). A GCN is a graph neural network designed to perform inference on data constructed as graphs. For example, graph neural networks can be directly applied to graphs to perform node-level, edge-level, and graph-level prediction tasks. A graph can typically have multiple nodes (e.g., vertices) and edges (e.g., connections between vertices) representing the relationship between one node and another. For example, edges can be associated with weights that can indicate the importance or cost associated with the edge.

[0068] To implement GCN, the array gain sphere can first be converted into a graphic data format. Various suitable techniques can be used to map the array gain sphere to the graphic. For example, each node of the graphic can be mapped to coordinates on the array gain sphere (e.g., phi and theta angular coordinates). Other techniques may include icosahedral sampling of the sphere or hierarchical equal-area equal-dimensional pixelation (HEALPix) techniques, as discussed above. Figure 6 , Figure 7 and Figure 8 More detailed description.

[0069] Figure 6 , Figure 7 and Figure 8 An example sampling technique is shown for mapping data to a graphical data structure.

[0070] For example, Figure 6 This illustrates the icosahedral sampling technique. As shown, the array gain sphere can be divided into a set of equilateral triangles. Each node in the diagram can be mapped to a vertex of one of the triangles (e.g., a vertex of triangle 602).

[0071] Figure 7 The HEALPix technology is illustrated. As shown, the array gain sphere can be divided into a hierarchical structure of equal-area curved quadrilaterals. Each quadrilateral can be further divided into multiple quadrilaterals in a layered manner. For example, quadrilateral 706 can be divided into quadrilaterals 708, 710, 712, and 714, and each of quadrilaterals 708, 710, 712, and 714 can be further divided into smaller quadrilaterals, and so on, as shown in the figure.

[0072] The original array gain sphere can be interpolated from the corner points of the codebook phi and theta to discrete points. For example... Figure 8 As shown, points on the sphere can then be transformed into a set of nodes (e.g., vertices (V), such as vertex 802) and edges (E) (e.g., edge 804), which defines a graph G(V, E). For example, when using the HEALPix technique, the sphere can include a total of 196,608 points for 7th-order discretization according to the following equation:

[0073] ,

[0074] in This represents the total number of points. Used for definition. The order of the differential equation can be 7 for a 7th order discretization. Figure 8 In the example, a graph edge is defined between each node and the set of adjacent faces (e.g., quadrilaterals). A node is defined as the centroid of each face (e.g., quadrilateral). Edge 804 can be defined between vertex 802 and vertex 806. Once the array gain sphere is mapped to the graph, the graph can be fed into one or more graph convolutional networks (GCNs) to generate embeddings, as described above. Figure 9 A more detailed description.

[0075] Figure 9Example techniques for implementing an autoencoder using a Graph Convolutional Network (GCN) according to certain aspects of this disclosure are illustrated. As described, an array gain sphere (or array gain sphere data) can be transformed into a graph and fed to the GCN to generate embeddings. For example, the graph can be processed by GCN 902, GCN 904, and GCN 906 to coarsen the graph and generate embedding 903, as shown. In other words, GCN 902 can transform a 3rd-order graph into a 2nd-order graph (e.g., representing a 3rd-order discretization), GCN 904 can transform a 2nd-order graph (e.g., representing a 2nd-order discretization) into a 1st-order graph (e.g., representing a 1st-order discretization), and GCN 906 can transform a 1st-order graph into a 0th-order graph. The 0th-order graph can be used to generate embedding 903. Embedding 903 is fed to a decoder implemented using GCN 908, GCN 910, and GCN 912 to reconstruct the input (to the input of GCN 902). GCN 908, 910, and 912 construct more detailed graphs. For example, GCN 908 converts a 0th-order graph to a 1st-order graph, GCN 910 converts a 1st-order graph to a 2nd-order graph, and GCN 912 converts a 2nd-order graph to a 3rd-order graph, and so on.

[0076] Figure 10 The diagram illustrates the generation of an embedding 1003 from a discretized sphere according to certain aspects of this disclosure. As described, the array gain sphere can be discretized using techniques such as HEALPix to generate a discretized sphere 1004, and then roughened to generate a roughened discretized sphere 1002. The discretized sphere 1002 can be flattened into a 1D vector to generate the embedding 1003, as shown. For example, the vertices of each quadrilateral on the roughened discretized sphere 1002 can be used to generate the embedding 1003.

[0077] Figure 11 This illustrates a denoising autoencoder 1100 according to certain aspects of this disclosure (e.g., corresponding to...). Figure 5 (Autoencoder 590). Adding noise to the data input has a regularization effect and thus improves the robustness of the ML model. As shown in the figure, at box 1102, noise can be added to the array gain sphere 1101 to generate a partially corrupted input 1104 (also referred to herein as a noise-enhanced input). The partially corrupted input 1104 can be provided to encoder 502 to generate an embedding, as described herein. Embedding 503 can be provided to decoder 504 to recover the original uncorrupted input (e.g., array gain sphere 1101).

[0078] Figure 12This disclosure illustrates techniques for generating embeddings using Fourier transforms, based on certain aspects of the present disclosure. For example, a 2D Fourier transform of beam gain from the phi and theta domains to the Fourier domain can be performed.

[0079] In the depicted example, the phi and theta corner points from the array gain sphere 1202 are input to the Fourier transform block 1204 to generate embedding 503. For example, Fourier coefficients as 1D vectors can be used as embedding 1203. The Fourier domain captures the dominant energy in a small set of points and thus generates a compressed representation of the array gain sphere 1202. Rotations in the phi and theta domains can be represented as phase rotations in the Fourier domain. The Fourier transform preserves rotation information, which may be useful in downstream tasks such as rotation estimation. Therefore, this is an alternative to using an autoencoder as described above to generate a low-dimensional representation of the array gain sphere 1202 data.

[0080] Techniques for spherical data processing of array gain

[0081] Some aspects of this disclosure relate to techniques for processing array gain spherical data using embedding lookup tables, which are generated using a trained encoder, for example, as per [reference to...]. Figure 5 As described, it can process array gain spheres to perform any appropriate target task, such as beam rotation, beam comparison, and Received Signal Received Power (RSRP) prediction, to name just a few.

[0082] Figure 13 Example operations for performing beam rotation according to certain aspects of this disclosure are shown. (See also: Regarding...) Figure 5 As described, encoder 502 can be used to generate values ​​for embedding lookup table 510, which indicates embeddings for one or more beams. The lookup table is then provided to the UE. In the depicted example, the UE can use lookup table 1310 to determine embedding 1314 for a specific beam (e.g., represented by array gain spherical 1350). As shown, embedding 1314 can be provided to rotary converter decoder 1302 (e.g., ...). Figure 5 (Example of ML model 516). The rotation transformer decoder 1302 can receive an angle rotation command 1306, which indicates the angle (e.g., Euler angle) by which the input array gain sphere will be rotated. The rotation transformer decoder 1302 then generates an array gain sphere 1304 rotated according to the angle rotation command 1306. The UE can use the rotated array gain sphere to facilitate communication with other nodes.

[0083] Figure 14Example operations for beam comparison according to certain aspects of this disclosure are illustrated. As shown, embeddings 1414 and 1415 associated with two beams (e.g., characterized by array gain spheres 1450 and 1452) can be generated and provided to a pairwise difference decoder 1402 (e.g., Figure 5 Another example of ML model 516 is used to generate point-by-point difference parameters that indicate the difference between two beams (e.g., the difference between array gain spheres 1454 and 1456). The UE can use the point-by-point difference parameters to facilitate communication with other nodes.

[0084] Figure 15 Example operations for RSRP prediction according to certain aspects of this disclosure are shown. In one example, UE RSRP can be predicted per synchronization signal block (SSB) based on beam embedding and UE orientation. For example, embedding 1514 can be provided to RSRP predictor 1502 (e.g., Figure 5 Another example of ML model 516). RSRP predictor 1502 may receive an indication of orientation 1506 for the UE. RSRP predictor 1502 may then determine (e.g., predict) the RSRP (e.g., each RSRP in RSRP 1508) for each of a plurality of synchronization signal blocks (SSBs) (e.g., SSB1 to SSBn, where n is an integer greater than 1) for a given orientation 1506.

[0085] In some respects, beam representations (e.g., representations of array gain spheres, such as array gain spheres 1350, 1450, 1452) can be provided to the encoder (e.g., encoder 602) to facilitate training of the encoder used to generate embeddings. For example, the UE can provide beam representations to the encoder at the UE local location (or at the BS), and the beam representations can be used to train or tune the encoder (e.g., to facilitate full end-to-end learning / fine-tuning of the encoder to improve embedding generation).

[0086] Example signal processing flow for data reuse

[0087] Figure 16 This is a flowchart illustrating an example operation 1600 for wireless communication according to a specific aspect of this disclosure. Operation 1600 can be performed, for example, by a UE (e.g., Figure 1 The wireless nodes, such as UE 104, in the wireless communication network 100 perform this action.

[0088] Operation 1600 can be implemented in one or more processors (e.g., Figure 2 The software components executed and running on the controller / processor 280. Furthermore, the transmission and reception of signals by the UE in operation 1600 can be, for example, by one or more antennas (e.g., Figure 2 This can be achieved via antenna 252. In some aspects, the UE's transmission and / or reception of signals can be achieved via a bus interface for acquiring and / or outputting signals from one or more processors (e.g., controller / processor 280 or beam processing component 281).

[0089] Operation 1600 begins at box 1602, where a wireless node (e.g., from a network) receives an embedding (e.g., representing a characterization associated with a beam) Figure 5 (Embedding 514). In some aspects, receiving the embedding may include receiving a lookup table including an indication of the embedding (e.g., Figure 5 (See lookup table 510). Characterizations can include spherical array gain data, array gain measurements, or codebook beam characterization metrics, to name just a few examples.

[0090] At box 1604, the wireless node can send data to a machine learning (ML) model (e.g., Figure 5 The ML model 516 provides embeddings.

[0091] At box 1606, the wireless node can generate one or more communication parameters for beam-based communication via an ML model based on embedding. For example, the ML model could be a rotation transformer decoder (e.g., Figure 13 (Rotator decoder 1302). In this case, generating one or more communication parameters may include determining the rotation of the beam used for communication based on the embedding via the rotation converter decoder. For example, one or more communication parameters may include based on the determined rotated beam. In some aspects, the wireless node may provide rotation commands to the rotation converter decoder (e.g., Figure 13 The rotation command 1306 can be used to determine the beam rotation.

[0092] In some respects, wireless nodes can be based on another beam (e.g., Figure 14 The array gain sphere 1452) is characterized to generate another embedding (e.g., Figure 14 Embedding 1415). Embedding (e.g., Figure 14 Embedding 1414) and other embeddings (e.g., Figure 14 The embedding 1415 is provided to the pointwise difference decoder (e.g., Figure 14 (The point-by-point difference decoder 1402). Then, the wireless node generates one or more communication parameters by predicting the point-by-point difference between another beam based on the embedding and another embedding via the point-by-point difference decoder.

[0093] In some respects, embedding (e.g., Figure 15 The embedded 1514) is provided to the Reference Signal Received Power (RSRP) decoder (e.g., Figure 15 RSRP predictor 1502). A wireless node can determine the RSRP of each of a plurality of synchronization signal blocks (SSBs) based on embedding via an RSRP decoder (e.g., Figure 15 Each RSRP in RSRP 1508 is used to generate one or more communication parameters. In some aspects, wireless nodes can provide directionality to wireless nodes (e.g., Figure 15 The RSRP is determined based on the orientation (1506).

[0094] At box 1608, a wireless node can communicate using one or more communication parameters.

[0095] In some respects, beam representations can be input into an encoder (e.g., encoder 602) to facilitate encoder training to generate embeddings. For example, the UE can provide beam representations to an encoder at the UE local location (or at the BS), and the beam representations can be used for encoder training or tuning to improve embedding generation.

[0096] Figure 17 This is a flowchart illustrating an example operation 1700 for wireless communication according to certain aspects of this disclosure. Operation 1700 can be performed, for example, by a network entity such as a BS (e.g., a network ISP). Figure 1 It is executed by BS102 in the wireless communication network 100.

[0097] Operation 1700 can be implemented in one or more processors (e.g., Figure 2 The software components that execute and run on the controller / processor 240. Furthermore, the transmission and reception of signals by the BS in operation 1700 can be, for example, by one or more antennas (e.g., Figure 2 This can be achieved via antenna 234. In some aspects, the BS can transmit and / or receive signals via a bus interface that acquires and / or outputs signals from one or more processors (e.g., controller / processor 240 or training component 241).

[0098] Operation 1700 begins at box 1702 with the network entity receiving a representation associated with the beam. The representation may include spherical array gain data, array gain measurements, or codebook beam characterization metrics, to name just a few.

[0099] At box 1704, the network entity generates an embedding based on the representation (e.g., embedding 503). In some aspects, the representation includes a spherical array gain, and the network entity can transform the spherical array gain into a graph. Graph-based graph convolutional networks can be used to generate the embedding.

[0100] In some aspects, network entities can receive one or more training codebooks for training an autoencoder. The encoder of the autoencoder can be used to generate embeddings. In some aspects, the representation is received as noise-enhanced input (e.g., Figure 11 The partially damaged input 1104), and can be used with a denoising autoencoder based on the noise-enhanced input (e.g., Figure 11 The partially corrupted input 1104) is used to generate the embedding. In some aspects, generating the embedding may include using Fourier transform (e.g., in...). Figure 12 The representation is downsampled at position 1204 of the Fourier transform block.

[0101] At box 1706, the network entity provides embedding to the radio node (e.g., UE). For example, the network entity may provide a lookup table indicating the embedding (e.g., lookup table 510).

[0102] Example training techniques

[0103] In some aspects of this disclosure, federated learning can be used to train autoencoders implemented at the network level, such as Figure 5 590. Federated learning is a framework that allows training a model across multiple nodes with local data samples, without requiring the nodes to share such data but only the parameters of their local models. Federated learning involves using iterative loops, where a coordinator (e.g., a BS) or a central server selects the model to be trained (e.g., with...). Figure 5 The coordinator generates an autoencoder 590 associated with a model and sends the model to a node (e.g., a UE). The node uses local data to train a local copy of its model. The coordinator then aggregates the local models and generates an aggregated model to be sent to the nodes. The main benefit of this approach is privacy, as no UE local data is shared across nodes or with the coordinator.

[0104] In some respects, UEs can send embeddings of their codebooks to the BS for various use cases. For example, embeddings can be used in models that will be trained at the BS, or used in conjunction with pre-trained models to generate inferences that run at the BS (e.g., using uplink (UL) measurements or reported downlink (DL) measurements). Information about codebook embeddings can include information for mapping such inputs to a specific ML model. This provides an opportunity to keep the actual codebook design proprietary, with only compressed versions of it being shared. These use cases can be combined with federated learning. A first set of UEs can receive an initial model and parameters. Such a set of UEs can train a model locally based on its codebook information, and each UE can report updated parameters back to the BS with or without its associated UE codebook embeddings. The BS can combine the parameters of these models with or without the corresponding UE codebook embeddings. The result will be a new model, which the BS can send to a second set of UEs. In some scenarios, the second set of UEs may include the first set of UEs. For example, a set of UEs with higher capabilities can train a model locally, and the associated parameters can be reused by different UEs with lower capabilities. In the context of a model that uses frequency range 2 (FR2) beam measurements as its input, codebook embedding provides invariance to leverage the learning performed by the UE set across other UEs. In other words, codebook embeddings can be derived such that the resulting encoder is codebook / UE independent. Once trained, an encoder trained based on FR2 beam measurements from one or more UEs can be used across any other UE operating on FR2. For a given ML model, a UE can transfer its ability to train the model locally and participate in a federated learning framework, or the lack thereof.

[0105] Example wireless communication device

[0106] Figure 18 An example communication device 1800 is depicted, which includes operations operable to, configured to, or adjusted to perform operations using the techniques disclosed herein (e.g., regarding...). Figure 17 The various components of the operation described. In some examples, the communication device 1800 may be, for example, about Figure 1 and 2 The base station 102 is described.

[0107] The communication device 1800 includes a processing system 1802 coupled to a transceiver 1808 (e.g., a transmitter and / or receiver). The transceiver 1808 is configured to transmit (or transmit) and receive signals for the communication device 1800 via an antenna 1810, such as the various signals described herein. The processing system 1802 may be configured to perform processing functions of the communication device 1800, including processing signals received and / or to be transmitted by the communication device 1800.

[0108] Processing system 1802 includes one or more processors 1820 coupled to computer-readable medium / memory 1830 via bus 1806. In some aspects, computer-readable medium / memory 1830 is configured to store instructions (e.g., computer-executable code) that, when executed by the one or more processors 1820, cause the one or more processors 1820 to perform actions... Figure 17 The operations shown herein, or other operations used to perform the various techniques discussed herein for processing beam characterization to generate embeddings.

[0109] In the depicted example, computer-readable medium / memory 1830 stores code 1831 for receiving, code 1832 for providing, code 1833 for generating, code 1834 for transmitting, code 1835 for encoding, and code 1837 for training. Code 1835 for encoding can be used to generate a lookup table 1836 stored in computer-readable medium / memory 1830.

[0110] In the depicted example, one or more processors 1820 include circuitry configured to implement code stored in computer-readable medium / memory 1830, including circuitry 1821 for receiving, circuitry 1822 for providing, circuitry 1823 for generating, circuitry 1824 for transmitting, circuitry 1825 for encoding, and circuitry 1826 for training. The circuitry 1825 for encoding can be used to generate a lookup table 1836 stored in computer-readable medium / memory 1830.

[0111] The various components of the communication device 1800 can provide for the execution of this document (including those concerning...) Figure 17 ( ) is a unit describing the method.

[0112] In some examples, the unit for transmitting or sending (or the unit for outputting for transmission) may include Figure 2 The transceiver 232 and / or antenna 234 of the base station 102 shown are shown. Figure 18 The transceiver 1808 and antenna 1810 of the communication equipment 1800.

[0113] In some examples, the unit for receiving (or the unit for obtaining) may include Figure 2 The base station transceiver 232 and / or antenna 234 shown in the figure Figure 18 The transceiver 1808 and antenna 1810 of the communication equipment 1800.

[0114] In some examples, the units for receiving, providing, generating, and sending can include various processing system components, such as: Figure 18 One or more processors 1820, or Figure 2 The various aspects of the base station 102 depicted include a receiver processor 238, a transmitter processor 220, a TX MIMO processor 230, and / or a controller / processor 240 (including a training component 241).

[0115] It should be noted that Figure 18 This is just an example, and many other examples and configurations of the communication device 1800 are possible.

[0116] Figure 19 Example communication device 1900 is depicted, which includes operations operable to, configured to, or adjusted to perform operations using the techniques disclosed herein (e.g., regarding...). Figure 16 The various components of the operation described. In some examples, the communication device 1900 may be, for example, about Figure 1 and 2 User equipment 104 is described.

[0117] The communication device 1900 includes a processing system 1902 coupled to a transceiver 1908 (e.g., a transmitter and / or receiver). The transceiver 1908 is configured to transmit (or transmit) and receive signals for the communication device 1900 via an antenna 1910, such as the various signals described herein. The processing system 1902 may be configured to perform processing functions of the communication device 1900, including processing signals received and / or to be transmitted by the communication device 1900.

[0118] Processing system 1902 includes one or more processors 1920 coupled to computer-readable medium / memory 1930 via bus 1906. In some aspects, computer-readable medium / memory 1930 is configured to store instructions (e.g., computer-executable code) that, when executed by the one or more processors 1920, cause the one or more processors 1920 to perform actions... Figure 16 The operations shown herein, or other operations used to perform the various techniques discussed herein for using machine learning to process embeddings for communication.

[0119] In the depicted example, computer-readable medium / memory 1930 stores code 1931 for receiving, code 1932 for generating, code 1933 for providing, code 1934 for converting, code 1935 for RSRP prediction, code 1936 for beam rotation, code 1937 for pairwise difference determination, and code 1939 for training. Code 1935 for RSRP prediction, code 1936 for beam rotation, and code 1937 for pairwise difference determination may be based on lookup table 1938 stored in computer-readable medium / memory 1930.

[0120] In the depicted example, one or more processors 1920 include circuitry configured to implement code stored in computer-readable medium / memory 1930, including circuitry 1921 for receiving, circuitry 1922 for generating, circuitry 1923 for providing, circuitry 1924 for converting, circuitry 1925 for RSRP prediction, circuitry 1926 for beam rotation, circuitry 1927 for pairwise difference determination, and circuitry 1929 for training. The circuitry 1925 for RSRP prediction, the circuitry 1926 for beam rotation, and the circuitry 1927 for pairwise difference determination may be based on a lookup table 1938 stored in computer-readable medium / memory 1930.

[0121] The various components of the communication device 1900 can provide for the execution of this document (including those concerning...) Figure 16 ( ) is a unit describing the method.

[0122] In some examples, the unit for transmitting or sending (or the unit for outputting for transmission) may include Figure 2 The transceiver 254 and / or antenna 252 of the user equipment 104 shown are shown. Figure 19 The communication equipment 1900 includes the transceiver 1908 and the antenna 1910.

[0123] In some examples, the unit for receiving (or the unit for acquiring) may include Figure 2 The transceiver 254 and / or antenna 252 of the user equipment 104 shown are shown. Figure 19 The communication equipment 1900 includes the transceiver 1908 and the antenna 1910.

[0124] In some examples, the unit for receiving, the unit for providing, the unit for generating, and the unit for providing can include various processing system components, such as: Figure 19 One or more processors 1920, or Figure 2The user equipment 104 depicted includes various aspects such as a receiver processor 258, a transmitter processor 264, a TX MIMO processor 266, and / or a controller / processor 280 (including a beam processing component 281).

[0125] It should be noted that Figure 19 This is just an example, and many other examples and configurations of the communication device 1900 are possible.

[0126] Example Terms

[0127] Examples of implementation methods are described in the following numbered clauses.

[0128] Clause 1. A method for wireless communication by a wireless node, comprising: receiving an embedding representing a representation associated with a beam; providing the embedding to a machine learning (ML) model; generating one or more communication parameters for communicating using the beam based on the embedding via the ML model; and communicating using the one or more communication parameters.

[0129] Clause 2. The method according to Clause 1, wherein receiving the embedding includes: receiving a lookup table indicating the embedding.

[0130] Clause 3. The method according to any one of Clauses 1-2, wherein: the ML model includes a rotator decoder; generating the one or more communication parameters includes determining, via the rotator decoder, based on the embedding, the rotation of the beam for the communication; and the one or more communication parameters include, based on the determined rotated beam.

[0131] Clause 4. The method according to Clause 3 further includes: providing a rotation command to the rotator decoder, wherein the rotation of the beam is determined based on the rotation command.

[0132] Clause 5. The method according to any one of Clauses 1-4, wherein the method further comprises: generating another embedding based on the characterization of another beam; the embedding and the other embedding being provided to a point-by-point difference decoder; and generating the one or more communication parameters comprising: predicting a point-by-point difference between the beam and the other beam via the point-by-point difference decoder based on the embedding and the other embedding.

[0133] Clause 6. The method according to any one of Clauses 1-5, wherein the embedding is provided to a Reference Signal Received Power (RSRP) decoder; and generating the one or more communication parameters includes: determining the RSRP of each of a plurality of Synchronization Signal Blocks (SSBs) based on the embedding via the RSRP decoder.

[0134] Clause 7. The method according to Clause 6 further includes: providing an indication of orientation of the wireless node, wherein the RSRP is determined based on the orientation.

[0135] Clause 8. The method according to any one of Clauses 1-7, wherein the characterization includes spherical array gain.

[0136] Clause 9. The method according to any one of Clauses 1-8, wherein the characterization includes an array gain measurement or a codebook beam characterization metric.

[0137] Clause 10. The method according to any one of Clauses 1-9 further comprises: training an encoder configured to generate an embedding using the representation associated with the beam.

[0138] Clause 11. A method for wireless communication, comprising: receiving a representation associated with a beam; generating an embedding based on the representation; and providing the embedding to a wireless node.

[0139] Clause 12, the method according to Clause 11, wherein providing the embedding includes: providing a lookup table indicating the embedding associated with the beam.

[0140] Clause 13. The method according to one of Clauses 11-12, wherein: the characterization includes a spherical array gain; the method further includes converting the spherical array gain into a graph; and the embedding is generated based on the graph using a graph convolutional network.

[0141] Clause 14. The method according to any one of Clauses 11-13 further comprises: receiving one or more training codebooks for training an autoencoder, wherein the embedding is generated using an encoder of the autoencoder.

[0142] Clause 15. The method according to one of Clauses 11-14, wherein the characterization includes an array gain measurement or a codebook beam characterization metric.

[0143] Clause 16, the method according to Clause 15, wherein the representation is received as noise-enhancing input, and wherein the embedding is generated using a denoising autoencoder based on the noise-enhancing input.

[0144] Clause 17. The method according to any one of Clauses 11-16, wherein generating the embedding comprises: downsampling the representation using a Fourier transform.

[0145] Clause 18. A method according to one of Clauses 11-17, wherein the characterization includes spherical array gain.

[0146] Clause 19. The method according to any one of Clauses 11-18 further comprises: training an autoencoder using a federated learning model, wherein the embedding is generated using an encoder of the autoencoder.

[0147] Clause 20: An apparatus comprising: a memory including executable instructions; and one or more processors configured to execute the executable instructions and cause the apparatus to perform the method according to any one of Clauses 1-19.

[0148] Clause 21: An apparatus comprising a unit for performing the method according to any one of Clauses 1-19.

[0149] Clause 23: A non-transitory computer-readable medium comprising executable instructions that, when executed by one or more processors of a device, cause the device to perform the method according to any one of Clauses 1-19.

[0150] Clause 24: A computer program product embodied on a computer-readable storage medium, said computer-readable storage medium including code for performing the method according to any one of Clauses 1-19.

[0151] Additional wireless communication network considerations

[0152] The techniques and methods described herein can be used in a variety of wireless communication networks (or wireless wide area networks (WWANs)) and radio access technologies (RATs). While terms commonly associated with 3G, 4G, and / or 5G (e.g., 5G New Radio (NR)) wireless technologies are used to describe aspects herein, the aspects of this disclosure are equally applicable to other communication systems and standards not expressly mentioned herein.

[0153] 5G wireless communication networks can support a variety of advanced wireless communication services, such as enhanced mobile broadband (eMBB), millimeter wave (mmWave), machine-type communication (MTC), and / or mission-critical services targeting ultra-reliable low-latency communication (URLLC). These and other services may include latency and reliability requirements.

[0154] return Figure 1 Various aspects of this disclosure can be implemented within the example wireless communication network 100.

[0155] In 3GPP, the term "cell" can refer to the coverage area of ​​a NodeB and / or the narrowband subsystem serving that coverage area, depending on the context in which the term is used. In NR systems, the term "cell" is used interchangeably with BS, next-generation NodeB (gNB or gNodeB), access point (AP), distributed unit (DU), carrier, or transmit / receive point. A BS can provide communication coverage for macrocells, picocells, femtocells, and / or other types of cells.

[0156] Macro cells typically cover a relatively large geographic area (e.g., a radius of several kilometers) and allow unrestricted access for UEs with service subscriptions. Pico cells cover a relatively small geographic area (e.g., a stadium) and allow unrestricted access for UEs with service subscriptions. Femto cells cover a relatively small geographic area (e.g., a residential area) and allow restricted access for UEs associated with the femto cell (e.g., UEs in a Closed Subscriber Group (CSG) and UEs for users in a residential building). A BS for a macro cell can be called a macro BS. A BS for a pico cell can be called a pico BS. A BS for a femto cell can be called a femto BS, a home BS, or a home NodeB.

[0157] Base station 102 configured for 4G LTE (collectively referred to as Evolved Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access Network (E-UTRAN)) can be connected to EPC 160 via a first backhaul link 132 (e.g., S1 interface). Base station 102 configured for 5G (e.g., 5G NR or Next Generation RAN (NG-RAN)) can be connected to 5GC 190 via a second backhaul link 184. BS 102 can communicate directly or indirectly with each other (e.g., via EPC 160 or 5GC 190) on a third backhaul link 134 (e.g., X2 interface). The third backhaul link 134 can typically be wired or wireless.

[0158] Small cell 102' can operate in licensed and / or unlicensed spectrum. When operating in unlicensed spectrum, small cell 102' can employ NR and use the same 5 GHz unlicensed spectrum as used by Wi-Fi AP 150. Small cell 102' employing NR in unlicensed spectrum can improve coverage of the access network and / or increase the capacity of the access network.

[0159] Some base stations (such as gNB 180) can operate in the conventional sub-6 GHz spectrum, in millimeter wave (mmWave) frequencies and / or near-mmWave frequencies, and communicate with UE 104. When gNB 180 operates in mmWave or near-mmWave frequencies, gNB 180 can be referred to as an mmWave base station.

[0160] The communication link 120 between base station 102 and, for example, UE 104 can be via one or more carriers. For example, base station 102 and UE 104 can use a spectrum of up to Y MHz (e.g., 5, 10, 15, 20, 100, 400, and other MHz) bandwidth per carrier allocated in carrier aggregation for transmission in each direction, up to a total of Yx MHz (x component carriers). These carriers may be adjacent to each other or may not be adjacent to each other. Carrier allocation can be asymmetric for DL ​​and UL (e.g., more or fewer carriers may be allocated to DL compared to UL). Component carriers may include primary component carriers and one or more secondary component carriers. The primary component carrier may be referred to as the primary cell (PCell), and the secondary component carriers may be referred to as secondary cells (SCells).

[0161] The wireless communication network 100 also includes a Wi-Fi access point (AP) 150 that communicates with a Wi-Fi station (STA) 152 via a communication link 154 in an unlicensed spectrum, such as 2.4 GHz and / or 5 GHz. When communicating in unlicensed spectrum, the STA 152 / AP 150 can perform a free channel assessment (CCA) to determine whether a channel is available before communication begins.

[0162] Some UEs 104 can communicate with each other using device-to-device (D2D) communication link 158. D2D communication link 158 can use DL / UL WWAN spectrum. D2D communication link 158 can use one or more sidelink channels, such as the Physical Sidelink Broadcast Channel (PSBCH), Physical Sidelink Discovery Channel (PSDCH), Physical Sidelink Shared Channel (PSSCH), and Physical Sidelink Control Channel (PSCCH). D2D communication can be conducted through various wireless D2D communication systems, including, for example, FlashLinQ, WiMedia, Bluetooth, ZigBee, Wi-Fi based on the IEEE 802.11 standard, 4G (e.g., LTE), or 5G (e.g., NR).

[0163] EPC 160 may include a Mobility Management Entity (MME) 162, other MMEs 164, a Serving Gateway 166, a Multimedia Broadcast Multicast Service (MBMS) Gateway 168, a Broadcast Multicast Service Center (BM-SC) 170, and a Packet Data Network (PDN) Gateway 172. MME 162 can communicate with the Home Subscriber Server (HSS) 174. MME 162 is the control node that handles signaling between UE 104 and EPC 160. Overall, MME 162 provides bearer and connectivity management.

[0164] Typically, user Internet Protocol (IP) packets are forwarded through Serving Gateway 166, which is itself connected to PDN Gateway 172. PDN Gateway 172 provides UE IP address allocation and other functions. PDN Gateway 172 and BM-SC 170 are connected to IP Service 176, which may include, for example, the Internet, intranet, IP Multimedia Subsystem (IMS), PS streaming service, and / or other IP services.

[0165] The BM-SC 170 provides functionality for MBMS user service provisioning and delivery. It can serve as an entry point for MBMS transmissions to content providers, authorize and initiate MBMS bearer services within a Public Land Mobile Network (PLMN), and schedule MBMS transmissions. The MBMS gateway 168 can distribute MBMS services to base stations 102 belonging to Multicast-Broadcast Single Frequency Network (MBSFN) areas specific to broadcast services, and handle session management (start / stop) and collection of eMBMS-related billing information.

[0166] 5GC 190 may include Access and Mobility Management Functions (AMF) 192, other AMFs 193, Session Management Functions (SMF) 194, and User Plane Functions (UPF) 195. AMF 192 may communicate with Unified Data Management (UDM) 196.

[0167] The AMF 192 is typically the control node that handles signaling between UE 104 and 5GC 190. Typically, the AMF 192 provides QoS flow and session management.

[0168] All user Internet Protocol (IP) packets are transmitted via UPF 195, which connects to IP service 197 and provides IP address allocation for the UE, as well as other functions for 5GC 190. IP service 197 may include, for example, the Internet, intranet, IP Multimedia Subsystem (IMS), PS streaming service, and / or other IP services.

[0169] return Figure 2 It describes various example components of BS 102 and UE 104 (e.g., Figure 1 A wireless communication network 100, which can be used to implement various aspects of the present disclosure.

[0170] At BS 102, the transmitting processor 220 can receive data from data source 212 and control information from controller / processor 240. The control information can be used for the Physical Broadcast Channel (PBCH), Physical Control Format Indicator Channel (PCFICH), Physical Hybrid ARQ Indicator Channel (PHICH), Physical Downlink Control Channel (PDCCH), Group Common PDCCH (GC PDCCH), and others. In some examples, the data can be used for the Physical Downlink Shared Channel (PDSCH).

[0171] The Media Access Control (MAC)-Control Element (MAC-CE) is a MAC layer communication structure used for exchanging control commands between wireless nodes. The MAC-CE can be carried in a shared channel (e.g., Physical Downlink Shared Channel (PDSCH), Physical Uplink Shared Channel (PUSCH), or Physical Sidelink Shared Channel (PSSCH)).

[0172] Processor 220 can process (e.g., encode and symbol map) data and control information to obtain data symbols and control symbols, respectively. Transmitter processor 220 can also generate reference symbols, such as those used for primary synchronization signal (PSS), secondary synchronization signal (SSS), PBCH demodulation reference signal (DMRS), and channel state information reference signal (CSI-RS).

[0173] The transmit (TX) multiple-input multiple-output (MIMO) processor 230 can perform spatial processing (e.g., precoding) on ​​data symbols, control symbols, and / or reference symbols (if applicable), and can provide an output symbol stream to the modulators (MODs) in transceivers 232a-232t. Each modulator in transceivers 232a-232t can process its corresponding output symbol stream (e.g., for OFDM) to obtain an output sample stream. Each modulator can further process (e.g., convert to analog, amplify, filter, and up-convert) the output sample stream to obtain a downlink signal. The downlink signal from the modulators in transceivers 232a-232t can be transmitted via antennas 234a-234t respectively.

[0174] At UE 104, antennas 252a-252r can receive downlink signals from BS 102 and can provide the received signals to demodulators (DEMODs) in transceivers 254a-254r respectively. Each demodulator in transceivers 254a-254r can adjust (e.g., filter, amplify, down-convert, and digitize) the corresponding received signal to obtain an input sample. Each demodulator can further process the input sample (e.g., for OFDM) to obtain the received symbols.

[0175] MIMO detector 256 can obtain received symbols from all demodulators in transceivers 254a-254r, perform MIMO detection on the received symbols (if applicable), and provide the detected symbols. Receiver processor 258 can process (e.g., demodulate, deinterleave, and decode) the detected symbols, provide decoded data for UE 104 to data sink 260, and provide decoded control information to controller / processor 280.

[0176] On the uplink, at UE 104, the transmit processor 264 can receive and process data from data source 262 (e.g., for the Physical Uplink Shared Channel (PUSCH)) and control information from controller / processor 280 (e.g., for the Physical Uplink Control Channel (PUCCH)). The transmit processor 264 can also generate reference symbols for reference signals (e.g., for Sounding Reference Signals (SRS)). Symbols from the transmit processor 264 can be precoded by the TX MIMO processor 266 (if applicable), further processed by modulators in transceivers 254a-254r (e.g., for SC-FDM, etc.), and transmitted to BS 102.

[0177] At BS 102, uplink signals from UE 104 can be received by antennas 234a-t, processed by demodulators in transceivers 232a-232t, detected by MIMO detector 236 (if applicable), and further processed by receiver processor 238 to obtain decoded data and control information transmitted by UE 104. Receiver processor 238 can provide the decoded data to data sink 239 and the decoded control information to controller / processor 240.

[0178] Memory 242 and 282 can store data and program code for BS 102 and UE 104, respectively.

[0179] Scheduler 244 can schedule data transmissions for the UE on the downlink and / or uplink.

[0180] 5G can utilize Orthogonal Frequency Division Multiplexing (OFDM) with a cyclic prefix (CP) on both the uplink and downlink. 5G can also use Time Division Duplex (TDD) to support half-duplex operation. OFDM and Single-Carrier Frequency Division Multiplexing (SC-FDM) divide the system bandwidth into multiple orthogonal subcarriers, often referred to as tones and frequency bands. Each subcarrier can be modulated with data. Modulation symbols can be transmitted in the frequency domain using OFDM and in the time domain using SC-FDM. The spacing between adjacent subcarriers can be fixed, and the total number of subcarriers can depend on the system bandwidth. In some examples, the minimum resource allocation (called a resource block (RB)) can be 12 consecutive subcarriers. The system bandwidth can also be divided into subbands. For example, a subband can cover multiple RBs. NR can support a basic subcarrier spacing (SCS) of 15 kHz, and other SCSs can be defined relative to the basic SCS (e.g., 30 kHz, 60 kHz, 120 kHz, 240 kHz, and others).

[0181] As mentioned above, Figures 3A-3D Describes the use of wireless communication networks (e.g.) Figure 1 Various examples of data structures in a wireless communication network 100.

[0182] In various aspects, the 5G frame structure can be Frequency Division Duplex (FDD), where for a specific set of subcarriers (carrier system bandwidth), subframes within that set are dedicated to either DL (Deep Length) or UL (Ultra-Length). The 5G frame structure can also be Time Division Duplex (TDD), where for a specific set of subcarriers (carrier system bandwidth), subframes within that set are dedicated to both DL and UL. Figure 3A and Figure 3C In the provided example, the 5G frame structure is assumed to be TDD, where subframe 4 is configured with slot format 28 (mostly DL), where D is DL, U is UL, and X is flexibly used between DL / UL, and subframe 3 is configured with slot format 34 (mostly UL). Although subframes 3 and 4 are shown with slot formats 34 and 28 respectively, any particular subframe can be configured with any of the various available slot formats 0-61. Slot formats 0 and 1 are all DL and all UL, respectively. Other slot formats 2-61 include a mixture of DL symbols, UL symbols, and flexible symbols. The UE is configured with a slot format via a received Slot Format Indicator (SFI) (dynamically configured via DL Control Information (DCI) or semi-statically / statically configured via Radio Resource Control (RRC) signaling). Note that the following description also applies to 5G frame structures that are TDD.

[0183] Other wireless communication technologies may have different frame structures and / or different channels. A frame (10 ms) can be divided into 10 equal-sized subframes (1 ms). Each subframe may include one or more time slots. Subframes may also include micro-time slots, which may include 7, 4, or 2 symbols. In some examples, depending on the time slot configuration, each time slot may include 7 or 14 symbols.

[0184] For example, for slot configuration 0, each slot can include 14 symbols, and for slot configuration 1, each slot can include 7 symbols. Symbols on the DL can be cyclic prefix (CP) OFDM (CP-OFDM) symbols. Symbols on the UL can be CP-OFDM symbols (for high-throughput scenarios) or Discrete Fourier Transform (DFT) Spread Spectrum OFDM (DFT-s-OFDM) symbols (also known as single-carrier frequency division multiple access (SC-FDMA) symbols) (for power-constrained scenarios; limited to a single stream transmission).

[0185] The number of time slots within a subframe is based on the time slot configuration and the digital scheme (numerology). For time slot configuration 0, different digital schemes (µ) 0 through 5 account for 1, 2, 4, 8, 16, and 32 time slots per subframe, respectively. For time slot configuration 1, different digital schemes 0 through 2 allow 2, 4, and 8 time slots per subframe, respectively. Therefore, for time slot configuration 0 and digital scheme µ, there are 14 symbols per time slot, and 2µ time slots / subframe per subframe. The subcarrier spacing and symbol length / duration are functions of the digital scheme. The subcarrier spacing can be equal to... kHz, where μ is the digital scheme from 0 to 5. Thus, digital scheme μ=0 has a subcarrier spacing of 15 kHz, and digital scheme μ=5 has a subcarrier spacing of 480 kHz. The symbol length / duration is inversely proportional to the subcarrier spacing. Figures 3A-3D Examples of slot configuration 0 with 14 symbols per slot and digital scheme µ=2 with 4 slots per subframe are provided. The slot duration is 0.25 ms, the subcarrier spacing is 60 kHz, and the symbol duration is approximately 16.67 μs.

[0186] A resource grid can be used to represent the frame structure. Each time slot includes a resource block (RB) (also called a physical RB (PRB)) extending for 12 consecutive subcarriers. The resource grid is divided into multiple resource elements (REs). The number of bits carried by each RE depends on the modulation scheme.

[0187] like Figure 3A As shown, some REs in the RE carry information for the UE (e.g., Figure 1 and 2The reference (pilot) signal (RS) for UE 104. The RS may include demodulation RS (DM-RS) (although indicated as Rx for a specific configuration, where 100x is the port number, but other DM-RS configurations are also possible) and channel state information reference signal (CSI-RS) for channel estimation at the UE. The RS may also include beam measurement RS (BRS), beam refinement RS (BRRS), and phase tracking RS (PT-RS).

[0188] Figure 3B Examples of various DL channels within a subframe of a frame are shown. The Physical Downlink Control Channel (PDCCH) carries the DCI in one or more Control Channel Elements (CCEs), each CCE comprising 9 RE Groups (REGs), each REG comprising 4 consecutive REs in an OFDM symbol.

[0189] The Primary Synchronization Signal (PSS) can be located within symbol 2 of a specific subframe of the frame. The PSS is generated by the UE (e.g., Figure 1 and 2 104) is used to determine subframe / symbol timing and physical layer identifier.

[0190] The secondary synchronization signal (SSS) can be located within symbol 4 of a specific subframe of the frame. The SSS is used by the UE to determine the physical layer cell identifier group number and radio frame timing.

[0191] Based on the Physical Layer Identifier and Physical Layer Cell Identifier Group Number, the UE can determine the Physical Cell Identifier (PCI). Based on the PCI, the UE can determine the location of the aforementioned DM-RS. The Physical Broadcast Channel (PBCH), carrying the Master Information Block (MIB), can be logically grouped with the PSS and SSS to form a Synchronization Signal (SS) / PBCH block. The MIB provides the number of RBs and the System Frame Number (SFN) in the system bandwidth. The Physical Downlink Shared Channel (PDSCH) carries user data, broadcast system information not transmitted via the PBCH (such as System Information Block (SIB)), and paging messages.

[0192] like Figure 3CAs shown, some REs in the REs carry DM-RS for channel estimation at the base station (indicated as R for a particular configuration, but other DM-RS configurations are possible). The UE can transmit DM-RS for the Physical Uplink Control Channel (PUCCH) and DM-RS for the Physical Uplink Shared Channel (PUSCH). PUSCH DM-RS can be transmitted in the first one or two symbols preceding the PUSCH. PUCCH DM-RS can be transmitted in different configurations depending on whether a short or long PUCCH is transmitted and the specific PUCCH format used. The UE can transmit a Sounding Reference Signal (SRS). SRS can be transmitted in the last symbol of a subframe. SRS can have a comb structure, and the UE can transmit SRS on one of the combs. The base station can use SRS for channel quality estimation to enable frequency-dependent scheduling on the UL.

[0193] Figure 3D Examples of various UL channels within a subframe of a frame are shown. The PUCCH can be positioned according to an indication in a configuration. The PUCCH carries uplink control information (UCI), such as scheduling requests, channel quality indicators (CQI), precoding matrix indicators (PMI), rank indicators (RI), and HARQ ACK / NACK feedback. The PUCCH carries data and can also be used to carry buffer status reports (BSR), power headroom reports (PHR), and / or UCI.

[0194] Additional Notes

[0195] The preceding description provides examples of techniques for generating and processing embedded representations of beams in a communication system. The foregoing description is provided to enable any person skilled in the art to implement the various aspects described herein. The examples discussed herein are not limited to the scope, applicability, or aspects set forth in the claims. Various modifications to these aspects will be apparent to those skilled in the art, and the general principles defined herein may also be applied to other aspects. For example, changes may be made to the function and arrangement of the elements discussed without departing from the scope of this disclosure. Various processes or components may be omitted, substituted, or added as appropriate in the examples. For example, the described methods may be performed in a different order than described, and steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined in some other examples. For example, any number of aspects set forth herein may be used to implement an apparatus or method of practice. Additionally, the scope of this disclosure is intended to cover such apparatus or methods implemented using other structures, functions, or structures and functions that complement or replace the various aspects of this disclosure set forth herein. It should be understood that any aspect of the disclosure herein may be embodied by one or more elements of the claims.

[0196] The techniques described in this article can be used in various wireless communication technologies, such as 5G (e.g., 5G NR), 3GPP Long Term Evolution (LTE), Improved LTE (LTE-A), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single Carrier Frequency Division Multiple Access (SC-FDMA), Time Division Synchronous Code Division Multiple Access (TD-SCDMA), and other networks. The terms "network" and "system" are often used interchangeably. CDMA networks can implement, for example, Universal Terrestrial Radio Access (UTRA), cdma2000, and other radio technologies. UTRA includes Wideband CDMA (WCDMA) and other variants of CDMA. cdma2000 encompasses the IS-2000, IS-95, and IS-856 standards. TDMA networks can implement radio technologies such as Global System for Mobile Communications (GSM). OFDMA networks can implement radio technologies such as NR (e.g., 5G RA), evolved UTRA (E-UTRA), Ultra Mobile Broadband (UMB), IEEE 802.11 (WiFi), IEEE 802.16 (WiMAX), IEEE 802.20, and Flash-OFDMA. UTRA and E-UTRA are part of the Universal Mobile Telecommunications System (UMTS). LTE and LTE-A are versions of UMTS using E-UTRA. UTRA, E-UTRA, UMTS, LTE, LTE-A, and GSM are described in documents from an organization called the 3rd Generation Partnership Project (3GPP), and cdma2000 and UMB are described in documents from an organization called the 3rd Generation Partnership Project 2 (3GPP2). NR is an emerging wireless communication technology under development.

[0197] The various illustrative logic blocks, modules, and circuits described in connection with this disclosure may be implemented or executed using a general-purpose processor, DSP, ASIC, field-programmable gate array (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but alternatively, the processor may be any commercially available processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors incorporating a DSP core, a system-on-a-chip (SoC), or any other such configuration.

[0198] If implemented in hardware, an example hardware configuration may include a processing system in a wireless node. This processing system can be implemented using a bus architecture. Depending on the specific application and overall design constraints of the processing system, the bus may include any number of interconnect buses and bridges. The bus can link various circuits together, including processors, machine-readable media, and bus interfaces. The bus interface can be used to connect network adapters, etc., to the processing system via the bus. The network adapter can be used to implement signal processing functions of the PHY layer. In the case of user equipment (see Figure 1), user interfaces (e.g., keyboards, displays, mice, joysticks, touchscreens, biometric sensors, proximity sensors, light-emitting elements, etc.) can also be connected to the bus. The bus can also link various other circuits, such as timing sources, peripherals, voltage regulators, power management circuits, etc., which are well known in the art and therefore will not be described further. The processor may be implemented using one or more general-purpose processors and / or special-purpose processors. Examples include microprocessors, microcontrollers, DSP processors, and other circuits capable of executing software. Those skilled in the art will recognize how best to implement the functions described for the processing system depends on the specific application and the overall design constraints imposed on the system.

[0199] If implemented in software, functionality can be stored on or transmitted via a computer-readable medium as one or more instructions or code. Software should be broadly interpreted as instructions, data, or any combination thereof, regardless of whether it is referred to as software, firmware, middleware, microcode, hardware description language, or other terms. Computer-readable media includes both computer storage media and communication media, with communication media encompassing any medium that facilitates the transfer of computer programs from one location to another. The processor may be responsible for managing the bus and general-purpose processing, including executing software modules stored on the machine-readable storage medium. The computer-readable storage medium may be coupled to the processor, allowing the processor to read information from and write information to it. Alternatively, the storage medium may be an integral part of the processor. By way of example, the machine-readable medium may include a transmission line, a carrier wave modulated by data, and / or a separate computer-readable storage medium on which instructions are stored, all of which may be accessible to the processor via a bus interface. Alternatively or additionally, the machine-readable medium or any portion thereof may be integrated into the processor, for example, in cases where it may have a cache and / or a general-purpose register file. As an example, examples of machine-readable storage media may include RAM (random access memory), flash memory, ROM (read-only memory), PROM (programmable read-only memory), EPROM (erasable programmable read-only memory), EEPROM (electrically erasable programmable read-only memory), registers, disks, optical disks, hard disks, or any other suitable storage media, or any combination thereof. Machine-readable media may be implemented in a computer program product.

[0200] Software modules can include single instructions or many instructions, and can be distributed across several different code segments, different programs, and across multiple storage media. Computer-readable media can include several software modules. A software module includes instructions that, when executed by a device such as a processor, cause the processing system to perform various functions. Software modules can include transmission modules and reception modules. Each software module can reside in a single storage device or be distributed across multiple storage devices. By way of example, when a triggering event occurs, a software module can be loaded from a hard disk drive into RAM. During the execution of a software module, the processor can load some instructions from the instruction set into a cache to improve access speed. One or more cache lines can then be loaded into a general-purpose register file for processor execution. When the functionality of a software module is referred to below, it will be understood that such functionality is implemented by the processor when executing instructions from that software module.

[0201] As used herein, the phrase “at least one of the items” refers to any combination of these items (including a single member). As an example, “at least one of a, b, or c” is intended to cover a, b, c, ab, ac, bc, and abc, as well as any combination of multiples of the same element (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other ordering of a, b, and c).

[0202] As used herein, the term "determine" encompasses a wide variety of actions. For example, "determine" can include calculation, operation, processing, deduction, investigation, lookup (e.g., searching in a table, database, or other data structure), ascertainment, etc. Moreover, "determine" can include receiving (e.g., receiving information), accessing (e.g., accessing data in memory), and so on. Furthermore, "determine" can include parsing, selecting, picking, building, etc.

[0203] The methods disclosed herein include one or more steps or actions for implementing the methods. The method steps and / or actions may be interchanged without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims. Furthermore, the various operations of the methods described above can be performed by any suitable unit capable of performing the corresponding function. Such units may include various hardware and / or software components and / or modules, including but not limited to circuits, application-specific integrated circuits (ASICs), or processors. Typically, where operations illustrated in the figures are present, those operations may have corresponding functional module components with similar numbering.

[0204] The appended claims are not intended to be limited to the aspects shown herein, but are given the full scope consistent with the wording of the claims. Within the claims, unless expressly stated otherwise, references to singular elements are not intended to mean “one and only one”, but rather “one or more.” Unless specifically stated otherwise, the term “some” refers to one or more. No claim element is to be interpreted pursuant to 35 USC §112(f) unless it is expressly recited using the phrase “for a unit of”, or, in the case of a method claim, using the phrase “for a step of”. All structural and functional equivalences of elements throughout the various aspects described in this disclosure that are known to a person skilled in the art or will be known later are expressly incorporated herein by reference and are intended to be covered by the claims. Furthermore, nothing disclosed herein is intended to be offered to the public, whether or not such disclosure is expressly referenced in the claims.

Claims

1. A method for wireless communication by a wireless node, comprising: Receive an embedding representing a low-dimensional representation of a beam, wherein the embedding is generated based on a representation associated with the beam; The embedding is then provided to the machine learning (ML) model; One or more communication parameters for communicating using the beam are generated based on the embedding via the ML model; and Communicate using one or more of the aforementioned communication parameters.

2. The method of claim 1, wherein, Receiving the embedding includes: receiving a lookup table indicating the embedding.

3. The method according to claim 1, wherein: The ML model includes a rotation transformer decoder; Generating the one or more communication parameters includes: determining the rotation of the beam for the communication based on the embedding via the rotation transformer decoder; as well as The one or more communication parameters include the determined rotated beam.

4. The method according to claim 3, further comprising: A rotation command is provided to the rotation converter decoder, wherein the rotation of the beam is determined based on the rotation command.

5. The method according to claim 1, wherein: The method further includes: receiving another embedding generated based on the representation of another beam; The embedding and the other embedding are provided to the point-to-point difference decoder; and Generating the one or more communication parameters includes: predicting the point-by-point difference between the beam and the other beam based on the embedding and the other embedding via the point-by-point difference decoder.

6. The method according to claim 1, wherein: The embedding is provided to the Reference Signal Received Power (RSRP) decoder; as well as Generating the one or more communication parameters includes determining the RSRP of each of a plurality of synchronization signal blocks (SSBs) based on the embedding via the RSRP decoder.

7. The method of claim 6, further comprising: The RSRP decoder is provided with an indication of the orientation of the wireless node, wherein the RSRP is determined based on the orientation.

8. The method of claim 1, wherein, The characterization includes the spherical array gain.

9. The method of claim 1, wherein, The characterization includes array gain measurement or codebook beam characterization metric.

10. The method of claim 1, further comprising: An encoder is trained, which is configured to generate an embedding using the representation associated with the beam.

11. A method for wireless communication, comprising: Receive the beam-associated representation; An embedding representing a low-dimensional representation of the beam is generated based on the representation; as well as The embedding is provided to the wireless node.

12. The method of claim 11, wherein, Providing the embedding includes providing a lookup table indicating the embedding associated with the beam.

13. The method according to claim 11, wherein: The characterization includes the spherical array gain; The method further includes: converting the spherical array gain into a graphic; and The embedding is generated based on the graph using a graph convolutional network.

14. The method of claim 11, further comprising: Receive one or more training codebooks for training an autoencoder, wherein the embedding is generated using an encoder of the autoencoder.

15. The method of claim 11, wherein, The characterization includes array gain measurement or codebook beam characterization metric.

16. The method according to claim 15, wherein, The representation is received as a noise-enhanced input, and the embedding is generated using a denoising autoencoder based on the noise-enhanced input.

17. The method of claim 11, wherein, Generating the embedding includes downsampling the representation using a Fourier transform.

18. The method of claim 11, wherein, The characterization includes the spherical array gain.

19. The method of claim 11, further comprising: A federated learning model is used to train the autoencoder, wherein the embedding is generated using the encoder of the autoencoder.

20. An apparatus for wireless communication by a wireless node, comprising: Includes memory for executing instructions; as well as One or more processors are configured to execute the executable instructions and cause the device to perform the following operations: Receive an embedding representing a low-dimensional representation of a beam, wherein the embedding is generated based on a representation associated with the beam; The embedding is then provided to the machine learning (ML) model; One or more communication parameters for communicating using the beam are generated based on the embedding via the ML model; and Communicate using one or more of the aforementioned communication parameters.

21. The apparatus of claim 20, wherein, When the device receives the embedding, the one or more processors are configured to cause the device to receive a lookup table indicating the embedding.

22. The apparatus according to claim 20, wherein: The ML model includes a rotation transformer decoder; When the device generates the one or more communication parameters, the one or more processors are configured to cause the device to determine the rotation of the beam for the communication based on the embedding via the rotation converter decoder; as well as The one or more communication parameters include the determined rotated beam.

23. The apparatus of claim 22, wherein, The one or more processors are further configured such that the device: provides rotation instructions to the rotation converter decoder, wherein the rotation of the beam is determined based on the rotation instructions.

24. The apparatus of claim 20, wherein: The one or more processors are further configured such that the device: receives another embedding generated based on a representation of another beam; The embedding and the other embedding are provided to the point-by-point difference decoder; as well as When the device generates the one or more communication parameters, the one or more processors are configured such that the device: predicts the point-by-point difference between the beam and the other beam based on the embedding and the other embedding via the point-by-point difference decoder.

25. The apparatus according to claim 20, wherein: The embedding is provided to the Reference Signal Received Power (RSRP) decoder; as well as When the device generates the one or more communication parameters, the one or more processors are configured such that the device: determines the RSRP of each of a plurality of synchronization signal blocks (SSBs) based on the embedding via the RSRP decoder.

26. The apparatus according to claim 25, wherein, The one or more processors are further configured to cause the device to: provide the RSRP decoder with an indication of orientation toward the wireless node, wherein the RSRP is determined based on the orientation.

27. The apparatus of claim 20, wherein, The characterization includes the spherical array gain.

28. The apparatus of claim 20, wherein, The characterization includes array gain measurement or codebook beam characterization metric.

29. The apparatus of claim 20, wherein, The one or more processors are further configured such that the device: trains an encoder, the encoder being configured to generate an embedding using the representation associated with the beam.

30. A non-transitory computer-readable medium including executable instructions that, when executed by one or more processors of a wireless node, cause the wireless node to perform the following operations: receiving an embedding representing a low-dimensional representation of a beam, wherein, The embedding is generated based on a representation associated with the beam; The embedding is then provided to the machine learning (ML) model; One or more communication parameters for communicating using the beam are generated based on the embedding via the ML model; and Communicate using one or more of the aforementioned communication parameters.

Images