Inter-vendor collaboration with temporal compression
Standardized dataset and model sharing between UE and network vendors improve CSI compression, addressing suboptimal inter-vendor collaboration issues and enhancing communication throughput.
Patent Information
- Application Number
- PCT/CN2024/112668
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2026-02-19
AI Technical Summary
Inter-vendor collaboration in machine learning-based channel state information (CSI) compression results in suboptimal compression and reconstruction, leading to inaccurate CSI delivery, increased latency, and reduced throughput due to lack of standardized data and model sharing between UE and network vendors.
Implementing a method for inter-vendor collaboration by sharing dataset and model information, including timing information and ML model parameters, to enhance spatial, temporal, and frequency compression of CSI, thereby improving CSI compression and decompression.
Enhances CSI compression and scheduled communications, increasing throughput and reducing latency by optimizing inter-vendor collaboration through standardized data and model sharing.
Smart Images

Figure CN2024112668_19022026_PF_FP_ABST
Abstract
Description
INTER-VENDOR COLLABORATION WITH TEMPORAL COMPRESSION
[0001] FIELD OF THE DISCLOSURE
[0002] Aspects of the present disclosure generally relate to wireless communication and specifically relate to techniques, apparatuses, and methods for inter-vendor collaboration with temporal compression.BACKGROUND
[0003] Wireless communication systems are widely deployed to provide various services that may include carrying voice, text, messaging, video, data, and / or other traffic. The services may include unicast, multicast, and / or broadcast services, among other examples. Typical wireless communication systems may employ multiple-access radio access technologies (RATs) capable of supporting communication with multiple users by sharing available system resources (for example, time domain resources, frequency domain resources, spatial domain resources, and / or device transmit power, among other examples) . Examples of such multiple-access RATs include code division multiple access (CDMA) systems, time division multiple access (TDMA) systems, frequency division multiple access (FDMA) systems, orthogonal frequency division multiple access (OFDMA) systems, single-carrier frequency division multiple access (SC-FDMA) systems, and time division synchronous code division multiple access (TD-SCDMA) systems.
[0004] The above multiple-access RATs have been adopted in various telecommunication standards to provide common protocols that enable different wireless communication devices to communicate on a municipal, national, regional, or global level. An example telecommunication standard is New Radio (NR) . NR, which may also be referred to as 5G, is part of a continuous mobile broadband evolution promulgated by the Third Generation Partnership Project (3GPP) . NR (and other mobile broadband evolutions beyond NR) may be designed to better support Internet of things (IoT) and reduced capability device deployments, industrial connectivity, millimeter wave (mmWave) expansion, licensed and unlicensed spectrum access, non-terrestrial network (NTN) deployment, sidelink and other device-to-device direct communication technologies (for example, cellular vehicle-to-everything (CV2X) communication) , massive multiple-input multiple-output (MIMO) , disaggregated network architectures and network topology expansions, multiple-subscriber implementations, high-precision positioning, and / or radio frequency (RF) sensing, among other examples. As the demand for mobile broadband access continues to increase, further improvements in NR may be implemented, and other radio access technologies such as 6G may be introduced, to further advance mobile broadband evolution.SUMMARY
[0005] Some aspects described herein relate to a method of wireless communication performed by a network entity. The method may include obtaining dataset information from a user equipment (UE) . The method may include training machine learning (ML) models based at least in part on the dataset information. The method may include transmitting model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of channel state information (CSI) .
[0006] Some aspects described herein relate to a method of wireless communication performed by a UE. The method may include transmitting dataset information. The method may include receiving model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information.
[0007] Some aspects described herein relate to an apparatus for wireless communication at a network entity. The apparatus may include one or more memories and one or more processors coupled to the one or more memories. The one or more processors may be individually or collectively configured to obtain dataset information from a UE. The one or more processors may be configured to training ML models based at least in part on the dataset information. The one or more processors may be individually or collectively configured to transmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI.
[0008] Some aspects described herein relate to an apparatus for wireless communication at a UE. The apparatus may include one or more memories and one or more processors coupled to the one or more memories. The one or more processors may be individually or collectively configured to transmit dataset information. The one or more processors may be individually or collectively configured to receive model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information.
[0009] Some aspects described herein relate to a non-transitory computer-readable medium that stores a set of instructions for wireless communication by a network entity. The set of instructions, when executed by one or more processors of the network entity, may cause the network entity to obtain dataset information from a UE. The set of instructions, when executed by one or more processors of the network entity, may cause the network entity to training ML models based at least in part on the dataset information. The set of instructions, when executed by one or more processors of the network entity, may cause the network entity to transmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI.
[0010] Some aspects described herein relate to a non-transitory computer-readable medium that stores a set of instructions for wireless communication by a UE. The set of instructions, when executed by one or more processors of the UE, may cause the UE to transmit dataset information. The set of instructions, when executed by one or more processors of the UE, may cause the UE to receive model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information.
[0011] Some aspects described herein relate to an apparatus for wireless communication. The apparatus may include means for obtaining dataset information from another apparatus. The apparatus may include training ML models based at least in part on the dataset information. The apparatus may include means for transmitting model information to the other apparatus that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI.
[0012] Some aspects described herein relate to an apparatus for wireless communication. The apparatus may include means for transmitting dataset information. The apparatus may include means for receiving model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information.
[0013] Aspects of the present disclosure may generally be implemented by or as a method, apparatus, system, computer program product, non-transitory computer-readable medium, user equipment, base station, network node, network entity, wireless communication device, and / or processing system as substantially described with reference to, and as illustrated by, the specification and accompanying drawings.
[0014] The foregoing paragraphs of this section have broadly summarized some aspects of the present disclosure. These and additional aspects and associated advantages will be described hereinafter. The disclosed aspects may be used as a basis for modifying or designing other aspects for carrying out the same or similar purposes of the present disclosure. Such equivalent aspects do not depart from the scope of the appended claims. Characteristics of the aspects disclosed herein, both their organization and method of operation, together with associated advantages, will be better understood from the following description when considered in connection with the accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS
[0015] The appended drawings illustrate some aspects of the present disclosure, but are not limiting of the scope of the present disclosure because the description may enable other aspects. Each of the drawings is provided for purposes of illustration and description, and not as a definition of the limits of the claims. The same or similar reference numbers in different drawings may identify the same or similar elements.
[0016] Fig. 1 is a diagram illustrating an example of a wireless communication network, in accordance with the present disclosure.
[0017] Fig. 2 is a diagram illustrating an example network node in communication with an example user equipment (UE) in a wireless network, in accordance with the present disclosure.
[0018] Fig. 3 is a diagram illustrating an example disaggregated base station architecture, in accordance with the present disclosure.
[0019] Fig. 4 is a diagram illustrating an example architecture of a functional framework for radio access network (RAN) intelligence enabled by data collection, in accordance with the present disclosure.
[0020] Fig. 5 is a diagram illustrating an example of encoding and decoding channel state information (CSI) , in accordance with the present disclosure.
[0021] Fig. 6 is a diagram illustrating an example of CSI feedback, in accordance with the present disclosure.
[0022] Fig. 7 is a diagram illustrating an example of inter-collaboration for machine learning CSI feedback with temporal compression, in accordance with the present disclosure.
[0023] Fig. 8 is a diagram illustrating an example of CSI compression and prediction, in accordance with the present disclosure.
[0024] Fig. 9 is a diagram illustrating an example of sharing modeling information for Option 4, in accordance with the present disclosure.
[0025] Fig. 10 is a diagram illustrating an example of sharing modeling information for Option 4, in accordance with the present disclosure.
[0026] Fig. 11 is a diagram illustrating an example of modeling information, in accordance with the present disclosure.
[0027] Fig. 12 is a diagram illustrating an example of compression and quantization, in accordance with the present disclosure.
[0028] Fig. 13 is a diagram illustrating an example of a refinement architecture, in accordance with the present disclosure.
[0029] Fig. 14 is a diagram illustrating an example process performed, for example, at a network entity or an apparatus of a network entity, in accordance with the present disclosure.
[0030] Fig. 15 is a diagram illustrating an example process performed, for example, at a UE or an apparatus of a UE, in accordance with the present disclosure.
[0031] Fig. 16 is a diagram of an example apparatus for wireless communication, in accordance with the present disclosure.
[0032] Fig. 17 is a diagram of an example apparatus for wireless communication, in accordance with the present disclosure.DETAILED DESCRIPTION
[0033] Various aspects of the present disclosure are described hereinafter with reference to the accompanying drawings. However, aspects of the present disclosure may be embodied in many different forms and is not to be construed as limited to any specific aspect illustrated by or described with reference to an accompanying drawing or otherwise presented in this disclosure. Rather, these aspects are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art. One skilled in the art may appreciate that the scope of the disclosure is intended to cover any aspect of the disclosure disclosed herein, whether implemented independently of or in combination with any other aspect of the disclosure. For example, an apparatus may be implemented or a method may be practiced using various combinations or quantities of the aspects set forth herein. In addition, the scope of the disclosure is intended to cover an apparatus having, or a method that is practiced using, other structures and / or functionalities in addition to or other than the structures and / or functionalities with which various aspects of the disclosure set forth herein may be practiced. Any aspect of the disclosure disclosed herein may be embodied by one or more elements of a claim.
[0034] Several aspects of telecommunication systems will now be presented with reference to various methods, operations, apparatuses, and techniques. These methods, operations, apparatuses, and techniques will be described in the following detailed description and illustrated in the accompanying drawings by various blocks, modules, components, circuits, steps, processes, or algorithms (collectively referred to as “elements” ) . These elements may be implemented using hardware, software, or a combination of hardware and software. Whether such elements are implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system.
[0035] A user equipment (UE) may measure reference signals on a channel and provide channel state information (CSI) feedback to assist a network entity with scheduling communications. CSI may be encoded and compressed at the UE (UE vendor) , and decompressed and decoded at the network entity (network vendor) , using machine learning (ML) . In such cross-node ML, a neural network may be split into two portions, where a first portion includes the encoder of the UE, and a second portion includes the decoder of the network entity. The encoder output of the UE is transmitted to the network entity as an input to the decoder. For example, the encoder at the UE may input CSI (e.g., downlink channel estimates) and output compressed CSI or another data signal, which is received as input at the decoder of the network entity. The decoder at the network entity may output a reconstructed CSI or another data signal, such as precoding vectors.
[0036] To alleviate or resolve issues related to inter-vendor collaboration of ML-based CSI compression, different options may be used. Option 1 may involve a fully standardized reference model (structure and parameters) . Option 2 may involve a standardized dataset. Option 3 may involve a standardized reference model structure and parameter exchange between the network-side and UE-side. Option 4 may involve a standardized data / dataset format and a dataset exchange between the network-side and the UE-side. Option 5 may involve a standardized model format and reference model exchange between the network-side and the UE-side.
[0037] CSI compression may include spatial compression and frequency compression. In some aspects, CSI compression may further include temporal compression, which involves multiple CSI samples over a period of time (e.g., observation time window) . Temporal compression of CSI may involve different cases. In an initial case (case 0) , the target CSI is in a present slot and encoding uses no past CSI information. For the UE, the past CSI information may include past model inputs and / or any information derived from them. For the network, the past CSI information may include past CSI feedback instances and / or any information derived from them.
[0038] In a first case (case 1) , the target CSI is in a present slot, the UE uses past CSI information for encoding CSI (generating CSI) , and the network entity does not use past CSI information for decoding (reconstructing CSI) . “Target CSI slot (s) ” refers to the slot (s) to which the CSI feedback in the report corresponds. “Present slot” refers to the slot of the most recent CSI reference signal (CSI-RS) measurement used to generate the CSI report. “Future slot (s) ” includes at least one slot after the present slot and may include the present slot as well.
[0039] In a second case (case 2) , the target CSI is in a present slot, the UE uses past CSI information for encoding, and the network entity uses past CSI information for decoding. In a third case (case 3) , the target CSI is in a future slot, the UE uses past CSI information, and the network entity does not use past CSI information. In a fourth case (case 4) , the target CSI is in a future slot, the UE uses past CSI information, and the network entity uses past CSI information. Case 4 may be a combination of case 2 and case 3. For cases 3 and case 4, the UE may perform CSI prediction as a separate step or jointly with CSI compression. Similarly, the network may perform CSI prediction as a separate step or jointly with CSI reconstruction. In a fifth case (case 5) , the target CSI is in a present slot, the UE does not use past CSI information, and the network entity uses past CSI information.
[0040] In some aspects, temporal compression may be used for cases 1, 2, and 5 using historical information to help with the compression. Temporal compression may be used for case 3, which involves compression and prediction. Temporal compression may be used for case 4, which involves prediction and compression. Historical information may also be used for case 4. For Options 3, 4, and 5, UE vendors may share data with network vendors, which use the data to train CSI compression and decompression ML models. However, with Options 3, 4, and 5, data is shared between vendors (e.g., UE and network entity) . If the vendors do not have information for sharing data and training models and / or parameters in the context of a specific Option and one of the cases, the inter-vendor collaboration may perform suboptimal compression and reconstruction. Suboptimal compression and reconstruction may deliver inaccurate CSI or delay the delivery of CSI, which may lead to a suboptimal scheduling of communications. Suboptimal scheduling increases latency, wastes signaling resources, and decreases throughput.
[0041] Various aspects relate generally to CSI feedback. Some aspects more specifically relate to a UE and a network entity that may collaborate to share dataset information and model information to improve CSI compression and decompression. In some aspects, the UE may transmit dataset information to the network entity. The dataset information may include timing information, such as information about time-tagged samples. The timing information may indicate a time window of the time-tagged samples and / or a quantity of the time-tagged samples.
[0042] In some aspects, the dataset information may include an input CSI (input of an encoder) that matches (is the same as) a target CSI, which may occur with ideal prediction (e.g., no prediction errors) . In some aspects, the dataset information may include an input CSI, a target CSI, and a predicted CSI. The network entity may train ML models based at least in part on the dataset information. The network entity may then transmit ML model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI. The model information may include a model format of a model structure and / or at least one parameter for the model structure.
[0043] Particular aspects of the subject matter described in this disclosure can be implemented to realize one or more of the following potential advantages. By receiving dataset information from the UE and providing ML model information that is based on training with the data set information, the network entity may improve inter-vendor collaboration for various combinations of Option and case. Inter-vendor collaboration may be simplified for CSI compression. As a result, CSI compression and scheduled communications improve, which increases throughput.
[0044] Multiple-access radio access technologies (RATs) have been adopted in various telecommunication standards to provide common protocols that enable wireless communication devices to communicate on a municipal, enterprise, national, regional, or global level. For example, 5G New Radio (NR) is part of a continuous mobile broadband evolution promulgated by the Third Generation Partnership Project (3GPP) . 5G NR supports various technologies and use cases including enhanced mobile broadband (eMBB) , ultra-reliable low-latency communication (URLLC) , massive machine-type communication (mMTC) , millimeter wave (mmWave) technology, beamforming, network slicing, edge computing, Internet of Things (IoT) connectivity and management, and network function virtualization (NFV) .
[0045] As the demand for broadband access increases and as technologies supported by wireless communication networks evolve, further technological improvements may be adopted in or implemented for 5G NR or future RATs, such as 6G, to further advance the evolution of wireless communication for a wide variety of existing and new use cases and applications. Such technological improvements may be associated with new frequency band expansion, licensed and unlicensed spectrum access, overlapping spectrum use, small cell deployments, non-terrestrial network (NTN) deployments, disaggregated network architectures and network topology expansion, device aggregation, advanced duplex communication, sidelink and other device-to-device direct communication, IoT (including passive or ambient IoT) networks, reduced capability (RedCap) UE functionality, industrial connectivity, multiple-subscriber implementations, high-precision positioning, radio frequency (RF) sensing, and / or artificial intelligence or machine learning (AI / ML) , among other examples. These technological improvements may support use cases such as wireless backhauls, wireless data centers, extended reality (XR) and metaverse applications, meta services for supporting vehicle connectivity, holographic and mixed reality communication, autonomous and collaborative robots, vehicle platooning and cooperative maneuvering, sensing networks, gesture monitoring, human-brain interfacing, digital twin applications, asset management, and universal coverage applications using non-terrestrial and / or aerial platforms, among other examples. The methods, operations, apparatuses, and techniques described herein may enable one or more of the foregoing technologies and / or support one or more of the foregoing use cases.
[0046] Fig. 1 is a diagram illustrating an example of a wireless communication network 100, in accordance with the present disclosure. The wireless communication network 100 may be or may include elements of a 5G (or NR) network or a 6G network, among other examples. The wireless communication network 100 may include multiple network nodes 110, shown as a network node (NN) 110a, a network node 110b, a network node 110c, and a network node 110d. The network nodes 110 may support communications with multiple UEs 120, shown as a UE 120a, a UE 120b, a UE 120c, a UE 120d, and a UE 120e.
[0047] The network nodes 110 and the UEs 120 of the wireless communication network 100 may communicate using the electromagnetic spectrum, which may be subdivided by frequency or wavelength into various classes, bands, carriers, and / or channels. For example, devices of the wireless communication network 100 may communicate using one or more operating bands. In some aspects, multiple wireless networks 100 may be deployed in a given geographic area. Each wireless communication network 100 may support a particular RAT (which may also be referred to as an air interface) and may operate on one or more carrier frequencies in one or more frequency ranges. Examples of RATs include a 4G RAT, a 5G / NR RAT, and / or a 6G RAT, among other examples. In some examples, when multiple RATs are deployed in a given geographic area, each RAT in the geographic area may operate on different frequencies to avoid interference with one another.
[0048] Various operating bands have been defined as frequency range designations FR1 (410 MHz through 7.125 GHz) , FR2 (24.25 GHz through 52.6 GHz) , FR3 (7.125 GHz through 24.25 GHz) , FR4a or FR4-1 (52.6 GHz through 71 GHz) , FR4 (52.6 GHz through 114.25 GHz) , and FR5 (114.25 GHz through 300 GHz) . Although a portion of FR1 is greater than 6 GHz, FR1 is often referred to (interchangeably) as a “Sub-6 GHz” band in some documents and articles. Similarly, FR2 is often referred to (interchangeably) as a “millimeter wave” band in some documents and articles, despite being different than the extremely high frequency (EHF) band (30 GHz through 300 GHz) , which is identified by the International Telecommunications Union (ITU) as a “millimeter wave” band. The frequencies between FR1 and FR2 are often referred to as mid-band frequencies, which include FR3. Frequency bands falling within FR3 may inherit FR1 characteristics or FR2 characteristics, and thus may effectively extend features of FR1 or FR2 into mid-band frequencies. Thus, “sub-6 GHz, ” if used herein, may broadly refer to frequencies that are less than 6 GHz, that are within FR1, and / or that are included in mid-band frequencies. Similarly, the term “millimeter wave, ” if used herein, may broadly refer to frequencies that are included in mid-band frequencies, that are within FR2, FR4, FR4-a or FR4-1, or FR5, and / or that are within the EHF band. Higher frequency bands may extend 5G NR operation, 6G operation, and / or other RATs beyond 52.6 GHz. For example, each of FR4a, FR4-1, FR4, and FR5 falls within the EHF band. In some examples, the wireless communication network 100 may implement dynamic spectrum sharing (DSS) , in which multiple RATs (for example, 4G / LTE and 5G / NR) are implemented with dynamic bandwidth allocation (for example, based on user demand) in a single frequency band. It is contemplated that the frequencies included in these operating bands (for example, FR1, FR2, FR3, FR4, FR4-a, FR4-1, and / or FR5) may be modified, and techniques described herein may be applicable to those modified frequency ranges.
[0049] A network node 110 may include one or more devices, components, or systems that enable communication between a UE 120 and one or more devices, components, or systems of the wireless communication network 100. A network node 110 may be, may include, or may also be referred to as an NR network node, a 5G network node, a 6G network node, a Node B, an eNB, a gNB, an access point (AP) , a transmission reception point (TRP) , a mobility element, a core, a network entity, a network element, a network equipment, and / or another type of device, component, or system included in a radio access network (RAN) .
[0050] A network node 110 may be implemented as a single physical node (for example, a single physical structure) or may be implemented as two or more physical nodes (for example, two or more distinct physical structures) . For example, a network node 110 may be a device or system that implements part of a radio protocol stack, a device or system that implements a full radio protocol stack (such as a full gNB protocol stack) , or a collection of devices or systems that collectively implement the full radio protocol stack. For example, and as shown, a network node 110 may be an aggregated network node (having an aggregated architecture) , meaning that the network node 110 may implement a full radio protocol stack that is physically and logically integrated within a single node (for example, a single physical structure) in the wireless communication network 100. For example, an aggregated network node 110 may consist of a single standalone base station or a single TRP that uses a full radio protocol stack to enable or facilitate communication between a UE 120 and a core network of the wireless communication network 100.
[0051] Alternatively, and as also shown, a network node 110 may be a disaggregated network node (sometimes referred to as a disaggregated base station) , meaning that the network node 110 may implement a radio protocol stack that is physically distributed and / or logically distributed among two or more nodes in the same geographic location or in different geographic locations. For example, a disaggregated network node may have a disaggregated architecture. In some deployments, disaggregated network nodes 110 may be used in an integrated access and backhaul (IAB) network, in an open radio access network (O-RAN) (such as a network configuration in compliance with the O-RAN Alliance) , or in a virtualized radio access network (vRAN) , also known as a cloud radio access network (C-RAN) , to facilitate scaling by separating base station functionality into multiple units that can be individually deployed.
[0052] The network nodes 110 of the wireless communication network 100 may include one or more central units (CUs) , one or more distributed units (DUs) , and / or one or more radio units (RUs) . A CU may host one or more higher layer control functions, such as radio resource control (RRC) functions, packet data convergence protocol (PDCP) functions, and / or service data adaptation protocol (SDAP) functions, among other examples. A DU may host one or more of a radio link control (RLC) layer, a medium access control (MAC) layer, and / or one or more higher physical (PHY) layers depending, at least in part, on a functional split, such as a functional split defined by the 3GPP. In some examples, a DU also may host one or more lower PHY layer functions, such as a fast Fourier transform (FFT) , an inverse FFT (iFFT) , beamforming, physical random access channel (PRACH) extraction and filtering, and / or scheduling of resources for one or more UEs 120, among other examples. An RU may host RF processing functions or lower PHY layer functions, such as an FFT, an iFFT, beamforming, or PRACH extraction and filtering, among other examples, according to a functional split, such as a lower layer functional split. In such an architecture, each RU can be operated to handle over the air (OTA) communication with one or more UEs 120.
[0053] In some aspects, a single network node 110 may include a combination of one or more CUs, one or more DUs, and / or one or more RUs. Additionally or alternatively, a network node 110 may include one or more Near-Real Time (Near-RT) RAN Intelligent Controllers (RICs) and / or one or more Non-Real Time (Non-RT) RICs. In some examples, a CU, a DU, and / or an RU may be implemented as a virtual unit, such as a virtual central unit (VCU) , a virtual distributed unit (VDU) , or a virtual radio unit (VRU) , among other examples. A virtual unit may be implemented as a virtual network function, such as associated with a cloud deployment.
[0054] Some network nodes 110 (for example, a base station, an RU, or a TRP) may provide communication coverage for a particular geographic area. In the 3GPP, the term “cell” can refer to a coverage area of a network node 110 or to a network node 110 itself, depending on the context in which the term is used. A network node 110 may support one or multiple (for example, three) cells. In some examples, a network node 110 may provide communication coverage for a macro cell, a pico cell, a femto cell, or another type of cell. A macro cell may cover a relatively large geographic area (for example, several kilometers in radius) and may allow unrestricted access by UEs 120 with service subscriptions. A pico cell may cover a relatively small geographic area and may allow unrestricted access by UEs 120 with service subscriptions. A femto cell may cover a relatively small geographic area (for example, a home) and may allow restricted access by UEs 120 having association with the femto cell (for example, UEs 120 in a closed subscriber group (CSG) ) . A network node 110 for a macro cell may be referred to as a macro network node. A network node 110 for a pico cell may be referred to as a pico network node. A network node 110 for a femto cell may be referred to as a femto network node or an in-home network node. In some examples, a cell may not necessarily be stationary. For example, the geographic area of the cell may move according to the location of an associated mobile network node 110 (for example, a train, a satellite base station, an unmanned aerial vehicle, or an NTN network node) .
[0055] The wireless communication network 100 may be a heterogeneous network that includes network nodes 110 of different types, such as macro network nodes, pico network nodes, femto network nodes, relay network nodes, aggregated network nodes, and / or disaggregated network nodes, among other examples. In the example shown in Fig. 1, the network node 110a may be a macro network node for a macro cell 130a, the network node 110b may be a pico network node for a pico cell 130b, and the network node 110c may be a femto network node for a femto cell 130c. Various different types of network nodes 110 may generally transmit at different power levels, serve different coverage areas, and / or have different impacts on interference in the wireless communication network 100 than other types of network nodes 110. For example, macro network nodes may have a high transmit power level (for example, 5 to 40 watts) , whereas pico network nodes, femto network nodes, and relay network nodes may have lower transmit power levels (for example, 0.1 to 2 watts) .
[0056] In some examples, a network node 110 may be, may include, or may operate as an RU, a TRP, or a base station that communicates with one or more UEs 120 via a radio access link (which may be referred to as a “Uu” link) . The radio access link may include a downlink and an uplink. “Downlink” (or “DL” ) refers to a communication direction from a network node 110 to a UE 120, and “uplink” (or “UL” ) refers to a communication direction from a UE 120 to a network node 110. Downlink channels may include one or more control channels and one or more data channels. A downlink control channel may be used to transmit downlink control information (DCI) (for example, scheduling information, reference signals, and / or configuration information) from a network node 110 to a UE 120. A downlink data channel may be used to transmit downlink data (for example, user data associated with a UE 120) from a network node 110 to a UE 120. Downlink control channels may include one or more physical downlink control channels (PDCCHs) , and downlink data channels may include one or more physical downlink shared channels (PDSCHs) . Uplink channels may similarly include one or more control channels and one or more data channels. An uplink control channel may be used to transmit uplink control information (UCI) (for example, reference signals and / or feedback corresponding to one or more downlink transmissions) from a UE 120 to a network node 110. An uplink data channel may be used to transmit uplink data (for example, user data associated with a UE 120) from a UE 120 to a network node 110. Uplink control channels may include one or more physical uplink control channels (PUCCHs) , and uplink data channels may include one or more physical uplink shared channels (PUSCHs) . The downlink and the uplink may each include a set of resources on which the network node 110 and the UE 120 may communicate.
[0057] Downlink and uplink resources may include time domain resources (frames, subframes, slots, and / or symbols) , frequency domain resources (frequency bands, component carriers, subcarriers, resource blocks, and / or resource elements) , and / or spatial domain resources (particular transmit directions and / or beam parameters) . Frequency domain resources of some bands may be subdivided into bandwidth parts (BWPs) . A BWP may be a continuous block of frequency domain resources (for example, a continuous block of resource blocks) that are allocated for one or more UEs 120. A UE 120 may be configured with both an uplink BWP and a downlink BWP (where the uplink BWP and the downlink BWP may be the same BWP or different BWPs) . A BWP may be dynamically configured (for example, by a network node 110 transmitting a DCI configuration to the one or more UEs 120) and / or reconfigured, which means that a BWP can be adjusted in real-time (or near-real-time) based on changing network conditions in the wireless communication network 100 and / or based on the specific requirements of the one or more UEs 120. This enables more efficient use of the available frequency domain resources in the wireless communication network 100 because fewer frequency domain resources may be allocated to a BWP for a UE 120 (which may reduce the quantity of frequency domain resources that a UE 120 is required to monitor) , leaving more frequency domain resources to be spread across multiple UEs 120. Thus, BWPs may also assist in the implementation of lower-capability UEs 120 by facilitating the configuration of smaller bandwidths for communication by such UEs 120.
[0058] As described above, in some aspects, the wireless communication network 100 may be, may include, or may be included in, an IAB network. In an IAB network, at least one network node 110 is an anchor network node that communicates with a core network. An anchor network node 110 may also be referred to as an IAB donor (or “IAB-donor” ) . The anchor network node 110 may connect to the core network via a wired backhaul link. For example, an Ng interface of the anchor network node 110 may terminate at the core network. Additionally or alternatively, an anchor network node 110 may connect to one or more devices of the core network that provide a core access and mobility management function (AMF) . An IAB network also generally includes multiple non-anchor network nodes 110, which may also be referred to as relay network nodes or simply as IAB nodes (or “IAB-nodes” ) . Each non-anchor network node 110 may communicate directly with the anchor network node 110 via a wireless backhaul link to access the core network, or may communicate indirectly with the anchor network node 110 via one or more other non-anchor network nodes 110 and associated wireless backhaul links that form a backhaul path to the core network. Some anchor network node 110 or other non-anchor network node 110 may also communicate directly with one or more UEs 120 via wireless access links that carry access traffic. In some examples, network resources for wireless communication (such as time resources, frequency resources, and / or spatial resources) may be shared between access links and backhaul links.
[0059] In some examples, any network node 110 that relays communications may be referred to as a relay network node, a relay station, or simply as a relay. A relay may receive a transmission of a communication from an upstream station (for example, another network node 110 or a UE 120) and transmit the communication to a downstream station (for example, a UE 120 or another network node 110) . In this case, the wireless communication network 100 may include or be referred to as a “multi-hop network. ” In the example shown in Fig. 1, the network node 110d (for example, a relay network node) may communicate with the network node 110a (for example, a macro network node) and the UE 120d in order to facilitate communication between the network node 110a and the UE 120d. Additionally or alternatively, a UE 120 may be or may operate as a relay station that can relay transmissions to or from other UEs 120. A UE 120 that relays communications may be referred to as a UE relay or a relay UE, among other examples.
[0060] The UEs 120 may be physically dispersed throughout the wireless communication network 100, and each UE 120 may be stationary or mobile. A UE 120 may be, may include, or may be included in an access terminal, another terminal, a mobile station, or a subscriber unit. A UE 120 may be, include, or be coupled with a cellular phone (for example, a smart phone) , a personal digital assistant (PDA) , a wireless modem, a wireless communication device, a handheld device, a laptop computer, a cordless phone, a wireless local loop (WLL) station, a tablet, a camera, a gaming device, a netbook, a smartbook, an ultrabook, a medical device, a biometric device, a wearable device (for example, a smart watch, smart clothing, smart glasses, a smart wristband, and / or smart jewelry, such as a smart ring or a smart bracelet) , an entertainment device (for example, a music device, a video device, and / or a satellite radio) , an XR device, a vehicular component or sensor, a smart meter or sensor, industrial manufacturing equipment, a Global Navigation Satellite System (GNSS) device (such as a Global Positioning System device or another type of positioning device) , a UE function of a network node, and / or any other suitable device or function that may communicate via a wireless medium.
[0061] A UE 120 and / or a network node 110 may include one or more chips, system-on-chips (SoCs) , chipsets, packages, or devices that individually or collectively constitute or comprise a processing system. The processing system includes processor (or “processing” ) circuitry in the form of one or multiple processors, microprocessors, processing units (such as central processing units (CPUs) , graphics processing units (GPUs) , neural processing units (NPUs) and / or digital signal processors (DSPs) ) , processing blocks, application-specific integrated circuits (ASIC) , programmable logic devices (PLDs) (such as field programmable gate arrays (FPGAs) ) , or other discrete gate or transistor logic or circuitry (all of which may be generally referred to herein individually as “processors” or collectively as “the processor” or “the processor circuitry” ) . One or more of the processors may be individually or collectively configurable or configured to perform various functions or operations described herein. A group of processors collectively configurable or configured to perform a set of functions may include a first processor configurable or configured to perform a first function of the set and a second processor configurable or configured to perform a second function of the set, or may include the group of processors all being configured or configurable to perform the set of functions.
[0062] The processing system may further include memory circuitry in the form of one or more memory devices, memory blocks, memory elements or other discrete gate or transistor logic or circuitry, each of which may include tangible storage media such as random-access memory (RAM) or read-only memory (ROM) , or combinations thereof (all of which may be generally referred to herein individually as “memories” or collectively as “the memory” or “the memory circuitry” ) . One or more of the memories may be coupled (for example, operatively coupled, communicatively coupled, electronically coupled, or electrically coupled) with one or more of the processors and may individually or collectively store processor-executable code (such as software) that, when executed by one or more of the processors, may configure one or more of the processors to perform various functions or operations described herein. Additionally or alternatively, in some examples, one or more of the processors may be preconfigured to perform various functions or operations described herein without requiring configuration by software. The processing system may further include or be coupled with one or more modems (such as a Wi-Fi (for example, IEEE compliant) modem or a cellular (for example, 3GPP 4G LTE, 5G, or 6G compliant) modem) . In some implementations, one or more processors of the processing system include or implement one or more of the modems. The processing system may further include or be coupled with multiple radios (collectively “the radio” ) , multiple RF chains, or multiple transceivers, each of which may in turn be coupled with one or more of multiple antennas. In some implementations, one or more processors of the processing system include or implement one or more of the radios, RF chains or transceivers. The UE 120 may include or may be included in a housing that houses components associated with the UE 120 including the processing system.
[0063] Some UEs 120 may be considered machine-type communication (MTC) UEs, evolved or enhanced machine-type communication (eMTC) , UEs, further enhanced eMTC (feMTC) UEs, or enhanced feMTC (efeMTC) UEs, or further evolutions thereof, all of which may be simply referred to as “MTC UEs” . An MTC UE may be, may include, or may be included in or coupled with a robot, an uncrewed aerial vehicle, a remote device, a sensor, a meter, a monitor, and / or a location tag. Some UEs 120 may be considered IoT devices and / or may be implemented as NB-IoT (narrowband IoT) devices. An IoT UE or NB-IoT device may be, may include, or may be included in or coupled with an industrial machine, an appliance, a refrigerator, a doorbell camera device, a home automation device, and / or a light fixture, among other examples. Some UEs 120 may be considered Customer Premises Equipment, which may include telecommunications devices that are installed at a customer location (such as a home or office) to enable access to a service provider's network (such as included in or in communication with the wireless communication network 100) .
[0064] Some UEs 120 may be classified according to different categories in association with different complexities and / or different capabilities. UEs 120 in a first category may facilitate massive IoT in the wireless communication network 100, and may offer low complexity and / or cost relative to UEs 120 in a second category. UEs 120 in a second category may include mission-critical IoT devices, legacy UEs, baseline UEs, high-tier UEs, advanced UEs, full-capability UEs, and / or premium UEs that are capable of URLLC, enhanced mobile broadband (eMBB) , and / or precise positioning in the wireless communication network 100, among other examples. A third category of UEs 120 may have mid-tier complexity and / or capability (for example, a capability between UEs 120 of the first category and UEs 120 of the second capability) . A UE 120 of the third category may be referred to as a reduced capacity UE ( “RedCap UE” ) , a mid-tier UE, an NR-Light UE, and / or an NR-Lite UE, among other examples. RedCap UEs may bridge a gap between the capability and complexity of NB-IoT devices and / or eMTC UEs, and mission-critical IoT devices and / or premium UEs. RedCap UEs may include, for example, wearable devices, IoT devices, industrial sensors, and / or cameras that are associated with a limited bandwidth, power capacity, and / or transmission range, among other examples. RedCap UEs may support healthcare environments, building automation, electrical distribution, process automation, transport and logistics, and / or smart city deployments, among other examples.
[0065] In some examples, two or more UEs 120 (for example, shown as UE 120a and UE 120e) may communicate directly with one another using sidelink communications (for example, without communicating by way of a network node 110 as an intermediary) . As an example, the UE 120a may directly transmit data, control information, or other signaling as a sidelink communication to the UE 120e. This is in contrast to, for example, the UE 120a first transmitting data in an UL communication to a network node 110, which then transmits the data to the UE 120e in a DL communication. In various examples, the UEs 120 may transmit and receive sidelink communications using peer-to-peer (P2P) communication protocols, device-to-device (D2D) communication protocols, vehicle-to-everything (V2X) communication protocols (which may include vehicle-to-vehicle (V2V) protocols, vehicle-to-infrastructure (V2I) protocols, and / or vehicle-to-pedestrian (V2P) protocols) , and / or mesh network communication protocols. In some deployments and configurations, a network node 110 may schedule and / or allocate resources for sidelink communications between UEs 120 in the wireless communication network 100. In some other deployments and configurations, a UE 120 (instead of a network node 110) may perform, or collaborate or negotiate with one or more other UEs to perform, scheduling operations, resource selection operations, and / or other operations for sidelink communications.
[0066] In various examples, some of the network nodes 110 and the UEs 120 of the wireless communication network 100 may be configured for full-duplex operation in addition to half-duplex operation. A network node 110 or a UE 120 operating in a half-duplex mode may perform only one of transmission or reception during particular time resources, such as during particular slots, symbols, or other time periods. Half-duplex operation may involve time-division duplexing (TDD) , in which DL transmissions of the network node 110 and UL transmissions of the UE 120 do not occur in the same time resources (that is, the transmissions do not overlap in time) . In contrast, a network node 110 or a UE 120 operating in a full-duplex mode can transmit and receive communications concurrently (for example, in the same time resources) . By operating in a full-duplex mode, network nodes 110 and / or UEs 120 may generally increase the capacity of the network and the radio access link. In some examples, full-duplex operation may involve frequency-division duplexing (FDD) , in which DL transmissions of the network node 110 are performed in a first frequency band or on a first component carrier and transmissions of the UE 120 are performed in a second frequency band or on a second component carrier different than the first frequency band or the first component carrier, respectively. In some examples, full-duplex operation may be enabled for a UE 120 but not for a network node 110. For example, a UE 120 may simultaneously transmit an UL transmission to a first network node 110 and receive a DL transmission from a second network node 110 in the same time resources. In some other examples, full-duplex operation may be enabled for a network node 110 but not for a UE 120. For example, a network node 110 may simultaneously transmit a DL transmission to a first UE 120 and receive an UL transmission from a second UE 120 in the same time resources. In some other examples, full-duplex operation may be enabled for both a network node 110 and a UE 120.
[0067] In some examples, the UEs 120 and the network nodes 110 may perform MIMO communication. “MIMO” generally refers to transmitting or receiving multiple signals (such as multiple layers or multiple data streams) simultaneously over the same time and frequency resources. MIMO techniques generally exploit multipath propagation. MIMO may be implemented using various spatial processing or spatial multiplexing operations. In some examples, MIMO may support simultaneous transmission to multiple receivers, referred to as multi-user MIMO (MU-MIMO) . Some RATs may employ advanced MIMO techniques, such as mTRP operation (including redundant transmission or reception on multiple TRPs) , reciprocity in the time domain or the frequency domain, single-frequency-network (SFN) transmission, or non-coherent joint transmission (NCJT) .
[0068] In some aspects, a network entity (e.g., a network node 110) may include a communication manager 150. As described in more detail elsewhere herein, the communication manager 150 may obtain dataset information from a UE; training ML models based at least in part on the dataset information; and transmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI. Additionally, or alternatively, the communication manager 150 may perform one or more other operations described herein.
[0069] In some aspects, a UE (e.g., a UE 120) may include a communication manager 140. As described in more detail elsewhere herein, the communication manager 140 may transmit dataset information; and receive model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information. Additionally, or alternatively, the communication manager 140 may perform one or more other operations described herein.
[0070] As indicated above, Fig. 1 is provided as an example. Other examples may differ from what is described with regard to Fig. 1.
[0071] Fig. 2 is a diagram illustrating an example network node 110 in communication with an example UE 120 in a wireless network, in accordance with the present disclosure.
[0072] As shown in Fig. 2, the network node 110 may include a data source 212, a transmit processor 214, a transmit (TX) MIMO processor 216, a set of modems 232 (shown as 232a through 232t, where t ≥ 1) , a set of antennas 234 (shown as 234a through 234v, where v ≥ 1) , a MIMO detector 236, a receive processor 238, a data sink 239, a controller / processor 240, a memory 242, a communication unit 244, a scheduler 246, and / or a communication manager 150, among other examples. In some configurations, one or a combination of the antenna (s) 234, the modem (s) 232, the MIMO detector 236, the receive processor 238, the transmit processor 214, and / or the TX MIMO processor 216 may be included in a transceiver of the network node 110. The transceiver may be under control of and used by one or more processors, such as the controller / processor 240, and in some aspects in conjunction with processor-readable code stored in the memory 242, to perform aspects of the methods, processes, and / or operations described herein. In some aspects, the network node 110 may include one or more interfaces, communication components, and / or other components that facilitate communication with the UE 120 or another network node.
[0073] The terms “processor, ” “controller, ” or “controller / processor” may refer to one or more controllers and / or one or more processors. For example, reference to “a / the processor, ” “a / the controller / processor, ” or the like (in the singular) should be understood to refer to any one or more of the processors described in connection with Fig. 2, such as a single processor or a combination of multiple different processors. Reference to “one or more processors” should be understood to refer to any one or more of the processors described in connection with Fig. 2. For example, one or more processors of the network node 110 may include transmit processor 214, TX MIMO processor 216, MIMO detector 236, receive processor 238, and / or controller / processor 240. Similarly, one or more processors of the UE 120 may include MIMO detector 256, receive processor 258, transmit processor 264, TX MIMO processor 266, and / or controller / processor 280.
[0074] In some aspects, a single processor may perform all of the operations described as being performed by the one or more processors. In some aspects, a first set of (one or more) processors of the one or more processors may perform a first operation described as being performed by the one or more processors, and a second set of (one or more) processors of the one or more processors may perform a second operation described as being performed by the one or more processors. The first set of processors and the second set of processors may be the same set of processors or may be different sets of processors. Reference to “one or more memories” should be understood to refer to any one or more memories of a corresponding device, such as the memory described in connection with Fig. 2. For example, operation described as being performed by one or more memories can be performed by the same subset of the one or more memories or different subsets of the one or more memories.
[0075] For downlink communication from the network node 110 to the UE 120, the transmit processor 214 may receive data ( “downlink data” ) intended for the UE 120 (or a set of UEs that includes the UE 120) from the data source 212 (such as a data pipeline or a data queue) . In some examples, the transmit processor 214 may select one or more modulation and coding schemes (MCSs) for the UE 120 in accordance with one or more channel quality indicators (CQIs) received from the UE 120. The network node 110 may process the data (for example, including encoding the data) for transmission to the UE 120 on a downlink in accordance with the MCS (s) selected for the UE 120 to generate data symbols. The transmit processor 214 may process system information (for example, semi-static resource partitioning information (SRPI) ) and / or control information (for example, CQI requests, grants, and / or upper layer signaling) and provide overhead symbols and / or control symbols. The transmit processor 214 may generate reference symbols for reference signals (for example, a cell-specific reference signal (CRS) , a demodulation reference signal (DMRS) , or a CSI-RS) and / or synchronization signals (for example, a primary synchronization signal (PSS) or a secondary synchronization signals (SSS) ) .
[0076] The TX MIMO processor 216 may perform spatial processing (for example, precoding) on the data symbols, the control symbols, the overhead symbols, and / or the reference symbols, if applicable, and may provide a set of output symbol streams (for example, T output symbol streams) to the set of modems 232. For example, each output symbol stream may be provided to a respective modulator component (shown as MOD) of a modem 232. Each modem 232 may use the respective modulator component to process (for example, to modulate) a respective output symbol stream (for example, for orthogonal frequency division multiplexing (OFDM) ) to obtain an output sample stream. Each modem 232 may further use the respective modulator component to process (for example, convert to analog, amplify, filter, and / or upconvert) the output sample stream to obtain a time domain downlink signal. The modems 232a through 232t may together transmit a set of downlink signals (for example, T downlink signals) via the corresponding set of antennas 234.
[0077] A downlink signal may include a DCI communication, a MAC control element (MAC-CE) communication, an RRC communication, a downlink reference signal, or another type of downlink communication. Downlink signals may be transmitted on a PDCCH, a PDSCH, and / or on another downlink channel. A downlink signal may carry one or more transport blocks (TBs) of data. A TB may be a unit of data that is transmitted over an air interface in the wireless communication network 100. A data stream (for example, from the data source 212) may be encoded into multiple TBs for transmission over the air interface. The quantity of TBs used to carry the data associated with a particular data stream may be associated with a TB size common to the multiple TBs. The TB size may be based on or otherwise associated with radio channel conditions of the air interface, the MCS used for encoding the data, the downlink resources allocated for transmitting the data, and / or another parameter. In general, the larger the TB size, the greater the amount of data that can be transmitted in a single transmission, which reduces signaling overhead. However, larger TB sizes may be more prone to transmission and / or reception errors than smaller TB sizes, but such errors may be mitigated by more robust error correction techniques.
[0078] For uplink communication from the UE 120 to the network node 110, uplink signals from the UE 120 may be received by an antenna 234, may be processed by a modem 232 (for example, a demodulator component, shown as DEMOD, of a modem 232) , may be detected by the MIMO detector 236 (for example, a receive (Rx) MIMO processor) if applicable, and / or may be further processed by the receive processor 238 to obtain decoded data and / or control information. The receive processor 238 may provide the decoded data to a data sink 239 (which may be a data pipeline, a data queue, and / or another type of data sink) and provide the decoded control information to a processor, such as the controller / processor 240.
[0079] The network node 110 may use the scheduler 246 to schedule one or more UEs 120 for downlink or uplink communications. In some aspects, the scheduler 246 may use DCI to dynamically schedule DL transmissions to the UE 120 and / or UL transmissions from the UE 120. In some examples, the scheduler 246 may allocate recurring time domain resources and / or frequency domain resources that the UE 120 may use to transmit and / or receive communications using an RRC configuration (for example, a semi-static configuration) , for example, to perform semi-persistent scheduling (SPS) or to configure a configured grant (CG) for the UE 120.
[0080] One or more of the transmit processor 214, the TX MIMO processor 216, the modem 232, the antenna 234, the MIMO detector 236, the receive processor 238, and / or the controller / processor 240 may be included in an RF chain of the network node 110. An RF chain may include one or more filters, mixers, oscillators, amplifiers, analog-to-digital converters (ADCs) , and / or other devices that convert between an analog signal (such as for transmission or reception via an air interface) and a digital signal (such as for processing by one or more processors of the network node 110) . In some aspects, the RF chain may be or may be included in a transceiver of the network node 110.
[0081] In some examples, the network node 110 may use the communication unit 244 to communicate with a core network and / or with other network nodes. The communication unit 244 may support wired and / or wireless communication protocols and / or connections, such as Ethernet, optical fiber, common public radio interface (CPRI) , and / or a wired or wireless backhaul, among other examples. The network node 110 may use the communication unit 244 to transmit and / or receive data associated with the UE 120 or to perform network control signaling, among other examples. The communication unit 244 may include a transceiver and / or an interface, such as a network interface.
[0082] The UE 120 may include a set of antennas 252 (shown as antennas 252a through 252r, where r ≥ 1) , a set of modems 254 (shown as modems 254a through 254u, where u ≥ 1) , a MIMO detector 256, a receive processor 258, a data sink 260, a data source 262, a transmit processor 264, a TX MIMO processor 266, a controller / processor 280, a memory 282, and / or a communication manager 140, among other examples. One or more of the components of the UE 120 may be included in a housing 284. In some aspects, one or a combination of the antenna (s) 252, the modem (s) 254, the MIMO detector 256, the receive processor 258, the transmit processor 264, or the TX MIMO processor 266 may be included in a transceiver that is included in the UE 120. The transceiver may be under control of and used by one or more processors, such as the controller / processor 280, and in some aspects in conjunction with processor-readable code stored in the memory 282, to perform aspects of the methods, processes, or operations described herein. In some aspects, the UE 120 may include another interface, another communication component, and / or another component that facilitates communication with the network node 110 and / or another UE 120.
[0083] For downlink communication from the network node 110 to the UE 120, the set of antennas 252 may receive the downlink communications or signals from the network node 110 and may provide a set of received downlink signals (for example, R received signals) to the set of modems 254. For example, each received signal may be provided to a respective demodulator component (shown as DEMOD) of a modem 254. Each modem 254 may use the respective demodulator component to condition (for example, filter, amplify, downconvert, and / or digitize) a received signal to obtain input samples. Each modem 254 may use the respective demodulator component to further demodulate or process the input samples (for example, for OFDM) to obtain received symbols. The MIMO detector 256 may obtain received symbols from the set of modems 254, may perform MIMO detection on the received symbols if applicable, and may provide detected symbols. The receive processor 258 may process (for example, decode) the detected symbols, may provide decoded data for the UE 120 to the data sink 260 (which may include a data pipeline, a data queue, and / or an application executed on the UE 120) , and may provide decoded control information and system information to the controller / processor 280.
[0084] For uplink communication from the UE 120 to the network node 110, the transmit processor 264 may receive and process data ( “uplink data” ) from a data source 262 (such as a data pipeline, a data queue, and / or an application executed on the UE 120) and control information from the controller / processor 280. The control information may include one or more parameters, feedback, one or more signal measurements, and / or other types of control information. In some aspects, the receive processor 258 and / or the controller / processor 280 may determine, for a received signal (such as received from the network node 110 or another UE) , one or more parameters relating to transmission of the uplink communication. The one or more parameters may include a reference signal received power (RSRP) parameter, a received signal strength indicator (RSSI) parameter, a reference signal received quality (RSRQ) parameter, a CQI parameter, or a transmit power control (TPC) parameter, among other examples. The control information may include an indication of the RSRP parameter, the RSSI parameter, the RSRQ parameter, the CQI parameter, the TPC parameter, and / or another parameter. The control information may facilitate parameter selection and / or scheduling for the UE 120 by the network node 110.
[0085] The transmit processor 264 may generate reference symbols for one or more reference signals, such as an uplink DMRS, an uplink sounding reference signal (SRS) , and / or another type of reference signal. The symbols from the transmit processor 264 may be precoded by the TX MIMO processor 266, if applicable, and further processed by the set of modems 254 (for example, for DFT-s-OFDM or CP-OFDM) . The TX MIMO processor 266 may perform spatial processing (for example, precoding) on the data symbols, the control symbols, the overhead symbols, and / or the reference symbols, if applicable, and may provide a set of output symbol streams (for example, U output symbol streams) to the set of modems 254. For example, each output symbol stream may be provided to a respective modulator component (shown as MOD) of a modem 254. Each modem 254 may use the respective modulator component to process (for example, to modulate) a respective output symbol stream (for example, for OFDM) to obtain an output sample stream. Each modem 254 may further use the respective modulator component to process (for example, convert to analog, amplify, filter, and / or upconvert) the output sample stream to obtain an uplink signal.
[0086] The modems 254a through 254u may transmit a set of uplink signals (for example, R uplink signals or U uplink symbols) via the corresponding set of antennas 252. An uplink signal may include a UCI communication, a MAC-CE communication, an RRC communication, or another type of uplink communication. Uplink signals may be transmitted on a PUSCH, a PUCCH, and / or another type of uplink channel. An uplink signal may carry one or more TBs of data. Sidelink data and control transmissions (that is, transmissions directly between two or more UEs 120) may generally use similar techniques as were described for uplink data and control transmission, and may use sidelink-specific channels such as a physical sidelink shared channel (PSSCH) , a physical sidelink control channel (PSCCH) , and / or a physical sidelink feedback channel (PSFCH) .
[0087] One or more antennas of the set of antennas 252 or the set of antennas 234 may include, or may be included within, one or more antenna panels, one or more antenna groups, one or more sets of antenna elements, or one or more antenna arrays, among other examples. An antenna panel, an antenna group, a set of antenna elements, or an antenna array may include one or more antenna elements (within a single housing or multiple housings) , a set of coplanar antenna elements, a set of non-coplanar antenna elements, or one or more antenna elements coupled with one or more transmission or reception components, such as one or more components of Fig. 2. As used herein, “antenna” can refer to one or more antennas, one or more antenna panels, one or more antenna groups, one or more sets of antenna elements, or one or more antenna arrays. “Antenna panel” can refer to a group of antennas (such as antenna elements) arranged in an array or panel, which may facilitate beamforming by manipulating parameters of the group of antennas. “Antenna module” may refer to circuitry including one or more antennas, which may also include one or more other components (such as filters, amplifiers, or processors) associated with integrating the antenna module into a wireless communication device.
[0088] In some examples, each of the antenna elements of an antenna 234 or an antenna 252 may include one or more sub-elements for radiating or receiving radio frequency signals. For example, a single antenna element may include a first sub-element cross-polarized with a second sub-element that can be used to independently transmit cross-polarized signals. The antenna elements may include patch antennas, dipole antennas, and / or other types of antennas arranged in a linear pattern, a two-dimensional pattern, or another pattern. A spacing between antenna elements may be such that signals with a desired wavelength transmitted separately by the antenna elements may interact or interfere constructively and destructively along various directions (such as to form a desired beam) . For example, given an expected range of wavelengths or frequencies, the spacing may provide a quarter wavelength, a half wavelength, or another fraction of a wavelength of spacing between neighboring antenna elements to allow for the desired constructive and destructive interference patterns of signals transmitted by the separate antenna elements within that expected range.
[0089] The amplitudes and / or phases of signals transmitted via antenna elements and / or sub-elements may be modulated and shifted relative to each other (such as by manipulating phase shift, phase offset, and / or amplitude) to generate one or more beams, which is referred to as beamforming. The term “beam” may refer to a directional transmission of a wireless signal toward a receiving device or otherwise in a desired direction. “Beam” may also generally refer to a direction associated with such a directional signal transmission, a set of directional resources associated with the signal transmission (for example, an angle of arrival, a horizontal direction, and / or a vertical direction) , and / or a set of parameters that indicate one or more aspects of a directional signal, a direction associated with the signal, and / or a set of directional resources associated with the signal. In some implementations, antenna elements may be individually selected or deselected for directional transmission of a signal (or signals) by controlling amplitudes of one or more corresponding amplifiers and / or phases of the signal (s) to form one or more beams. The shape of a beam (such as the amplitude, width, and / or presence of side lobes) and / or the direction of a beam (such as an angle of the beam relative to a surface of an antenna array) can be dynamically controlled by modifying the phase shifts, phase offsets, and / or amplitudes of the multiple signals relative to each other.
[0090] Different UEs 120 or network nodes 110 may include different numbers of antenna elements. For example, a UE 120 may include a single antenna element, two antenna elements, four antenna elements, eight antenna elements, or a different number of antenna elements. As another example, a network node 110 may include eight antenna elements, 24 antenna elements, 64 antenna elements, 128 antenna elements, or a different number of antenna elements. Generally, a larger number of antenna elements may provide increased control over parameters for beam generation relative to a smaller number of antenna elements, whereas a smaller number of antenna elements may be less complex to implement and may use less power than a larger number of antenna elements. Multiple antenna elements may support multiple-layer transmission, in which a first layer of a communication (which may include a first data stream) and a second layer of a communication (which may include a second data stream) are transmitted using the same time and frequency resources with spatial multiplexing.
[0091] While blocks in Fig. 2 are illustrated as distinct components, the functions described above with respect to the blocks may be implemented in a single hardware, software, or combination component or in various combinations of components. For example, the functions described with respect to the transmit processor 264, the receive processor 258, and / or the TX MIMO processor 266 may be performed by or under the control of the controller / processor 280.
[0092] Fig. 3 is a diagram illustrating an example disaggregated base station architecture 300, in accordance with the present disclosure. One or more components of the example disaggregated base station architecture 300 may be, may include, or may be included in one or more network nodes (such one or more network nodes 110) . The disaggregated base station architecture 300 may include a CU 310 that can communicate directly with a core network 320 via a backhaul link, or that can communicate indirectly with the core network 320 via one or more disaggregated control units, such as a Non-RT RIC 350 associated with a Service Management and Orchestration (SMO) Framework 360 and / or a Near-RT RIC 370 (for example, via an E2 link) . The CU 310 may communicate with one or more DUs 330 via respective midhaul links, such as via F1 interfaces. Each of the DUs 330 may communicate with one or more RUs 340 via respective fronthaul links. Each of the RUs 340 may communicate with one or more UEs 120 via respective RF access links. In some deployments, a UE 120 may be simultaneously served by multiple RUs 340.
[0093] Each of the components of the disaggregated base station architecture 300, including the CUs 310, the DUs 330, the RUs 340, the Near-RT RICs 370, the Non-RT RICs 350, and the SMO Framework 360, may include one or more interfaces or may be coupled with one or more interfaces for receiving or transmitting signals, such as data or information, via a wired or wireless transmission medium.
[0094] In some aspects, the CU 310 may be logically split into one or more CU user plane (CU-UP) units and one or more CU control plane (CU-CP) units. A CU-UP unit may communicate bidirectionally with a CU-CP unit via an interface, such as the E1 interface when implemented in an O-RAN configuration. The CU 310 may be deployed to communicate with one or more DUs 330, as necessary, for network control and signaling. Each DU 330 may correspond to a logical unit that includes one or more base station functions to control the operation of one or more RUs 340. For example, a DU 330 may host various layers, such as an RLC layer, a MAC layer, or one or more PHY layers, such as one or more high PHY layers or one or more low PHY layers. Each layer (which also may be referred to as a module) may be implemented with an interface for communicating signals with other layers (and modules) hosted by the DU 330, or for communicating signals with the control functions hosted by the CU 310. Each RU 340 may implement lower layer functionality. In some aspects, real-time and non-real-time aspects of control and user plane communication with the RU (s) 340 may be controlled by the corresponding DU 330.
[0095] The SMO Framework 360 may support RAN deployment and provisioning of non-virtualized and virtualized network elements. For non-virtualized network elements, the SMO Framework 360 may support the deployment of dedicated physical resources for RAN coverage requirements, which may be managed via an operations and maintenance interface, such as an O1 interface. For virtualized network elements, the SMO Framework 360 may interact with a cloud computing platform (such as an open cloud (O-Cloud) platform 390) to perform network element life cycle management (such as to instantiate virtualized network elements) via a cloud computing platform interface, such as an O2 interface. A virtualized network element may include, but is not limited to, a CU 310, a DU 330, an RU 340, a non-RT RIC 350, and / or a Near-RT RIC 370. In some aspects, the SMO Framework 360 may communicate with a hardware aspect of a 4G RAN, a 5G NR RAN, and / or a 6G RAN, such as an open eNB (O-eNB) 380, via an O1 interface. Additionally or alternatively, the SMO Framework 360 may communicate directly with each of one or more RUs 340 via a respective O1 interface. In some deployments, this configuration can enable each DU 330 and the CU 310 to be implemented in a cloud-based RAN architecture, such as a vRAN architecture.
[0096] The Non-RT RIC 350 may include or may implement a logical function that enables non-real-time control and optimization of RAN elements and resources, AI / ML workflows including model training and updates, and / or policy-based guidance of applications and / or features in the Near-RT RIC 370. The Non-RT RIC 350 may be coupled to or may communicate with (such as via an A1 interface) the Near-RT RIC 370. The Near-RT RIC 370 may include or may implement a logical function that enables near-real-time control and optimization of RAN elements and resources via data collection and actions via an interface (such as via an E2 interface) connecting one or more CUs 310, one or more DUs 330, and / or an O-eNB with the Near-RT RIC 370.
[0097] In some aspects, to generate AI / ML models to be deployed in the Near-RT RIC 370, the Non-RT RIC 350 may receive parameters or external enrichment information from external servers. Such information may be utilized by the Near-RT RIC 370 and may be received at the SMO Framework 360 or the Non-RT RIC 350 from non-network data sources or from network functions. In some examples, the Non-RT RIC 350 or the Near-RT RIC 370 may tune RAN behavior or performance. For example, the Non-RT RIC 350 may monitor long-term trends and patterns for performance and may employ AI / ML models to perform corrective actions via the SMO Framework 360 (such as reconfiguration via an O1 interface) or via creation of RAN management policies (such as A1 interface policies) .
[0098] As indicated above, Fig. 3 is provided as an example. Other examples may differ from what is described with regard to Fig. 3.
[0099] The network node 110, the controller / processor 240 of the network node 110, the UE 120, the controller / processor 280 of the UE 120, the CU 310, the DU 330, the RU 340, or any other component (s) of Figs. 1, 2, or 3 may implement one or more techniques or perform one or more operations associated with inter-vendor collaboration for encoding CSI feedback, as described in more detail elsewhere herein. For example, the controller / processor 240 of the network node 110, the controller / processor 280 of the UE 120, any other component (s) of Fig. 2, the CU 310, the DU 330, or the RU 340 may perform or direct operations of, for example, process 1400 of Fig. 14, process 1500 of Fig. 15, or other processes as described herein (alone or in conjunction with one or more other processors) . The memory 242 may store data and program codes for the network node 110, the network node 110, the CU 310, the DU 330, or the RU 340. The memory 282 may store data and program codes for the UE 120. In some examples, the memory 242 or the memory 282 may include a non-transitory computer-readable medium storing a set of instructions (for example, code or program code) for wireless communication. The memory 242 may include one or more memories, such as a single memory or multiple different memories (of the same type or of different types) . The memory 282 may include one or more memories, such as a single memory or multiple different memories (of the same type or of different types) . For example, the set of instructions, when executed (for example, directly, or after compiling, converting, or interpreting) by one or more processors of the network node 110, the UE 120, the CU 310, the DU 330, or the RU 340, may cause the one or more processors to perform process 1400 of Fig. 14, process 1500 of Fig. 15, , or other processes as described herein. In some examples, executing instructions may include running the instructions, converting the instructions, compiling the instructions, and / or interpreting the instructions, among other examples.
[0100] In some aspects, a network entity (e.g., a network node 110) includes means for obtaining dataset information from a UE; training ML models based at least in part on the dataset information; and / or means for transmitting model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI. In some aspects, the means for the network entity to perform operations described herein may include, for example, one or more of communication manager 150, transmit processor 214, TX MIMO processor 216, modem 232, antenna 234, MIMO detector 236, receive processor 238, controller / processor 240, memory 242, or scheduler 246.
[0101] In some aspects, a UE (e.g., a UE 120) includes means for transmitting dataset information; and / or means for receiving model information that is associated with using the information for ML training of spatial, temporal, and frequency compression of channel state information. The means for the UE to perform operations described herein may include, for example, one or more of communication manager 140, antenna 252, modem 254, MIMO detector 256, receive processor 258, transmit processor 264, TX MIMO processor 266, controller / processor 280, or memory 282.
[0102] Fig. 4 is a diagram illustrating an example architecture 400 of a functional framework for RAN intelligence enabled by data collection, in accordance with the present disclosure. In some scenarios, the functional framework for RAN intelligence may be enabled by further enhancement of data collection through use cases and / or examples. For example, principles or algorithms for RAN intelligence enabled by AI / ML and the associated functional framework (e.g., the AI functionality and / or the input / output of the component for AI enabled optimization) have been utilized or studied to identify the benefits of AI enabled RAN through possible use cases (e.g., beam management, energy saving, load balancing, mobility management, and / or coverage optimization, among other examples) . In one example, as shown by the architecture 400, a functional framework for RAN intelligence may include multiple logical entities, such as a model training host 402, a model inference host 404, data sources 406, and an actor 408.
[0103] The model inference host 404 may be configured to run an AI / ML model based on inference data provided by the data sources 406, and the model inference host 404 may produce an output (e.g., a prediction) with the inference data input to the actor 408. The actor 408 may be an element or an entity of a core network or a RAN. For example, the actor 408 may be a UE, a network node, base station (e.g., a gNB) , a CU, a DU, and / or an RU, among other examples. In addition, the actor 408 may also depend on the type of tasks performed by the model inference host 404, type of inference data provided to the model inference host 404, and / or type of output produced by the model inference host 404. For example, if the output from the model inference host 404 is associated with beam management, then the actor 408 may be a UE, a DU or an RU. In other examples, if the output from the model inference host 404 is associated with Tx / Rx scheduling, then the actor 408 may be a CU or a DU.
[0104] After the actor 408 receives an output from the model inference host 404, the actor 408 may determine whether to act based on the output. For example, if the actor 408 is a DU or an RU and the output from the model inference host 404 is associated with beam management, the actor 408 may determine whether to change / modify a Tx / Rx beam based on the output. If the actor 408 determines to act based on the output, the actor 408 may indicate the action to at least one subject of action 410. For example, if the actor 408 determines to change / modify a Tx / Rx beam for a communication between the actor 408 and the subject of action 410 (e.g., a UE 120) , then the actor 408 may transmit a beam (re-) configuration or a beam switching indication to the subject of action 410. The actor 408 may modify its Tx / Rx beam based on the beam (re-) configuration, such as switching to a new Tx / Rx beam or applying different parameters for a Tx / Rx beam, among other examples. As another example, the actor 408 may be a UE and the output from the model inference host 404 may be associated with beam management. For example, the output may be one or more predicted measurement values for one or more beams. The actor 408 (e.g., a UE) may determine that a measurement report (e.g., a Layer 1 (L1) RSRP report) is to be transmitted to a network node 110.
[0105] The data sources 406 may also be configured for collecting data that is used as training data for training an ML model or as inference data for feeding an ML model inference operation. For example, the data sources 406 may collect data from one or more core network and / or RAN entities, which may include the subject of action 410, and provide the collected data to the model training host 402 for ML model training. For example, after a subject of action 410 (e.g., a UE 120) receives a beam configuration from the actor 408, the subject of action 410 may provide performance feedback associated with the beam configuration to the data sources 406, where the performance feedback may be used by the model training host 402 for monitoring or evaluating the ML model performance, such as whether the output (e.g., prediction) provided to the actor 408 is accurate. In some examples, if the output provided by the actor 408 is inaccurate (or the accuracy is below an accuracy threshold) , then the model training host 402 may determine to modify or retrain the ML model used by the model inference host, such as via an ML model deployment / update.
[0106] As indicated above, Fig. 4 is provided as an example. Other examples may differ from what is described with regard to Fig. 4.
[0107] Fig. 5 is a diagram illustrating an example 500 of encoding and decoding CSI, in accordance with the present disclosure.
[0108] A network entity (e.g., a base station or gNB) may transmit CSI-RSs to a UE. The CSI-RSs may be configured to be periodic (e.g., using RRC signaling) , semi-persistent (e.g., using MAC-CE signaling) , and / or aperiodic (e.g., using DCI) . The UE may measure the CSI-RSs and report CSI to the network entity. In some examples, the network entity may transmit CSI-RSs on different transmit beams using different receive beams to support selection of transmit beam / receive beam pairs and to further refine beam selection.
[0109] CSI may be encoded at a UE and decoded at a network entity using ML. In such cross-node ML, a neural network may be split into two portions, where a first portion includes the encoder of the UE, and a second portion includes the decoder of the network entity. The encoder output of the UE is transmitted to the network entity as an input to the decoder. For example, as shown in example 400 of Fig. 5, the encoder 502 at the UE may input CSI (e.g., downlink channel estimates) and output compressed CSI or another data signal, which is received as input at the decoder 504 of the network entity. The decoder at the network entity may output a reconstructed CSI or another data signal, such as precoding vectors.
[0110] In multi-vendor training, each vendor (e.g., UE vendor, base station vendor) may be associated with a corresponding server that participates in offline training. The UE vendor server (s) communicate with the network entity vendor server (s) (e.g., base station (BS) servers) during the training using server-to-server connections. The UE vendor server may also be referred to as a UE-associated network entity or a UE server.
[0111] To evaluate the ML based CSI compression use cases, one or more different types of quantization or dequantization methods may be used, such as vector quantization or scalar quantization. In CSI compression using two-sided model use cases, multiple ML model trainings may be used. In some instances, a two-sided model may be jointly trained at a single side / entity (e.g., UE-sided or network-sided) . In some instances, the two-sided model may be jointly trained at a network side and a UE side, respectively. In some instances, models may be separately trained at a network side and a UE side, where the UE side CSI generation part and the network side CSI reconstruction part are trained by the UE side and the network side, respectively. “Joint training” may refer to the generation model and reconstruction model being trained in the same loop for forward propagation and backward propagation. Joint training may be done both at a single node or across multiple nodes (e.g., through gradient exchange between nodes) . Separate training may include sequential training starting with the UE side training, or sequential training starting with the network side training, or parallel training at the UE and the network.
[0112] Multi-vendor training may involve UE-network entity pairs (e.g., UE-BS pairs or UE-gNB pairs) . For example, a first BS (gNB 1) and a second BS (gNB 2) provide a respective cell, and multiple UEs (e.g., UE 1, UE 2, UE 3, UE 4) within the coverage region of gNB 1 or gNB 2. Vin is received by the encoder of UE 1 and is compressed. The output Z of the encoder is transmitted to the decoder of gNB 1, where gNB 1 decodes Z to reconstruct the Vin as Vout.
[0113] In instances without multi-vendor training, each UE-base station pair would be expected to utilize different encoder-decoder pairs. Multi-vendor training eliminates the need to utilize different encoder-decoder pairs for each UE-base station pairing. For example, in instances of multi-UE vendors with one base station vendor, a common base station decoder may be trained to work with multiple UE encoders. As such, the base station does not need to maintain a separate decoder model for each UE in its cell. In instances of a single-UE vendor with multi-base station vendors, a common UE encoder may be trained to work with multiple base station decoders. In such instances, the UE does not need to maintain a separate encoder model for each base station. In instances of multi-UE vendors with multi-base station vendors, the UE encoder may be trained to work with multiple base station decoders, while the base station decoder may be trained to work with multiple UE encoder. In an example, the respective encoders of UE 1 and UE 2 may be trained to work with the decoder of gNB 1, while the encoder of UE 4 may be trained to work with the decoder of gNB 2. However, UE 3 may be at a cell edge and between gNB 1 and gNB 2, such that the encoder of UE 3 may be trained to work with the decoder of either gNB 1 or gNB 2.
[0114] In one-sided concurrent training (e.g., offline) , both the encoder and the decoder may be trained jointly, such that the model weights of the encoder and decoder can be both optimized jointly. In offline concurrent training, models may be trained offline and may be provided to either the base station (e.g., gNB) or the UE. However, one-sided concurrent training may allow for the trained models to be exposed to the base station or the UE. Joint training may occur at the UE server or the base station (BS) server. For example, a UE vendor may train both the encoder and decoder models using its own dataset and may share the trained decoder model with the base station vendor, that is a different vendor than the UE vendor. The decoder shared with the other vendor may reveal or provide relevant information related to implementation details of the UE vendor’s modem. This information may be revealed due in part to symmetry that typically exists between the encoder and the decoder. As a result, the trained encoder and decoder may be a trade secret or include proprietary information that a vendor may not want to reveal to a competitor.
[0115] In sequential training, instead of revealing the neural network or model architecture, as in one-sided concurrent training, sequential training allows for the UE or base station to keep the trained models private. In base station driven sequential training, the base station decoder may be trained first at the BS server with an encoder selected by the base station. The UE encoder may be trained based on a dataset shared by the base station, for example. The dataset shared with the UE may include the original input Vin and the output Z of the encoder. In some aspects, this may include CSI and a representation of compressed CSI.
[0116] Multiple UE encoders may be trained based on the trained base station decoder. The base station decoder may be trained, and then share the dataset with each of the UEs (e.g., UE 1, UE 2) so that the respective UE encoders may be trained based on the dataset shared by the base station. In some instances, the original input Vin used as input at the base station encoder may comprise a precoder vector V. The BS server may train the base station decoder and generate a sequential training dataset (e.g., Z, Vin) which is shared with the UE server. Each UE server trains the respective UE encoder based on the sequential training dataset. In some instances, training the UE encoder may be achieved by minimizing a loss between Z (e.g., the output of the base station encoder) with Zue which is the output of the UE encoder.
[0117] In vector quantization, each input vector may be quantized and mapped to one of the vectors in a quantization codebook. In some instances, a quantization codebook may comprise vectors of size 2 or 4 where each entry may be represented by 2 bits. However, in other instances, the quantization codebook may comprise vectors of different sizes, or vector sizes other than 2 or 4. In addition, the entries may be represented by any size bits.
[0118] For example, an input Vin may be inputted into the encoder, which produces an encoder output Ze. The encoder output Ze may be quantized to produce a quantized output Zq. The quantized output Zq may be processed by the decoder in an effort to reconstruct the Vin, where the decoder output is Vout. To perform the quantization, a vector quantization (VQ) may receive the encoder output Ze and divide Ze into sub-vectors of size d-subset (e.g., 2 or 4) . A sub-vector (e.g., Ze0, Ze1) is quantized based on a quantization codebook to produce a quantized sub-vector (e.g., Zq0, Zq1) , where the quantized sub-vector is mapped to one of the vectors in the codebook. To perform the mapping based on the codebook, the quantizer maps the values of the quantized sub-vector to two values of the codebook (e.g., one of K values of the codebook) . The VQ may map inputs to the closest quantized value of the codebook. The quantized sub-vectors are then merged to form the quantized output Zq.
[0119] To alleviate or resolve issues related to inter-vendor training collaboration of AI / ML-based CSI compression using a two-sided model, different options may be used. Option 1 may involve a fully standardized reference model (structure and parameters) . Option 2 may involve a standardized dataset. Option 3 may involve a standardized reference model structure and parameter exchange between the network-side and UE-side. Option 4 may involve a standardized data / dataset format and a dataset exchange between the network-side and the UE-side. Option 5 may involve a standardized model format and reference model exchange between the network-side and the UE-side.
[0120] For Option 3, two sub-options may include Option 3a and Option 3b. For Option 3a, where parameters received at the UE or the UE-side proceeds through offline engineering at the UE-side (e.g., UE-side over-the-top (OTT) server) , including potential re-training, re-development of a different model, and / or offline testing. For Option 3b, parameters received at the UE are directly used for inference at the UE without offline engineering, potentially with on-device operations. For Option 5, two sub-options may include Option 5a and Option 5b. For Option 5a, a model received at the UE or the UE-side proceeds through offline engineering at the UE-side (e.g., UE-side OTT server) , including potential re-training, re-development of a different model, and / or offline testing. For Option 5b, a model received at the UE is directly used for inference at the UE without offline engineering, potentially with on-device operations. For Option 4, a dataset received at the UE or the UE-side may proceed through offline engineering at the UE-side (e.g., UE-side OTT server) , including model training or offline testing.
[0121] For Option 3a / 5a, the model (5a) / parameter (3a) exchange originates from the network-side and ends at the UE-side. Model (5a) / parameters (3a) exchanged from the network-side to UE-side is either CSI generation or reconstruction part or both. An encoder may be known as a “CSI generation model, ” while a decoder may be known as a “CSI reconstruction model. ”
[0122] For Option 3a-1 / 5a-1, the model / parameters exchanged from the network-side to UE-side is a CSI generation part (for encoder) . For Option 3a-2 / 5a-2, the model / parameters exchanged from the network-side to UE-side is a CSI reconstruction part (for decoder) . For Option 3a-3 / 5a-3, the model / parameters exchanged from the network-side to UE-side are both the CSI generation part and the CSI reconstruction part. Some additional information, if necessary, may be shared from the network-side to help the UE-side with offline engineering and to provide performance guidance. The additional information may include a performance target, or a dataset or information related to collecting the dataset.
[0123] For Option 3b, the method of exchanging is over the air-interface via model transfer / delivery case. The parameter exchange is from the network entity to the UE. Parameters exchanged from the network-side to the UE-side is a CSI generation part. For Option 5b, the method of exchanging is over the air-interface via model transfer / delivery Case z4, assuming that the model structure is aligned based on offline inter-vendor collaboration. The model exchange is from the network entity to the UE. The model is exchanged from the network-side to the UE-side is a CSI generation part.
[0124] For Option 4, the dataset exchange originates from the network-side and ends at the UE-side. For Option 4-1, the dataset is exchanged from the network-side to the UE-side and includes target CSI and CSI feedback. For Option 4-2, the dataset is exchanged from the network-side to the UE-side and includes CSI feedback and a reconstructed target CSI. For Option 4-3, the dataset is exchanged from the network-side to the UE-side and includes a target CSI, CSI feedback, and a reconstructed target CSI. Some additional information, if necessary, may be shared from the network-side to help UE-side offline engineering and to provide performance guidance (e.g., performance target) .
[0125] As indicated above, Fig. 5 is provided as an example. Other examples may differ from what is described with regard to Fig. 5.
[0126] Fig. 6 is a diagram illustrating an example 600 of CSI feedback, in accordance with the present disclosure.
[0127] Example 600 shows CSI feedback generated by an encoder from target CSI using a CSI-side CSI compression model. CSI feedback may include precoder feedback. CSI feedback may include a CSI report with a precoding matrix that the UE prefers that the network entity use, based on an observed channel. The network entity may use a network-side reconstruction model at a decoder to reconstruct the target CSI. For the reconstruction of the target CSI to be accurate, the UE-side and network-side ML models are expected to be trained in a collaborative manner so that the compressed representation created by the UE-side model is interpreted and decoded correctly by the network-side model. If this requirement is satisfied, then the pair of models is considered to be compatible to each other.
[0128] CSI compression may include spatial compression and frequency compression. In some aspects, CSI compression may further include temporal compression, which involves multiple CSI samples over a period of time (e.g., observation time window) . Temporal compression of CSI may involve different cases. In an initial case (case 0) , the target CSI is in a present slot and encoding uses no past CSI information. For the UE, the past CSI information may include past model inputs and / or any information derived from them. For the network, the past CSI information may include past CSI feedback instances and / or any information derived from them.
[0129] In a first case (case 1) , the target CSI is in a present slot, the UE uses past CSI information for encoding CSI (generating CSI) , and the network entity does not use past CSI information for decoding (reconstructing CSI) . “Target CSI slot (s) ” refers to the slot (s) to which the CSI feedback in the report corresponds. “Present slot” refers to the slot of the most recent CSI-RS measurement used to generate the CSI report. “Future slot (s) ” includes at least one slot after the present slot and may include the present slot as well.
[0130] In a second case (case 2) , the target CSI is in a present slot, the UE uses past CSI information for encoding, and the network entity uses past CSI information for decoding. In a third case (case 3) , the target CSI is in a future slot, the UE uses past CSI information, and the network entity does not use past CSI information. In a fourth case (case 4) , the target CSI is in a future slot, the UE uses past CSI information, and the network entity uses past CSI information. Case 4 may be a combination of case 2 and case 3. For cases 3 and case 4, the UE may perform CSI prediction as a separate step or jointly with CSI compression. Similarly, the network may perform CSI prediction as a separate step or jointly with CSI reconstruction. In a fifth case (case 5) , the target CSI is in a present slot, the UE does not use past CSI information, and the network entity uses past CSI information.
[0131] Past CSI information may be based on an observation window of size m. K, where K is the number of CSI-RS occasions in the window and m is the time between two adjacent CSI-RSs. This definition is applicable for both periodic and semi-periodic (SP) CSI-RS resources and for an aperiodic CSI-RS burst.
[0132] The past information outside the observation window should not impact the report. The past CSI impact is limited to the number of CSI reports based on cumulative historical CSI information with an unbounded duration. The historical CSI will reset based on an explicit command for the node (e.g., when an SP CSI report is being reactivated) .
[0133] In some aspects, temporal compression may be used for cases 1, 2, and 5 using historical information to help with the compression. Temporal compression may be used for case 3, which involves compression and prediction. Temporal compression may be used for case 4, which involves prediction and compression. Historical information may also be used for case 4. For cases 3, 4, and 5, UE vendors may share data with network vendors, which use the data to train CSI compression and decompression ML models. However, with Options 3, 4, and 5, data is shared between vendors (e.g., UE and network entity) . If the vendors do not have information for sharing data and training models and / or parameters in the context of a specific Option and one of the cases, the inter-vendor collaboration may perform suboptimal compression and reconstruction. Suboptimal compression and reconstruction may deliver inaccurate CSI or delay the delivery of CSI, which may lead to a suboptimal scheduling of communications. Suboptimal scheduling increases latency, wastes signaling resources, and decreases throughput.
[0134] As indicated above, Fig. 6 is provided as an example. Other examples may differ from what is described with regard to Fig. 6
[0135] Fig. 7 is a diagram illustrating an example 700 of inter-collaboration for ML CSI feedback with temporal compression, in accordance with the present disclosure. As shown in Fig. 7, a network entity 710 (e.g., a network node 110) and a UE 720 (e.g., UE 120) may communicate with one another via a wireless network (e.g., wireless communication network 100) .
[0136] According to various aspects described herein, a UE and a network entity may collaborate to share dataset information and model information to improve CSI compression and decompression. In some aspects, a network entity may obtain dataset information from a UE. As shown by reference number 725, the UE 720 may transmit dataset information to the network entity 710. The dataset information may include timing information, such as information about time-tagged samples. The timing information may indicate a time window of the time-tagged samples and / or a quantity of the time-tagged samples.
[0137] In some aspects, the dataset information may include an input CSI (input of an encoder) that matches (is the same as) a target CSI, which may occur with ideal prediction (e.g., no prediction errors) . In some aspects, the dataset information may include an input CSI, a target CSI, and a predicted CSI. With realistic prediction, there may be prediction errors between the predicted CSI and the actual observed target CSI.
[0138] In some aspects, the dataset information may include a dataset of only input CSI samples within an observation window. In some aspects, the dataset information may include a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window. In some aspects, the dataset information may include a dataset of only UE-predicted CSI samples. In some aspects, the dataset information may include a dataset of only target CSI samples and predicted CSI samples.
[0139] As shown by reference number 730, the network entity 710 may train ML models based at least in part on the dataset information. The ML models may be for an encoder, a reference encoder, and / or a decoder. The ML models may be for the prediction of a target CSI. Target CSI may be in a present window for case 2 and in a future window for cases 3 and 4.
[0140] The ML models may be trained using past CSI information. Past CSI feedback may be used for training for cases 3 and 4, as the observation window for cases 3 and 4 may include multiple time samples. For example, an observation window may be a size of 2, which includes one past CSI information and one current CSI. Past CSI information may represent time samples inside the observation window. This observation window is used to generate a CSI report (Report1) . For generating Report 2, there may be past CSI information in the observation window, in addition to the past CSI feedback (Report 1) .
[0141] As shown by reference number 735, the network entity may transmit ML model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI. The model information may include a model format of a model structure and / or at least one parameter for the model structure. By receiving dataset information from the UE and providing ML model information that is based on training with the data set information, the network entity may improve inter-vendor collaboration for various combinations of Option and case. Inter-vendor collaboration may be simplified for CSI compression. As a result, CSI compression and scheduled communications improve, which increases throughput.
[0142] In some aspects, the model information may include a model format for a CSI generation model that is associated with a reference encoder that is for compression only. Some or all of the cases may apply for the reference encoder. For example, for case 2, the compression may use target CSI in a present slot and past CSI feedback. Past CSI feedback may include CSI feedback of a past observation window. For case 3, the compression may use target CSI in a future slot and no past CSI feedback. For case 4, the compression may use target CSI in a future slot and past CSI feedback.
[0143] In some aspects, the model information may include a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only. The decompression may use target CSI in a present slot and past CSI feedback. The decompression may use target CSI in a future slot and no past CSI feedback. The decompression may use target CSI in a future slot and past CSI feedback.
[0144] In some aspects, the model information may include a model format for a CSI generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression. The training may assume realistic prediction associated with a target CSI and a prediction output. The training may include joint prediction and precoder compression using a ML model for both prediction and compression. The prediction and compression may use target CSI in a future slot and no past CSI feedback. The prediction and compression may use target CSI in a future slot and past CSI feedback.
[0145] In some aspects, the model information may include a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction. The model information may include a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction. The model information may include a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.
[0146] In some aspects, the model information may include a dataset or a dataset format associated with a model structure. The dataset may include target CSI in a present slot or a future slot and past CSI feedback. The dataset may include past CSI feedback. When ideal prediction is expected for the training, the dataset may further include a reconstructed target CSI that corresponds to an input CSI time window. Realistic prediction may be expected for the training, and the dataset may further include an input CSI that corresponds to an observation window and / or a reconstructed target CSI that corresponds to a prediction window. In some aspects, the model information may indicate a base quantization codebook and / or differential codebooks.
[0147] In some aspects, such as for Options 5a and 3a, the model information may include reference model information. For Options 5b and 3b, the model information may indicate or include an actual encoder.
[0148] As indicated above, Fig. 7 is provided as an example. Other examples may differ from what is described with regard to Fig. 7.
[0149] Fig. 8 is a diagram illustrating an example 800 of CSI compression and prediction, in accordance with the present disclosure.
[0150] In some aspects, collaborating vendors may use separate prediction and compression (SPC) . For channel prediction and precoder compression, prediction may be in a channel domain using ML or non-ML approaches (e.g., Auto-regressive (AR) models) , and compression in a precoder domain may use ML. For precoder prediction and precoder compression, prediction in the precoder domain may use ML or non-ML.
[0151] In some aspects, the vendors may use joint prediction and compression (JPC) . One ML module may be used for both prediction and compression in the precoder domain. Prediction may be based at least in part on a UE-side or joint prediction between UE and network models. That is, prediction may be performed by a UE encoder (UE-Enc) only or a network decoder (NW-Dec) also contributes to the prediction.
[0152] Cases 1, 2, and 5 involve temporal compression without prediction. The input CSI may be the same as the target CSI. That is, target CSI may be CSI without compression or prediction errors. The target CSI may be in a current window. The dataset (in dataset information) shared by the UE may include a time domain (TD) window with N samples, such as {Vt1, Vt2} for N = 2. The darker squares represent a slot with CSI-RS (CSI observation for a sample) .
[0153] Example 800 shows observation and prediction windows for realistic prediction and for ideal prediction. Input CSI 802 may be obtained from samples within an observation window. In some aspects, for cases 3 and 4 that involve temporal compression with prediction, the input CSI 802 is in the observation window 808 (V_in) , and the target CSI 804 is in the prediction window 810 (V_target) . The network entity may reconstruct the CSI (reconstructed target CSI 806) . Predicted CSI is an input to the compression module for SPC (V_pred) . In some aspects, the dataset shared by the UE may include only {V_in} samples, and the network-side training is focused on compression. The dataset be more applicable if the quantity of TD samples in the observation window 808 and the prediction window 810 is the same.
[0154] In some aspects, the dataset may include only {V_in, V_target} samples. In some aspects, the dataset may include only {V_pred} samples, where network-side training is focused on compression, and the network entity attempts to reconstruct the V_pred. In some aspects, the dataset may include only {V_target, V_pred} . The network-side training may be focused on compression.
[0155] As indicated above, Fig. 8 is provided as an example. Other examples may differ from what is described with regard to Fig. 8.
[0156] Fig. 9 is a diagram illustrating an example 900 of sharing modeling information for Option 4, in accordance with the present disclosure.
[0157] For a data sample, the TD window for a target CSI 902 or a reconstructed target CSI 904 may be 4 TD samples. The latent (compressed) representation of the TD window may be transmitted as a latent message from the UE to the network entity.
[0158] The network entity may share modeling information with the UE as part of Option 4. The modeling information may include a dataset 906 shared for SPC. In some aspects, network-side training may be performed assuming ideal prediction. That is, the target CSI 902 and the CSI input of the auto-encoder (AE) model corresponds to the same TD samples. The impact of the prediction error is not considered when training the network side AE. This simplifies model management for compression as it is decoupled from the prediction.
[0159] In some aspects, the dataset 906 includes target CSI 902 that is the same as input CSI (same TD window is used for input CSI and the target CSI 902) . The dataset 906 may include a reconstructed target CSI 904 that corresponds to the TD window of the input CSI. The dataset 906 may be in addition to CSI feedback. The UE may use the latent Z to be able to train its encoder. This is sequential network-first training, where UE is training the Enc based at least in part on the dataset shared by the network entity (Option 4) .
[0160] In some aspects, for dataset (model information) sharing for JPC, the network-side training may be performed assuming ideal prediction. The UE reference decoder (ref-Dec) may be trained for decompression only. The prediction operation may be carried out by the UE encoder (Enc) . This may be considered UE-side prediction, and the network entity does not contribute to the prediction operation. The modeling information (e.g., dataset) may include a target CSI that is the same as input CSI (i.e., same TD window is used for input and the target) and a reconstructed target CSI that corresponds to the input CSI TD window.
[0161] In some aspects, the network-side training may be performed assuming realistic prediction, That is, the CSI input to the AE model is the output of the prediction module and the target CSI. Both UE-Enc and NW-Dec may contribute to the prediction. The modeling information (e.g., dataset) may include both input CSI and target CSI. In some aspects, the modeling information may optionally include a reconstructed target CSI. The input CSI ma correspond to the observation window, and the reconstructed target CSI may correspond to the prediction window. UE ref-Dec may be trained based at least in part on the reconstructed target CSI. The UE Enc may be trained using input CSI and target CSI.
[0162] In example 900, each highlighted square may represent a CSI sample that has size nSB (number of CSI subbands in frequency) × nPort (number of ports in the precoding vector) . nTD may represent the quantity of time samples in the window (2 samples in example 900) .
[0163] As indicated above, Fig. 9 is provided as an example. Other examples may differ from what is described with regard to Fig. 9.
[0164] Fig. 10 is a diagram illustrating an example 1000 of sharing modeling information for Option 4, in accordance with the present disclosure.
[0165] For Option 4, the network-side training may be performed assuming realistic prediction. The input CSI 1002 to the AE model is the output of the prediction module and the target CSI 1004. The realistic prediction algorithm is up to network implementation. In some aspects, the modeling information (e.g., dataset) may include a target CSI and / or a reconstructed target CSI 1006 that can be used to train a reference decoder (ref-Dec) at the UE-side (doing decompression only) . Note that a reconstruction error in Vout may be impacted by both prediction and compression modules.
[0166] In some aspects, network-side training may be based at least in part on the UE prediction output. The UE may share the target CSI and the predicted CSI in dataset information. The predicted CSI may be used as an input to compression module, which is trained for reconstruction of the target CSI. The prediction error may be implicitly accounted for during network-side training. The modeling information (e.g., dataset) may include a reconstructed target CSI. In some aspects, the modeling information may include a UE-predicted CSI (input to a compression Enc module) .
[0167] As indicated above, Fig. 10 is provided as an example. Other examples may differ from what is described with regard to Fig. 10.
[0168] Fig. 11 is a diagram illustrating an example 1100 of modeling information, in accordance with the present disclosure.
[0169] In some aspects, a UE may perform spatial-frequency-temporal compression. For Options 3 and 5, some neural network structures (e.g., model structures, model parameters, or model formats) may be shared for spatial-frequency-temporal compression. Some neural network design flavors may be used for multi-vendor training and monitoring. An AE may be trained for a base time sample. Differential AE may be used for the remaining time samples with input augmented with temporal information.
[0170] In some aspects for modeling information associated with SPC, network-side training may be performed assuming ideal prediction. This naturally decouples the prediction and compression operations. Inter-vendor only may involve the exchange of compression models, and prediction may be up to UE implementation. In some aspects associated with Option 3a-1 and 5a-1, the model and / or parameters shared from the network-side to the UE-side may include a CSI generation part (e.g., CSI generation model and / or parameters for an encoder) . The Ref-Enc shared by the network may be for compression only (no prediction operation) . In some aspects associated with Option 3a-2 and 5a-2, the model and / or parameters exchanged from the network-side to the UE-side may be a CSI reconstruction part (e.g., CSI reconstruction model and / or parameters for a decoder) . A Ref-Dec shared by the network entity may include for decompression only (no prediction operation) . In some aspects associated with Option 3a-3 and 5a-3, the model and / or parameters exchanged from the network-side to the UE-side may include both a CSI generation part and a CSI reconstruction part. The Ref-Enc and Ref-Dec shared by the network may be for compression only or decompression only (no prediction operation) .
[0171] In some aspects for modeling information associated with SPC, network-side training may be performed assuming its own realistic prediction module. If feasible, additional information describing the Ref-Enc and / or Ref-Dec may include whether realistic prediction is used during training and / or a type (e.g., precoder or channel) and a general description of the prediction algorithm. Implementation details may be up to the UE.
[0172] In some aspects for modeling information associated with JPC, network-side training may be performed assuming ideal prediction. The ref-Enc and ref-Dec may be only for compression. Inter-vendor operation may be the same regardless of whether the UE implements SPC or JPC.
[0173] In some aspects for modeling information associated with JPC, network-side training may be performed assuming its own realistic prediction module. In some aspects associated with Option 3a-1 and 5a-1, the model and / or parameters shared from the network-side to the UE-side may include the CSI generation part. The general description of the prediction algorithm may make these options possible. The CSI generation part may include both prediction and compression modules. In some aspects associated with Option 3a-2 and 5a-2, the model and / or parameters shared from the network-side to the UE-side may include the CSI reconstruction part. The Ref-Dec shared by the network entity may be for decompression and may contribute to the prediction. In some aspects associated with Option 3a-3 and 5a-3, the model and / or parameters shared from the network-side to the UE-side may include both CSI generation part and CSI reconstruction part. The UE may determine whether or not to use ref-Enc (which includes a prediction component) .
[0174] In some aspects for modeling information (e.g., for the encoder) associated with JPC, network-side training may be performed assuming ideal prediction. For Option 3b and 5b, the model and / or parameters shared from the network-side to the UE-side may include the CSI generation part, including an actual encoder or an indication of an actual encoder. CSI generation may be trained based at least in part on realistic prediction.
[0175] In some aspects also associated with JPC, the network-side training may be performed assuming a realistic prediction module. This may involve the specification of details related to the prediction module inside JPC. The Enc model and / or parameters shared by the network may include both prediction and compression components. This may limit the prediction scope to precoder prediction.
[0176] As indicated above, Fig. 11 is provided as an example. Other examples may differ from what is described with regard to Fig. 11.
[0177] Fig. 12 is a diagram illustrating an example 1200 of compression and quantization, in accordance with the present disclosure.
[0178] Example 1200 shows an input (Vin) that is encoded and quantized. The resulting latent message is dequantized and decoded to produce decoder output (Vout) .
[0179] In some aspects, inter-vendor collaboration proposals for case 4 may follow case 2 and case 3. There may be two different architecture types for case 2. In some aspects, a refinement architecture may include an agnostic encoder or decoder with respect to case 0 or case 2. There may be two quantization modules: a base sample quantizer and a differential-sample quantizer.
[0180] Other architectures may include an encoder or decoder and possibly a quantizer that is adapted to case 2. Enc and Dec may be specific to case 2. Other architectures may include Long Short-Term Memory (LSTM) and convolutional LSTM (ConvLSTM) -based architectures.
[0181] In some aspects for Option 4, the modeling information from the network entity may include a dataset that is organized in a time window format, such as {Vt1,Vt2, Vt3, Vt4} for 4 TD samples for the target CSI, and the same for CSI feedback and reconstructed CSI. A quantization codebook may be designed and shared by the network entity. The UE-vendor may also design and share the quantization codebook (base / differential) as part of the dataset information. The network entity may have a separate quantization module for each UE-vendor or design a common quantization module based at least in part on the aggregated CBs. In some aspects, each UE-vendor may design a quantization codebook that is shared with the network entity. The network entity may maintain separate quantization codebooks for each UE-vendor.
[0182] In some aspects for Option 4, the interpretation of CSI feedback is based at least in part on the quantization codebook. The dataset information may include information describing the quantization method and whether it is TD invariant or not. The dataset information may include a quantization codebooks for base and differential samples. If the quantization is time dependent, then the quantization codebooks for different TD samples may be shared.
[0183] In some aspects for Options 3 and 5, the reference Enc / Dec model / parameters may be the same as for case 0. For Options 3a / 5a, the Ref-Enc and Ref-Dec for case 2 may be the same as for case 0. For Options 3b / 5b, the UE-Enc for case 2 may be the same as for case 0. In some aspects, only the quantization block may be time dependent. Such aspects may simply inter-vendor collaboration for case 2.
[0184] In some aspects for Options 3 and 5, the reference Enc / Dec model / parameters may be different than for case 0. For Options 3a / 5a, the Ref-Enc / Ref-Dec may be trained to take historical information into account. The Ref-Enc / Ref-Dec may include time dependent components (LSTM, ConvLSTM, ..., etc. ) . In some aspects for Options 3b and 5b, the UE-Enc may be trained to take historical information into account. The UE-Enc may include time dependent components (LSTM, ConvLSTM, ..., etc. )
[0185] In some aspects, inter-vendor collaboration may be facilitated for ML CSI feedback (CSF) with temporal domain compression. In some aspects, there may be a preference for SPC, inter-vendor collaboration may proceed as case 0 (ideal prediction) . For example, prediction and compression may be decoupled. The prediction module may be up to UE implementation, and inter-vendor collaboration may be focused on the compression module.
[0186] As indicated above, Fig. 12 is provided as an example. Other examples may differ from what is described with regard to Fig. 12.
[0187] Fig. 13 is a diagram illustrating an example 1300 of a refinement architecture, in accordance with the present disclosure.
[0188] In some aspects, a refined architecture may involve VQ. The same model structure may be used, but with different model parameters. The NN designed for spatial-frequency compression may be reused at each time of a temporal block. There may be differential time steps.
[0189] At the UE-side, VQ may be applied on an updated amount of the compressed CSI, where the updated amount is with respect to the previous CSI feedback. The network entity may recover the compressed CSI by accumulating the received updated amount of the compressed CSI into the previously accumulated CSI feedback. The same or different VQ codebooks may be used at different time steps.
[0190] As indicated above, Fig. 13 is provided as an example. Other examples may differ from what is described with regard to Fig. 13.
[0191] Fig. 14 is a diagram illustrating an example process 1400 performed, for example, at a network entity or an apparatus of a network entity, in accordance with the present disclosure. Example process 1400 is an example where the apparatus or the network entity (e.g., network entity 710) performs operations associated with inter-vendor collaboration with temporal compression.
[0192] As shown in Fig. 14, in some aspects, process 1400 may include obtaining dataset information from a UE (block 1410) . For example, the network entity (e.g., using reception component 1702 and / or communication manager 1706, depicted in Fig. 17) may obtain dataset information from a UE, as described above.
[0193] As further shown in Fig. 14, in some aspects, process 1400 may include training ML models based at least in part on the dataset information (block 1420) . For example, the network entity (e.g., using communication manager 1706, depicted in Fig. 17) may training ML models based at least in part on the dataset information, as described above.
[0194] As further shown in Fig. 14, in some aspects, process 1400 may include transmitting model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI (block 1430) . For example, the network entity (e.g., using transmission component 1704 and / or communication manager 1706, depicted in Fig. 17) may transmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI, as described above.
[0195] Process 1400 may include additional aspects, such as any single aspect or any combination of aspects described below and / or in connection with one or more other processes described elsewhere herein.
[0196] In a first aspect, the dataset information includes timing information of the CSI.
[0197] In a second aspect, alone or in combination with the first aspect, the timing information includes one or more of a time window and a quantity of time-tagged samples.
[0198] In a third aspect, alone or in combination with one or more of the first and second aspects, the dataset information includes an input CSI that matches a target CSI.
[0199] In a fourth aspect, alone or in combination with one or more of the first through third aspects, the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.
[0200] In a fifth aspect, alone or in combination with one or more of the first through fourth aspects, the dataset information includes a dataset of only input CSI samples within an observation window.
[0201] In a sixth aspect, alone or in combination with one or more of the first through fifth aspects, the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.
[0202] In a seventh aspect, alone or in combination with one or more of the first through sixth aspects, the dataset information includes a dataset of only UE-predicted CSI samples.
[0203] In an eighth aspect, alone or in combination with one or more of the first through seventh aspects, the dataset information includes a dataset of only target CSI samples and predicted CSI samples.
[0204] In a ninth aspect, alone or in combination with one or more of the first through eighth aspects, the model information includes a model format of a model structure and / or at least one parameter for the model structure.
[0205] In a tenth aspect, alone or in combination with one or more of the first through ninth aspects, the training includes separate channel prediction and precoder compression.
[0206] In an eleventh aspect, alone or in combination with one or more of the first through tenth aspects, the training assumes ideal prediction associated with a target CSI and time domain samples.
[0207] In a twelfth aspect, alone or in combination with one or more of the first through eleventh aspects, the training includes separate precoder prediction and precoder compression.
[0208] In a thirteenth aspect, alone or in combination with one or more of the first through twelfth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.
[0209] In a fourteenth aspect, alone or in combination with one or more of the first through thirteenth aspects, the compression uses target CSI in a present slot and past CSI feedback.
[0210] In a fifteenth aspect, alone or in combination with one or more of the first through fourteenth aspects, the compression uses target CSI in a future slot and no past CSI feedback.
[0211] In a sixteenth aspect, alone or in combination with one or more of the first through fifteenth aspects, the compression uses target CSI in a future slot and past CSI feedback.
[0212] In a seventeenth aspect, alone or in combination with one or more of the first through sixteenth aspects, the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.
[0213] In an eighteenth aspect, alone or in combination with one or more of the first through seventeenth aspects, the decompression uses target CSI in a present slot and past CSI feedback.
[0214] In a nineteenth aspect, alone or in combination with one or more of the first through eighteenth aspects, the decompression uses target CSI in a future slot and no past CSI feedback.
[0215] In a twentieth aspect, alone or in combination with one or more of the first through nineteenth aspects, the decompression uses target CSI in a future slot and past CSI feedback.
[0216] In a twenty-first aspect, alone or in combination with one or more of the first through twentieth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.
[0217] In a twenty-second aspect, alone or in combination with one or more of the first through twenty-first aspects, the training assumes realistic prediction associated with a target CSI and a prediction output.
[0218] In a twenty-third aspect, alone or in combination with one or more of the first through twenty-second aspects, the training includes joint prediction and precoder compression using a ML model for both prediction and compression.
[0219] In a twenty-fourth aspect, alone or in combination with one or more of the first through twenty-third aspects, the prediction and compression use target CSI in a present slot and past CSI feedback.
[0220] In a twenty-fifth aspect, alone or in combination with one or more of the first through twenty-fourth aspects, the prediction and compression use target CSI in a future slot and no past CSI feedback.
[0221] In a twenty-sixth aspect, alone or in combination with one or more of the first through twenty-fifth aspects, the prediction and compression use target CSI in a future slot and past CSI feedback.
[0222] In a twenty-seventh aspect, alone or in combination with one or more of the first through twenty-sixth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.
[0223] In a twenty-eighth aspect, alone or in combination with one or more of the first through twenty-seventh aspects, the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.
[0224] In a twenty-ninth aspect, alone or in combination with one or more of the first through twenty-eighth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.
[0225] In a thirtieth aspect, alone or in combination with one or more of the first through twenty-ninth aspects, the model information includes a dataset or a dataset format associated with a model structure.
[0226] In a thirty-first aspect, alone or in combination with one or more of the first through thirtieth aspects, the dataset includes target CSI in a present slot or a future slot and past CSI feedback.
[0227] In a thirty-second aspect, alone or in combination with one or more of the first through thirty-first aspects, the dataset includes past CSI feedback.
[0228] In a thirty-third aspect, alone or in combination with one or more of the first through thirty-second aspects, ideal prediction is expected for the training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.
[0229] In a thirty-fourth aspect, alone or in combination with one or more of the first through thirty-third aspects, realistic prediction is expected for the training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.
[0230] In a thirty-fifth aspect, alone or in combination with one or more of the first through thirty-fourth aspects, the model information indicates base and differential quantization codebooks.
[0231] Although Fig. 14 shows example blocks of process 1400, in some aspects, process 1400 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in Fig. 14. Additionally, or alternatively, two or more of the blocks of process 1400 may be performed in parallel.
[0232] Fig. 15 is a diagram illustrating an example process 1500 performed, for example, at a UE or an apparatus of a UE, in accordance with the present disclosure. Example process 1500 is an example where the apparatus or the UE (e.g., UE 720) performs operations associated with inter-vendor collaboration with temporal compression.
[0233] As shown in Fig. 15, in some aspects, process 1500 may include transmitting dataset information (block 1510) . For example, the UE (e.g., using transmission component 1604 and / or communication manager 1606, depicted in Fig. 16) may transmit dataset information, as described above.
[0234] As further shown in Fig. 15, in some aspects, process 1500 may include receiving model information that is associated with using the dataset information for ML training of spatial, temporal, and frequency compression of channel state information (block 1520) . For example, the UE (e.g., using reception component 1602 and / or communication manager 1606, depicted in Fig. 16) may receive model information that is associated with using the dataset information for ML training of spatial, temporal, and frequency compression of channel state information, as described above.
[0235] Process 1500 may include additional aspects, such as any single aspect or any combination of aspects described below and / or in connection with one or more other processes described elsewhere herein.
[0236] In a first aspect, the model information includes reference model information.
[0237] In a second aspect, alone or in combination with the first aspect, the model information indicates an encoder.
[0238] In a third aspect, alone or in combination with one or more of the first and second aspects, the dataset information includes timing information of the CSI.
[0239] In a fourth aspect, alone or in combination with one or more of the first through third aspects, the timing information includes one or more of a time window and a quantity of time-tagged samples.
[0240] In a fifth aspect, alone or in combination with one or more of the first through fourth aspects, the dataset information includes an input CSI that matches a target CSI.
[0241] In a sixth aspect, alone or in combination with one or more of the first through fifth aspects, the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.
[0242] In a seventh aspect, alone or in combination with one or more of the first through sixth aspects, the dataset information includes a dataset of only input CSI samples within an observation window.
[0243] In an eighth aspect, alone or in combination with one or more of the first through seventh aspects, the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.
[0244] In a ninth aspect, alone or in combination with one or more of the first through eighth aspects, the dataset information includes a dataset of only UE-predicted CSI samples.
[0245] In a tenth aspect, alone or in combination with one or more of the first through ninth aspects, the dataset information includes a dataset of only target CSI samples and predicted CSI samples.
[0246] In an eleventh aspect, alone or in combination with one or more of the first through tenth aspects, the model information includes a model format of a model structure and / or at least one parameter for the model structure.
[0247] In a twelfth aspect, alone or in combination with one or more of the first through eleventh aspects, the model information is associated with training for separate channel prediction and precoder compression.
[0248] In a thirteenth aspect, alone or in combination with one or more of the first through twelfth aspects, the model information is associated with training for a target CSI and time domain samples.
[0249] In a fourteenth aspect, alone or in combination with one or more of the first through thirteenth aspects, the training includes separate precoder prediction and precoder compression.
[0250] In a fifteenth aspect, alone or in combination with one or more of the first through fourteenth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.
[0251] In a sixteenth aspect, alone or in combination with one or more of the first through fifteenth aspects, the compression uses target CSI in a present slot and past CSI feedback.
[0252] In a seventeenth aspect, alone or in combination with one or more of the first through sixteenth aspects, the compression uses target CSI in a future slot and no past CSI feedback.
[0253] In an eighteenth aspect, alone or in combination with one or more of the first through seventeenth aspects, the compression uses target CSI in a future slot and past CSI feedback.
[0254] In a nineteenth aspect, alone or in combination with one or more of the first through eighteenth aspects, the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.
[0255] In a twentieth aspect, alone or in combination with one or more of the first through nineteenth aspects, the decompression uses target CSI in a present slot and past CSI feedback.
[0256] In a twenty-first aspect, alone or in combination with one or more of the first through twentieth aspects, the decompression uses target CSI in a future slot and no past CSI feedback.
[0257] In a twenty-second aspect, alone or in combination with one or more of the first through twenty-first aspects, the decompression uses target CSI in a future slot and past CSI feedback.
[0258] In a twenty-third aspect, alone or in combination with one or more of the first through twenty-second aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.
[0259] In a twenty-fourth aspect, alone or in combination with one or more of the first through twenty-third aspects, the model information is associated with training that assumes realistic prediction associated with a target CSI and a prediction output.
[0260] In a twenty-fifth aspect, alone or in combination with one or more of the first through twenty-fourth aspects, the model information is associated with training that includes joint prediction and precoder compression using a ML model for both prediction and compression.
[0261] In a twenty-sixth aspect, alone or in combination with one or more of the first through twenty-fifth aspects, the prediction and compression use target CSI in a present slot and past CSI feedback.
[0262] In a twenty-seventh aspect, alone or in combination with one or more of the first through twenty-sixth aspects, the prediction and compression use target CSI in a future slot and no past CSI feedback.
[0263] In a twenty-eighth aspect, alone or in combination with one or more of the first through twenty-seventh aspects, the prediction and compression use target CSI in a future slot and past CSI feedback.
[0264] In a twenty-ninth aspect, alone or in combination with one or more of the first through twenty-eighth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.
[0265] In a thirtieth aspect, alone or in combination with one or more of the first through twenty-ninth aspects, the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.
[0266] In a thirty-first aspect, alone or in combination with one or more of the first through thirtieth aspects, the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.
[0267] In a thirty-second aspect, alone or in combination with one or more of the first through thirty-first aspects, the model information includes a dataset or a dataset format associated with a model structure.
[0268] In a thirty-third aspect, alone or in combination with one or more of the first through thirty-second aspects, the dataset includes target CSI in a present slot or a future slot and past CSI feedback.
[0269] In a thirty-fourth aspect, alone or in combination with one or more of the first through thirty-third aspects, the dataset includes past CSI feedback.
[0270] In a thirty-fifth aspect, alone or in combination with one or more of the first through thirty-fourth aspects, ideal prediction is expected for the training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.
[0271] In a thirty-sixth aspect, alone or in combination with one or more of the first through thirty-fifth aspects, realistic prediction is expected for the training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.
[0272] In a thirty-seventh aspect, alone or in combination with one or more of the first through thirty-sixth aspects, the model information indicates base quantization codebooks and differential quantization codebooks.
[0273] Although Fig. 15 shows example blocks of process 1500, in some aspects, process 1500 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in Fig. 15. Additionally, or alternatively, two or more of the blocks of process 1500 may be performed in parallel.
[0274] Fig. 16 is a diagram of an example apparatus 1600 for wireless communication, in accordance with the present disclosure. The apparatus 1600 may be a UE, or a UE may include the apparatus 1600. In some aspects, the apparatus 1600 includes a reception component 1602, a transmission component 1604, and / or a communication manager 1606, which may be in communication with one another (for example, via one or more buses and / or one or more other components) . In some aspects, the communication manager 1606 is the communication manager 140 described in connection with Fig. 1. As shown, the apparatus 1600 may communicate with another apparatus 1608, such as a UE or a network node (such as a CU, a DU, an RU, or a base station) , using the reception component 1602 and the transmission component 1604.
[0275] In some aspects, the apparatus 1600 may be configured to perform one or more operations described herein in connection with Figs. 1-13. Additionally, or alternatively, the apparatus 1600 may be configured to perform one or more processes described herein, such as process 1500 of Fig. 15. In some aspects, the apparatus 1600 and / or one or more components shown in Fig. 16 may include one or more components of the UE described in connection with Fig. 1 and Fig. 2. Additionally, or alternatively, one or more components shown in Fig. 16 may be implemented within one or more components described in connection with Fig. 1 and Fig. 2. Additionally, or alternatively, one or more components of the set of components may be implemented at least in part as software stored in one or more memories. For example, a component (or a portion of a component) may be implemented as instructions or code stored in a non-transitory computer-readable medium and executable by one or more controllers or one or more processors to perform the functions or operations of the component.
[0276] The reception component 1602 may receive communications, such as reference signals, control information, data communications, or a combination thereof, from the apparatus 1608. The reception component 1602 may provide received communications to one or more other components of the apparatus 1600. In some aspects, the reception component 1602 may perform signal processing on the received communications (such as filtering, amplification, demodulation, analog-to-digital conversion, demultiplexing, deinterleaving, de-mapping, equalization, interference cancellation, or decoding, among other examples) , and may provide the processed signals to the one or more other components of the apparatus 1600. In some aspects, the reception component 1602 may include one or more antennas, one or more modems, one or more demodulators, one or more MIMO detectors, one or more receive processors, one or more controllers / processors, one or more memories, or a combination thereof, of the UE described in connection with Fig. 1 and Fig. 2.
[0277] The transmission component 1604 may transmit communications, such as reference signals, control information, data communications, or a combination thereof, to the apparatus 1608. In some aspects, one or more other components of the apparatus 1600 may generate communications and may provide the generated communications to the transmission component 1604 for transmission to the apparatus 1608. In some aspects, the transmission component 1604 may perform signal processing on the generated communications (such as filtering, amplification, modulation, digital-to-analog conversion, multiplexing, interleaving, mapping, or encoding, among other examples) , and may transmit the processed signals to the apparatus 1608. In some aspects, the transmission component 1604 may include one or more antennas, one or more modems, one or more modulators, one or more transmit MIMO processors, one or more transmit processors, one or more controllers / processors, one or more memories, or a combination thereof, of the UE described in connection with Fig. 1 and Fig. 2. In some aspects, the transmission component 1604 may be co-located with the reception component 1602 in one or more transceivers.
[0278] The communication manager 1606 may support operations of the reception component 1602 and / or the transmission component 1604. For example, the communication manager 1606 may receive information associated with configuring reception of communications by the reception component 1602 and / or transmission of communications by the transmission component 1604. Additionally, or alternatively, the communication manager 1606 may generate and / or provide control information to the reception component 1602 and / or the transmission component 1604 to control reception and / or transmission of communications.
[0279] The transmission component 1604 may transmit dataset information. The reception component 1602 may receive model information that is associated with using the dataset information for ML training of spatial, temporal, and frequency compression of channel state information.
[0280] The number and arrangement of components shown in Fig. 16 are provided as an example. In practice, there may be additional components, fewer components, different components, or differently arranged components than those shown in Fig. 16. Furthermore, two or more components shown in Fig. 16 may be implemented within a single component, or a single component shown in Fig. 16 may be implemented as multiple, distributed components. Additionally, or alternatively, a set of (one or more) components shown in Fig. 16 may perform one or more functions described as being performed by another set of components shown in Fig. 16.
[0281] Fig. 17 is a diagram of an example apparatus 1700 for wireless communication, in accordance with the present disclosure. The apparatus 1700 may be a network entity, or a network entity may include the apparatus 1700. In some aspects, the apparatus 1700 includes a reception component 1702, a transmission component 1704, and / or a communication manager 1706, which may be in communication with one another (for example, via one or more buses and / or one or more other components) . In some aspects, the communication manager 1706 is the communication manager 150 described in connection with Fig. 1. As shown, the apparatus 1700 may communicate with another apparatus 1708, such as a UE or a network node (such as a CU, a DU, an RU, or a base station) , using the reception component 1702 and the transmission component 1704.
[0282] In some aspects, the apparatus 1700 may be configured to perform one or more operations described herein in connection with Figs. 1-13. Additionally, or alternatively, the apparatus 1700 may be configured to perform one or more processes described herein, such as process 1400 of Fig. 14. In some aspects, the apparatus 1700 and / or one or more components shown in Fig. 17 may include one or more components of the network entity described in connection with Fig. 1 and Fig. 2. Additionally, or alternatively, one or more components shown in Fig. 17 may be implemented within one or more components described in connection with Fig. 1 and Fig. 2. Additionally, or alternatively, one or more components of the set of components may be implemented at least in part as software stored in one or more memories. For example, a component (or a portion of a component) may be implemented as instructions or code stored in a non-transitory computer-readable medium and executable by one or more controllers or one or more processors to perform the functions or operations of the component.
[0283] The reception component 1702 may receive communications, such as reference signals, control information, data communications, or a combination thereof, from the apparatus 1708. The reception component 1702 may provide received communications to one or more other components of the apparatus 1700. In some aspects, the reception component 1702 may perform signal processing on the received communications (such as filtering, amplification, demodulation, analog-to-digital conversion, demultiplexing, deinterleaving, de-mapping, equalization, interference cancellation, or decoding, among other examples) , and may provide the processed signals to the one or more other components of the apparatus 1700. In some aspects, the reception component 1702 may include one or more antennas, one or more modems, one or more demodulators, one or more MIMO detectors, one or more receive processors, one or more controllers / processors, one or more memories, or a combination thereof, of the network entity described in connection with Fig. 1 and Fig. 2.
[0284] The transmission component 1704 may transmit communications, such as reference signals, control information, data communications, or a combination thereof, to the apparatus 1708. In some aspects, one or more other components of the apparatus 1700 may generate communications and may provide the generated communications to the transmission component 1704 for transmission to the apparatus 1708. In some aspects, the transmission component 1704 may perform signal processing on the generated communications (such as filtering, amplification, modulation, digital-to-analog conversion, multiplexing, interleaving, mapping, or encoding, among other examples) , and may transmit the processed signals to the apparatus 1708. In some aspects, the transmission component 1704 may include one or more antennas, one or more modems, one or more modulators, one or more transmit MIMO processors, one or more transmit processors, one or more controllers / processors, one or more memories, or a combination thereof, of the network entity described in connection with Fig. 1 and Fig. 2. In some aspects, the transmission component 1704 may be co-located with the reception component 1702 in one or more transceivers.
[0285] The communication manager 1706 may support operations of the reception component 1702 and / or the transmission component 1704. For example, the communication manager 1706 may receive information associated with configuring reception of communications by the reception component 1702 and / or transmission of communications by the transmission component 1704. Additionally, or alternatively, the communication manager 1706 may generate and / or provide control information to the reception component 1702 and / or the transmission component 1704 to control reception and / or transmission of communications.
[0286] The reception component 1702 may obtain dataset information from a UE. The communication manager 1706 may training ML models based at least in part on the dataset information. The transmission component 1704 may transmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of CSI.
[0287] The number and arrangement of components shown in Fig. 17 are provided as an example. In practice, there may be additional components, fewer components, different components, or differently arranged components than those shown in Fig. 17. Furthermore, two or more components shown in Fig. 17 may be implemented within a single component, or a single component shown in Fig. 17 may be implemented as multiple, distributed components. Additionally, or alternatively, a set of (one or more) components shown in Fig. 17 may perform one or more functions described as being performed by another set of components shown in Fig. 17.
[0288] The following provides an overview of some Aspects of the present disclosure:
[0289] Aspect 1: A method of wireless communication performed by a network entity, comprising: obtaining dataset information from a user equipment (UE) ; training machine learning models based at least in part on the dataset information; and transmitting model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of channel state information (CSI) .
[0290] Aspect 2: The method of Aspect 1, wherein the dataset information includes timing information of the CSI.
[0291] Aspect 3: The method of Aspect 2, wherein the timing information includes one or more of a time window and a quantity of time-tagged samples.
[0292] Aspect 4: The method of any of Aspects 1-3, wherein the dataset information includes an input CSI that matches a target CSI.
[0293] Aspect 5: The method of any of Aspects 1-4, wherein the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.
[0294] Aspect 6: The method of Aspect 5, wherein the dataset information includes a dataset of only input CSI samples within an observation window.
[0295] Aspect 7: The method of Aspect 5, wherein the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.
[0296] Aspect 8: The method of Aspect 5, wherein the dataset information includes a dataset of only UE-predicted CSI samples.
[0297] Aspect 9: The method of Aspect 5, wherein the dataset information includes a dataset of only target CSI samples and predicted CSI samples.
[0298] Aspect 10: The method of any of Aspects 1-9, wherein the model information includes a model format of a model structure and / or at least one parameter for the model structure.
[0299] Aspect 11: The method of any of Aspects 1-10, wherein the training includes separate channel prediction and precoder compression.
[0300] Aspect 12: The method of any of Aspects 1-11, wherein the training assumes ideal prediction associated with a target CSI and time domain samples.
[0301] Aspect 13: The method of any of Aspects 1-12, wherein the training includes separate precoder prediction and precoder compression.
[0302] Aspect 14: The method of Aspect 13, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.
[0303] Aspect 15: The method of Aspect 14, wherein the compression uses target CSI in a present slot and past CSI feedback.
[0304] Aspect 16: The method of Aspect 14, wherein the compression uses target CSI in a future slot and no past CSI feedback.
[0305] Aspect 17: The method of Aspect 14, wherein the compression uses target CSI in a future slot and past CSI feedback.
[0306] Aspect 18: The method of Aspect 13, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.
[0307] Aspect 19: The method of Aspect 18, wherein the decompression uses target CSI in a present slot and past CSI feedback.
[0308] Aspect 20: The method of Aspect 18, wherein the decompression uses target CSI in a future slot and no past CSI feedback.
[0309] Aspect 21: The method of Aspect 18, wherein the decompression uses target CSI in a future slot and past CSI feedback.
[0310] Aspect 22: The method of Aspect 13, wherein the model information includes a model format for a channel state information (CSI) generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.
[0311] Aspect 23: The method of any of Aspects 1-22, wherein the training assumes realistic prediction associated with a target CSI and a prediction output.
[0312] Aspect 24: The method of any of Aspects 1-23, wherein the training includes joint prediction and precoder compression using a machine learning model for both prediction and compression.
[0313] Aspect 25: The method of Aspect 24, wherein the prediction and compression use target CSI in a present slot and past CSI feedback.
[0314] Aspect 26: The method of Aspect 24, wherein the prediction and compression use target CSI in a future slot and no past CSI feedback.
[0315] Aspect 27: The method of Aspect 24, wherein the prediction and compression use target CSI in a future slot and past CSI feedback.
[0316] Aspect 28: The method of Aspect 24, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.
[0317] Aspect 29: The method of Aspect 24, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.
[0318] Aspect 30: The method of Aspect 24, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.
[0319] Aspect 31: The method of any of Aspects 1-30, wherein the model information includes a dataset or a dataset format associated with a model structure.
[0320] Aspect 32: The method of Aspect 31, wherein the dataset includes target CSI in a present slot or a future slot and past CSI feedback.
[0321] Aspect 33: The method of Aspect 32, wherein the dataset includes past CSI feedback.
[0322] Aspect 34: The method of Aspect 32, wherein ideal prediction is expected for the training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.
[0323] Aspect 35: The method of Aspect 32, wherein realistic prediction is expected for the training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.
[0324] Aspect 36: The method of Aspect 32, wherein the model information indicates base and differential quantization codebooks.
[0325] Aspect 37: A method of wireless communication performed by a user equipment (UE) , comprising: transmitting dataset information; and receiving model information that is associated with using the dataset information for machine learning training of spatial, temporal, and frequency compression of channel state information.
[0326] Aspect 38: The method of Aspect 37, wherein the model information includes reference model information.
[0327] Aspect 39: The method of any of Aspects 37-38, wherein the model information indicates an encoder.
[0328] Aspect 40: The method of any of Aspects 37-39, wherein the dataset information includes timing information of the CSI.
[0329] Aspect 41: The method of Aspect 40, wherein the timing information includes one or more of a time window and a quantity of time-tagged samples.
[0330] Aspect 42: The method of any of Aspects 37-41, wherein the dataset information includes an input CSI that matches a target CSI.
[0331] Aspect 43: The method of any of Aspects 37-42, wherein the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.
[0332] Aspect 44: The method of Aspect 43, wherein the dataset information includes a dataset of only input CSI samples within an observation window.
[0333] Aspect 45: The method of Aspect 43, wherein the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.
[0334] Aspect 46: The method of Aspect 43, wherein the dataset information includes a dataset of only UE-predicted CSI samples.
[0335] Aspect 47: The method of Aspect 43, wherein the dataset information includes a dataset of only target CSI samples and predicted CSI samples.
[0336] Aspect 48: The method of any of Aspects 37-47, wherein the model information includes a model format of a model structure and / or at least one parameter for the model structure.
[0337] Aspect 49: The method of any of Aspects 37-48, wherein the model information is associated with training for separate channel prediction and precoder compression.
[0338] Aspect 50: The method of any of Aspects 37-49, wherein the model information is associated with training for a target CSI and time domain samples.
[0339] Aspect 51: The method of any of Aspects 37-50, wherein the training includes separate precoder prediction and precoder compression.
[0340] Aspect 52: The method of Aspect 51, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.
[0341] Aspect 53: The method of Aspect 52, wherein the compression uses target CSI in a present slot and past CSI feedback.
[0342] Aspect 54: The method of Aspect 52, wherein the compression uses target CSI in a future slot and no past CSI feedback.
[0343] Aspect 55: The method of Aspect 52, wherein the compression uses target CSI in a future slot and past CSI feedback.
[0344] Aspect 56: The method of Aspect 52, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.
[0345] Aspect 57: The method of Aspect 56, wherein the decompression uses target CSI in a present slot and past CSI feedback.
[0346] Aspect 58: The method of Aspect 56, wherein the decompression uses target CSI in a future slot and no past CSI feedback.
[0347] Aspect 59: The method of Aspect 56, wherein the decompression uses target CSI in a future slot and past CSI feedback.
[0348] Aspect 60: The method of Aspect 47, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.
[0349] Aspect 61: The method of any of Aspects 37-60, wherein the model information is associated with training that assumes realistic prediction associated with a target CSI and a prediction output.
[0350] Aspect 62: The method of any of Aspects 37-61, wherein the model information is associated with training that includes joint prediction and precoder compression using a machine learning model for both prediction and compression.
[0351] Aspect 63: The method of Aspect 62, wherein the prediction and compression use target CSI in a present slot and past CSI feedback.
[0352] Aspect 64: The method of Aspect 62, wherein the prediction and compression use target CSI in a future slot and no past CSI feedback.
[0353] Aspect 65: The method of Aspect 62, wherein the prediction and compression use target CSI in a future slot and past CSI feedback.
[0354] Aspect 66: The method of Aspect 62, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.
[0355] Aspect 67: The method of Aspect 62, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.
[0356] Aspect 68: The method of Aspect 62, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.
[0357] Aspect 69: The method of any of Aspects 37-68, wherein the model information includes a dataset or a dataset format associated with a model structure.
[0358] Aspect 70: The method of Aspect 69, wherein the dataset includes target CSI in a present slot or a future slot and past CSI feedback.
[0359] Aspect 71: The method of Aspect 70, wherein the dataset includes past CSI feedback.
[0360] Aspect 72: The method of Aspect 70, wherein ideal prediction is expected for the training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.
[0361] Aspect 73: The method of Aspect 70, wherein realistic prediction is expected for the training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.
[0362] Aspect 74: The method of Aspect 70, wherein the model information indicates base quantization codebooks and differential quantization codebooks.
[0363] Aspect 75: An apparatus for wireless communication at a device, the apparatus comprising one or more processors; one or more memories coupled with the one or more processors; and instructions stored in the one or more memories and executable by the one or more processors to cause the apparatus to perform the method of one or more of Aspects 1-74.
[0364] Aspect 76: An apparatus for wireless communication at a device, the apparatus comprising one or more memories and one or more processors coupled to the one or more memories, the one or more processors configured to cause the device to perform the method of one or more of Aspects 1-74.
[0365] Aspect 77: An apparatus for wireless communication, the apparatus comprising at least one means for performing the method of one or more of Aspects 1-74.
[0366] Aspect 78: A non-transitory computer-readable medium storing code for wireless communication, the code comprising instructions executable by one or more processors to perform the method of one or more of Aspects 1-74.
[0367] Aspect 79: A non-transitory computer-readable medium storing a set of instructions for wireless communication, the set of instructions comprising one or more instructions that, when executed by one or more processors of a device, cause the device to perform the method of one or more of Aspects 1-74.
[0368] Aspect 80: A device for wireless communication, the device comprising a processing system that includes one or more processors and one or more memories coupled with the one or more processors, the processing system configured to cause the device to perform the method of one or more of Aspects 1-74.
[0369] Aspect 81: An apparatus for wireless communication at a device, the apparatus comprising one or more memories and one or more processors coupled to the one or more memories, the one or more processors individually or collectively configured to cause the device to perform the method of one or more of Aspects 1-74.
[0370]
[0371] The foregoing disclosure provides illustration and description but is not intended to be exhaustive or to limit the aspects to the precise forms disclosed. Modifications and variations may be made in light of the above disclosure or may be acquired from practice of the aspects.
[0372] As used herein, the term “component” is intended to be broadly construed as hardware or a combination of hardware and at least one of software or firmware. “Software” shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, or functions, among other examples, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise. As used herein, a “processor” is implemented in hardware or a combination of hardware and software. It will be apparent that systems or methods described herein may be implemented in different forms of hardware or a combination of hardware and software. The actual specialized control hardware or software code used to implement these systems or methods is not limiting of the aspects. Thus, the operation and behavior of the systems or methods are described herein without reference to specific software code, because those skilled in the art will understand that software and hardware can be designed to implement the systems or methods based, at least in part, on the description herein. A component being configured to perform a function means that the component has a capability to perform the function, and does not require the function to be actually performed by the component, unless noted otherwise.
[0373] As used herein, “satisfying a threshold” may, depending on the context, refer to a value being greater than the threshold, greater than or equal to the threshold, less than the threshold, less than or equal to the threshold, equal to the threshold, or not equal to the threshold, among other examples.
[0374] As used herein, a phrase referring to “at least one of” a list of items refers to any combination of those items, including single members. As an example, “at least one of: a, b, or c” is intended to cover a, b, c, a + b, a + c, b + c, and a + b + c, as well as any combination with multiples of the same element (for example, a + a, a + a + a, a + a + b, a + a + c, a + b + b, a + c + c, b + b, b + b + b, b + b + c, c + c, and c + c + c, or any other ordering of a, b, and c) .
[0375] No element, act, or instruction used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” are intended to include one or more items and may be used interchangeably with “one or more. ” Further, as used herein, the article “the” is intended to include one or more items referenced in connection with the article “the” and may be used interchangeably with “the one or more. ” Furthermore, as used herein, the terms “set” and “group” are intended to include one or more items and may be used interchangeably with “one or more. ” Where only one item is intended, the phrase “only one” or similar language is used. Also, as used herein, the terms “has, ” “have, ” “having, ” and similar terms are intended to be open-ended terms that do not limit an element that they modify (for example, an element “having” A may also have B) . Further, the phrase “based on” is intended to mean “based on or otherwise in association with” unless explicitly stated otherwise. Also, as used herein, the term “or” is intended to be inclusive when used in a series and may be used interchangeably with “and / or, ” unless explicitly stated otherwise (for example, if used in combination with “either” or “only one of” ) . It should be understood that “one or more” is equivalent to “at least one. ”
[0376] Even though particular combinations of features are recited in the claims or disclosed in the specification, these combinations are not intended to limit the disclosure of various aspects. Many of these features may be combined in ways not specifically recited in the claims or disclosed in the specification. The disclosure of various aspects includes each dependent claim in combination with every other claim in the claim set.
Claims
1.An apparatus for wireless communication at a network entity, comprising:one or more memories; andone or more processors, coupled to the one or more memories, individually or collectively configured to cause the network entity to:obtain dataset information from a user equipment (UE) ;train machine learning models based at least in part on the dataset information; andtransmit model information to the UE that is based at least in part on the training and that is associated with spatial, temporal, and frequency compression of channel state information (CSI) .2.The apparatus of claim 1, wherein the one or more processors are further configured to cause the network entity to time information of the CSI.3.The apparatus of claim 2, wherein the timing information includes one or more of a time window and a quantity of time-tagged samples.4.The apparatus of claim 1, wherein the dataset information includes an input CSI that matches a target CSI.5.The apparatus of claim 1, wherein the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.6.The apparatus of claim 5, wherein the dataset information includes a dataset of only input CSI samples within an observation window.7.The apparatus of claim 5, wherein the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.8.The apparatus of claim 5, wherein the dataset information includes a dataset of only UE-predicted CSI samples.9.The apparatus of claim 5, wherein the dataset information includes a dataset of only target CSI samples and predicted CSI samples.10.The apparatus of claim 1, wherein the model information includes a model format of a model structure and / or at least one parameter for the model structure.11.The apparatus of claim 1, wherein to train the machine learning models, the one or more processors are individually or collectively configured to cause the network entity to train the machine learning models using separate channel prediction and precoder compression.12.The apparatus of claim 1, wherein to train the machine learning models, the one or more processors are individually or collectively configured to cause the network entity to train the machine learning models assuming ideal prediction associated with a target CSI and time domain samples.13.The apparatus of claim 1, wherein the one or more processors are individually or collectively configured to cause the network entity to separate precoder prediction and precoder compression.14.The apparatus of claim 13, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.15.The apparatus of claim 14, wherein the compression uses target CSI in a present slot and past CSI feedback.16.The apparatus of claim 14, wherein the compression uses target CSI in a future slot and no past CSI feedback.17.The apparatus of claim 14, wherein the compression uses target CSI in a future slot and past CSI feedback.18.The apparatus of claim 13, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.19.The apparatus of claim 18, wherein the decompression uses target CSI in a present slot and past CSI feedback.20.The apparatus of claim 18, wherein the decompression uses target CSI in a future slot and no past CSI feedback.21.The apparatus of claim 18, wherein the decompression uses target CSI in a future slot and past CSI feedback.22.The apparatus of claim 13, wherein the model information includes a model format for a channel state information (CSI) generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.23.The apparatus of claim 1, wherein to train the machine learning models, the one or more processors are individually or collectively configured to cause the network entity to train the machine learning models assuming realistic prediction associated with a target CSI and a prediction output.24.The apparatus of claim 1, wherein to train the machine learning models, the one or more processors are individually or collectively configured to cause the network entity to train the machine learning models using joint prediction and precoder compression and using a machine learning model for both prediction and compression.25.The apparatus of claim 24, wherein the prediction and compression use target CSI in a present slot and past CSI feedback.26.The apparatus of claim 24, wherein the prediction and compression use target CSI in a future slot and no past CSI feedback.27.The apparatus of claim 24, wherein the prediction and compression use target CSI in a future slot and past CSI feedback.28.The apparatus of claim 24, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.29.The apparatus of claim 24, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.30.The apparatus of claim 24, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.31.The apparatus of claim 1, wherein the model information includes a dataset or a dataset format associated with a model structure.32.The apparatus of claim 31, wherein the one or more processors are individually or collectively configured to cause the network entity to target CSI in a present slot or a future slot and past CSI feedback.33.The apparatus of claim 32, wherein the dataset includes past CSI feedback.34.The apparatus of claim 32, wherein ideal prediction is expected for training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.35.The apparatus of claim 32, wherein realistic prediction is expected for training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.36.The apparatus of claim 32, wherein the model information indicates base quantization codebooks and differential quantization codebooks.37.An apparatus for wireless communication at a user equipment (UE) , comprising:one or more memories; andone or more processors, coupled to the one or more memories, individually or collectively configured to cause the UE to:transmit dataset information; andreceive model information that is associated with using the information for machine learning training of spatial, temporal, and frequency compression of channel state information.38.The apparatus of claim 37, wherein the one or more processors are individually or collectively configured to cause the UE to reference model information.39.The apparatus of claim 37, wherein the model information indicates an encoder.40.The apparatus of claim 37, wherein the one or more processors are individually or collectively configured to cause the UE to time information of the CSI.41.The apparatus of claim 40, wherein the timing information includes one or more of a time window and a quantity of time-tagged samples.42.The apparatus of claim 37, wherein the dataset information includes an input CSI that matches a target CSI.43.The apparatus of claim 37, wherein the dataset information of the CSI includes an input CSI, a target CSI, and a predicted CSI.44.The apparatus of claim 43, wherein the dataset information includes a dataset of only input CSI samples within an observation window.45.The apparatus of claim 43, wherein the dataset information includes a dataset of only input CSI samples within an observation window and target CSI samples within a prediction window.46.The apparatus of claim 43, wherein the dataset information includes a dataset of only UE-predicted CSI samples.47.The apparatus of claim 43, wherein the dataset information includes a dataset of only target CSI samples and predicted CSI samples.48.The apparatus of claim 37, wherein the model information includes a model format of a model structure and / or at least one parameter for the model structure.49.The apparatus of claim 37, wherein the model information is associated with training for separate channel prediction and precoder compression.50.The apparatus of claim 37, wherein the model information is associated with training for a target CSI and time domain samples.51.The apparatus of claim 37, wherein the one or more processors are individually or collectively configured to cause the UE to separate precoder prediction and precoder compression.52.The apparatus of claim 51, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression only.53.The apparatus of claim 52, wherein the compression uses target CSI in a present slot and past CSI feedback.54.The apparatus of claim 52, wherein the compression uses target CSI in a future slot and no past CSI feedback.55.The apparatus of claim 52, wherein the compression uses target CSI in a future slot and past CSI feedback.56.The apparatus of claim 52, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression only.57.The apparatus of claim 56, wherein the decompression uses target CSI in a present slot and past CSI feedback.58.The apparatus of claim 56, wherein the decompression uses target CSI in a future slot and no past CSI feedback.59.The apparatus of claim 56, wherein the decompression uses target CSI in a future slot and past CSI feedback.60.The apparatus of claim 37, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and a CSI reconstruction model that is associated with a reference decoder for decompression.61.The apparatus of claim 37, wherein the model information is associated with training that assumes realistic prediction associated with a target CSI and a prediction output.62.The apparatus of claim 37, wherein the model information is associated with training that includes joint prediction and precoder compression using a machine learning model for both prediction and compression.63.The apparatus of claim 62, wherein the prediction and compression use target CSI in a present slot and past CSI feedback.64.The apparatus of claim 62, wherein the prediction and compression use target CSI in a future slot and no past CSI feedback.65.The apparatus of claim 62, wherein the prediction and compression use target CSI in a future slot and past CSI feedback.66.The apparatus of claim 62, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder that is for compression and prediction.67.The apparatus of claim 62, wherein the model information includes a model format for a CSI reconstruction model that is associated with a reference decoder that is for decompression and prediction.68.The apparatus of claim 62, wherein the model information includes a model format for a CSI generation model that is associated with a reference encoder for compression and prediction, and a CSI reconstruction model that is associated with a reference decoder for decompression and prediction.69.The apparatus of claim 37, wherein the model information includes a dataset or a dataset format associated with a model structure.70.The apparatus of claim 69, wherein the one or more processors are individually or collectively configured to cause the UE to target CSI in a present slot or a future slot and past CSI feedback.71.The apparatus of claim 70, wherein the dataset includes past CSI feedback.72.The apparatus of claim 70, wherein ideal prediction is expected for the training, and wherein the dataset further includes a reconstructed target CSI that corresponds to an input CSI time window.73.The apparatus of claim 70, wherein realistic prediction is expected for the training, and wherein the dataset further includes one or more of an input CSI that corresponds to an observation window and a reconstructed target CSI that corresponds to a prediction window.74.The apparatus of claim 70, wherein the model information indicates base quantization codebooks and differential quantization codebooks.
Citation Information
Patent Citations
Functionality based two-sided machine learning operations
US20240276241A1
Method and apparatus for supporting ai and ML operation in a wireless communication system
WO2024158175A1