Machine learning of channel state feedback encoder based on indicated reference decoder

Abstract

A method for wireless communication by a user equipment (UE) includes encoding a network-based machine learning channel state feedback (CSF) based on information specifying a CSF decoder in order to generate a channel state information (CSI) report. The method also includes transmitting the CSI report to a network device. A method for wireless communication by a network device includes transmitting information specifying a network-based machine learning (CSF) decoder to a UE. The method also includes receiving a CSI report from the UE. The method further includes decoding the CSI report with the network-based machine learning (CSF) decoder.

Description

Technical Field

[0001] The present disclosure relates generally to wireless communications, and more particularly to specifications for a machine learning (ML) channel state feedback (CSF) reference decoder. Background Art

[0002] Wireless communication systems are widely deployed to provide a variety of telecommunication services, such as telephony, video, data, messaging, and broadcasts. Typical wireless communication systems employ multiple-access technologies that support communication with multiple users by sharing available system resources (e.g., bandwidth, transmit power, etc.). Examples of such multiple-access technologies include code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal frequency division multiple access (OFDMA), single-carrier frequency division multiple access (SC-FDMA), time division synchronous code division multiple access (TD-SCDMA), and long-term evolution (LTE). LTE / LTE-Advanced is a set of enhancements to the Universal Mobile Telecommunications System (UMTS) mobile standard promulgated by the Third Generation Partnership Project (3GPP). Narrowband (NB) Internet of Things (IoT) and enhanced machine-type communications (eMTC) are enhancements to LTE for machine-type communications.

[0003] A wireless communication network may include multiple base stations (BSs), which can support communication for multiple user equipment (UEs). User equipment (UEs) can communicate with the base stations (BSs) via downlinks and uplinks. The downlink (or forward link) refers to the communication link from the BS to the UE, while the uplink (or reverse link) refers to the communication link from the UE to the BS. As will be described in more detail, a BS may be referred to as a Node B, an evolved Node B (eNB), a gNB, an access point (AP), a radio head, a transmit and receive point (TRP), a new radio (NR) BS, a 5G Node B, etc.

[0004] The above multiple access technologies have been adopted in various telecommunications standards to provide a common protocol that enables different user equipment to communicate at the city, national, regional, and even global levels. New Radio (NR), also known as 5G, is a set of enhancements to the LTE mobile standard released by the Third Generation Partnership Project (3GPP). NR is designed to better support mobile broadband internet access by improving spectral efficiency, reducing costs, improving services, utilizing new spectrum, and better integrating with other open standards by using orthogonal frequency division multiplexing (OFDM) with a cyclic prefix (CP) (CP-OFDM) on the downlink (DL) and CP-OFDM and / or SC-FDM (e.g., also known as discrete Fourier transform spread OFDM (DFT-s-OFDM)) on the uplink (UL). It also supports beamforming, multiple-input multiple-output (MIMO) antenna technology, and carrier aggregation.

[0005] An artificial neural network may include a group of interconnected artificial neurons (e.g., a neuron model). An artificial neural network may be a computing device or represented as a method executed by a computing device. A convolutional neural network (such as a deep convolutional neural network) is a type of feedforward artificial neural network. A convolutional neural network may include layers of neurons that can be arranged in a tiled receptive field. It would be desirable to apply neural network processing to wireless communications to achieve higher efficiency. Summary of the Invention

[0006] In various aspects of the present disclosure, a method for wireless communication by a user equipment (UE) includes encoding a network-based machine learning channel state feedback (CSF) decoder based on information specifying the CSF to generate a channel state information (CSI) report. The method also includes transmitting the CSI report to a network device.

[0007] In various aspects of the present disclosure, a method for wireless communication by a network device includes sending information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE). The method also includes receiving a channel state information (CSI) report from the UE. The method also includes decoding the CSI report using the network-based machine learning channel state feedback (CSF) decoder.

[0008] Other aspects of the present disclosure relate to an apparatus having a memory and one or more processors coupled to the memory. The processors are configured to encode a network-based machine learning channel state feedback (CSF) based on information specifying a CSF decoder to generate a channel state information (CSI) report. The processors are further configured to transmit the CSI report to a network device.

[0009] Other aspects of the present disclosure relate to an apparatus having a memory and one or more processors coupled to the memory. The processors are configured to send information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE). The processors are further configured to receive a channel state information (CSI) report from the UE and decode the CSI report using the network-based machine learning channel state feedback (CSF) decoder.

[0010] Aspects generally include methods, apparatus, systems, computer program products, non-transitory computer-readable media, user equipment, base stations, wireless communication devices, and processing systems substantially as described with reference to and as illustrated in the accompanying drawings and description.

[0011] The features and technical advantages of the examples according to the present disclosure have been outlined quite broadly above so that the detailed description that follows may be better understood. Additional features and advantages will be described. The concepts and specific examples disclosed may be readily utilized as a basis for modifying or designing other structures for achieving the same purposes of the present disclosure. Such equivalent constructions do not depart from the scope of the appended claims. The characteristics of the disclosed concepts, both in terms of their organization and method of operation, and the associated advantages will be better understood by considering the following description in conjunction with the accompanying drawings. Each of the figures in the accompanying drawings is provided for the purpose of illustration and description and not as a definition of limitations of the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order that the features of the present disclosure may be understood in detail, a more particular description may be given with reference to various aspects, some of which are illustrated in the accompanying drawings. It should be noted, however, that the drawings illustrate only certain aspects of the present disclosure and are not to be considered limiting of its scope, as the description may admit to other equally effective aspects. The same reference numerals in different drawings may identify the same or similar elements.

[0013] Figure 1 is a block diagram conceptually illustrating an example of a wireless communication network in accordance with various aspects of the present disclosure.

[0014] Figure 2 is a block diagram conceptually illustrating an example of a base station communicating with a user equipment (UE) in a wireless communication network according to various aspects of the present disclosure.

[0015] Figure 3 is a block diagram illustrating an example decomposed base station architecture in accordance with various aspects of the present disclosure.

[0016] Figure 4 An example implementation of designing a neural network using a system on a chip (SOC) including a general-purpose processor according to certain aspects of the present disclosure is illustrated.

[0017] Figure 5A 、 Figure 5B and Figure 5C is a diagram illustrating a neural network according to aspects of the present disclosure.

[0018] Figure 5D is a diagram illustrating an exemplary deep convolutional network (DCN) according to aspects of the present disclosure.

[0019] Figure 6 is a block diagram illustrating an exemplary deep convolutional network (DCN) according to aspects of the present disclosure.

[0020] Figure 7 is a block diagram illustrating codebook-based channel state feedback (CSF).

[0021] Figure 8 is a block diagram illustrating machine learning based channel state feedback (CSF) according to various aspects of the present disclosure.

[0022] Figure 9 is a block diagram illustrating a wireless communication system with a reference decoder according to various aspects of the present disclosure.

[0023] Figure 10 is a block diagram illustrating a wireless communication system with a cell-specific encoder according to various aspects of the present disclosure.

[0024] Figure 11 is a block diagram illustrating a base decoder of a reference decoder according to various aspects of the present disclosure.

[0025] Figure 12 is a block diagram illustrating batches of decoder output according to various aspects of the present disclosure.

[0026] Figure 13 and Figure 14 is a block diagram illustrating decoder output for batches with subband and layer definitions according to various aspects of the present disclosure.

[0027] Figure 15 is a block diagram illustrating decoder output for different numbers of layers according to various aspects of the present disclosure.

[0028] Figure 16 A table of layer and subband based decoder identification (ID) according to various aspects of the present disclosure is illustrated.

[0029] Figure 17 A table showing mapping between bandwidth part (BWP) (in units of physical resource blocks (PRBs)) and subband size (in units of PRBs) for codebook-based channel state feedback (CSF) is illustrated according to various aspects of the present disclosure.

[0030] Figure 18 Illustrated is a table showing the mapping between bandwidth part (BWP) (in units of physical resource blocks (PRBs)), subband size (in units of PRBs), and the number of precoders per subband according to various aspects of the present disclosure.

[0031] Figure 19 A table showing the mapping between bandwidth part (BWP) (in PRBs), subband size (in PRBs), and the number of precoders per subband is illustrated according to various aspects of the present disclosure.

[0032] Figure 20 is a block diagram illustrating subband overlapping according to various aspects of the present disclosure.

[0033] Figure 21 is a block diagram illustrating a first option for in-phase and quadrature (I / Q) mapping according to various aspects of the present disclosure.

[0034] Figure 22 is a block diagram illustrating a second option for I / Q mapping according to various aspects of the present disclosure.

[0035] Figure 23 is a flow diagram illustrating an example process performed, for example, by a user equipment (UE) according to various aspects of the present disclosure.

[0036] Figure 24 is a flow diagram illustrating an example process, eg, performed by a network device, according to various aspects of the present disclosure. DETAILED DESCRIPTION

[0037] Various aspects of the present disclosure are described more fully below with reference to the accompanying drawings. However, the present disclosure can be embodied in many different forms and should not be construed as limited to any specific structure or function presented throughout the present disclosure. Rather, these aspects are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of protection of the present disclosure to those skilled in the art. Based on the teachings, those skilled in the art should recognize that the scope of the present disclosure is intended to cover any aspect of the present disclosure, whether that aspect is implemented independently of any other aspect of the present disclosure or implemented in combination with any other aspect. For example, a device may be implemented or a method may be practiced using any number of the aspects set forth. In addition, the scope of the present disclosure is intended to cover such devices or methods practiced using other structures, functionality, or structure and functionality as a supplement to the various aspects of the present disclosure set forth or in addition. It should be understood that any aspect of the present disclosure disclosed may be embodied by one or more elements of the claims.

[0038] Several aspects of telecommunications systems will now be presented with reference to various devices and techniques. These devices and techniques are described in the following detailed description and illustrated in the accompanying drawings by various blocks, modules, components, circuits, steps, processes, algorithms, etc. (collectively, "elements"). These elements can be implemented using hardware, software, or a combination thereof. Whether such elements are implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system.

[0039] It should be noted that while various aspects may be described using terminology typically associated with 5G and later generation wireless technologies, various aspects of the present disclosure may be applied in communication systems based on other generations, such as and including 3G and / or 4G technologies.

[0040] The UE measures reference signals from the network and estimates the channel based on the reference signal measurements. These estimates are provided to network equipment (such as the base station) as channel state feedback (CSF) via channel state information (CSI) reports. For codebook-based CSF, the UE calculates a precoder and maps it to the CSF payload.

[0041] Machine learning (ML) techniques can be used for CSF communication. Using ML techniques, the UE receives a channel state information reference signal (CSI-RS) from a base station (e.g., a gNB) and estimates the channel based on CSI-RS measurements. A machine learning encoder (e.g., a neural network) can be used to compress (e.g., encode) the channel. The machine learning encoder also maps the output to a CSF payload. The UE sends an ML CSI report to the base station. The base station decodes the ML CSI report using a machine learning decoder (e.g., a neural network decoder). The decoder at the base station reconstructs the pre-decoder (V) based on machine learning processing of the CSF payload. out ), and can choose to use the pre-decoder to send downlink signals.

[0042] In order for the encoder to communicate with the decoder, a common understanding of the encoding and decoding processes is required. For example, a neural network encoder and a neural network decoder can be trained together, and the device performing the training can share the trained model (or just one of the trained models) with one or more other devices. Training the encoder and decoder together and then sharing the model is inefficient. For example, sharing the model requires communication resources. Other solutions are desirable to enable the encoder to encode data in a way that can be successfully decoded on the network side. The network decoder may be referred to as a reference decoder.

[0043] According to various aspects of the present disclosure, a reference decoder is specified, for example, in a wireless standard. The structure of the reference decoder can be specified using fixed or configurable parameters. In other aspects of the present disclosure, the format of the decoder input and / or output is known to the UE. In still other aspects, the UE receives a reference decoder identification (ID) and / or other information to enable correct operation of the encoder. Information about the reference decoder (e.g., decoder specifications, input / output specifications, and / or signaling for the reference decoder) can allow the UE to train the encoder without having to train the decoder.

[0044] Figure 1 Figure 1 is a diagram illustrating a network 100 in which various aspects of the present disclosure may be practiced. Network 100 may be a 5G or NR network, or some other wireless network (such as an LTE network). Wireless network 100 may include multiple base stations (BSs) 110 (illustrated as BS 110a, BS 110b, BS 110c, and BS 110d) and other network entities. A BS is an entity that communicates with user equipment (UE) and may also be referred to as a base station, NR BS, Node B, gNB, 5G Node B, access point, transmit and receive point (TRP), network node, network entity, etc. A BS may be implemented as a converged base station, a disaggregated base station, an integrated access and backhaul (IAB) node, a relay node, a sidelink node, etc. A BS may be implemented in a converged or monolithic base station architecture, or alternatively in a disaggregated base station architecture, and may include one or more of a central unit (CU), a distributed unit (DU), a radio unit (RU), a near real-time (near-RT) RAN intelligent controller (RIC), or a non-real-time (non-RT) RIC. Each BS can provide communication coverage for a specific geographical area. In 3GPP, the term "cell" can refer to the coverage area of ​​a BS and / or a BS subsystem serving the coverage area, depending on the context in which the term is used.

[0045] A BS may provide communication coverage for a macro cell, a pico cell, a femto cell, and / or another type of cell. A macro cell may cover a relatively large geographic area (e.g., a radius of several kilometers) and may allow unrestricted access by UEs with service subscriptions. A pico cell may cover a relatively small geographic area and may allow unrestricted access by UEs with service subscriptions. A femto cell may cover a relatively small geographic area (e.g., a home) and may allow restricted access by UEs associated with the femto cell (e.g., UEs in a closed subscriber group (CSG)). A BS for a macro cell may be referred to as a macro BS. A BS for a pico cell may be referred to as a pico BS. A BS for a femto cell may be referred to as a femto BS or a home BS. In Figure 1In the example shown in FIG, BS 110a may be a macro BS for macrocell 102a, BS 110b may be a pico BS for picocell 102b, and BS 110c may be a femto BS for femtocell 102c. A BS may support one or more (e.g., three) cells. The terms "eNB," "base station," "NR BS," "gNB," "AP," "Node B," "5G NB," "TRP," and "cell" may be used interchangeably.

[0046] In some aspects, the cells may not necessarily be stationary, and the geographic area of ​​the cells may move depending on the location of the mobile BS. In some aspects, the BSs may interconnect with each other and / or with one or more other BSs or network nodes (not shown) in the wireless network 100 via various types of backhaul interfaces (e.g., direct physical connections, virtual networks, etc.) using any suitable transport network.

[0047] The wireless network 100 may also include a relay station. A relay station is an entity that can receive data transmissions from an upstream station (e.g., a BS or a UE) and transmit the data transmissions to a downstream station (e.g., a UE or a BS). A relay station may also be a UE that can relay transmissions for other UEs. Figure 1 In the example shown in FIG, a relay station 110d may communicate with a macro BS 110a and a UE 120d to facilitate communication between the BS 110a and the UE 120d. A relay station may also be referred to as a relay BS, a relay base station, a relay, or the like.

[0048] The wireless network 100 may be a heterogeneous network including different types of BSs (e.g., macro BSs, pico BSs, femto BSs, relay BSs, etc.). These different types of BSs may have different transmit power levels, different coverage areas, and different impacts on interference in the wireless network 100. For example, a macro BS may have a high transmit power level (e.g., 5 watts to 40 watts), while a pico BS, a femto BS, and a relay BS may have a lower transmit power level (e.g., 0.1 watt to 2 watts).

[0049] The network controller 130 may be coupled to a group of BSs and may provide coordination and control for these BSs. The network controller 130 may communicate with the BSs via backhaul. The BSs may also communicate with each other via wireless or wired backhaul (eg, directly or indirectly).

[0050] UEs 120 (e.g., 120a, 120b, 120c) may be dispersed throughout wireless network 100, and each UE may be stationary or mobile. A UE may also be referred to as an access terminal, terminal, mobile station, subscriber unit, station, etc. A UE may be a cellular phone (e.g., a smartphone), 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 computer, a camera, a gaming device, a netbook, a smartbook, an ultrabook, a medical device or equipment, a biometric sensor / device, a wearable device (e.g., a smart watch, smart clothing, smart glasses, a smart wristband, smart jewelry (e.g., a smart ring, a smart bracelet)), an entertainment device (e.g., a music or video device or satellite radio), a component or sensor of a vehicle, a smart meter / sensor, industrial manufacturing equipment, a global positioning system device, or any other suitable device configured to communicate via a wireless or wired medium.

[0051] Some UEs may be considered machine-type communication (MTC) or evolved or enhanced machine-type communication (eMTC) UEs. For example, MTC and eMTC UEs include robots, drones, remote devices, sensors, meters, monitors, location tags, and the like that can communicate with a base station, another device (e.g., a remote device), or some other entity. A wireless node may provide connectivity to or with a network (e.g., a wide area network such as the Internet or a cellular network) via a wired or wireless communication link. Some UEs may be considered Internet of Things (IoT) devices and / or may be implemented as NB-IoT (Narrowband Internet of Things) devices. Some UEs may be considered customer premises equipment (CPE). UE 120 may be included in a housing that houses components of UE 120 (e.g., a processor component, a memory component, etc.).

[0052] Generally speaking, any number of wireless networks can be deployed in a given geographic area. Each wireless network can support a specific radio access technology (RAT) and can operate on one or more frequencies. A RAT may also be referred to as a radio technology, air interface, etc. A frequency may also be referred to as a carrier, frequency channel, etc. Each frequency in a given geographic area may support a single RAT to avoid interference between wireless networks of different RATs. In some cases, NR or 5G RAT networks may be deployed.

[0053] In some aspects, two or more UEs 120 (e.g., shown as UE 120a and UE 120e) may communicate directly (e.g., without using base station 110 as an intermediary) using one or more sidelink channels. For example, UEs 120 may communicate using peer-to-peer (P2P) communication, device-to-device (D2D) communication, vehicle-to-everything (V2X) protocols (e.g., which may include vehicle-to-vehicle (V2V) protocols, vehicle-to-infrastructure (V2I) protocols, etc.), mesh networks, and the like. In such cases, UEs 120 may perform scheduling operations, resource selection operations, and / or other operations described elsewhere herein, performed by base station 110. For example, base station 110 may configure UEs 120 via downlink control information (DCI), radio resource control (RRC) signaling, medium access control-control elements (MAC-CEs), or via system information (e.g., system information blocks (SIBs)).

[0054] The UE 120 may include a machine learning (ML) configuration module 140. For simplicity, only one UE 120d is shown as including the machine learning (ML) configuration module 140. The ML configuration module 140 may encode a network-based machine learning channel state feedback (CSF) decoder based on information specifying the CSF to generate a channel state information (CSI) report. The ML configuration module 140 may also send the CSI report to a network device.

[0055] Base station 110 may include a machine learning (ML) configuration module 138. For simplicity, only one base station 110a is shown as including the machine learning (ML) configuration module 138. The ML configuration module 138 may send information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE). The ML configuration module 138 may also receive channel state information (CSI) reports from the UE. The ML configuration module 138 may further utilize the network-based machine learning channel state feedback (CSF) decoder to decode the CSI reports.

[0056] As indicated above, Figure 1 is provided as an example only. Other examples can be found in the Figure 1 The examples described are different.

[0057] Figure 2 A block diagram shows a design 200 of a base station 110 and a UE 120, which may be Figure 1 One of the base stations in Figure 1Base station 110 may be equipped with T antennas 234a through 234t, and UE 120 may be equipped with R antennas 252a through 252r, where in general T ≥ 1 and R ≥ 1.

[0058] At base station 110, transmit processor 220 may receive data for one or more UEs from data source 212, select one or more modulation and coding schemes (MCS) for each UE based at least in part on a channel quality indicator (CQI) received from that UE, process (e.g., encode and modulate) the data for each UE based at least in part on the selected MCS for that UE, and provide data symbols for all UEs. Reducing the MCS lowers throughput but increases transmission reliability. Transmit processor 220 may also process system information (e.g., for semi-static resource allocation information (SRPI)) and control information (e.g., CQI requests, grants, upper layer signaling, etc.), and provide overhead symbols and control symbols. Transmit processor 220 may also generate reference symbols for reference signals (e.g., cell-specific reference signals (CRS)) and synchronization signals (e.g., primary synchronization signals (PSS) and secondary synchronization signals (SSS)). The transmit (TX) multiple-input, multiple-output (MIMO) processor 230 may perform spatial processing (e.g., precoding) on ​​data symbols, control symbols, overhead symbols, and / or reference symbols, as applicable, and may provide T output symbol streams to T modulators (MODs) 232a through 232t. Each modulator 232 may process a corresponding output symbol stream (e.g., for orthogonal frequency division multiplexing (OFDM) or the like) to obtain an output sample stream. Each modulator 232 may further process (e.g., convert to analog, amplify, filter, and upconvert) the output sample stream to obtain a downlink signal. The T downlink signals from modulators 232a through 232t may be transmitted via T antennas 234a through 234t, respectively. According to various aspects described in greater detail below, position coding may be utilized to generate synchronization signals to convey additional information.

[0059] At UE 120, antennas 252a through 252r may receive downlink signals from base station 110 and / or other base stations and may provide received signals to demodulators (DEMODs) 254a through 254r, respectively. Each demodulator 254 may condition (e.g., filter, amplify, downconvert, and digitize) the received signal to obtain input samples. Each demodulator 254 may further process the input samples (e.g., for OFDM, etc.) to obtain received symbols. A MIMO detector 256 may obtain received symbols from all R demodulators 254a through 254r, perform MIMO detection on the received symbols (if applicable), and provide detected symbols. A receive processor 258 may process (e.g., demodulate and decode) the detected symbols, provide decoded data for UE 120 to a data sink 260, and provide decoded control information and system information to a controller / processor 280. The channel processor may determine reference signal received power (RSRP), received signal strength indicator (RSSI), reference signal received quality (RSRQ), channel quality indicator (CQI), etc. In some aspects, one or more components of UE 120 may be included in a housing.

[0060] On the uplink, at UE 120, a transmit processor 264 may receive data from a data source 262 and control information (e.g., for reports including RSRP, RSSI, RSRQ, CQI, etc.) from a controller / processor 280 and process the data and control information. The transmit processor 264 may also generate reference symbols for one or more reference signals. The symbols from the transmit processor 264 may be pre-decoded by a TX MIMO processor 266, if applicable, further processed by modulators 254a through 254r (e.g., for discrete Fourier transform spread OFDM (DFT-s-OFDM), CP-OFDM, etc.), and transmitted to the base station 110. At the base station 110, uplink signals from UE 120 and other UEs may be received by antennas 234, processed by demodulators 254, detected by MIMO detector 236 (if applicable), and further processed by receive processor 238 to obtain decoded data and control information transmitted by UE 120. The receive processor 238 may provide the decoded data to a data sink 239 and the decoded control information to a controller / processor 240. The base station 110 may include a communication unit 244 and communicate with the network controller 130 via the communication unit 244. The network controller 130 may include a communication unit 294, a controller / processor 290, and a memory 292.

[0061] The controller / processor 240 of the base station 110, the controller / processor 280 of the UE 120, and / or Figure 2 Any other component of the UE 120 may perform one or more techniques associated with the machine learning CSF reference decoder, as described in more detail elsewhere. For example, the controller / processor 240 of the base station 110, the controller / processor 280 of the UE 120, and / or Figure 2 Any other component of may perform or direct e.g. Figure 23 and Figure 24 1 and / or other processes as described. Memory 242 and memory 282 may store data and program codes for base station 110 and UE 120, respectively. Scheduler 246 may schedule UEs for data transmission on the downlink and / or uplink.

[0062] In some aspects, the UE 120 may include means for encoding, means for transmitting, means for training, means for computing, and means for receiving. In some aspects, the base station 110 may include means for transmitting, means for receiving, and means for decoding. Such means may include means for encoding, transmitting, receiving, and decoding. Figure 2 One or more components of a UE 120 or base station 110 are depicted.

[0063] As indicated above, Figure 2 It is provided as an example only. Other examples can be found in the reference Figure 2 The examples described are different.

[0064] In some cases, different types of devices supporting different types of applications and / or services can coexist in a cell. Examples of different types of devices include UEs, customer premises equipment (CPE), vehicles, and Internet of Things (IoT) devices. Examples of different types of applications include ultra-reliable low-latency communications (URLLC), massive machine-type communications (mMTC), enhanced mobile broadband (eMBB), and vehicle-to-everything (V2X). Furthermore, in some cases, a single device can simultaneously support different applications or services.

[0065] The deployment of a communication system, such as a 5G New Radio (NR) system, can be arranged in a variety of ways using various components or constituent parts. In a 5G NR system or network, a network node, a network entity, a mobility element of a network, a radio access network (RAN) node, a core network node, a network element, or network equipment (such as a base station (BS)), or one or more units (or one or more components) performing base station functions can be implemented in a converged or disaggregated architecture. For example, a base station (such as a node B (NB), an evolved NB (eNB), an NR base station, a 5G NB, an access point (AP), a transmit and receive point (TRP), or a cell) can be implemented as a converged base station (also known as a standalone base station or a monolithic base station) or a disaggregated base station.

[0066] A converged base station may be configured to utilize a radio protocol stack that is physically or logically integrated within a single RAN node. A disaggregated base station may be configured to utilize a protocol stack that is physically or logically distributed across two or more units, such as one or more central or centralized units (CUs), one or more distributed units (DUs), or one or more radio units (RUs). In some aspects, a CU may be implemented within a RAN node, and one or more DUs may be co-located with the CU, or alternatively, may be geographically or virtually distributed across one or more other RAN nodes. A DU may be implemented to communicate with one or more RUs. Each of the CU, DU, and RU may also be implemented as a virtual unit (e.g., a virtual central unit (VCU), a virtual distributed unit (VDU), or a virtual radio unit (VRU)).

[0067] Base station type operation or network design may take into account the aggregated nature of base station functionality. For example, a disaggregated base station may be used in an integrated access backhaul (IAB) network, an open radio access network (O-RAN (such as the network configuration initiated by the O-RAN Alliance)), or a virtualized radio access network (vRAN, also known as a cloud radio access network (C-RAN)). Disaggregation can include distributing functionality across two or more units at various physical locations, as well as virtually distributing the functionality of at least one unit, which enables flexibility in network design. The various units of a disaggregated base station or disaggregated RAN architecture can be configured for wired or wireless communication with at least one other unit.

[0068] Figure 3 A diagram illustrating an example disaggregated base station 300 architecture is shown. The disaggregated base station 300 architecture may include one or more central units (CUs) 310, which may communicate directly with a core network 320 via a backhaul link, or indirectly with the core network 320 through one or more disaggregated base station units, such as a near real-time (near-RT) RAN intelligent controller (RIC) 325 via an E2 link, or a non-real-time (non-RT) RIC 315 associated with a service management and orchestration (SMO) framework 305, or both. The CUs 310 may communicate with one or more distributed units (DUs) 330 via corresponding midhaul links, such as the F1 interface. The DUs 330 may communicate with one or more radio units (RUs) 340 via corresponding fronthaul links. The RUs 340 may communicate with corresponding UEs 120 via one or more radio frequency (RF) access links. In some implementations, a UE 120 may be served simultaneously by multiple RUs 340.

[0069] Each of these units (e.g., CU 310, DU 330, RU 340, as well as near-RT RIC 325, non-RTRIC 315, and SMO framework 305) may include one or more interfaces, or may be coupled to one or more interfaces configured to receive or transmit signals, data, or information (collectively, signals) via a wired or wireless transmission medium. Each of the units, or an associated processor or controller that provides instructions to the communication interface of these units, may be configured to communicate with one or more of the other units via the transmission medium. For example, these units may include a wired interface that is configured to receive or transmit signals to one or more of the other units via the wired transmission medium. Additionally, these units may include a wireless interface that may include a receiver, transmitter, or transceiver (such as a radio frequency (RF) transceiver) that is configured to receive signals or transmit signals to one or more of the other units, or both, via the wireless transmission medium.

[0070] In some aspects, the CU 310 may host one or more higher-layer control functions. Such control functions may include radio resource control (RRC), packet data convergence protocol (PDCP), service data adaptation protocol (SDAP), etc. Each control function may be implemented using an interface configured to communicate signals with other control functions hosted by the CU 310. The CU 310 may be configured to handle user plane functions (e.g., central unit-user plane (CU-UP)), control plane functions (e.g., central unit-control plane (CU-CP)), or a combination thereof. In some implementations, the CU 310 may be logically split into one or more CU-UP units and one or more CU-CP units. When implemented in an O-RAN configuration, the CU-UP units may communicate bidirectionally with the CU-CP units via an interface (such as the E1 interface). As needed, the CU 310 may be implemented to communicate with the DU 330 for network control and signaling.

[0071] The DU 330 may correspond to a logical unit that includes one or more base station functions for controlling the operation of one or more RUs 340. In some aspects, the DU 330 may host one or more of a radio link control (RLC) layer, a medium access control (MAC) layer, and one or more higher physical (PHY) layers (such as modules for forward error correction (FEC) encoding and decoding, scrambling, modulation and demodulation, etc.), depending at least in part on a functional partitioning (such as that defined by the Third Generation Partnership Project (3GPP)). In some aspects, the DU 330 may also host one or more lower PHY layers. Each layer (or module) may be implemented using an interface configured to communicate signals with other layers (and modules) hosted by the DU 330 or with control functions hosted by the CU 310.

[0072] Lower layer functionality may be implemented by one or more RUs 340. In some deployments, a RU 340 controlled by a DU 330 may correspond to a logical node that hosts RF processing functionality or low-PHY layer functionality (such as performing fast Fourier transforms (FFTs), inverse FFTs (iFFTs), digital beamforming, physical random access channel (PRACH) extraction and filtering, etc.), or both, based at least in part on a functional split (such as a lower layer functional split). In such an architecture, a RU 340 may be implemented to handle over-the-air (OTA) communications with one or more UEs 120. In some implementations, both real-time and non-real-time aspects of control and user plane communications with the RU 340 may be controlled by the corresponding DU 330. In some scenarios, this configuration may enable the implementation of the DU 330 and CU 310 in a cloud-based RAN architecture, such as a vRAN architecture.

[0073] The SMO framework 305 can be configured to support RAN deployment and provisioning of both non-virtualized and virtualized network elements. For non-virtualized network elements, the SMO framework 305 can be configured to support the deployment of dedicated physical resources for RAN coverage requirements, which can be managed via an operations and maintenance interface (such as the O1 interface). For virtualized network elements, the SMO framework 305 can be configured to interact with a cloud computing platform (such as the Open Cloud (O-Cloud) platform 390) to perform network element lifecycle management (such as instantiating virtualized network elements) via cloud computing platform interfaces (such as the O2 interface). Such virtualized network elements may include, but are not limited to, the CU 310, DU 330, RU 340, and near-RT RIC 325. In some implementations, the SMO framework 305 can communicate with 4G RAN hardware (such as the Open eNB (O-eNB) 311) via the O1 interface. Additionally, in some implementations, the SMO framework 305 can communicate directly with one or more RUs 340 via the O1 interface. The SMO framework 305 may also include a non-RT RIC 315 configured to support the functionality of the SMO framework 305 .

[0074] The non-RT RIC 315 may be configured to include logic that enables non-real-time control and optimization of RAN elements and resources, artificial intelligence / machine learning (AI / ML) workflows including model training and updating, or policy-based guidance of applications / features in the near-RT RIC 325. The non-RT RIC 315 may be coupled to or in communication with the near-RT RIC 325 (e.g., via an A1 interface). The near-RT RIC 325 may be configured to include logic that enables near-real-time control and optimization of RAN elements and resources via data collection and actions over an interface (e.g., via an E2 interface) that connects one or more CUs 310, one or more DUs 330, or both, and the O-eNB 311 with the near-RT RIC 325.

[0075] In some implementations, to generate AI / ML models to be deployed in the near-RT RIC 325, the non-RT RIC 315 may receive parameters or external enrichment information from an external server. Such information may be utilized by the near-RT RIC 325 and may be received from non-network data sources or from network functions at the SMO framework 305 or the non-RT RIC 315. In some examples, the non-RT RIC 315 or the near-RT RIC 325 may be configured to tune RAN behavior or performance. For example, the non-RT RIC 315 may monitor long-term trends and patterns in performance and employ AI / ML models to execute corrective actions through the SMO framework 305 (e.g., via reconfiguration of O1) or through the creation of RAN management policies (e.g., A1 policies).

[0076] Figure 4 An example implementation of a system-on-chip (SOC) 400, according to certain aspects of the present disclosure, is illustrated. The system-on-chip (SOC) 400 may include a central processing unit (CPU) 402 or a multi-core CPU configured to generate gradients for neural network training. The SOC 400 may be included in a base station 110 or a user equipment (UE) 120. Variables (e.g., neural signals and synaptic weights), system parameters associated with a computing device (e.g., a neural network with weights), latency, frequency bin information, and task information may be stored in a memory block associated with a neural processing unit (NPU) 408, a memory block associated with the CPU 402, a memory block associated with a graphics processing unit (GPU) 404, a memory block associated with a digital signal processor (DSP) 406, a memory block 418, or may be distributed across multiple blocks. Instructions executed by the CPU 402 may be loaded from a program memory associated with the CPU 402 or from the memory block 418.

[0077] The SOC 400 may also include additional processing blocks tailored for specific functions, such as a GPU 404, a DSP 406, a connectivity block 410 (which may include fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, etc.), and a multimedia processor 412 that may, for example, detect and recognize gestures. In one embodiment, the NPU is implemented in the CPU, DSP, and / or GPU. The SOC 400 may also include a sensor processor 414, an image signal processor (ISP) 416, and / or a navigation module 420, which may include a global positioning system.

[0078] SOC 400 may be based on the ARM instruction set. In one aspect of the present disclosure, the instructions loaded into general-purpose processor 402 may include code for encoding a network-based machine learning channel state feedback (CSF) decoder based on information specifying the CSF to generate a channel state information (CSI) report. General-purpose processor 402 may also include code for sending the CSI report to a network device.

[0079] In other aspects of the present disclosure, the instructions loaded into the general purpose processor 402 may include code for sending information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE). The general purpose processor 402 may also include code for receiving a channel state information (CSI) report from the UE. The general purpose processor 402 may also include code for decoding the CSI report using the network-based machine learning channel state feedback (CSF) decoder.

[0080] Deep learning architectures can perform object recognition tasks by learning to represent input at successively higher levels of abstraction at each layer, thereby constructing useful feature representations of the input data. In this way, deep learning addresses a major bottleneck in traditional machine learning. Before the advent of deep learning, machine learning approaches to object recognition problems could rely heavily on human-designed features, perhaps in conjunction with shallow classifiers. Shallow classifiers could be, for example, two-class linear classifiers, where the weighted sum of the feature vector components is compared to a threshold to predict which class the input belongs to. Human-designed features could be templates or kernels customized for a specific problem domain by engineers with domain expertise. In contrast, while deep learning architectures can learn to represent features similar to those that human engineers might design, they do so through training. Furthermore, deep networks can learn to represent and recognize new types of features that humans might not have considered.

[0081] Deep learning architectures can learn hierarchies of features. For example, if presented with visual data, the first layer might learn to recognize relatively simple features in the input stream, such as edges. In another example, if presented with auditory data, the first layer might learn to recognize spectral power at specific frequencies. The second layer, taking the output of the first layer as input, might learn to recognize combinations of features, such as simple shapes in visual data or combinations of sounds in auditory data. For example, higher layers might learn to represent complex shapes in visual data or words in auditory data. Even higher layers might learn to recognize common visual objects or spoken phrases.

[0082] Deep learning architectures perform particularly well when applied to problems with a natural hierarchical structure. For example, the classification of motorized vehicles can benefit from first learning to recognize wheels, windshields, and other features. These features can then be combined in different ways at higher levels to identify cars, trucks, and airplanes.

[0083] Neural networks can be designed with a variety of connection patterns. In a feedforward network, information is passed from lower layers to higher layers, with each neuron in a given layer communicating with neurons in higher layers. As described above, hierarchical representations can be constructed in successive layers of a feedforward network. Neural networks can also have recurrent or feedback (also known as top-down) connections. In a recurrent connection, the output from a neuron in a given layer can be communicated to another neuron in the same layer. Recurrent architectures can help identify patterns that span more than one block of input data sequentially delivered to the neural network. Connections from neurons in a given layer to neurons in lower layers are called feedback (or top-down) connections. Networks with many feedback connections can be helpful when the recognition of high-level concepts can assist in discerning specific low-level features of the input.

[0084] The connections between the layers of a neural network can be fully connected or partially connected. Figure 5A Illustrated is an example of a fully connected neural network 502. In a fully connected neural network 502, a neuron in a first layer may communicate its output to every neuron in a second layer, such that every neuron in the second layer will receive input from every neuron in the first layer. Figure 5B An example of a locally connected neural network 504 is illustrated. In the locally connected neural network 504, neurons in a first layer may be connected to a limited number of neurons in a second layer. More generally, the locally connected layers of the locally connected neural network 504 may be configured such that each neuron in the layer will have the same or similar connection pattern, but the connection strengths may have different values ​​(e.g., 510, 512, 514, and 516). The locally connected connectivity pattern may produce spatially different receptive fields in higher layers because higher layer neurons in a given region may receive inputs that have been tuned through training to the characteristics of a limited portion of the total input to the network.

[0085] An example of a locally connected neural network is a convolutional neural network. Figure 5C An example of a convolutional neural network 506 is illustrated. The convolutional neural network 506 can be configured such that the connection strengths associated with the inputs to each neuron in the second layer are shared (e.g., 508). Convolutional neural networks may be well suited for problems where the spatial location of the inputs is meaningful.

[0086] One type of convolutional neural network is the deep convolutional network (DCN). Figure 5D A detailed example of a DCN 500 is shown, which is designed to recognize visual features from an image 526 input from an image capture device 530 (such as an onboard camera). The DCN 500 of the current example can be trained to identify traffic signs and the numbers provided on traffic signs. Of course, the DCN 500 can be trained for other tasks, such as identifying lane markings or traffic lights.

[0087] DCN 500 can be trained using supervised learning. During training, DCN 500 can be presented with an image, such as image 526 of a speed limit sign, and can subsequently compute a forward pass to produce output 522. DCN 500 can include a feature extraction portion and a classification portion. Upon receiving image 526, convolutional layer 532 can apply a convolution kernel (not shown) to image 526 to generate a first set of feature maps 518. As an example, the convolution kernel of convolutional layer 532 can be a 5×5 kernel that generates a 28×28 feature map. In this example, because four different feature maps are generated in first set of feature maps 518, four different convolution kernels are applied to image 526 at convolutional layer 532. Convolution kernels can also be referred to as filters or convolution filters.

[0088] The first set of feature maps 518 may be subsampled by a max pooling layer (not shown) to generate a second set of feature maps 520. The max pooling layer reduces the size of the first set of feature maps 518. That is, the size of the second set of feature maps 520 (e.g., 14×14) is smaller than the size of the first set of feature maps 518 (e.g., 28×28). This reduced size provides similar information to subsequent layers while reducing memory consumption. The second set of feature maps 520 may be further convolved via one or more subsequent convolutional layers (not shown) to generate one or more subsequent sets of feature maps (not shown).

[0089] exist Figure 5D In the example of , second set of feature maps 520 are convolved to generate a first feature vector 524. Furthermore, first feature vector 524 is further convolved to generate a second feature vector 528. Each feature of second feature vector 528 may include a number corresponding to a possible feature of image 526, such as "sign," "60," and "100." A softmax function (not shown) may convert the numbers in second feature vector 528 into probabilities. Thus, output 522 of DCN 500 may be a probability that image 526 includes one or more features.

[0090] In this example, the probability of "logo" and "60" in output 522 is higher than the probability of other numbers in output 522, such as "30," "40," "50," "70," "80," "90," and "100." Before training, output 522 generated by DCN 500 may be incorrect. Therefore, the error between output 522 and the target output can be calculated. The target output is the true value of image 526 (e.g., "logo" and "60"). The weights of DCN 500 can then be adjusted so that output 522 of DCN 500 is closer to the target output.

[0091] To adjust a weight, the learning algorithm calculates a gradient vector for the weight. The gradient indicates how much the error will increase or decrease if the weight is adjusted. At the top layer, the gradient may directly correspond to the value of the weight connecting the activated neurons in the penultimate layer to the neurons in the output layer. In lower layers, the gradient may depend on the value of the weight and the calculated error gradient of the higher layer. The weight can then be adjusted to reduce the error. This method of adjusting weights is called "backpropagation" because it involves a "backward pass" through the neural network.

[0092] In practice, the error gradient of the weights can be calculated over a small number of examples, so that the calculated gradient approximates the true error gradient. This approximation method is referred to as stochastic gradient descent. Stochastic gradient descent can be repeated until the achievable error rate of the entire system stops decreasing or until the error rate reaches a target level. After learning, a new image (e.g., a speed limit sign in image 526) can be presented to DCN 500, and a forward pass through DCN 500 can produce output 522, which can be considered an inference or prediction of DCN 500.

[0093] A deep belief network (DBN) is a probabilistic model consisting of multiple layers of hidden nodes. A DBN can be used to extract a hierarchical representation of a training dataset. A DBN can be obtained by stacking layers of restricted Boltzmann machines (RBMs). RBMs are a type of artificial neural network that learns a probability distribution over a set of inputs. Because RBMs can learn a probability distribution without having information about the class to which each input should be classified, they are often used for unsupervised learning. Using a hybrid supervised and unsupervised paradigm, the bottom RBM of a DBN can be trained in an unsupervised manner and used as a feature extractor, while the top RBM can be trained in a supervised manner (on the joint distribution of inputs from the previous layer and the target class) and used as a classifier.

[0094] A DCN is a network of convolutional networks configured with additional pooling and normalization layers. DCN has achieved state-of-the-art performance on many tasks. DCN can be trained using supervised learning, where both the input and output targets are known for many examples and are used to modify the network's weights using gradient descent.

[0095] A DCN can be a feedforward network. Furthermore, as described above, connections from neurons in the first layer of a DCN to a set of neurons in the next higher layer are shared across the neurons in the first layer. The feedforward and shared connections of a DCN can be used for fast processing. For example, the computational burden of a DCN can be much smaller than that of a similarly sized neural network that includes recurrent or feedback connections.

[0096] The processing of each layer of a convolutional network can be thought of as a spatially invariant template or basis projection. If the input is first decomposed into multiple channels, such as the red, green, and blue channels of a color image, then the convolutional network trained on this input can be thought of as three-dimensional, with two spatial dimensions along the axes of the image and the third dimension capturing color information. The output of the convolutional connection can be thought of as forming a feature map in the subsequent layer, where each element in this feature map (e.g., 520) receives input from a certain range of neurons in the previous layer (e.g., feature map 518) and from each of the multiple channels. The values ​​in the feature map can be further processed using nonlinearities (such as rectification, maximum value (max) (0, x)). Values ​​from neighboring neurons can be further pooled, which corresponds to downsampling and can provide additional local invariance and dimensionality reduction. Normalization corresponding to whitening can also be applied through lateral inhibition between neurons in the feature map.

[0097] The performance of deep learning architectures can increase as more labeled data points become available or as computing power increases. Modern deep neural networks are often trained with computing resources thousands of times greater than those available to typical researchers just fifteen years ago. New architectures and training paradigms can further improve deep learning performance. Rectified linear units can alleviate the training problem known as vanishing gradients. New training techniques can reduce overfitting and, therefore, enable larger models to achieve better generalization. Encapsulation techniques can extract data within a given receptive field and further improve overall performance.

[0098] Figure 6 FIG is a block diagram illustrating a deep convolutional network 650. Based on connections and weight sharing, the deep convolutional network 650 may include multiple different types of layers. Figure 6As shown, the deep convolutional network 650 includes convolution blocks 654A and 654B. Each of the convolution blocks 654A and 654B can be configured with a convolution layer (CONV) 656, a normalization layer (LNorm) 658, and a maximum pooling layer (MAXPOOL) 660. Although only two convolution blocks 654A and 654B are shown, the present disclosure is not limited thereto, and instead, any number of convolution blocks 654A and 654B can be included in the deep convolutional network 650 according to design preferences.

[0099] The convolution layer 656 may include one or more convolution filters that may be applied to the input data to generate a feature map. The normalization layer 658 may normalize the output of the convolution filter. For example, the normalization layer 658 may provide whitening or lateral suppression. The max pooling layer 660 may provide spatial downsampling aggregation to achieve local invariance and dimensionality reduction.

[0100] For example, a parallel filter bank of a deep convolutional network can be loaded into the SOC 400 (e.g., Figure 4 ) on the CPU 402 or GPU 404 of the SOC 400 to achieve high performance and low power consumption. In an alternative embodiment, the parallel filter bank can be loaded onto the DSP 406 or ISP 416 of the SOC 400. In addition, the deep convolutional network 650 can access other processing blocks that may exist on the SOC 400, such as the sensor processor 414 and the navigation module 420 dedicated to sensors and navigation, respectively.

[0101] The deep convolutional network 650 may also include one or more fully connected layers 662 (FC1 and FC2). The deep convolutional network 650 may also include a logistic regression (LR) layer 664. Each layer 656, 658, 660, 662, and 664 of the deep convolutional network 650 has weights (not shown) to be updated between them. The output of each of these layers (e.g., 656, 658, 660, 662, and 664) may be used as input to subsequent layers in the deep convolutional network 650 to learn a hierarchical feature representation from the input data 652 (e.g., images, audio, video, sensor data, and / or other input data) supplied at the first convolutional block 654A. The output of the deep convolutional network 650 is a classification score 666 for the input data 652. The classification score 666 may be a set of probabilities, where each probability is a probability of the input data including a feature from a feature set.

[0102] As indicated above, Figures 4 to 6 are provided as examples. Other examples may be relevant to Figure 3 The example described to FIG5 is different.

[0103] The UE can measure reference signals from the network and estimate the channel based on the reference signal measurements. These estimates can be provided to network equipment (such as the base station) as channel state feedback (CSF) in channel state information (CSI) reports. For codebook-based CSF, the UE calculates a precoder and maps it to the CSF payload.

[0104] Figure 7 is a block diagram illustrating codebook-based channel state feedback (CSF). Figure 7 In the example of FIG. 1 , at block 702, UE 120 receives a channel state information reference signal (CSI-RS) from a base station (e.g., gNB) 110 or a network node and estimates the channel. At block 704, UE 120 calculates the channel state information and sends a legacy CSI report to base station 110. At block 706, a network device (such as base station 110) reconstructs a precoder (V out The reconstruction mapping between the CSF payload and the pre-decoder (eg, codebook) is specified in wireless communication standards (such as 3GPP standards). If the base station 110 chooses to use the pre-decoder, the base station 110 may utilize the pre-decoder (V out ) to send downlink signals.

[0105] Machine learning (ML) techniques can be used for CSF. Figure 8 is a block diagram illustrating a machine learning based channel state feedback (CSF) according to various aspects of the present disclosure. Figure 8 In the example of FIG. 8 , at block 802, UE 120 receives a channel state information reference signal (CSI-RS) from base station 110 and estimates the channel. At block 804, UE 120 compresses the channel state via a machine learning encoder (e.g., a neural network encoder). The machine learning encoder also maps the output to a CSF payload. UE 120 sends an ML CSI report to base station 110. At block 806, base station 110 decodes the ML CSI report using a machine learning decoder (e.g., a neural network decoder). The decoder at base station 110 reconstructs a pre-decoder (V) based on machine learning processing of the CSF payload in the ML CSI report. out ), and optionally utilizes a pre-decoder (V out ) to send downlink signals.

[0106] In order for the encoder to decoder to communicate, a common understanding of the encoding and decoding process is required. For example, a neural network encoder and decoder can be trained together, where the device performing the training shares the trained model (or just one of the trained models) with one or more other devices.

[0107] Training the encoder and decoder together and then sharing the model is inefficient. For example, sharing the model requires communication resources. Other solutions are expected to enable the encoder to encode the data in a way that can be decoded at the network side. The decoder at the network can be called a reference decoder.

[0108] According to various aspects of the present disclosure, a reference decoder is specified, for example, in a wireless standard. In other aspects of the present disclosure, the format of the decoder input and / or output is known to the UE. In still other aspects, the UE receives a reference decoder identification (ID) and / or other information. Information about the reference decoder at the UE (e.g., decoder specifications, input / output specifications, and / or signaling for the reference decoder) allows the UE to train the encoder without having to also train the decoder.

[0109] If a reference decoder is specified, for example, in a wireless standard, the structure of the reference decoder may be specified using fixed or configurable parameters. The UE may use the structure of the reference decoder and the fixed parameters when training the encoder and when calculating the channel quality indicator (CQI) component of the CSF. In some aspects, multiple fixed, well-defined decoders (e.g., multiple reference neural networks) may be defined. For example, the UE may train different encoders for different cells to improve performance. In some aspects, the neural network may be based on the number of beams in the horizontal and vertical directions (e.g., the values ​​N1 and N2 in the existing codebook specification). The neural network may also be based on the precoder structure (e.g., W1).

[0110] Figure 9 is a block diagram illustrating a wireless communication system with a reference decoder according to various aspects of the present disclosure. Figure 9 In the example of FIG. 1 , at block 902, UE 120 receives a channel state information reference signal (CSI-RS) from a base station (e.g., gNB) 110 and performs channel estimation. At block 904, UE 120 compresses the channel using a neural network encoder. The encoder also maps the output to a CSF payload. According to aspects of the present disclosure, UE 120 trains the compression encoder based on specifications of a reference decoder. These specifications may indicate a neural network structure and a fixed or configurable set of parameters.

[0111] UE 120 sends an ML CSI report to base station 110. At block 906, base station 110 optionally reconstructs the precoder (V out). For example, the reconstruction mapping (e.g., codebook) between the CSF payload and the pre-decoder is specified in the wireless standard. Alternatively, at block 908, the base station 110 utilizes a reference decoder to decode the ML CSI report. With this option, the reference decoder at the base station 110 uses machine learning techniques based on the CSF payload to reconstruct the pre-decoder (V out ).

[0112] Figure 10 is a block diagram illustrating a wireless communication system with a cell-specific encoder according to various aspects of the present disclosure. Figure 10 In the example of , base station 110 stores and operates a reference decoder. A first UE (UE 1) 120-1 and a second UE (UE 2) 120-1 train their encoders based on the specifications of the reference decoder. For example, when first UE 120-1 and second UE 120-2 are in a first cell (cell 1), UEs 120-1 and 120-2 operate a first encoder (encoder 1), which has been trained based on the characteristics of the first cell (cell 1) and the specifications of the reference decoder. When first UE 120-1 and second UE 120-2 are served by a second cell (cell 2), UEs 120-1 and 120-2 operate a second encoder (encoder 2), which has been trained based on the characteristics of the second cell (cell 2) and the specifications of the reference decoder.

[0113] If a reference decoder structure is specified and the parameters are configurable, the base station may update the parameters. The base station shares the parameters with the UE in an offline or online manner to enable the UE to update its neural network encoder. In addition, the base station may indicate to the UE that the base station will update the parameters. Thus, the UE trains the encoder under the assumption that the decoder structure is fixed. In some aspects, different parameter sets are used for different cells, allowing the UE to use different encoders when served by different cells, such as Figure 10 shown.

[0114] In some aspects of the present disclosure, instead of or in addition to the reference decoder structure and parameters, the formats of the decoder input and / or decoder output are also specified. The dimensions of the reference decoder input and / or output may include batch size, number of layers, number of subbands, number of precoders per subband, number of ports, and / or in-phase and quadrature (I / Q) data. The batch size (B) refers to how the decoder input is partitioned. The input can be partitioned into B parts. In some aspects, a specific base decoder will be used for each batch part.

[0115] Figure 11 is a block diagram illustrating a base decoder of a reference decoder according to various aspects of the present disclosure. Figure 11In the example, reference decoder 1102 includes multiple basic decoders 1104-0...1104-B-1. Each basic decoder 1104-0...1104-B-1 receives one of B total batches (Batch 0...Batch B-1) and outputs one of B different outputs (Output 0...Output B-1). By employing multiple basic decoders 1104-0...1104-B-1, reference decoder 1102 is more flexible. For example, if the number of layers or subbands used for a particular output format changes, the number of basic decoders 1104-0...1104-B-1 can be changed accordingly. The number of layers refers to the rank of the CSF. The number of subbands refers to the number of subbands into which the entire bandwidth is divided.

[0116] The number of ports dimension may refer to the number of CSI-RS ports associated with the decoder input / output. The I / Q dimension accounts for the fact that the encoder may operate with real numbers, while complex numbers represent I / Q data. Therefore, the I and Q components of the complex numbers can be considered two different inputs. In other implementations, amplitude and phase may also be used.

[0117] Some of the decoder dimensions can be combined. For example, if the same decoder is applied to every layer and subband, the number of layers multiplied by the number of subbands defines the batch size. In this case, the reference decoder dimensions are the batch size multiplied by the number of precoders per subband multiplied by the number of ports multiplied by the I / Q dimensions. Each batch (e.g., a base decoder) has the dimensions of the number of precoders per subband multiplied by the number of ports multiplied by the I / Q dimensions.

[0118] In another implementation, the same decoder is applied to each layer or each subband. In this implementation, the batch size is defined as the number of layers or the number of subbands. The dimensions of the reference decoder are the batch size multiplied by the number of layers multiplied by the number of pre-decoders per subband multiplied by the number of ports multiplied by the I / Q dimensions, or the batch size multiplied by the number of subbands multiplied by the number of pre-decoders per subband multiplied by the number of ports multiplied by the I / Q dimensions.

[0119] In a third implementation, a large decoder is applied directly to all layers and subbands. In other words, no basic decoder is deployed. In this implementation, the batch size is equal to one. The dimensions of the reference decoder are the number of layers multiplied by the number of subbands multiplied by the number of pre-decoders per subband multiplied by the number of ports multiplied by the I / Q dimensions.

[0120] Now refer to Figure 12 Describes how to map to an input or output. Figure 12 is a block diagram illustrating batches of decoder outputs according to various aspects of the present disclosure. Figure 12 In the examples, decoder outputs are shown, but the concepts apply equally to decoder inputs. Figure 12In the example of , there are four layers and the batch size is defined based on the number of layers. Therefore, four batches are defined. In this example, a single neural network ID with different settings is used. After compression, the UE can place the data into four parts corresponding to the four layers or batches. The data for layer one is processed as batch 0 by the first basic decoder, the data for layer two is processed as batch 1 by the second basic decoder, the data for layer three is processed as batch 2 by the third basic decoder, and the data for layer four is processed as batch 3 by the fourth basic decoder. Although Figure 12 The examples are described with respect to batches defined based on the number of layers, but the concepts are equally applicable to batches defined with respect to the number of subbands or with respect to the number of layers and the number of subbands (eg, the second specific implementation described previously).

[0121] Figure 13 and Figure 14 is a block diagram illustrating decoder output for batches with subband and layer definitions according to various aspects of the present disclosure. Figure 13 and Figure 14 In the examples, decoder outputs are shown, but the concepts apply equally to decoder inputs. Figure 13 and Figure 14 In the examples shown, there are four subbands and two layers, and the batch size is defined based on the number of layers multiplied by the number of subbands. Therefore, eight batches are defined. In these examples, only a single neural network ID (NNID) is required for each of the different settings. After compression, the UE can place the data into eight payloads corresponding to the eight batches.

[0122] exist Figure 13 In the example of , data for layer 1 and subband 0 is processed as batch 0 by the first basic decoder, and data for layer 2 and subband 0 is processed as batch 1 by the second basic decoder. Data for layer 1 and subband 1 is processed as batch 2 by the third basic decoder, and data for layer 2 and subband 1 is processed as batch 3 by the fourth basic decoder. Data for layer 1 and subband 2 is processed as batch 4 by the fifth basic decoder, and data for layer 2 and subband 2 is processed as batch 5 by the sixth basic decoder. Data for layer 1 and subband 3 is processed as batch 6 by the seventh basic decoder, and data for layer 2 and subband 3 is processed as batch 7 by the eighth basic decoder.

[0123] exist Figure 14In the example of , data for layer 1 and subband 0 is processed as batch 0 by the first basic decoder, and data for layer 1 and subband 1 is processed as batch 1 by the second basic decoder. Data for layer 1 and subband 2 is processed as batch 2 by the third basic decoder, and data for layer 1 and subband 3 is processed as batch 3 by the fourth basic decoder. Data for layer 2 and subband 0 is processed as batch 4 by the fifth basic decoder, and data for layer 2 and subband 1 is processed as batch 5 by the sixth basic decoder. Data for layer 2 and subband 2 is processed as batch 6 by the seventh basic decoder, and data for layer 2 and subband 3 is processed as batch 7 by the eighth basic decoder.

[0124] If the number of layers or subbands is used to indicate the decoder (e.g., the third embodiment described previously), only a single decoder processes all layers (or subbands). In the example with four layers, four decoder IDs (or neural network IDs (NNIDs)) are used. Figure 15 is a block diagram illustrating decoder output for different numbers of layers according to various aspects of the present disclosure. Figure 15 In the examples, decoder outputs are shown, but the concepts apply equally to decoder inputs. Figure 15 In the example of , the first decoder (identified by NNID 0) decodes data for layer one. The second decoder (identified by NNID 1) decodes data for layers one and two. The third decoder (identified by NNID 2) decodes data for layers one, two, and three. The fourth decoder (identified by NNID 3) decodes data for layers one, two, three, and four. The first decoder, the second decoder, the third decoder, and the fourth decoder are different from each other. In some aspects, the decoders can be nested. Although Figure 15 The examples are described with respect to multiple layers, but the concepts apply equally to multiple sub-bands.

[0125] If the number of layers and subbands is used to specify a decoder (e.g., the third implementation described previously), a single decoder processes all layers and subbands. The number of layers (e.g., rank) multiplied by the number of subband combinations defines the number of decoders. In the example with four layers and three different subband combinations, twelve decoder IDs (or neural network IDs (NNIDs)) are used. Figure 16 exemplifies a table of decoder IDs based on layers and subbands according to various aspects of the present disclosure. Figure 16As shown, the single-layer (rank-1) decoder with neural network IDs 0, 1, and 2 corresponds to the input / output for one subband, two subbands, and four subbands, respectively. The two-layer (rank-2) decoder with neural network IDs 3, 4, and 5 corresponds to the decoder input / output for one subband, two subbands, and four subbands, respectively. The three-layer (rank-3) decoder with neural network IDs 6, 7, and 8 corresponds to the decoder input / output for one subband, two subbands, and four subbands, respectively. The four-layer (rank-4) decoder with neural network IDs 9, 10, and 11 corresponds to the decoder input / output for one subband, two subbands, and four subbands, respectively.

[0126] The format of the decoder output (or input) is now described with respect to the subband size. Figure 17 A table illustrating a mapping between a bandwidth part (BWP) (in units of physical resource blocks (PRBs)) and a subband size (in units of PRBs) for codebook-based channel state feedback (CSF) according to various aspects of the present disclosure is illustrated. According to aspects of the present disclosure, the mapping is used for machine learning CSF. For example, if the BWP is 24 PRBs, the subband size can be four PRBs or eight PRBs. Radio resource control (RRC) signaling can indicate which value to use. If the subband size is four PRBs, three subbands will be used for a BWP of 24 PRBs. Assuming that when using Figure 17 When the table is used, there is only one pre-decoder per sub-band. The sub-band size can be used to calculate the pre-decoder.

[0127] Since the machine learning CSF can have a more flexible payload than the codebook-based CSF, the number of subbands and the number of precoders per subband do not need to be limited. Therefore, alternative mappings are considered. Figure 18 Illustrated is a table showing the mapping between bandwidth part (BWP) (in units of physical resource blocks (PRBs)), subband size (in units of PRBs), and the number of precoders per subband according to various aspects of the present disclosure. Figure 19 A table illustrating the mapping between bandwidth part (BWP) (in units of PRBs), number of subbands, and number of precoders per subband according to various aspects of the present disclosure is illustrated. In some aspects, the predefined table is used for all UEs. In other aspects, the predefined table is selected based on UE capabilities. Figure 18 In the table, the subband size is different from Figure 17 In addition, the number of precoders per subband can be based on the subband size (e.g., two precoders per subband) or can be one. Figure 19In the table shown, the number of subbands is provided, rather than the subband size. Therefore, the compression rate and payload can be varied based on the number of subbands. In addition, the number of precoders per subband can be based on the subband size (e.g., two precoders per subband) or can be one. Figure 18 and Figure 19 The table in reduces the number of subbands and therefore the number of base decoders. Too many base decoders may degrade performance.

[0128] Figure 20 is a block diagram illustrating subband overlap according to various aspects of the present disclosure. In order to reuse the same neural network framework, the same subband size may be required. If the bandwidth part (BWP) is not a multiple of the subband size, overlapping mapping may be required, such as Figure 20 In some aspects, the resource blocks (RBs) that will overlap may be predefined. Return to Reference Figure 18 , if the BWP size is 272, the subband size can be 48 PRBs. Figure 20 In , 94 subbands (0-93) are shown. Subbands 0-47 and 48...92 cannot be used because if a basic decoder is used, the same data size is expected for the decoder input and output. Therefore, overlap is used. Figure 20 In the example of , subbands 46 and 47 are used for both subband zero and subband one because for 94 subbands or PRBs, the size of the first part is 48 subbands or PRBs and the size of the remaining subbands or PRBs is 46 (94-48=46) instead of 48.

[0129] Now let's discuss the in-phase and quadrature (I / Q) mapping used for the decoder output (and input) format. For real neural networks, complex numbers cannot be used as input or output. Therefore, I / Q mapping is required for the pre-decoder output, which is a complex number corresponding to the channel. Figure 21 is a block diagram illustrating a first option for I / Q mapping according to various aspects of the present disclosure. Figure 21 In the example of , all I values ​​are used first, followed by all Q values. Although not shown, the present disclosure contemplates switching I and Q such that the Q value is followed by the I value. Figure 22 is a block diagram illustrating a second option for I / Q mapping according to various aspects of the present disclosure. Figure 22 In the example above, the I values ​​are interleaved with the Q values. By having the same understanding of the I / Q mapping, the decoder is able to reconstruct the channel sent by the encoder.

[0130] According to various aspects of the present disclosure, the neural network ID and / or other information can be signaled by the base station. In a first option, only the neural network ID is indicated. In this case, the number of subbands and the number of layers are mapped to the neural network ID. This option is beneficial because neural networks are optimized for specific shapes or sizes. If the number of subbands or layers changes, the neural network changes.

[0131] In the second option, the neural network ID is signaled along with the number of subbands and / or layers. That is, the base decoder neural network ID is indicated. In this case, the number of subbands and layers is used as the batch size. In the third option, for example, when a single base decoder is used, the base station indicates only the number of subbands and / or layers.

[0132] In some aspects, a radio resource control (RRC) parameter indicates the neural network ID. For example, the reportQuantity in the CSI-ReportConfig can be used. In other aspects, dynamic scheduling is used. For example, downlink control information (DCI) can switch batch sizes for the same base neural network, for example, changing the number of subbands or the number of layers. For example, DCI can be used with the third option of using a single base decoder. DCI can also be used for neural network switching. Even if the neural network ID is different (e.g., the first option indicating only the neural network ID), as long as the base neural network is the same, processing time for switching is not a consideration. If there are multiple neural network IDs (e.g., a group of NNIDs), processing time for switching is required in order to switch between networks, for example, because a different neural network may be in another processor and the different neural network may need to be loaded before it is used.

[0133] Figure 23 2 is a flow chart illustrating an example process 2300 performed, for example, by a user equipment (UE) according to various aspects of the present disclosure. Example process 2300 is an example of utilizing a machine learning (ML) channel state feedback (CSF) reference decoder. The operations of process 2300 may be implemented by UE 120.

[0134] At block 2302, the UE encodes a network-based machine-learned channel state feedback (CSF) decoder based on information specifying the CSF to generate a channel state information (CSI) report. For example, the UE (e.g., using the controller / processor 280, the memory 282, etc.) may encode the CSF. In some aspects, the information may include a decoder neural network model having a fixed parameter set or a configurable parameter set. In other aspects, the information includes the format of the output from the machine-learned CSF decoder and / or the input to the machine-learned CSF decoder. In still other aspects, the information includes a neural network identification (ID), or the number of layers and / or the number of subbands.

[0135] At block 2304, the UE sends a CSI report to the network device. For example, the UE (eg, using antennas 252, DEMOD / MOD 254, TX MIMO processor 266, transmit processor 264, controller / processor 280, memory 282, etc.) may send a CSI report.

[0136] Figure 24 2 is a flow chart illustrating an example process 2400, for example, performed by a network device, according to various aspects of the present disclosure. Example process 2400 is an example of utilizing a machine learning (ML) channel state feedback (CSF) reference decoder. The operations of process 2400 may be implemented by base station 110.

[0137] At block 2402, a base station transmits information specifying a network-based machine-learning channel state feedback (CSF) decoder to a user equipment (UE). For example, the base station (e.g., using antenna 234, MOD / DEMOD 232, TX MIMO processor 230, transmit processor 220, controller / processor 240, memory 242, etc.) may transmit the information. In some aspects, the information may be a decoder neural network model with a fixed parameter set or a configurable parameter set. In other aspects, the information includes the format of the output from and / or input to the machine-learning CSF decoder. In still other aspects, the information includes a neural network identification (ID), or the number of layers and / or the number of subbands.

[0138] At block 2404, the base station receives a channel state information (CSI) report from the UE. For example, the base station (e.g., using antenna 234, MOD / DEMOD 232, MIMO detector 236, receive processor 238, controller / processor 240, memory 242, etc.) may receive the CSI report. At block 2406, the base station decodes the CSI report using a network-based machine learning CSI decoder. For example, the base station (e.g., using controller / processor 240, memory 242, etc.) may decode the CSI report.

[0139] Example aspects

[0140] Aspect 1: A method of wireless communication by a user equipment (UE), the method comprising: encoding a network-based machine learning channel state feedback (CSF) based on information specifying a CSF decoder to generate a channel state information (CSI) report; and sending the CSI report to a network device.

[0141] Aspect 2: The method according to aspect 1, wherein the information includes a decoder neural network model with a fixed set of parameters.

[0142] Aspect 3: The method according to aspect 1 or 2 further includes: using the information to train an encoder based on at least one cell serving the UE; and using the encoder to calculate a channel quality index (CQI) for the CSI report.

[0143] Aspect 4: The method according to aspect 1, wherein the information includes a decoder neural network model having a configurable parameter set.

[0144] Aspect 5: The method according to any one of aspects 1 or 4, further comprising: receiving an update for the set of configurable parameters from the network device.

[0145] Aspect 6: The method according to aspect 1, wherein the information includes the format of the output from the machine learning CSF decoder and / or the input to the machine learning CSF decoder.

[0146] Aspect 7: The method according to aspect 1 or 6, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband.

[0147] Aspect 8: The method according to aspect 1, wherein the information includes a neural network identification (ID).

[0148] Aspect 9: The method according to aspect 1, wherein the information includes the number of layers and / or the number of subbands.

[0149] Aspect 10: A method for wireless communication by a network device, the method comprising: sending information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE); receiving a channel state information (CSI) report from the UE; and utilizing the network-based machine learning CSF decoder to decode the CSI report.

[0150] Aspect 11: The method according to Aspect 10, wherein the information includes a decoder neural network model with a fixed set of parameters.

[0151] Aspect 12: The method according to Aspect 10, wherein the information includes a decoder neural network model having a configurable parameter set.

[0152] Aspect 13: The method according to any one of aspects 10 or 12 further comprises: sending an update for the configurable parameter set to the UE.

[0153] Aspect 14: The method according to aspect 10, wherein the information includes the format of the output from the machine learning CSF decoder and / or the input to the machine learning CSF decoder.

[0154] Aspect 15: The method according to any one of aspects 10 or 14, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband.

[0155] Aspect 16: The method according to aspect 10, wherein the information includes a neural network identification (ID).

[0156] Aspect 17: The method according to aspect 10, wherein the information includes the number of layers and / or the number of subbands.

[0157] Aspect 18: An apparatus for wireless communication, the apparatus comprising: a memory; and at least one processor coupled to the memory, the at least one processor configured to: encode a network-based machine learning channel state feedback (CSF) based on information specifying a CSF decoder to generate a channel state information (CSI) report; and send the CSI report to a network device.

[0158] Aspect 19: The apparatus of aspect 18, wherein the information comprises a decoder neural network model having a fixed set of parameters.

[0159] Aspect 20: The apparatus according to aspect 18 or 19, wherein the at least one processor is further configured to: train an encoder using the information based on at least one cell serving the UE; and calculate a channel quality index (CQI) for the CSI report using the encoder.

[0160] Aspect 21: The apparatus of aspect 18, wherein the information comprises a decoder neural network model having a configurable set of parameters.

[0161] Aspect 22: The apparatus of any of aspects 18 or 21, wherein the at least one processor is further configured to: receive an update to the set of configurable parameters from the network device.

[0162] Aspect 23: The apparatus of aspect 18, wherein the information comprises a format of an output from and / or an input to the machine learning CSF decoder.

[0163] Aspect 24: The apparatus according to aspect 18 or 23, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband.

[0164] Aspect 25: The apparatus according to aspect 18, wherein the information includes a neural network identification (ID).

[0165] Aspect 26: The apparatus according to aspect 18, wherein the information includes the number of layers and / or the number of subbands.

[0166] Aspect 27: An apparatus for wireless communication, the apparatus comprising: a memory; and at least one processor coupled to the memory, the at least one processor configured to: send information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE); receive a channel state information (CSI) report from the UE; and decode the CSI report using the network-based machine learning CSF decoder.

[0167] Aspect 28: The apparatus of aspect 27, wherein the information comprises a decoder neural network model having a fixed set of parameters.

[0168] Aspect 29: The apparatus of aspect 27, wherein the information comprises a decoder neural network model having a configurable set of parameters.

[0169] Aspect 30: The apparatus according to any one of aspects 27 or 29, wherein the at least one processor is further configured to: send an update for the set of configurable parameters to the UE.

[0170] 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 are possible in light of the above disclosure or may be acquired from practice of these aspects.

[0171] As used, the term "component" is intended to be broadly interpreted as hardware, firmware, and / or a combination of hardware and software. As used, a processor is implemented using hardware, firmware, and / or a combination of hardware and software.

[0172] Some aspects are described in conjunction with thresholds. As used, satisfying a threshold can mean a value is greater than a threshold, greater than or equal to a threshold, less than a threshold, less than or equal to a threshold, equal to a threshold, not equal to a threshold, etc., depending on the context.

[0173] It will be apparent that the described systems and / or methods can be implemented in various forms of hardware, firmware, and / or a combination of hardware and software. The actual dedicated control hardware or software code used to implement these systems and / or methods does not limit the aspects. Therefore, the operation and performance of these systems and / or methods are described without reference to specific software code, and it should be understood that software and hardware used to implement these systems and / or methods can be designed based at least in part on these descriptions.

[0174] Although particular combinations of features are set forth in the claims and / or disclosed in the specification, these combinations are not intended to limit the disclosure of the various aspects. In fact, many of these features can be combined in ways not specifically set forth in the claims and / or disclosed in the specification. Although each dependent claim listed below may directly depend on only one claim, the disclosure of the various aspects includes each dependent claim in combination with every other claim in the claim set. A phrase referring to "at least one" of a list of items refers to any combination of those items, including individual members. For example, "at least one of a, b, or c" is intended to cover a, b, c, ab, ac, bc, and abc, as well as any combination with multiple identical elements (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other ordering of a, b, and c).

[0175] The elements, actions, or instructions used should not be construed as critical or essential unless explicitly described as such. Furthermore, as used, the articles "a" and "an" are intended to include one or more items and can be used interchangeably with "one or more." Furthermore, as used, the terms "set" and "group" are intended to include one or more items (e.g., related items, unrelated items, combinations of related and unrelated items, etc.) and can be used interchangeably with "one or more." If only one item is intended to be referred to, the phrase "only one" or similar terms are used. Furthermore, as used, the terms "having" and the like are intended to be open-ended terms. Furthermore, the phrase "based on" is intended to mean "based at least in part on," unless explicitly stated otherwise.

Claims

1. A method for wireless communication by a user equipment (UE), the method comprising: encoding a channel state feedback (CSF) based on information specifying a network-based machine learning channel state feedback (CSF) decoder to generate a channel state information (CSI) report; as well as The CSI report is sent to a network device.

2. The method of claim 1 , wherein the information comprises a decoder neural network model having a fixed set of parameters.

3. The method according to claim 2, further comprising: training an encoder using the information based on at least one cell serving the UE; as well as A channel quality index (CQI) for the CSI report is calculated using the encoder.

4. The method of claim 1 , wherein the information comprises a decoder neural network model having a configurable set of parameters.

5. The method according to claim 4, further comprising: An update to the set of configurable parameters is received from the network device.

6. The method of claim 1, wherein the information includes a format of an output from and / or an input to the machine learning CSF decoder.

7. The method of claim 6, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband. The method of claim 1 , wherein the information includes a neural network identification (ID).

9. The method according to claim 1, wherein the information includes the number of layers and / or the number of subbands.

10. A method for wireless communication by a network device, the method comprising: sending information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE); receiving a channel state information (CSI) report from the UE; as well as The CSI report is decoded using the network-based machine learning CSF decoder.

11. The method of claim 10, wherein the information comprises a decoder neural network model having a fixed set of parameters.

12. The method of claim 10, wherein the information comprises a decoder neural network model having a configurable set of parameters.

13. The method according to claim 12, further comprising: An update for the configurable parameter set is sent to the UE.

14. The method of claim 10, wherein the information includes a format of an output from and / or an input to the machine learning CSF decoder.

15. The method of claim 14, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband.

16. The method of claim 10, wherein the information includes a neural network identification (ID). The method according to claim 10 , wherein the information includes the number of layers and / or the number of subbands.

18. An apparatus for wireless communication, the apparatus comprising: Memory; and at least one processor coupled to the memory, the at least one processor configured to: encoding a channel state feedback (CSF) based on information specifying a network-based machine learning channel state feedback (CSF) decoder to generate a channel state information (CSI) report; as well as The CSI report is sent to a network device.

19. The apparatus of claim 18, wherein the information comprises a decoder neural network model having a fixed set of parameters.

20. The apparatus of claim 19, wherein the at least one processor is further configured to: Using the information to train an encoder based on at least one cell serving the UE; and A channel quality index (CQI) for the CSI report is calculated using the encoder.

21. The apparatus of claim 18, wherein the information comprises a decoder neural network model having a configurable set of parameters.

22. The apparatus of claim 21, wherein the at least one processor is further configured to receive an update to the set of configurable parameters from the network device.

23. The apparatus of claim 18, wherein the information comprises a format of an output from and / or an input to the machine learning CSF decoder.

24. The apparatus of claim 23, wherein the machine learning CSF decoder comprises a plurality of base decoders, each base decoder being associated with at least one of a layer or a subband.

25. The apparatus of claim 18, wherein the information comprises a neural network identification (ID).

26. The apparatus according to claim 18, wherein the information comprises the number of layers and / or the number of subbands.

27. An apparatus for wireless communication, the apparatus comprising: Memory; and at least one processor coupled to the memory, the at least one processor configured to: sending information specifying a network-based machine learning channel state feedback (CSF) decoder to a user equipment (UE); receiving a channel state information (CSI) report from the UE; as well as The CSI report is decoded using the network-based machine learning CSF decoder.

28. The apparatus of claim 27, wherein the information comprises a decoder neural network model having a fixed set of parameters.

29. The apparatus of claim 27, wherein the information comprises a decoder neural network model having a configurable set of parameters.

30. The apparatus of claim 29, wherein the at least one processor is further configured to send an update to the set of configurable parameters to the UE.

Images