Configurable uplink control information (UCI) reporting based on artificial intelligence (AI)

Abstract

A user equipment (UE), a base station, a baseband processor, or other network device may operate in a network to handle a network configuration of an uplink control information (UCI) payload based on artificial intelligence (AI) channel state information (CSI) feedback. The network configuration includes indications of the UCI configuration and the AI model in order to generate an AI-based UCI report. In accordance with the network configuration, the AI model is associated with at least one of a layer common configuration, a layer specific configuration, or a rank indication (RI) specific configuration.

Description

[0001] Citation of Related Applications

[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 485,450, filed on February 16, 2023, the contents of which are hereby incorporated by reference in their entirety. Technical Field

[0003] The present disclosure relates to wireless communication networks and mobile device capabilities. Background Art

[0004] Wireless communication networks and wireless communication services are becoming increasingly dynamic, complex and ubiquitous. For example, some wireless communication networks may be developed to implement fifth-generation (5G) or new radio (NR) technology, sixth-generation (6G) technology, and the like. Such technologies may include solutions for enabling user equipment (UE) and network equipment (such as base stations) to communicate with each other. Large-scale multiple-input multiple-output (MIMO) equipping base stations (BSs) with many antennas can significantly improve system performance. However, the benefits of large-scale MIMO are based on the understanding of channel state information (CSI) feedback. In frequency division duplex (FDD) large-scale MIMO systems, due to the lack of channel reciprocity, UEs operate to feed back downlink CSI to the BS via the uplink. Due to the high dimensionality of CSI in large-scale MIMO systems, the feedback overhead may be large. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 Illustrated are examples of signaling diagrams for configuration of UCI reporting according to various aspects.

[0006] Figure 2 The structure of CSI feedback based on automatic encoder / decoder is illustrated.

[0007] Figure 3 Illustrated are examples of AI configurations / methods for AI UCI-based reporting according to various aspects.

[0008] Figure 4 AI model data defined according to various aspects is exemplified.

[0009] Figure 5 Illustrated are examples of AI configurations / methods for AI UCI-based reporting according to various aspects.

[0010] Figure 6 Illustrated are examples of AI configurations / methods for AI UCI-based reporting according to various aspects.

[0011] Figure 7 Illustrated are examples of AI configurations / methods for AI UCI-based reporting according to various aspects.

[0012] Figure 8 Examples of CSI feedback in AI-based UCI reporting according to various aspects are illustrated.

[0013] Figure 9 Examples of AI model ID lists according to various aspects are illustrated.

[0014] Figure 10 An exemplary block diagram illustrating an example of a user equipment (UE) communicatively coupled with a network component as a peer device via a network that can be used in conjunction with various embodiments (aspects) described herein.

[0015] Figure 11 Illustrated is an example simplified block diagram of a user equipment (UE) wireless communication device or other network device / component (e.g., eNB, gNB) according to various aspects. DETAILED DESCRIPTION

[0016] The following detailed description refers to the accompanying drawings. The same reference numerals in different drawings may identify the same or similar features, elements, operations, etc. Additionally, the present disclosure is not limited to the following description, as other specific implementations may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure.

[0017] Various aspects include methods and configurations for uplink (UL) control information (UCI) reporting for channel state information (CSI) feedback using artificial intelligence (AI). Compared to codebook-based and compressed sensing (CS) feedback algorithms, AI models automatically compress and reconstruct CSI to significantly improve feedback accuracy. AI provides machines or systems with the ability to dynamically simulate human intelligence and behavior. Unlike conventional methods supported by domain knowledge and theoretical proof, AI-based CSI methods (including deep learning and reinforcement training) automatically learn or extract features from training datasets. To efficiently learn features from datasets, neural network (NN) architectures (such as dense, convolutional, and recurrent NNs) are being developed and introduced into wireless communications, particularly to integrate AI into air interfaces. AI and machine learning (ML) are used interchangeably, with ML being considered a subfield of AI research. Typical implementations of AI / ML utilize NNs, such as conventional neural networks (CNNs), recurrent / recurrent neural networks (RNNs), and generative adversarial networks (GANs). The following description may use a neural network as an example of an AI / ML model; however, it should be understood that the AI / ML models discussed herein may not necessarily be limited thereto, and any other model that performs inference on the UE or network side of the network is possible.

[0018] In one aspect, a UE may be operable to receive a network configuration for a UCI payload for AI-based CSI feedback, as provided by a base station. The network configuration may include an indication of a UCI configuration and an AI model. These indications provided by the network configuration may be used to generate an AI-based UCI report for AI-based CSI feedback. Specifically, the network configuration indicates to the UE how to report UCI bits through the UCI configuration and AI model. The UCI configuration indicates the number of UCI bits, and the AI ​​model indicates how the UCI bits are encoded by an encoder for signaling to the network or base station. The UCI bits of the AI-based UCI report are then decoded by a decoder component at the base station. The AI-based UCI report is generated by the UE based on the network configuration and then provided to the base station to enable efficient handling of CSI for wireless transmission. The network configuration may also indicate an AI method or configuration for encoding UCI for compressed AI-based CSI feedback in the AI-based UCI report. For example, the AI ​​method or AI configuration used for AI-based UCI reporting may include a layer-specific configuration, a layer-common configuration, or a rank-specific configuration.

[0019] In another aspect, a network device or base station may provide a network configuration for a UCI payload to the UE. The network configuration may indicate the UCI configuration and the AI ​​model. For example, the network configuration may include an AI model ID indicating the UCI configuration and the AI ​​model. The network may then receive an AI-based UCI report for AI-based CSI feedback based on the provided network configuration.

[0020] The AI ​​model for AI-based CSI feedback compression in AI-based UCI reporting described herein can be configured as a two-sided AI model for AI-based wireless communications. Compared to codebook-based and compressed sensing-based feedback algorithms, the two-sided model can significantly improve feedback accuracy and is also used to potentially replace the old e-Type II codebook-based encoding and decoding with neural network (NN)-based encoders and decoders, respectively. Therefore, the existing CSI feedback standard may face major revisions, and the need for UCI encoding rules for AI-based CSI feedback compression still exists along with the structure and priority of UCI omission rules for managing UCI payload bit size.

[0021] Other aspects and details of the disclosure are further described below with respect to the accompanying drawings.

[0022] Figure 1An example signal flow 100 is illustrated for configuring AI-based UCI reporting for CSI feedback or AI-based CSI feedback between a UE 110 and a network device or base station 122 (e.g., a gNB or other network component) to enable efficient wireless communication between antennas. The UE 110 may include a smartphone (e.g., a handheld, touchscreen mobile computing device capable of connecting to one or more wireless communication networks). Additionally or alternatively, the UE 110 may include other types of mobile or non-mobile computing devices configured for wireless communication, such as a personal data assistant (PDA), a pager, a laptop, a desktop computer, a wireless handset, etc. The UE 110 may include one or more UEs 110, such as Internet of Things (IoT) devices (or IoT UEs). Additionally or alternatively, an IoT UE may utilize one or more types of technologies, such as machine-to-machine (M2M) communication or machine-type communication (MTC) (e.g., for exchanging data with an MTC server or other device via a public land mobile network (PLMN)), proximity services (ProSe) or device-to-device (D2D) communication, sensor networks, IoT networks, etc. Additionally, UE 110 may include a vehicle UE, a pedestrian UE, or a vehicle-to-everything (V2X) UE.

[0023] The UE may perform an estimation of the channel for CSI feedback after precoding (or beamforming). A codebook for CSI is typically supported for CSI beam selection, co-phasing between polarizations, and beam combining operations. CSI may include channel quality information (CQI), rank indication (RI), and precoding matrix indication (PMI). CQI may, for example, provide information about the signal to noise and interference ratio (SINR) of a transmission channel or layer. Based on the CSI recommended by the UE, the base station applies an appropriate modulation and coding scheme (MCS) to subsequent UE transmissions. A low SINR for one transmission channel may result in the use of a low MCS relative to another transmission channel with a higher or better SINR. The base station and UE then communicate using the recommended MCS, at least until the channel conditions between them change. Other components including PMI and RI are associated with multiple-input multiple-output (MIMO) transmission, where MIMO generally refers to multiple sets of antennas on either or both of a receiver / transmitter in a wireless cellular system.

[0024] Typically, the MIMO feedback framework includes an e-Type II codebook design, where UCI coding and omission rules are specified for the following two parts of CSI: CSI Part 1 and CSI Part 2. CSI Part 1 includes RI, wideband CQI, subband CQI, and the total number of non-zero coefficients (K NZ,TOT). CSI part 2 includes three groups of data with the following priority: group 1> group 1> group 2. When the specified UCI bit threshold is exceeded, the omission rules for omitting UCI bits operate according to the group priority, including layer> spatial basis> frequency basis. In the MIMO framework, multiple beams (also referred to as spatial spatial layers, MIMO layers or transmission layers in this article) can be sent on the same time and frequency resources to maximize spectral efficiency. The transmission layer or MIMO spatial layer in this article may refer to the codebook layer of the MIMO spatial and / or frequency precoder used for transmission. RI indicates the number of MIMO spatial layers that can be sent to the UE simultaneously. Thus, for example, if the UE is within the line of sight to the base station, the base station can send two signals simultaneously on the same time and frequency resources by using, for example, two orthogonal polarizations. Due to their orthogonality, the two signals will not interfere with each other even if the same time and frequency resources are used. AI-based UCI reporting of CSI feedback (or as AI-based CSI feedback compression) is expected to improve performance and reduce complexity. AI-based CSI feedback as mentioned in this article includes deep learning to automatically compress and reconstruct CSI while significantly improving feedback accuracy. Specifically, for AI-based UCI reporting, UCI encoding and omission rules remain to be specified, which may be beneficial for utilizing AI-based models for efficient and effective management of massive MIMO feedback. AI-based UCI coding does not utilize non-zero coefficient processes and does not implement a clear structure on a frequency basis or a spatial basis for adaptive control of beam patterning.

[0025] Signaling flow 100 may initiate the provision of a UE capability report 102 by UE 110 to base station 122. UE capability report 102 may indicate support for one or more AI models, one or more AI model IDs associated with one or more AI models, or one or more UCI configurations for generating AI-based UCI reports for AI-based CSI feedback. Prior to deploying AI-enabled CSI feedback processes and systems, many issues can be addressed by training with datasets from various UEs specific to a particular owner and network vendor. The specific AI model used to configure CSI feedback, and therefore AI-based UCI reporting, will likely be left to individual design. Therefore, the design of the AI ​​model and its specific operation will be determined by each vendor's partners and, from a signaling perspective, may involve pre-training AI models prior to deployment, involving trials with paired encoder and decoder components. During deployment, when UE 110 is processing CSI feedback with interference processes, depending on which model format is pre-trained or supported, the corresponding AI model may be communicated between UE 110 and network base station 122 at 102 to establish which paired link encoder-decoder process to execute. Therefore, statistics related to dataset generation, the dataset itself, AI model design, method training, last function and other optimization processes can be implementation specific and need not be specified or disclosed except for UE capability reporting.

[0026] Prior to deployment, the UE 110 and the base station 122 may perform offline training of a set of paired models for the encoder (UE 110) and the decoder (network) base station 122. Each AI model may be associated with a unique model ID. The offline UE vendor and the NW vendor may have agreed to support one or more AI model methods / configurations during training. These AI model configurations may be layer-common configurations, layer-specific configurations, rank-specific configurations, or any combination thereof. The UE capability report signaling 102 may include one or more supported AI model IDs, respectively, associated with the AI ​​models.

[0027] The AI ​​model ID herein may include a UE vendor ID, a network device vendor ID, a use case ID, a sub-use case ID, a model ID for the use case, an input / output format, a UCI configuration for the number of UCI bits, an AI model associated with an AI model ID or an AI model index, and an AI model configuration. The supported UCI configuration may not support all UCI configurations, but rather a set of UCI configurations that supports one or more UCI configurations in a set of specified UCI configurations (e.g., 60 bits, 80 bits, 90 bits, 120 bits, 240 bits, or other amounts of UCI bits). The AI ​​configuration may refer to a layer-common configuration, a layer-specific configuration, and an RI or rank-specific configuration, or any combination thereof. Depending on the model ID definition, the AI ​​model ID or network configuration may represent or indicate any one or more of the above-mentioned information associated with the AI ​​model ID. Therefore, each change in the number of UCI bits of a UCI configuration or a change in the AI ​​configuration may be associated with a unique model ID. In one aspect, the network may therefore provide only the AI ​​model ID or the model ID for providing one or more indications of the UCI configuration, the AI ​​model, the AI ​​configuration, and one or more of the following: UE vendor ID, network equipment vendor ID, use case ID, sub-use case ID, model ID of the use case, or input and output formats.

[0028] Base station 122 provides network configuration signaling 104, which may be based on or derived from a UE capability report, for example. The network configuration may include an AI model ID, a UCI configuration, or an AI model to configure AI-based UCI reporting at UE 110. The network configuration may be based on dynamic configuration of network cells or analysis of the environment of the network with UE 110. Given that wireless channels depend on the propagation environment, the channels of a particular cell may exhibit specific characteristics that can be considered as environmental knowledge. AI can be a tool that can be used to effectively assist in extracting and utilizing environmental knowledge to facilitate CS I feedback via end-to-end learning.

[0029] In response to receiving a network configuration associated with AI-based UCI reporting, UE 110 generates an AI-based UCI report 106 for CSI feedback to facilitate wireless communication with base station 122. The AI ​​model is used to generate UCI bits for AI-based UCI reporting with CSI feedback according to the AI ​​configuration / method. If a maximum UCI bit payload is provided, UE 110 may further configure UCI omission rules to omit some bits but not others based on a UCI omission rule threshold.

[0030] CSI feedback reduction methods may lead to excessive feedback overhead, especially in massive MIMO implementations. Autoencoder / decoder-based CSI feedback enhancement is an example of an approach used to address this challenge. Figure 2The abstract level structure of CSI feedback based on automatic encoder / decoder is illustrated. Figure 2 As shown, on the UE side, the pre-processed CSI input is encoded by an encoder component, which can be an AI model. The CSI input is then quantized by a quantizer and sent to the network. On the network side, the CSI feedback is dequantized by a dequantizer and decoded by a decoder to calculate the pre-decoder, which can also be an AI model.

[0031] An autoencoder / decoder-based approach trains the entire encoder and decoder NNs via deep learning to minimize a total loss function between the decoder output and the encoder input. Encoder / decoder training is centralized, while inference functionality is split between the UE and the NG-RAN node (e.g., gNB), with encoder inference at the UE 110 and decoder inference at the gNB or base station 122. To achieve this, UE-gNB coordination and over-the-air model transfer, as discussed before deployment, can be facilitated in the signaling flow 100.

[0032] In this example, the NN, which includes both an encoder and a decoder, is a two-sided model. If the NN is trained and owned by the network equipment vendor, for example, a portion of the NN (i.e., the autoencoder for inference at the UE) is downloaded to the UE 110. If the NN is trained and owned by the UE vendor, for example, a portion of the NN (i.e., the autodecoder for inference at the gNB) is uploaded to the gNB or base station 122. Alternatively, the NN may be trained and owned by a third party, such that the two portions of the NN are delivered to the UE 110 and the base station 122, respectively.

[0033] Alternatively, the autoencoder or autodecoder can be trained separately as a single-sided model. For example, the UE vendor can train only the encoder NN based on downlink measurement data in different cells, and the network equipment vendor can train only the decoder NN based on uplink data for different UEs. In this case, UE 110 and gNB 122 can obtain the corresponding NN from the UE vendor's server or the network equipment vendor's server, respectively.

[0034] On the one hand, the network may have different deployments, such as indoor, Umi or Uma deployment, different numbers of antennas deployed in the cell, single TRP (sTRP) or multiple TRP (mTRP), and therefore, multiple NNs may be trained to achieve flexible adaptive codebook design to optimize system performance. On the other hand, UE 110 may have different individual AI capabilities or memory limitations, and therefore, multiple NNs may be trained with various AI models to adapt to UE differences.

[0035] Figure 3 Examples of various AI configurations 300 for generating AI-based UCI reports based on network configuration are illustrated. The network configuration indicates one or more of the following: an AI model ID, an AI model, or a UCI configuration. Various AI models are illustrated based on hash tag differentiation, including, for example, AI models 1 to 4, but a fewer or greater number of AI models may be utilized and configured depending on the network configuration performed by the base station 122. AI models 1 to 4 are hashed in different ways and illustrated according to corresponding AI configurations and corresponding ranks. The AI ​​configuration may include, for example, a layer-common configuration 310, a layer-specific configuration 320, or an RI-specific configuration 330 illustrated from top to bottom, and corresponding ranks or RIs for RI=1 350, RI=2 360, RI=3 370, and RI=4 380.

[0036] Each AI model includes an encoder and a decoder as an encoder-decoder pair, where one or more inputs such as V# (e.g., V1, etc.) provide one or more outputs such as primed V numbers (e.g., V1'). Once the channel is acquired via a cell-specific reference signal (CS-RS), the subband covariance matrix can be determined based on the subband size configuration. For example, if there are 72 antennas at the network, a 72×72 covariance matrix for each subband can be determined, and then a factor V of the covariance matrix can be calculated, which factor V can be represented by the inputs (e.g., V1, V2, V3, V4, etc.) and the outputs such as primed V (e.g., V1', V2', V3', V4', etc.). However, this calculation process may depend on the specific implementation, and any specific AI model may be configured according to an AI / ML or deep learning process, as represented by an AI model with different hash tags. Therefore, how the UE 110 performs measurements and calculates the factors V or V' may depend on various specific implementations.

[0037] Each AI model or AI model ID may include a model index to indicate a specific AI model associated with the UCI configuration and the indicated UCI configuration. For example, the AI ​​model ID of model 1-1 includes, for example, a first indication 312 indicating a first AI model (or AI model 1) associated with a second indication 314 (such as UCI configuration index 1 indicating a specific UCI configuration associated with the AI ​​model). The UCI configuration 314 indicates the number of UCI bits for the UCI payload of the AI-based UCI report or the number of UCI bits for UE feedback based on the network configuration. For example, UCI configuration 1 may be 60 bits, UCI configuration 2 may be 80 or 90 bits, and configuration 3 may indicate, for example, 120 bits. These are examples of UCI configuration indications 314, and the model index indication 312 of the associated UCI configuration 314 indicates, for example, for 60 bits, whether the UE executes one AI model, two AI models, or multiple AI models according to the model index indication 312 in its AI-based UCI report.

[0038] To report AI-based UCI reports, UE 110 may configure a layer-common configuration 310 based on the network configuration. Based on the AI ​​configuration, the layer-common configuration 310 is configured to correspond to one AI model for each UCI payload. Thus, each MIMO layer or transmission layer reported in an AI-based UCI report uses the layer-common configuration 310 for the AI ​​configuration used for each RI 350 to 380. The layer-common configuration 310 indicates that for any given RI 350 to 380, the same AI model will be used across any number of transmission layers. For example, when RI=1 at RI 350, the transmission layer uses AI model 1, which is associated with a specific AI model for generating UCI bits for reporting that transmission layer in an AI-based UCI report. When RI=2 at RI 360, all transmission layers use the same AI model 1. When R=3 at RI 370, all three transmission layers use the same AI model 1. Similarly, when R=4 at RI=380, all four transmission layers use the same AI model (AI model 1).

[0039] When a layer-common configuration 310 is utilized to generate an AI-based UCI report, where RI=1 is fed back at 350, the output of the V1 factor may be passed through the AI ​​model once to feed back data V1' in the AI-based UCI report. Similarly, if a rank value RI=2 is indicated, the same AI model is used to feed back two layers of V factors (V1' and V2') by first processing through a first V factor or V1 and then through a second V factor. The outputs may be summed, and all UCI bit outputs may be based on the same format. Similarly, the same process may be applied to 3 layers and 4 layers, as additionally associated with RI=3 and 4, respectively. Compared to the layer-specific configuration 320 and the rank or RI-specific configuration 330, the layer-common configuration may be particularly less complex, which may have advantages in terms of performance or memory savings.

[0040] Alternatively or additionally, to report AI-based UCI reporting, the UE 110 may configure a layer-specific configuration 320 based on a network configuration received from the base station 122. The layer-specific configuration 320 for AI-based UCI reporting may be a variation of the layer-common configuration, except that the layer-specific configuration includes using the same AI model for one transmission layer per UCI configuration. For example, the encoder-decoder pair of the layer-specific configuration 320 with RI=1 at 350 utilizes AI model 1, while the UCI configuration here may be, for example, 60 bits for AI-based UCI reporting. Similarly, each layer 1 of each RI 350 to 380 (V1, V2' as representative input and output) also includes the same AI model for the same UCI configuration (e.g., 60 UCI bits). If the AI-based UCI feedback has two layers, the first layer is passed through a model dedicated to and trained for the first factor (V1), and then for the second MIMO spatial layer, another AI model is dedicated to and trained for the second factor (V2). Therefore, the layer-specific operations include different encoder-decoder pairs having different statistical operations for each layer. Training different AI models for each layer can maximize the performance of generating AI-based CSI feedback in AI-based UCI reports. Depending on the selected MIMO spatial layer (V1, V2, V3 or V4 input of the encoder-decoder pair), different AI models can perform inference operations for different AI model IDs, and the output of each AI model can then be processed or compressed into a UCI report according to the UCI report format.

[0041] Alternatively or additionally, in order to report AI-based UCI reports, the UE 110 may configure a rank or RI-specific configuration 330 based on the network configuration received from the base station 122. The RI-specific configuration 330 includes using a different AI model per RI to generate an AI-based UCI report. In the RI-specific configuration, the input is not a V factor of a neural network that takes into account different levels of rank and outputs them together (e.g., V1' to V4'). Thus, if the input is an RI of 1, one AI model is used; for RI=2, a separate AI model is used, and the same is true for ranks 3 or 4, or even additionally, other different AI models are used. In this case, the rank or RI may be decided / selected by the UE 110, and therefore, the AI ​​model is then determined by the UE itself.

[0042] The AI ​​models (e.g., NN) available to UE 110 or base station 122 (e.g., gNB) can be trained by different entities. For example, UE 110 or base station 122 can receive multiple different models and store them in local memory. One or more of these models can be activated for appropriate use. For example, the network can activate, deactivate, or switch AI models at UE 110 via signaling. Alternatively, UE 110 can select the AI ​​model to use and notify the network of its selection in the UE capability report or AI-based UC I report signaling.

[0043] refer to Figure 4 , illustrates an example of AI model data 400, a model description using metadata, and a model ID after offline training between a UE and a NW. The AI ​​model ID is reported in a network configuration, a UE capability report, or together with an AI-based UCI report. In one aspect, a unique model ID or AI model ID may be assigned to each of the AI ​​models. The model ID is used to unambiguously identify the AI ​​model, for example, within a public land mobile network (PLMN) or across several PLMNs. The AI ​​model ID may include an AI configuration or method, which includes one or more of the following: a network equipment vendor identifier, a UE vendor identifier, a PLMN ID, a use case ID, a number of an AI model for the use case, and a UCI configuration for the number of UCI bits to be used for CSI reporting or feedback, a specific AI model associated with the indicated UCI configuration, and an AI method or AI configuration for reporting (e.g., a layer-common configuration, a layer-specific configuration, or an RI-specific configuration).

[0044] The network equipment vendor identifier can indicate the network vendor that has trained the AI ​​model, and the UE vendor identifier can indicate the UE vendor that has trained the AI ​​model. The PLMN ID can indicate the operator network to which the AI ​​model is applied. In addition, the use case ID can indicate the use case for which the AI ​​model is targeted. If there is more than one AI model or AI configuration for a specific use case, the number of the AI ​​model used for that use case can be used to distinguish these AI models. It should be understood that not all of the above items are currently required or available. The definition of the AI ​​model ID can be specific to the operator network for local differentiation, or it can be provided in the specification for global differentiation.

[0045] AI models can be stored and transmitted as model data, which includes a model file associated with an AI model ID and metadata. Metadata is generated to describe the corresponding AI model and may indicate various information about the AI ​​model, including but not limited to: training status: the network being trained and tested, and an indication of the AI ​​model's potential training dataset; AI model functionality / objects, inputs / outputs; one or more latency benchmarks, memory requirements, and accuracy for the AI ​​model; compression status of the AI ​​model; inference / operation conditions: urban, indoor, dense macro; and pre- and post-processing of measurements of AI inputs / outputs. The model file or model ID may contain or indicate model parameters used to construct the AI ​​model or UCI report, respectively, based on the UCI payload of UCI bits output by the AI ​​configuration / method. In the case of a deep neural network, the model file may include the layers and weights / biases of the neural network. The model is saved in a file depending on the machine learning framework used. For example, a first machine learning framework may save a file in a first format, while a second machine learning framework may save a file in a different format to represent the ML model.

[0046] Due to the varying model formats currently used in the AI ​​industry, it is expected that models trained by different vendors may have different formats. Model files may need to be reformatted before, during, or after being delivered to a UE or gNB. Assuming that an AI model is stored in a first format after training, but the UE or gNB may support a second format different from the first, format conversion may be necessary. For example, the server storing the model may convert the model file format to the second format before sending the model. In another example, a network function (NF) in the core of an operator's network may convert the model file format before forwarding the model to a gNB. In another example, the gNB may be responsible for converting the format of the model destined for the gNB or UE. In the latter case, the gNB then forwards the reformatted model to the UE. In another example, the UE converts the model file format based on the UE's supported capabilities. AI model data may be compressed for storage and / or migration, for example, using the standard compression methods provided in ISO-IEC 15938-17 or any other possible compression methods, which are not described in detail here.

[0047] Figure 5 An example of an AI configuration or method for generating an AI-based UCI report based on a layer-common configuration 500 is illustrated. In the case where only one model is trained per UCI configuration (or the number of configured UCI bits), although each MIMO spatial layer (or transmit layer) can be configured with a different payload, it results in a different AI model. From the perspective of the AI ​​model, this can be classified as a layer-common configuration, as in the present disclosure. Alternatively, from the perspective of the UCI configuration, this can also be considered a layer-specific configuration. If there is only one model per UCI configuration, the layer-common configuration can be viewed from different perspectives. From the perspective of the AI ​​model, it is layer-common, but from the perspective of the configuration, there can still be different AI models. For example, the first MIMO layer may use 60 bits as the UCI configuration, and then the second layer uses 90 bits. From the perspective of the configuration, there are still two unique network model IDs associated with the configuration of each layer.

[0048] Specifically, layer-common configuration involves using the same AI model across multiple transmission layers of a UCI configuration, which is applicable to each of the AI ​​configuration methods 510 to 540 illustrating example layer-common configurations. Layer-common configuration refers to the case where a single / identical AI model is trained for all layers of each UCI configuration. A UCI configuration has a single UCI bit size per layer. For example, a UCI configuration can be selected from a set of UCI payload sizes (e.g., 60 bits, 80 bits, 90 bits, 120 bits, 240 bits, or another number of UCI bits for a layer). The number of bits indicated by the UCI configuration is not limited to the illustrated example but can also be any other number of UCI bits or set of UCI bits. Here, a particular UCI configuration may only have a single model associated with it. Therefore, if a 60-bit UCI configuration exists anywhere, the same AI model can be utilized. Similarly, if the UCI configuration is 120 bits, the same AI model is utilized regardless of where 120 bits are used. Consequently, for 240 bits, the same AI model is associated, which is different from the AI ​​models of other UCI configurations with different bit sizes. From an AI perspective, this may mean that for each of the configurations, an associated AI model is used in the AI-based UCI reporting.

[0049] In one aspect, for example, a first layer-common configuration 510 for AI-based UCI reporting may span one or more RI configurations with RI values ​​of 1 to 4 across ranks. The base station 122 or NW may configure one AI model ID (i.e., the same UCI bits for all transmission layers), depending on the UE rank selection. The UCI payload size may increase linearly with increasing RI. For example, the UE 110 may be operable to generate UCI bits for AI-based UCI reporting based on a network-configured AI model ID. The AI ​​model ID may indicate a layer-common configuration for configuring a UCI payload size that increases linearly with increasing RI values ​​associated with the UCI configuration. In the first AI configuration, only one AI model is configured, and the UE 110 may select an RI. Depending on the selected RI, the UE 110 may process the same AI model across each transmission layer individually, and linearly increase the payload or UCI payload based on the RI index or value. From a configuration perspective, the UE provides the UCI bits of the UCI configuration in the UCIPUCCH feedback; therefore, a sufficiently large UCI payload is enabled so that the UE has the flexibility to adapt those bits, or configure omission rules to handle the limitation or omission of UCI bits that meet or exceed the threshold.

[0050] Alternatively or additionally, the second layer common configuration 520 for AI-based UCI reporting may be configured across one or more RI configurations with RI indexes or values. For example, the base station 122 or NW may configure one AI model ID for a maximum (max) RI = 2 and another model ID for a max RI = 4, as signaled via network configuration. Depending on the UE rank selection, the UE 110 may select a corresponding model based on the rank, such as a first AI model ID or another second model ID. For example, if the UE 110 selects rank 1 or 2, it will select the first AI model ID, and if RI = 3 or 4, the UE 110 selects another second model ID using rank 3 or 4. The specific AI model ID selected will apply to all transmission layers / MIMO spatial layers within the rank. Since there may be different AI model IDs assigned to a set of ranks, such as one AI model ID for max RI = 2 and another AI model ID for max RI = 4, the UCI configuration or number of bits may be doubled across ranks. For example, the UCI bits for rank 2 may be twice the size of those for rank 1. For RI=3 and 4, the same number or a higher number of UCI bits relative to RI=2 or when compared to RI=2 can be configured based on the network configuration. Thus, for example, rank 4 and rank 2 can have the same or exactly the same payload size. In general, the AI ​​configuration for the second layer common configuration 520 can minimize the maximum number of UCI bits that the UE 110 can transmit on the uplink in the UCIPUCCH, especially for higher RIs or ranks with more MIMO spatial layers (e.g., layers 1 to 4). Therefore, this AI configuration aspect can provide some flexibility and is consistent with the legacy design to constrain the overhead of higher layers. Compared to the first layer common configuration 510, where the UCI payload increases linearly, the second layer common configuration can reduce the amount of overhead of using ranks 3 and 4.

[0051] In the case of the second layer common configuration 520, UE 110 may select an AI model ID indicated in the network configuration from the AI ​​model IDs based on the RI and UCI configuration to generate UCI bits for AI-based UCI reporting. The AI ​​model ID may indicate the layer common configuration to configure the UCI payload size in the network configuration, such as through one or more indications or by utilizing the AI ​​model ID. The AI ​​model ID may include a first AI model ID associated with a first set of RIs and a second AI model ID associated with a second set of RIs having a greater RI value than the first set of RIs. The first AI model ID and the second model ID may each indicate different AI models that constrain the UCI payload size of the second set of RIs relative to the first set of RIs.

[0052] Alternatively or additionally, the third layer common configuration 530 for AI-based UCI reporting can be configured across one or more RIs with RI indexes or values. The base station 122 can configure an AI model per RI. Depending on the RI, the UE 110 can select an AI model for encoding each transmission layer. The UCI for each rank can be different. Therefore, for RI=1, the network can configure or the UE selects one AI model, and for RI=2, the network can configure or the UE selects another AI model, wherein for RI=2, RI=3, and RI=4, the same AI model can be configured for each layer. Therefore, the third layer common configuration 530 provides a certain degree of flexibility and adaptability to make the UCI payload size more constant. Here, the UE 110 is configured to select an AI model to encode each transmission layer based on the RI with the layer common configuration indicated by the network configuration. The UCI configuration is different between RIs and is associated with multiple UCI bits together with the AI ​​model. Therefore, each RI is configured with the same AI model and UCI configuration.

[0053] Alternatively or additionally, the fourth layer common configuration 540 for AI-based UCI reporting can be configured across one or more RIs. Base station 122 can configure an AI model per layer per RI. Depending on the RI, different UCI bits per layer are possible. Here, based on the network configuration, UE 110 can encode each transmission layer based on the AI ​​model and the RI. The UCI configuration may or may not vary across transmission layers of the RI.

[0054] As described above, the network or base station 122 may control the AI ​​model to be used by the UE 110 based on one or more indications of the network configuration, including an AI model ID, a UCI configuration, an AI model, an AI configuration / method, or other indications. Alternatively or additionally, the network or base station 122 may also configure and indicate a maximum UCI configuration size (or UCI payload size). The UE 110 may then be enabled to select an AI model or AI configuration / method based on the maximum UCI payload size. Thus, in the network configuration, the AI ​​model ID may be configured to the UE, but in other alternatives, the network may configure the maximum UCI payload size so that the UE 110 can select the AI ​​model to be processed based on the model ID, with the selected number of UCI bits being the UCI configuration. The UE 110 may be configured to select from the layer common configurations 510 to 540 for generating AI-based UCI based on the UCI payload size or the maximum UCI payload size. For example, if the base station 122 configures a maximum UCI payload size of 240 bits for the third layer common configuration 530, the UE 110 may choose to transmit layer 1 as 240 bits, 120 bits, or 60 bits or less. The UE 110 may then indicate to the network the information being used, such as the UCI configuration it has selected to configure the AI-based UCI reporting, for example, in the AI-based UCI report itself or in separate signaling.

[0055] Figure 6 An example of an AI configuration or method 600 for generating an AI-based UCI report based on one or more layer-specific configurations 610 to 630 is illustrated. Layer-specific configuration means that different AI models are utilized to generate the output of the AI-based UCI report regardless of the UCI configuration. The complexity here can be high due to the number of AI models. Assuming a maximum of four MIMO spatial layers per UE, the storage can be four times that of a layer-common configuration. For example, one AI model is trained per layer per configuration.

[0056] In one aspect, for example, a first layer-specific configuration 610 for AI-based UCI reporting may be configured based on one or more RIs having RI values ​​of 1 to 4 in a rank. The base station 122 may send a network configuration to the UE 110 with one or more indications as discussed herein, such as an AI model ID, a UCI configuration of the number of UCI bits used to generate AI-based UCI reporting, an AI model, an AI configuration or method (e.g., AI configuration 610, 620, or 630), or other indications as discussed herein. For the first layer-specific configuration 610, the same UCI configuration (e.g., 60 bits) is utilized across RIs, but different AI models are used to generate results for each transmission layer.

[0057] Through network configuration, the network or base station 122 configures one AI model ID per transmission layer or MIMO spatial layer (e.g., layer 1, layer 2, layer 3, or layer 4), which is the case for each layer-specific configuration 610, 620, and 630. In the case of the first layer-specific configuration 610, the same UCI configuration or number of UCI bits is used for each transmission layer. For example, a UCI configuration of 60, 90, or 120 bits may be configured for each transmission layer in transmission layers 1 / 2 / 3 / 4 for RI. Even with the same configuration (e.g., 60 bits for all layers), the encoder-decoder pair still executes different AI models for different transmission layers under that UCI configuration, such as Figure 3 exemplified by the different hash tags discussed.

[0058] The second layer-specific configuration 620 may configure different numbers of UC bits or different UCI configurations per transmission layer. For example, UCI configurations of 60 / 90 / 120 may be configured for transmission layers 1 and 2, and other UCI configurations of 40 / 60 / 80 may be configured for layers 3 and 4. Thus, a larger UCI payload may be configured for transmission layers 1 and 2, as transmission layer 2 may have significantly higher power, while the UCI bits for transmission layers 3 and 4 are lower and less important than for transmission layers 1 and 2. The UCI payload size may then increase linearly from one RI to another as the value increases, depending on the UE rank selection or RI selected by the UE. UE 110 may then utilize the AI ​​model to generate an AI-based UCI report based on the layer-specific configuration, in which the AI ​​model varies based on the transmission layer. The UCI configuration may have the same or different numbers of UCI bits across the transmission layers associated with the RI, while the UCI payload size may increase linearly based on the UE rank selection of the RI.

[0059] The third layer-specific configuration 630 can configure the UCI report based on the AI ​​model ID list or the data set per RI per layer. For example, RI=1 can be associated with model ID 11 of layer 1, and RI=2 can be associated with model ID 12 of layer 1 and model ID 22 of layer 2; RI=3 can be associated with model ID 13 of layer 1, model ID 23 of layer 2, and model ID 33 of layer 3. Therefore, the network configuration can provide all relevant AI model lists or AI model ID lists per RI, so that the UE selects the AI ​​model list or AI model ID list corresponding to each transmission layer and RI from them. AI-based UCI reporting using the AI ​​model is based on RI and layer-specific configuration. The AI ​​model corresponds to one transmission layer and can vary between transmission layers based on the AI ​​model ID list that can be different or the same for each transmission layer or RI.

[0060] Figure 7An example of an AI configuration or method 700 for generating an AI-based UCI report based on one or more RI-specific configurations 700 is illustrated. Here, the network configuration may configure one AI model ID per rank or RI. The number of feedback bits per transmission layer is not separable as part of the AI ​​model optimization. The network configuration may configure the same number of output UCI bits per RI or different bits per RI. For example, on the left, RI=1 is configured with a first AI model for a 240-bit UCI configuration, while on the right, RI=1 is configured with a first AI model for a 120-bit UCI configuration for generating UCI bits in an AI-based UCI report.

[0061] From a configuration perspective, AI configuration 700 is considered RI-specific. Therefore, a single model ID is used per rank, and the feedback bits for each layer are inseparable. The UE outputs an AI model for multiple bits for layer 1 and multiple UCI bits for layer 2, which are effectively inseparable because they are all mixed together in the output. When the decoder processes their reconstructions together, the NW can configure the output bits to be the same for each RI or different for each RI.

[0062] refer to Figure 8 , illustrates example UCI reporting for UCI according to various AI configurations / methods discussed herein (such as layer-common configuration, layer-specific configuration, and rank-specific configuration), which may be based on an AI-based CSI pattern for the UCI bits. The UCI report 802 may be multiplexed with various subchannels 804 for different CSI parts (specifically, including CSI part 1 806 and CSI part 2 808). The base station 122 requires CSI part 1 to be able to decode CSI part 2 carrying CSI. Other symbols may include uplink scheduling (UL-SCH) information, hybrid automatic repeat request (HA RQ) ACK feedback, or demodulation reference signaling (DM-RS).

[0063] In one aspect, AI-based CSI compression can be configured to provide similar information. For both layer-common and layer-specific configurations, UE 110 can utilize UCI Part 1 using RI, while the UCI configuration explicitly configures the number of bits per layer. RI indirectly indicates the number of bits required for Part 2 and is based on the UCI configuration, as the UCI Part 2 payload size scales with RI. RI, wideband CQI, and subband CQI can also be configured on UCI Part 1 or CSI Part 1, similar to the legacy design. The AI ​​model ID can also be carried by or be part of UCI Part 1.

[0064] In one aspect, UCI part 1 may include the AI ​​model ID. If the AI ​​model ID is configured by the network, then once the indication of the AI ​​model ID is signaled to the UE, the model ID does not need to be included in the UCI. Alternatively or additionally, if UE 110 determines which AI model to use, the AI ​​model ID may be included as part of the UCI (e.g., in UCI part 1). Alternatively or additionally, if the network is still in control, but now provides or indicates a list of AI model IDs for the UE to choose from (e.g., as in Figure 6 ), and one AI model ID corresponds to one AI output size per layer, the UE selects one AI model ID and reports it back via the index of the AI ​​model in the list. For example, if two model IDs are configured for selection by the UE and the UE selects one model ID, one bit may be signaled to indicate that the first AI model ID or the second AI model ID has been selected. Alternatively or additionally, if the AI ​​model ID represents an AI model with adaptation layers, and different adaptation layers can generate different output sizes, the UE selects one adaptation layer / output bit size and reports the selection together with the index associated with it. These aspects are applicable to both layer-common and layer-specific configurations.

[0065] For rank-specific configurations, UCI part 1 may also include the RI. Since the AI ​​model is linked to the RI, this also indicates the model ID when only one model ID is configured per rank. Wideband and subband CQI are common across all instances. If multiple IDs are configured per rank, UE 110 may report the model ID index within the configured list. For example, for RI=1, model IDs xx11, xx12 are configured. The UE then additionally reports one bit to indicate the model being used for RI=1.

[0066] Additionally, the RI indication and the model ID selection may be jointly signaled, for example, the RI indication and the model ID selection are network configured via RRC or MAC CE. In one example, with a 3-bit signaling overhead, the RI indication and the model ID selection may be signaled via a radio resource control (RRC) / medium access control (MAC) control element (MAC CE) such as Figure 9 The table is configured as shown.

[0067] In one aspect, UCI part 1 can be a fixed size with known corresponding bits, where the information corresponds to which bits, but the size of UCI part 2 can be flexible and can be larger or smaller depending on the RI. The size of UCI part 2 can be derived from UCI part 1. With all the information from UCI part 1, the base station can know exactly how large UCI part 2 is and what information and AI configuration / method it contains.

[0068] For layer-common configurations / methods, UCI part 2 may include CSI generation model outputs for each transmit layer. Depending on the RI, this may start with layer 1, followed by layer 2, and so on, each of which may be performed one by one. Then, depending on the RI, for the format of UCI part 2, the format may start with layer 1, followed by layer 2, and so on, up to layer 3 and layer 4. For layer-specific configurations, the format may be similar to the layer-common configuration / method, depending on UCI part 1, layer 1 may be the first layer, layer 2 is the second layer, and then layer 3, and then the AI ​​model operating on each layer is different, but the format is the same. UCI part 2 includes CSI generation model outputs for each transmit layer. Depending on the RI, the output for transmit layer 1 uses its corresponding model, the output for layer 2 uses its corresponding AI model, and the same is true for layers 3 and 4 across UCI part 2.

[0069] For rank specificity, transmission layers 1, 2, 3, and 4 may all be mixed together, making it difficult to know which is which across UCI part 2. The CSI generation model output may place all outputs into UCI part 2 bit-by-bit or in bit-by-bit order based on the AI ​​model output itself.

[0070] In one aspect, UE 110 may further operate with restricted omission rules based on layer-common and layer-specific configurations. The CSI generation model output is for layer 1 > layer 2 > layer 3. Omission of UCI bits for CSI part 2 may then be based on a priority order associated with the layer-common or layer-specific configuration, with the priority order decreasing from higher layers to lower layers. If UCI bits are restricted (such as by a network configuration that provides a maximum UCI payload size), UE 110 may have to drop or omit some UCI bits to accommodate AI-based UCI reporting. In this case, the transmitted layer 3 or 4 bits may be dropped first, followed by the transmitted layer 2, and finally the transmitted layer 1. For rank-specific configurations, no omission rules are configured because all information is mixed together. Therefore, dropping any portion would mean inaccurate decoding of the transmitted layer. Therefore, there is no omission process for AI-based UCI reporting for rank-specific configurations of the output UCI bits. Instead, UE 110 may select a rank to suit the payload size.

[0071] Figure 10 1 is an example network 1000 according to one or more implementations described herein. Example network 1000 may include UE 110-1, UE 110-2, etc. (collectively, "UE 110" and individually, "UE 110"), a radio access network (RAN) 1020, a core network (CN) 1030, an application server 1040, and external networks 1050.

[0072] UE 110 may communicate with and establish a connection (communicatively coupled) to RAN 1020, which may involve one or more radio channels 1014-1 and 1014-2, each of which may include a physical communication interface / layer. In some implementations, the UE may be configured with dual connectivity (DC) as multiple radio access technologies (multi-RAT) or multi-radio dual connectivity (MR-DC), wherein a UE supporting multiple receive and transmit (Rx / Tx) signals may use resources provided by different network nodes or base stations 122 (e.g., 122-1 and 122-2), which may be connected via non-ideal backhaul (e.g., one network node provides NR access while the other network node provides E-UTRA for LTE or NR access for 5G). In such a scenario, one network node may operate as a master node (MN) and the other node may operate as a secondary node (SN). The MN and SN may be connected via a network interface, and at least the MN may be connected to CN 1030. Additionally, at least one of the MN or the SN may operate via shared spectrum channel access, and the functionality specified for the UE 110 may be used for an integrated access and backhaul mobile terminal (IAB-MT). Similar to the UE 110, the IAB-MT may access the network using one network node or using two different nodes with an enhanced dual connectivity (EN-DC) architecture, a new radio dual connectivity (NR-DC) architecture, or other direct connections such as a SL communication channel as the SL interface 112.

[0073] In some implementations, a base station (as described herein) may be an example of a network node 122. As shown, the UE 110 may additionally or alternatively be connected to an access point (AP) 1016 via a connection interface 1018, which may include an air interface that enables the UE 110 to be communicatively coupled to the AP 1016. The AP 1016 may include a wireless local area network (WLAN), a WLAN node, a WLAN endpoint, etc. The connection 1018 may include a local wireless connection, such as a connection consistent with any IEEE 702.11 protocol, and the AP 1016 may include a wireless fidelity (Wi-Fi) protocol. The AP 1016 may also be connected to another network (eg, the Internet) instead of being connected to the RAN 1020 or the CN 1030.

[0074] The RAN 1020 may also include one or more RAN nodes 122-1 and 122-2 (collectively, RAN nodes 122 and individually, RAN nodes 122) that enable establishing channels 1014-1 and 1014-2 between the UE 110 and the RAN 1020. The RAN node 122 may include a network access point configured to provide radio baseband functionality for data and / or voice connections between a user and a network based on one or more of the communication technologies described herein (e.g., 2G, 3G, 4G, 5G, WiFi, etc.). Thus, as an example, the RAN node may be an E-UTRAN Node B (e.g., an enhanced Node B, eNodeB, eNB, 4G base station, etc.), a next-generation base station (e.g., a 5G base station, a NR base station, a next-generation eNB (gNB), etc.). The RAN node 122 may include a roadside unit (RSU), a transmit / receive point (TRxP or TRP), and one or more other types of ground stations (e.g., a terrestrial access point). In some scenarios, the RAN node 122 may be a dedicated physical device such as a macrocell base station or a low-power (LP) base station for providing a femtocell, picocell, or other similar cell with a smaller coverage area, smaller user capacity, or higher bandwidth than a macrocell. As described below, in some implementations, the satellite 160 may operate as a base station (e.g., a RAN node 122) relative to the UE 110. Therefore, references herein to a base station, RAN node 122, etc. may relate to implementations in which the base station, RAN node 122, etc. is a terrestrial network node, as well as to implementations in which the base station, RAN node 122, etc. is a non-terrestrial network node.

[0075] Some or all of the RAN nodes 122 may be implemented as one or more software entities running on a server computer as part of a virtual network, which may be referred to as a centralized RAN (CRAN) or a virtual baseband unit pool (vBBUP). In these implementations, the CRAN or vBBUP may implement RAN functional splits, such as a packet data convergence protocol (PDCP) split, where the radio resource control (RRC) and PDCP layers may be operated by the CRAN / vBBUP, and other layer 2 (L2) protocol entities may be operated by a separate RAN node 122; a medium access control (MAC) / physical (PHY) layer split, where the RRC, PDCP, radio link control (RLC), and MAC layers may be operated by the CRAN / vBBUP, and the PHY layer may be operated by a separate RAN node 122; or a "lower PHY" split, where the RRC, PDCP, RLC, MAC layer, and upper portions of the PHY layer may be operated by the CRAN / vBBUP, and the lower portions of the PHY layer may be operated by a separate RAN node 122. The virtualization framework may allow idle processor cores of the RAN node 122 to execute or implement, for example, other virtualized applications.

[0076] In some implementations, the individual RAN nodes 122 may represent individual gNB distributed units (DUs) connected to a gNB control unit (CU) via respective F1 interfaces. In such implementations, the gNB-DUs may include one or more remote radio heads or radio frequency (RF) front-end modules (RFEMs), and the gNB-CUs may be operated by a server (not shown) located in the RAN 1020 or by a server pool (e.g., a group of servers configured to share resources) in a manner similar to a CRAN / vBBUP. Additionally or alternatively, one or more of the RAN nodes 122 may be next-generation eNBs (i.e., gNBs), which may provide Evolved Universal Terrestrial Radio Access (E-UTRA) user and control plane protocol terminations to the UE 110 and may be connected to the 5G core network (5GC) 1030 via the next generation (NG) interface 1024.

[0077] Any of the RAN nodes 122 can serve as an endpoint for an air interface protocol and can be the first point of contact for the UE 110. In some implementations, any of the RAN nodes 122 can perform various logical functions of the RAN 1020, including, but not limited to, functions of a radio network controller (RNC), such as radio bearer management, uplink and downlink dynamic radio resource management, data packet scheduling, and mobility management. The UEs 110 can be configured to communicate with each other or with any of the RAN nodes 122 using orthogonal frequency division multiplexing (OFDM) communication signals over multi-carrier communication channels according to various communication technologies, such as, but not limited to, OFDMA communication technologies (e.g., for downlink communications) or single-carrier frequency division multiple access (SC-FDMA) communication technologies (e.g., for uplink and ProSe or sidelink (SL) communications), although the scope of such implementations may not be limited in this respect. OFDM signals may include multiple orthogonal subcarriers.

[0078] The physical downlink shared channel (PDSCH) may carry user data and higher layer signaling to UE 110. The physical downlink control channel (PDCCH) may carry information regarding, among other things, the transport format and resource allocation associated with the PDSCH channel. The PDCCH may also inform UE 110 about the transmit format, resource allocation, and hybrid automatic repeat request (HARQ) information associated with the uplink shared channel. Typically, downlink scheduling (assignment of control and shared channel resource blocks to UE 110-2 within a cell) may be performed on any of RAN nodes 122 based on channel quality information fed back from any of UEs 110. Downlink resource assignment information may be transmitted on the PDCCH for (e.g., assigned to) each of UEs 110.

[0079] PDCCH uses control channel elements (CCE) to transmit control information, where multiple (e.g., 6 or other numbers) of CCEs may be composed of resource element groups (REGs), where REGs are defined as physical resource blocks (PRBs) in OFDM symbols. For example, before being mapped to resource elements, PDCCH complex-valued symbols may first be organized into quadruplets, which may then be arranged using a sub-block interleaver for rate matching. Each PDCCH may be sent using one or more of these CCEs, where each CCE may correspond to nine sets of four physical resource elements, referred to as REGs. Four quadrature phase shift keying (QPSK) symbols may be mapped to each REG. Depending on the size of the DCI and channel conditions, one or more CCEs may be used to send the PDCCH. There may be four or more different PDCCH formats with different numbers of CCEs (e.g., aggregation levels, L=1, 2, 4, 8, or 16).

[0080] The RAN nodes 122 may be configured to communicate with each other via an interface 1023. In a specific implementation where the system is an LTE system, the interface 1023 may be an X2 interface. In an LTE network, the X2 interface and the S1 interface are defined as interfaces between RAN nodes and between the RAN and the core network. 5G can operate in two modes, a non-standalone mode and an independent mode. For non-standalone operation, the specification defines extensions of the S1 and X2 interfaces, and for independent operation, the specification defines the interface between the RAN nodes 122 as X2 / Xn and the interface 1024 between the RAN 120 and the CN 1030 as S1 / NG. The interface 1024 may be defined between two or more RAN nodes 122 (e.g., two or more eNBs / gNBs or a combination thereof) connected to an evolved packet core (EPC) or a CN 1030, or between eNBs connected to the EPC. In some implementations, the X2 / Xn interface may include an X2 / Xn user plane interface (X2-U / Xn-U) and an X2 control plane interface (X2-C / Xn-C). X2-U / Xn-U may provide a flow control mechanism for user data packets transmitted over the X2 / Xn interface and may be used to convey information regarding the delivery of user data between eNBs or gNBs. For example, X2-U / Xn-U may provide specific sequence number information regarding user data transmitted from a master eNB (MeNB) to a secondary eNB (SeNB); information regarding successful in-sequence delivery of PDCP packet data units (PDUs) for user data from the SeNB to the UE 110; information regarding PDCP PDUs that were not delivered to the UE 110; information regarding the current minimum expected buffer size at the SeNB for sending user data to the UE; and the like. X2-C / Xn-C may provide intra-LTE access mobility functionality (e.g., including context transfer from a source eNB to a target eNB, user plane transmission control, etc.), load management functionality, and inter-cell interference coordination functionality.

[0081] Alternatively or additionally, the RAN 1020 may also be connected (e.g., communicatively coupled) to the CN 1030 via a next generation (NG) interface as interface 1024. The NG interface 1024 may be divided into two parts: a next generation (NG) user plane (NG-U) interface 1026, which carries traffic data between the RAN node 122 and a user plane function (UPF); and an S1 control plane (NG-C) interface 1028, which is a signaling interface between the RAN node 122 and an access and mobility management function (AMF).

[0082] The CN 1030 may include a plurality of network elements 1032 configured to provide various data and telecommunication services to customers / subscribers (e.g., users of the UE 110) connected to the CN 1030 via the RAN 1020. In some implementations, the CN 1030 may include an evolved packet core (EPC), a 5G CN, and / or one or more additional or alternative types of CNs. Components of the CN 1030 may be implemented in one physical node or in separate physical nodes, including components for reading and executing instructions from a machine-readable or computer-readable medium (e.g., a non-transitory machine-readable storage medium).

[0083] As shown, CN 1030, application server 1040, and external network 1050 may be connected to each other via interfaces 1034, 1036, and 1038, which may include IP network interfaces. Application server 1040 may include one or more server devices or network elements (e.g., virtual network functions (VNFs)) that provide applications (e.g., Universal Mobile Telecommunications System Packet Service (UMTS PS) domain, LTE-PS data services, etc.) that use IP bearer resources with CN 1030. Application server 1040 may also or alternatively be configured to support one or more communication services (e.g., IP voice (VoIP sessions, push-to-talk (PTT) sessions, group communication sessions, social networking services, etc.) for UE 110 via CN 1030. Similarly, external network 1050 may include one or more of various networks (including the Internet), thereby providing access to various additional services, information, interconnectivity, and other network features to mobile communication networks and UEs 110 of the network.

[0084] In one aspect, UE 110 may operate via processing circuitry by receiving a network configuration associated with AI-based UCI reporting. The network configuration may include one or more indications of a UCI configuration and an AI model to generate the AI-based UCI report. The UE sends the AI-based UCI report based on the network configuration. UE 110 may utilize the AI ​​model to generate the AI-based UCI report. Based on the indication of the AI ​​model and the UCI configuration, the AI ​​model may be associated with at least one of: a layer-common configuration, a layer-specific configuration, or a rank indication (RI)-specific configuration. The AI ​​model may include a bilateral model that performs AI-based CSI compression for the AI-based UCI report. A layer-common configuration includes using the same AI model across multiple transmission layers of a UCI configuration. A layer-specific configuration includes using the same AI model for each transmission layer of each UCI configuration. An RI-specific configuration includes using a different AI model for each RI. The network configuration may include an AI model ID, which indicates at least one of the following: a UCI configuration associated with multiple UCI bits, an AI model including an AI model index and a UCI configuration, an AI configuration or an AI method, a UE vendor identifier, a network device vendor identifier, a use case ID of a use case, or a model number of a use case ID.

[0085] refer to Figure 11 , illustrates a block diagram of a UE device 110 (e.g., UE 110-1 or 110-2) or other network device / component (e.g., V-UE / P-UE, IoT, gNB, eNB, base station 122, or other participating network entity / component) 1100. Device 1100 includes: one or more processors 1110 (e.g., one or more baseband processors) including processing circuitry and associated interfaces; transceiver circuitry 1120 (e.g., including RF circuitry, which may include transmitter circuitry (e.g., associated with one or more transmit chains) and / or receiver circuitry (e.g., associated with one or more receive chains), which may employ common circuit elements, different circuit elements, or a combination thereof); and memory 1130 (which may include any of a variety of storage media and may store instructions and / or data associated with one or more of processor 1110 or transceiver circuitry 1120).

[0086] Memory 1130 (and other memory components discussed herein, such as memory, data storage devices, etc.) may include one or more machine-readable media containing instructions that, when executed by a machine or component herein, cause the machine or other device to perform the actions of a method, apparatus, or system for communicating using a variety of communication technologies according to the aspects, embodiments, and examples described herein. It should be understood that the aspects described herein may be implemented by hardware, software, firmware, or any combination thereof. When implemented in software, the functionality may be stored as one or more instructions or codes on a computer-readable medium (e.g., a memory or other storage device described herein) or sent via a computer-readable medium. Computer-readable media include both computer storage media and communication media, including any media that facilitates transferring a computer program from one place to another. Storage media or computer-readable storage devices may be any available media that can be accessed by a general-purpose computer or a special-purpose computer. By way of example only and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM, or other optical disk storage devices, magnetic disk storage devices, or other magnetic storage devices, or other tangible and / or non-transitory media that can be used to carry or store desired information or executable instructions. Any connection may also be referred to as a computer-readable medium.

[0087] The memory 1130 may include executable instructions and may be integrated into or communicatively coupled to the processor or processing circuit 1110. The executable instructions of the memory 1130 may cause the processing circuit 1110 to receive / process instructions to initiate a U2U relay path to the destination UE through the first relay UE by providing a direct communication request to the first relay UE. U2U relay reselection may be performed on the second relay UE in response to a trigger condition. The trigger condition may be based on at least one of the following: a measurement of the first channel link or the second channel link to the first relay UE being lower than a (pre)configured threshold, detection of a radio link failure (RLF) on the first channel link or the second channel link, notification based on a channel link between the first relay UE and the destination UE or between the source UE and the first relay UE, or reception of a release message. Then, a U2U relay to the destination UE may be further established through the second relay UE, as well as other aspects described in the present disclosure.

[0088] The memory 1130 may include executable instructions and may be integrated into or communicatively coupled to the processor or processing circuit 1110. The executable instructions of the memory 1130 may cause the processing circuit 1110 to

[0089] The device 1100 is configured to process, execute, generate, communicate or cause execution of the present invention or related Figures 1 to 10Any one or more combined aspects described in association with any of the figures.

[0090] Although the method described in the present disclosure is illustrated and described as a series of actions or events in this article, it should be understood that the order of such actions or events shown should not be interpreted as having a limiting meaning. For example, some actions can occur in different orders and / or simultaneously with other actions or events other than those illustrated and / or described herein. In addition, it may not be necessary for all illustrated actions to implement one or more aspects or embodiments of this specification. In addition, one or more actions in the actions depicted herein can be performed in one or more separate actions and / or stages. For ease of description, reference can be made to the above-mentioned accompanying drawings. However, the method is not limited to any specific embodiment, aspect or example provided in the present disclosure, and can be applied to any one of the systems / devices / components disclosed herein.

[0091] It is understood that the use of personally identifiable information should be subject to privacy policies and practices that are generally recognized to meet or exceed industry or government requirements for maintaining user privacy. Specifically, personally identifiable information data should be managed and processed to minimize the risk of unintentional or unauthorized access or use, and the nature of authorized use should be clearly stated to users.

[0092] The present disclosure will now be described with reference to the accompanying drawings, wherein throughout the text, like reference numerals are used to refer to like elements, and wherein illustrated structures and devices need not be drawn to scale. As used herein, the terms "component", "system", "interface", etc. are intended to refer to entities, hardware, software (e.g., software in execution), and / or firmware related to a computer. For example, a component can be a processor (e.g., a microprocessor, a controller or other processing device), a process running on a processor, a controller, an object, an executable file, a program, a storage device, a computer, a tablet computer, and / or a user equipment (e.g., a mobile phone, etc.) with a processing device. In an illustrative manner, an application program and a server running on a server can also be a component. One or more components can reside in a process, and a component can be located on a computer and / or distributed between two or more computers. This article can describe a set of elements or other sets of components, wherein the term "set" can be interpreted as "one or more".

[0093] In addition, these components can execute from various computer-readable storage media having various data structures stored thereon, such as using modules. Components can communicate via local and / or remote processes, such as according to signals having one or more data packets (e.g., data from one component interacts with another component in a local system, a distributed system, and / or an entire network, such as the Internet, a local area network, a wide area network, or a similar network with other systems via signals).

[0094] As another example, a component may be a device that has a specific function, where the specific function is provided by mechanical components that operate through electrical or electronic circuits, where the electrical or electronic circuits can be operated by software applications or firmware applications executed by one or more processors. The one or more processors can be internal or external to the device and can execute at least a portion of the software application or firmware application. As another example, a component may be a device that provides a specific function through electronic components without the need for mechanical components; the electronic components may include one or more processors to execute at least a portion of the software and / or firmware that provides the electronic components with their functions.

[0095] The use of the word "exemplary" is intended to present concepts in a concrete manner. As used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified or clear from the context, "X employs A or B" is intended to mean any of the natural inclusive permutations. That is, if X employs A; X employs B; or X employs both A and B, then "X employs A or B" is satisfied in any of the foregoing cases. In addition, the articles "a" and "an" as used in this application and the appended claims should generally be construed to mean "one or more," unless otherwise specified or clear from the context to be directed to the singular. Furthermore, to the extent that the terms "including," "comprising," "having," "having," "with," or variations thereof are used in the detailed description and claims, such terms are intended to be inclusive in a manner similar to the term "comprising." In addition, where one or more numbered items are discussed (e.g., "a first X," "a second X," etc.), generally, the one or more numbered items can be different or they can be the same, but in some cases, the context may indicate that they are different or that they are the same.

[0096] As used herein, the term "circuit" may refer to, may be a part of, or may include an application-specific integrated circuit (ASIC), an electronic circuit, a processor (shared, dedicated, or group), or associated memory (shared, dedicated, or group) operably coupled to the circuit that executes one or more software or firmware programs, a combinational logic circuit, or other suitable hardware components that provide the described functionality. In some embodiments, the circuit may be implemented in one or more software or firmware modules, or the functionality associated with the circuit may be implemented by one or more software or firmware modules. In some embodiments, the circuit may include logic that is at least partially operable in hardware.

[0097] As used in this specification, the term "processor" may refer to substantially any computational processing unit or device, including but not limited to a single-core processor; a single processor with software multi-threaded execution capability; a multi-core processor; a multi-core processor with software multi-threaded execution capability; a multi-core processor with hardware multi-threading technology; a parallel platform; and a parallel platform with distributed shared memory. In addition, a processor may refer to an integrated circuit, an application-specific integrated circuit, a digital signal processor, a field programmable gate array, a programmable logic controller, a complex programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions and / or processes described herein. The processor may utilize nanoscale architectures, such as, but not limited to, molecular and quantum dot-based transistors, switches, and gates, in order to optimize space usage or enhance performance of mobile devices. The processor may also be implemented as a combination of computational processing units.

[0098] Examples (implementations) may include subject matter such as methods, apparatus for performing the actions or blocks of the method, and at least one machine-readable medium comprising instructions that, when executed by a machine (e.g., a processor with memory, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), etc.), cause the machine to perform the actions of a method or apparatus or system for concurrent communication using multiple communication technologies according to the embodiments and examples described herein.

[0099] A first embodiment is an apparatus in a user equipment (UE), the apparatus comprising: a memory; a processing circuit coupled to the memory, the processing circuit being configured to, when executing instructions stored in the memory, cause the UE to: receive a network configuration associated with an artificial intelligence (AI)-based uplink control information (UCI) report, wherein the network configuration includes one or more indications of a UCI configuration and an AI model to generate the AI-based UCI report; and send the AI-based UCI report based on the network configuration.

[0100] A second embodiment may include the first embodiment, wherein the instructions, when executed by the processing circuit, further configure the device: generate the AI-based UCI report using the AI ​​model, wherein based on the indication of the AI ​​model and the UCI configuration, the AI ​​model is associated with at least one of the following: a layer-common configuration, a layer-specific configuration, or a rank indication (RI)-specific configuration, wherein the AI ​​model includes a bilateral model that performs AI-based CSI compression for the AI-based UCI report.

[0101] A third embodiment may include the first embodiment or the second embodiment, wherein the layer-common configuration includes using the same AI model across multiple transmission layers of the UCI configuration, wherein the layer-specific configuration includes using the same AI model for one transmission layer per UCI configuration, and wherein the RI-specific configuration includes using a different AI model per RI.

[0102] The fourth embodiment may include any one or more embodiments of the first to third embodiments, wherein the network configuration further indicates an AI model ID, and the AI ​​model ID indicates at least one of the following: the UCI configuration associated with multiple UCI bits, the AI ​​model including the AI ​​model index and the UCI configuration, the AI ​​configuration or the AI ​​method, the UE vendor identifier, the network device vendor identifier, the use case ID of the use case, or the model number of the use case ID.

[0103] The fifth embodiment may include any one or more embodiments of the first to fourth embodiments, wherein the instruction, when executed by the processing circuit, further configures the processing circuit to: generate UCI bits for the AI-based UCI report based on the AI ​​model ID of the network configuration, wherein the AI ​​model ID indicates a layer common configuration to configure a UCI payload size that increases linearly with an increase in the RI value associated with the UCI configuration.

[0104] A sixth embodiment may include any one or more embodiments of the first to fifth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit: select an AI model ID indicated in the network configuration from a plurality of AI model IDs based on the RI and the UCI configuration to generate UCI bits for the AI-based UCI report; and wherein the AI ​​model ID indicates a layer common configuration to configure the UCI payload size, wherein the plurality of AI model IDs include a first AI model ID associated with a first set of RIs and a second AI model ID associated with a second set of RIs having a greater RI value than the first set of RIs, and wherein the first AI model ID and the second model ID respectively include different AI models that constrain the UCI payload size of the second set of RIs relative to the first set of RIs.

[0105] The seventh embodiment may include any one or more embodiments of the first to sixth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to: select the AI ​​model to encode each transmission layer based on the RI having a layer common configuration indicated by the network configuration, wherein the UCI configuration is different between RIs and is associated with multiple UCI bits.

[0106] The eighth embodiment may include any one or more embodiments of the first to seventh embodiments, wherein the instruction, when executed by the processing circuit, further configures the processing circuit to: encode each transmitting layer based on the AI ​​model and the RI having the layer common configuration indicated by the network configuration, wherein the UCI configuration varies or does not vary between transmitting layers based on the RI.

[0107] The ninth embodiment may include any one or more embodiments of the first to eighth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to select the AI ​​model based on the maximum UCI payload size indicated by the network configuration.

[0108] The tenth embodiment may include any one or more embodiments of the first to ninth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to: provide a UE capability report, the UE capability report indicating support for at least one of the following: one or more AI models, one or more AI model IDs associated with the one or more AI models, or one or more UCI configurations for the AI-based UCI report.

[0109] The eleventh embodiment may include any one or more embodiments of the first to tenth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to: generate the AI-based UCI report using the AI ​​model based on a layer-specific configuration, wherein the AI ​​model varies based on a transmission layer, wherein the UCI configuration is the same number or different number of UCI bits between transmission layers associated with the RI, and the UCI payload size increases linearly based on the UE rank selection of the RI.

[0110] The twelfth embodiment may include any one or more embodiments of the first to eleventh embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to generate the AI-based UCI report using the AI ​​model based on the RI and a layer-specific configuration, wherein the AI ​​model corresponds to a transmission layer and varies between transmission layers and based on a different AI model ID list for each transmission layer and RI.

[0111] The thirteenth embodiment may include any one or more embodiments of the first to twelfth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to generate the AI-based UCI report using the AI ​​model based on an RI-specific configuration including different AI models associated with different RIs, respectively, according to a network configuration, wherein the network configuration indicates the same number or different number of output bits per RI.

[0112] A fourteenth embodiment may include any one or more of the first to thirteenth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to: generate the AI-based UCI report based on the network configuration, wherein a channel state information (CSI) part 1 of the AI-based UCI report indicates the size of the CSI part 2 by an AI model ID based on whether only the network configuration determines or the UE also determines which AI model is used to generate the UCI with a layer-common configuration or a layer-specific configuration; or generate the AI-based UCI report based on the network configuration, wherein the AI-based UCI report indicates the size of the CSI part 2 by an AI model ID. The CSI part 1 of the CI report indicates the AI ​​model ID based on multiple AI model IDs selected by the UE, wherein the AI ​​model ID is associated with the UE selection for generating UCI bits with a layer-common configuration or a layer-specific configuration; or the AI-based UCI report is generated based on the network configuration, wherein the CSI part 1 of the AI-based UCI report indicates the output bit size of the AI ​​model with the adaptation layer based on multiple adaptation layers selected by the UE, wherein the AI ​​model ID and the corresponding adaptation layer are associated with the UE selection for generating UCI bits for the layer-common configuration or the layer-specific configuration.

[0113] The fifteenth embodiment may include any one or more embodiments of the first to sixteenth embodiments, wherein the instruction, when executed by the processing circuit, further configures the processing circuit to: generate the AI-based UCI report based on the network configuration for a rank-specific configuration, wherein the UCI part 1 of the AI-based UCI report provides an RI, the RI indicates an AI model ID associated with the RI, and in response to more than one AI model ID being associated with the RI, the UCI part 1 also includes an AI model ID index, or the RI and the AI ​​model ID are signaled via the network configuration through radio resource control (RRC) signaling or medium access control (MAC) control element (MAC CE) signaling.

[0114] The sixteenth embodiment may include any one or more embodiments of the first to seventeenth embodiments, wherein the instruction, when executed by the processing circuit, further configures the processing circuit to: generate the AI-based UCI report having a CSI part 2, wherein the CSI part 2 includes: for a layer-common configuration, a CSI generation model output for each transmitting layer depending on the selected RI in the sequential order of the transmitting layer; for a layer-specific configuration, a CSI generation model output for each transmitting layer depending on the selected RI and the corresponding AI model output; or generate the CSI generation model output for a rank-specific configuration.

[0115] The seventeenth embodiment may include any one or more embodiments of the first to sixteenth embodiments, wherein the instructions, when executed by the processing circuit, further configure the processing circuit to generate UCI bit omission for the CSI part 2 based on a priority order associated with a layer-common configuration or the layer-specific configuration, wherein the priority order decreases from a higher layer to a lower layer.

[0116] The eighteenth embodiment may be a method of a user equipment (UE), the method comprising: receiving, via a processing circuit, a network configuration associated with an uplink control information (UCI) payload of an artificial intelligence (AI)-based channel state information (CSI) feedback, wherein the network configuration comprises one or more indications of a UCI configuration and an AI model to generate an AI-based UCI report; and sending the AI-based UCI report based on the network configuration.

[0117] The nineteenth embodiment may include the eighteenth embodiment, and further include: generating the AI-based UCI report using the AI ​​model, wherein based on the one or more indications of the UCI configuration and the AI ​​model, the AI ​​model is associated with at least one of the following: a layer-common configuration, a layer-specific configuration, or a rank indication (RI)-specific configuration, wherein the AI ​​model includes a bilateral model that performs AI-based CSI compression for the AI-based UCI report, wherein the layer-common configuration includes using the same AI model across multiple transmission layers of the UCI configuration, wherein the layer-specific configuration includes using the same AI model for one transmission layer per UCI configuration, and wherein the RI-specific configuration includes using a different AI model per RI.

[0118] The twentieth embodiment may be a base station, the device comprising: a memory; a processing circuit, the processing circuit being coupled to the memory, the processing circuit being configured to cause the base station, when executing instructions stored in the memory, to: send a network configuration for artificial intelligence (AI)-based UCI reporting, wherein the network configuration includes an indication of a UCI configuration and an AI model; and receive the AI-based UCI report based on the network configuration.

[0119] The twenty-first embodiment may include the twentieth embodiment, wherein the network configuration indicates an AI model ID having a layer-common configuration for configuring UCI bits of a UCI payload size in the AI-based UCI report, and wherein the UCI payload size increases linearly with an increase in the RI associated with the UCI configuration including the number of UCI bits for each transmission layer, or wherein according to an indication of the network configuration, the UCI configuration is different between RIs having one AI model associated with each RI.

[0120] The twenty-second embodiment may include any one or more embodiments of the twentieth to twenty-first embodiments, wherein the network configuration indicates an AI model ID associated with a layer common configuration for configuring UCI bits of the UCI payload size in the AI-based UCI report, and wherein the network configuration further indicates a maximum RI associated with the AI ​​model ID and another maximum RI associated with another AI model ID, wherein the other AI model ID is for a different AI model and has the same number or a higher number of UCI bits as compared to the AI ​​model ID, or wherein the UCI configuration varies between transmission layers of the RI.

[0121] The twenty-third embodiment may include any one or more embodiments of the twentieth to twenty-second embodiments, wherein the network configuration indicates a layer-specific configuration, so that the AI ​​model varies based on the transmission layer of the RI, and further indicates that the UCI configuration is the same number or different number of UCI bits between the transmission layers associated with the RI, so that the UCI payload size increases linearly with the increase of the RI value of the RI.

[0122] The twenty-fourth embodiment may include any one or more embodiments of the twentieth to twenty-third embodiments, wherein the network configuration indicates a layer-specific configuration to be used for the AI-based UCI reporting, wherein the AI ​​model varies between different transmission layers according to a different AI model list for each transmission layer and RI, or wherein the network configuration indication includes RI-specific configurations of different AI models associated with different RIs, respectively, and indicates the same number or different number of output bits per RI.

[0123] The twenty-fifth embodiment may include any one or more embodiments of the twentieth to twenty-fourth embodiments, wherein the network configuration indicates a maximum UCI payload size to enable the UE to select the AI ​​model and the UC configuration based on the indication of the AI ​​model and the UC configuration, based on at least one of the following: a layer common configuration, a layer specific configuration, or a rank indication (RI) specific configuration, wherein the AI ​​model includes a two-sided model that performs AI-based CSI compression for the AI-based UCI report.

[0124] In addition, the various aspects or features described herein can be implemented as methods, devices or products using standard programming and / or engineering techniques. As used herein, the term "product" is intended to cover computer programs that can be accessed from any computer-readable device, carrier or medium. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., EPROMs, cards, sticks, key drives, etc.). In addition, the various storage media described herein may represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing and / or carrying instructions and / or data. In addition, a computer program product may include a computer-readable medium having one or more instructions or codes that are operable to cause a computer to perform the functions described herein.

[0125] Communication media embodies computer-readable instructions, data structures, program modules, or other structured or unstructured data in a data signal, such as a modulated data signal, such as a carrier wave, or other transport mechanism, and includes any information delivery or transmission medium. The term "modulated data signal" or signal refers to a signal that has one or more of its characteristics set or changed in such a manner as to encode information in one or more signals. By way of example, and not limitation, communication media includes wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, RF, infrared, and other wireless media.

[0126] An exemplary storage medium may be coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. In an alternative, the storage medium may be integrated with the processor. Furthermore, in some aspects, the processor and the storage medium may reside in an ASIC. Additionally, the ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components. Additionally, in some aspects, the processes and / or actions of the method or algorithm may reside on a machine-readable medium and / or computer-readable medium as one or any combination or set of codes and / or instructions and may be incorporated into a computer program product.

[0127] In this regard, although the subject matter disclosed herein has been described in conjunction with various embodiments and corresponding drawings, it should be understood that other similar embodiments may be used, or modifications and additions may be made to the described embodiments, where applicable, to perform the same, similar, alternative, or alternative functions of the disclosed subject matter without departing from the described embodiments. Accordingly, the disclosed subject matter should not be limited to any single embodiment described herein, but rather should be construed in accordance with the breadth and scope of the claims appended hereto.

[0128] In particular, with respect to the various functions performed by the aforementioned components (assemblies, devices, circuits, systems, etc.), unless otherwise indicated, terms used to describe such components (including references to "means") are intended to correspond to any component or structure that performs the specified function of the described component (e.g., functionally equivalent), even if not structurally equivalent to the disclosed structure that performs the function in the exemplary embodiments of the present disclosure illustrated herein. In addition, although particular features have been disclosed with respect to only one of several embodiments, for any given or particular application, such features may be combined with one or more other features of other embodiments, as may be desirable and advantageous.

Claims

1. A user equipment (UE), comprising: Memory; a processing circuit coupled to the memory, the processing circuit being configured to, when executing instructions stored in the memory, cause the UE to: receiving a network configuration associated with artificial intelligence (AI)-based uplink control information (UCI) reporting, wherein the network configuration includes one or more indications of a UCI configuration and an AI model to generate the AI-based UCI reporting; and The AI-based UCI report is sent based on the network configuration.

2. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI-based UCI report is generated using the AI ​​model, wherein based on the indication of the AI ​​model and the UCI configuration, the AI ​​model is associated with at least one of the following: a layer-common configuration, a layer-specific configuration, or a rank indication (RI)-specific configuration, wherein the AI ​​model includes a bilateral model that performs AI-based CSI compression for the AI-based UCI report, wherein the layer-common configuration includes using the same AI model across multiple transmission layers of the UCI configuration, wherein the layer-specific configuration includes using the same AI model for one transmission layer per UCI configuration, and wherein the RI-specific configuration includes using a different AI model per RI.

3. The UE according to claim 1, wherein the network configuration further indicates an AI model ID, wherein the AI ​​model ID indicates at least one of the following: the UCI configuration associated with multiple UCI bits, the AI ​​model including an AI model index and the UCI configuration, an AI configuration or an AI method, a UE vendor identifier, a network device vendor identifier, a use case ID of a use case, or a model number of the use case ID.

4. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: UCI bits for the AI-based UCI report are generated based on an AI model ID of the network configuration, wherein the AI ​​model ID indicates a layer common configuration for configuring a UCI payload size that increases linearly with an increase in an RI value associated with the UCI configuration.

5. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: Selecting an AI model ID indicated in the network configuration from a plurality of AI model IDs based on the RI and the UCI configuration to generate UCI bits for the AI-based UCI report; and The AI ​​model ID indicates a layer common configuration to configure a UCI payload size, wherein the multiple AI model IDs include a first AI model ID associated with a first set of RIs and a second AI model ID associated with a second set of RIs having a greater RI value than the first set of RIs, and wherein the first AI model ID and the second model ID respectively include different AI models that constrain the UCI payload size of the second set of RIs relative to the first set of RIs.

6. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: For a layer-common configuration indicated by the network configuration, the AI ​​model is selected based on the RI to encode each transmission layer, wherein the UCI configuration is different between RIs and is associated with the number of UCI bits.

7. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: For a layer-common configuration indicated by the network configuration, each transmission layer is encoded based on the AI ​​model and the RI, wherein the UCI configuration varies between transmission layers or does not vary based on the RI.

8. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI ​​model is selected based on a maximum UCI payload size indicated by the network configuration.

9. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: A UE capability report is provided, the UE capability report indicating support for at least one of: one or more AI models, one or more AI model IDs associated with the one or more AI models, or one or more UCI configurations for the AI-based UCI reporting.

10. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI-based UCI report is generated using the AI ​​model based on a layer-specific configuration, wherein the AI ​​model varies based on the transmission layer, wherein the UCI configuration is the same number or a different number of UCI bits between the transmission layers associated with the RI, and the UCI payload size increases linearly based on the UE rank selection of the RI.

11. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI-based UCI report is generated using the AI ​​model based on RI and layer-specific configuration, in which the AI ​​model corresponds to one transmission layer and varies between transmission layers based on a different AI model ID list for each transmission layer and RI.

12. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI-based UCI report is generated using the AI ​​model based on an RI-specific configuration, wherein the RI-specific configuration includes different AI models associated with different RIs, respectively, according to a network configuration, wherein the network configuration indicates a same number or a different number of output bits for each RI.

13. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: generating the AI-based UCI report based on the network configuration, wherein a channel state information (CSI) part 1 of the AI-based UCI report indicates a size of a CSI part 2 by an AI model ID based on whether only the network configuration determines or the UE also determines which AI model to utilize to generate UCI bits having a layer-common configuration or a layer-specific configuration; or generating the AI-based UCI report based on the network configuration, wherein the CSI part 1 of the AI-based UCI report indicates an AI model ID based on a plurality of AI model IDs for selection by the UE, wherein the AI ​​model ID is associated with the UE selection for generating UCI bits having a layer-common configuration or a layer-specific configuration; or The AI-based UCI report is generated based on the network configuration, wherein the CSI part 1 of the AI-based UCI report indicates an output bit size of an AI model with an adaptation layer based on multiple adaptation layers for UE selection, wherein the AI ​​model ID and the corresponding adaptation layer are associated with the UE selection for generating UCI bits for the layer-common configuration or the layer-specific configuration.

14. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: The AI-based UCI report is generated based on the network configuration for a rank-specific configuration, wherein a UCI part 1 of the AI-based UCI report provides an RI, the RI indicates an AI model ID associated with the RI, and in response to more than one AI model ID being associated with the RI, the UCI part 1 further includes an AI model ID index, or the RI and the AI ​​model ID are signaled via radio resource control (RRC) signaling or medium access control (MAC) control element (MAC CE) signaling via the network configuration.

15. The UE of claim 1 , wherein the instructions, when executed by the processing circuit, further cause the UE to: Generating the AI-based UCI report having a CSI part 2, wherein the CSI part 2 includes: For layer-common configuration, the CSI generation model output for each transmission layer depends on the RI selected in the sequential order of the transmission layers; For layer-specific configurations, the CSI generation model output for each transmit layer depends on the selected RI and the corresponding AI model output; generating the CSI generation model output for a rank-specific configuration; or UCI bit omission for the CSI part 2 is generated based on a priority order associated with a layer-common configuration or the layer-specific configuration, wherein the priority order decreases from a higher layer to a lower layer.

16. A baseband processor, the baseband processor being configured to perform operations comprising: receiving, via the processing circuitry, a network configuration associated with an uplink control information (UCI) payload of artificial intelligence (AI)-based channel state information (CSI) feedback, wherein the network configuration includes one or more indications of a UCI configuration and an AI model to generate an AI-based UCI report; and The AI-based UCI report is sent based on the network configuration.

17. The baseband processor according to claim 16, further comprising: The AI-based UCI report is generated using the AI ​​model, wherein based on the one or more indications of the UCI configuration and the AI ​​model, the AI ​​model is associated with at least one of the following: a layer-common configuration, a layer-specific configuration, or a rank indication (RI)-specific configuration, wherein the AI ​​model includes a bilateral model that performs AI-based CSI compression for the AI-based UCI report, wherein the layer-common configuration includes using the same AI model across multiple transmission layers of the UCI configuration, wherein the layer-specific configuration includes using the same AI model for one transmission layer per UCI configuration, and wherein the RI-specific configuration includes using a different AI model per RI.

18. A base station, comprising: Memory; a processing circuit coupled to the memory, the processing circuit being configured to, when executing instructions stored in the memory, cause the base station to: sending a network configuration for artificial intelligence (AI)-based UCI reporting, wherein the network configuration includes an indication of a UCI configuration and an AI model; and The AI-based UCI report is received based on the network configuration.

19. The base station according to claim 18, wherein the network configuration indicates an AI model ID having a layer-common configuration for configuring UCI bits of a UCI payload size in the AI-based UCI report, and wherein the UCI payload size increases linearly with an increase in the RI associated with the UCI configuration including the number of UCI bits for each transmission layer, or wherein according to the indication of the network configuration, the UCI configuration is different between RIs, each RI being associated with an AI model.

20. The base station according to claim 18, wherein the network configuration indicates an AI model ID associated with a layer common configuration for configuring UCI bits of a UCI payload size in the AI-based UCI report, and wherein the network configuration further indicates a maximum RI associated with the AI ​​model ID and another maximum RI associated with another AI model ID, wherein the other AI model ID is for a different AI model and has the same number or a higher number of UCI bits than the AI ​​model ID, or wherein the UCI configuration varies between transmission layers of the RI.

Images