Channel feature extraction via model-based neural networks

Abstract

A method of wireless communication by a receiving device includes receiving from a transmitting device a latent representation of a channel sequence for a wireless signal, wherein a decoder applies a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. patent application Ser. No. 17 / 352,922, filed June 21, 2021, entitled "CHANNEL FEATURE EXTRACTION VIA MODEL-BASED NEURAL NETWORKS," the entire disclosure of which is expressly incorporated by reference into this specification.

[0002] Aspects of the present disclosure relate generally to wireless communication, and more specifically, to techniques and apparatus for channel feature extraction via a physical propagation channel model-based neural network. [Background technology]

[0003]

[0003] Wireless communication systems have been widely deployed to provide various telecommunication services, such as telephony, video, data, messaging, and broadcast. A typical wireless communication system may employ multiple access technologies capable of supporting 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) systems, time division multiple access (TDMA) systems, frequency-division multiple access (FDMA) systems, orthogonal frequency-division multiple access (OFDMA) systems, single-carrier frequency-division multiple access (SC-FDMA) systems, time division synchronous code division multiple access (TD-SCDMA) systems, and long term evolution (LTE). LTE / LTE-Advanced is a set of extensions to the universal mobile telecommunications system (UMTS) mobile standard promulgated by the Third Generation Partnership Project (3GPP®).

[0004]

[0004] A wireless communication network may include several base stations (BS) that can support communication for several user equipment (UE). The user equipment (UE) may communicate with the base stations (BS) via a downlink and an uplink. The downlink (or forward link) refers to the communication link from the BS to the UE, and the uplink (or reverse link) refers to the communication link from the UE to the BS. As described in more detail, the BS may be referred to as a Node B, 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.

[0005]

[0005] The above multiple access technologies have been adopted in various telecommunications standards to provide common protocols that allow different user equipment to communicate on a city, national, regional, or even global scale. New Radio (NR), sometimes referred to as 5G, is a set of enhancements to the LTE mobile standard promulgated by the 3rd Generation Partnership Project (3GPP). NR is designed to improve spectral efficiency, reduce costs, improve services, take advantage of new spectrum, and better support mobile broadband Internet access by using orthogonal frequency division multiplexing (OFDM) with 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), to better integrate with other open standards, as well as support beamforming, multiple-input multiple-output (MIMO) antenna technology, and carrier aggregation.

[0006]

[0006] An artificial neural network may comprise an interconnected group of artificial neurons (e.g., neuron models). An artificial neural network may be a computing device or may be represented as a method to be executed by a computing device. A convolutional neural network, such as a deep convolutional neural network, is a type of feed-forward artificial neural network. A convolutional neural network may include layers of neurons that can be arranged into tiled receptive fields. It would be desirable to apply neural network processing to wireless communications to achieve greater efficiency. Summary of the Invention

[0007]

[0007] The present disclosure is set out in respective independent claims. Some aspects of the disclosure are set out in dependent claims.

[0008]

[0008] In one aspect of the disclosure, a method of wireless communication by a receiving device is provided, the method including receiving from a transmitting device a latent representation of a channel sequence for a wireless signal, and applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal.

[0009]

[0009] In one aspect of the disclosure, an apparatus is provided for wireless communication by a receiving device. The apparatus includes a memory and one or more processors coupled to the memory. The processor(s) are configured to receive a latent representation of a channel sequence for a wireless signal from a transmitting device. The processor(s) are also configured to apply, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal.

[0010]

[0010] In one aspect of the disclosure, an apparatus for wireless communication by a receiving device is provided. The apparatus includes means for receiving a latent representation of a channel sequence for a wireless signal from a transmitting device. The apparatus also includes means for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal.

[0011]

[0011] In one aspect of the disclosure, a non-transitory computer readable medium is provided. The computer readable medium has encoded thereon program code for wireless communication by a receiving device. The program code, when executed by a processor, includes code for receiving from a transmitting device a latent representation of a channel sequence for a wireless signal. The program code also includes code for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal.

[0012]

[0012] In one aspect of the disclosure, a method of wireless communication by a transmitting device is provided. The method includes receiving, via an encoder, an input comprising a channel sequence for a wireless signal. The method also includes processing, via the encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal. The latent representation includes one or more elements that map to parameters of a physical propagation channel model. Additionally, the method includes transmitting the latent representation to a receiving device.

[0013]

[0013] In one aspect of the disclosure, an apparatus for wireless communication by a transmitting device is provided. The apparatus includes a memory and one or more processors coupled to the memory. The processor(s) is configured to receive, via an encoder, an input comprising a channel sequence for a wireless signal. The processor(s) is also configured to process, via the encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal. The latent representation includes one or more elements that map to parameters of a physical propagation channel model. Additionally, the processor(s) is configured to transmit the latent representation to a receiving device.

[0014]

[0014] In one aspect of the disclosure, an apparatus for wireless communication by a transmitting device is provided. The apparatus includes means for receiving, via an encoder, an input comprising a channel sequence for a wireless signal. The apparatus also includes means for processing, via the encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal. The latent representation includes one or more elements that map to parameters of a physical propagation channel model. Additionally, the apparatus includes means for transmitting the latent representation to a receiving device.

[0015]

[0015] In one aspect of the disclosure, a non-transitory computer-readable medium is provided. The computer-readable medium has encoded thereon program code for wireless communication by a transmitting device. The program code is executed by a processor and includes code for receiving, via an encoder, an input comprising a channel sequence for a wireless signal. The program code also includes code for processing, via the encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal. The latent representation includes one or more elements that map to parameters of a physical propagation channel model. Additionally, the program code includes code for transmitting the latent representation to a receiving device.

[0016]

[0016] Aspects generally include methods, apparatus, systems, computer program products, non-transitory computer-readable media, user equipment, base stations, wireless communication devices, and processing systems as fully described and illustrated in the accompanying drawings and this specification.

[0017]

[0017] The foregoing has outlined rather broadly the features and technical advantages of the embodiments of the present disclosure in order that the following "Description of the Preferred Embodiments" may be better understood. Additional features and advantages are described below. The concepts and examples disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes of the present disclosure. Such equivalent structures do not depart from the scope of the appended claims. The properties of the disclosed concepts, both their construction and method of operation, together with associated advantages, will be better understood from the following description when considered in conjunction with the accompanying figures. Each of the figures is provided for the purpose of illustration and explanation, and not as a definition of the limits of the claims. [Brief description of the drawings]

[0018]

[0018] In order to allow the features of the present disclosure to be understood in detail, a specific description may be made by referring to embodiments, some of which are shown in the accompanying drawings. However, since this description may admit of other equally effective embodiments, it should be noted that the accompanying drawings show only some embodiments of the present disclosure and therefore should not be considered as limiting its scope. The same reference numbers in different drawings may identify the same or similar elements.

[0019] [Figure 1] 1 is a block diagram conceptually illustrating one embodiment of a wireless communication network in accordance with various aspects of the present disclosure.

[0020] [Diagram 2] 1 is a block diagram conceptually illustrating one embodiment of a base station in communication with user equipment (UE) in a wireless communication network, in accordance with various aspects of the present disclosure.

[0021] [Diagram 3] 1 illustrates an example implementation of designing a neural network using a system-on-a-chip (SOC) that includes a general-purpose processor, according to some aspects of the present disclosure.

[0022] [Figure 4A] FIG. 1 illustrates a neural network according to an aspect of the present disclosure. [Figure 4B] FIG. 1 illustrates a neural network according to an aspect of the present disclosure. [Figure 4C] FIG. 1 illustrates a neural network according to an aspect of the present disclosure.

[0023] [Figure 4D] FIG. 1 illustrates an example deep convolutional network (DCN), in accordance with aspects of the present disclosure.

[0024] [Diagram 5] FIG. 1 is a block diagram illustrating an example deep convolutional network (DCN), in accordance with aspects of the present disclosure.

[0025] [Figure 6] FIG. 1 is a block diagram showing a conventional architecture for extracting features of a channel sequence.

[0026] [Figure 7] FIG. 1 illustrates an example architecture for model-based channel feature extraction, according to aspects of the present disclosure.

[0027] [Figure 8] 1 is a flow diagram illustrating an example process performed, for example, by a receiving device, in accordance with various aspects of the present disclosure.

[0028] [Figure 9] 4 is a flow diagram illustrating an example process performed, for example, by a transmitting device, in accordance with various aspects of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019]

[0029] Various aspects of the present disclosure are described more fully below with reference to the accompanying drawings. However, the present disclosure may 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 this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art. Based on the teachings, those skilled in the art should understand that the scope of the present disclosure is intended to encompass any aspect of the present disclosure, whether implemented independently of or in combination with any other aspect of the present disclosure. For example, an apparatus may be implemented or a method may be practiced using any number of the aspects described. In addition, the scope of the present disclosure is intended to encompass such an apparatus or method that is practiced using other structures, functions, or structures and functions in addition to or other than the various aspects of the present disclosure described. It should be understood that any aspect of the present disclosure disclosed may be embodied by one or more elements of a claim.

[0020]

[0030] Several aspects of a telecommunications system are now presented with reference to various devices and techniques that are described in the following detailed description and illustrated in the accompanying drawings by various blocks, modules, components, circuits, steps, processes, algorithms, etc. (collectively referred to as "elements"). These elements may be implemented using hardware, software, or a combination thereof. Whether such elements are implemented as hardware or software depends on the particular application and design constraints imposed on the overall system.

[0021]

[0031] Although aspects may be described using terminology commonly associated with 5G and beyond wireless technologies, it should be noted that aspects of the present disclosure may be applied in other generation-based communication systems such as and including 3G and / or 4G technologies.

[0022]

[0032] Observed wireless propagation channels (e.g., downlink, uplink, or sidelink) use high-dimensional data to describe the channel, e.g., with a three-dimensional tensor across Tx antennas, Rx antennas, and subcarriers. It is desirable to extract a low-dimensional latent representation of the channel underlying the high-dimensional channel. For example, extracting a compact representation can be useful for applications such as channel prediction, channel inference, channel compression, and channel generation. Channel prediction can be useful for predicting fading channels, as well as reference signal received power (RSRP). In addition, the compact representation can be useful for generating inferences about unknown channels. For example, a channel can be observed for a first set of antennas, which can provide inferences of the channel for a second set of antennas, beams, frequencies, or locations. Furthermore, channel modeling can be improved for synthetic channel generation, augmentation of measured channels, and differentiable channel modeling for link and system simulations.

[0023]

[0033] Conventional techniques can utilize artificial neural networks to generate compact representations. In such conventional techniques, the artificial neural network can be implemented as an autoencoder to map channels to latent representations and vice versa. Doing so is a nonlinear process that can be expressed as follows:

[0024]

number

[0025]

[0034] To address these and other deficiencies, aspects of the present disclosure are directed to channel feature extraction via physical propagation model-based neural networks.

[0026]

[0035] FIG. 1 illustrates a network 100 in which aspects of the disclosure can be practiced. The network 100 may be a 5G network or a NR network, or some other wireless network, such as an LTE network. The wireless network 100 may include several BSs 110 (shown 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 (NB), access point, transmit reception point (TRP), etc. Each BS may provide communication coverage for a particular geographic area. In 3GPP, the term "cell" may refer to a coverage area of ​​a BS and / or a BS subsystem serving this coverage area, depending on the context in which the term is used.

[0027]

[0036] 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 a service subscription. A pico cell may cover a relatively small geographic area and may allow unrestricted access by UEs with a service subscription. A femto cell may cover a relatively small geographic area (e.g., a home) and may allow restricted access by UEs that have an association 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 the embodiment shown in Figure 1, BS 110a may be a macro BS for macro cell 102a, BS 110b may be a pico BS for pico cell 102b, and BS 110c may be a femto BS for femto cell 102c. A BS may support one or multiple (e.g., three) cells. The terms "eNB", "base station", "NR BS", "gNB", "TRP", "AP", "Node B", "5G NB", and "cell" may be used interchangeably.

[0028]

[0037] In some aspects, the cells may not necessarily be fixed and the geographic area of ​​the cells may move according to the location of the mobile BS. In some aspects, the BSs may be interconnected to each other and / or to one or more other BSs or network nodes (not shown) in wireless network 100 through various types of backhaul interfaces, such as direct physical connections, virtual networks, etc., using any suitable transport network.

[0029]

[0038] The wireless network 100 may also include relay stations. 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. In the embodiment shown in FIG. 1, 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, etc.

[0030]

[0039] 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 susceptibility to interference within the wireless network 100. For example, a macro BS may have a high transmit power level (e.g., 5-40 watts), while a pico BS, femto BS, and relay BS may have a lower transmit power level (e.g., 0.1-2 watts).

[0031]

[0040] A network controller 130 may couple to a set of BSs and provide coordination and control for these BSs. The network controller 130 may communicate with the BSs via a backhaul. The BSs may also communicate with each other, e.g., directly or indirectly via wireless or wired backhaul.

[0032]

[0041] The UEs 120 (e.g., 120a, 120b, 120c) may be distributed throughout the wireless network 100, and each UE may be fixed 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 telephone (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, a camera, a gaming device, a netbook, a smartbook, an ultrabook, a medical device or equipment, a biometric sensor / device, a wearable device (smart watch, smart clothing, smart glasses, smart wristband, smart jewelry (e.g., smart ring, smart bracelet)), an entertainment device (e.g., music or video device, or satellite radio), a vehicular component or sensor, a smart meter / sensor, industrial manufacturing equipment, a global positioning system device, or any other suitable device configured to communicate over a wireless or wired medium.

[0033]

[0042] Some UEs may be considered as machine-type communications (MTC) UEs or evolved or enhanced machine-type communications (eMTC) UEs. MTC UEs and eMTC UEs include, for example, a robot, a remote device, a sensor, a meter, a monitor, a location tag, etc. that may communicate with a base station, another device (e.g., a remote device), or some other entity. A wireless node may provide connectivity for or to a network (e.g., a wide area network such as the Internet or a cellular network), for example, via a wired or wireless communication link. Some UEs may be considered as Internet-of-Things (IoT) devices and / or may be implemented as narrowband internet of things (NB-IoT) devices. Some UEs may be considered as customer premises equipment (CPE). The UE 120 may be included within a housing that houses components of the UE 120, such as a processor component, a memory component, etc.

[0034]

[0043] In general, any number of wireless networks may be deployed in a given geographic area. Each wireless network may support a particular RAT and may operate on one or more frequencies. The RAT may also be referred to as a radio technology, an air interface, etc. The frequencies may also be referred to as a carrier, a frequency channel, etc. Each frequency may support a single RAT in a given geographic area to avoid interference between wireless networks of different RATs. In some cases, NR networks or 5G RAT networks may be deployed.

[0035]

[0044] In some aspects, two or more UEs 120 (e.g., shown as UE 120a and UE 120e) may communicate directly (e.g., without using the base station 110 as an intermediary to communicate with each other) using one or more sidelink channels. For example, the UEs 120 may communicate using peer-to-peer (P2P) communications, device-to-device (D2D) communications, vehicle-to-everything (V2X) protocols (which may include, e.g., vehicle-to-vehicle (V2V) protocols, vehicle-to-infrastructure (V2I) protocols, etc.), mesh networks, etc. In this case, the UEs 120 may perform scheduling operations, resource selection operations, and / or other operations described elsewhere as being performed by the base station 110. For example, the base station 110 may configure the UE 120 via downlink control information (DCI), radio resource control (RRC) signaling, media access control-control element (MAC-CE), or via system information (e.g., a system information block (SIB)).

[0036]

[0045] 1 , in some aspects, a UE, such as UE 120, can extract features of the channel sequence. UE 120 can include a feature extraction component 140 configured to extract features of the channel sequence based on neural network processing. A base station, such as base station 110, can also extract features of the channel sequence. Base station 110 can include a feature extraction component 150 configured to extract features of the channel sequence based on neural network processing.

[0037]

[0046] As noted above, Figure 1 is provided as an example only, and other examples may differ from those described with respect to Figure 1.

[0038]

[0047] 2 shows a block diagram of a design 200 of a base station 110 and a UE 120, which may be one of the base stations and one of the UEs in FIG. 1. Base 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.

[0039]

[0048] At the base station 110, a transmit processor 220 may receive data for one or more UEs from a data source 212, select one or more modulation and coding schemes (MCS) for each UE based at least in part on channel quality indicators (CQI) received from the UE, process (e.g., code and modulate) the data for each UE based at least in part on the MCS(es) selected for the UE, and provide data symbols to all UEs. Reducing the MCS reduces throughput but increases reliability of transmission. The transmit processor 220 may also process system information (e.g., for semi-static resource partitioning information (SRPI), etc.) and control information (e.g., CQI requests, grants, higher layer signaling, etc.) and provide overhead symbols and control symbols. The transmit processor 220 may also generate reference symbols for a reference signal (e.g., a cell-specific reference signal (CRS)) and synchronization signals (e.g., a primary synchronization signal (PSS) and a secondary synchronization signal (SSS)). A transmit (TX) multiple-input multiple-output (MIMO) processor 230 may perform spatial processing (e.g., precoding) on ​​the data symbols, control symbols, overhead symbols, and / or reference symbols, if applicable, and may provide T output symbol streams to T modulators (MOD) 232a through 232t. Each modulator 232 may process a respective output symbol stream (e.g., for OFDM, etc.) 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.T downlink signals from modulators 232a through 232t may be transmitted via T antennas 234a through 234t, respectively. According to various aspects described in more detail below, synchronization signals can be generated using position coding to convey additional information.

[0040]

[0049] At UE 120, antennas 252a-252r may receive downlink signals from base station 110 and / or other base stations and may provide received signals to demodulators (DEMODs) 254a-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 the received symbols from all R demodulators 254a-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 and provide decoded data for UE 120 to a data sink 260 and provide decoded control and system information to controller / processor 280. The channel processor may determine a reference signal received power (RSRP), a received signal strength indicator (RSSI), a reference signal received quality (RSRQ), a channel quality indicator (CQI), etc. In some aspects, one or more components of the UE 120 may be included in a housing.

[0041]

[0050] On the uplink, at the UE 120, a transmit processor 264 may receive and process data from a data source 262 and control information (e.g., for reports comprising RSRP, RSSI, RSRQ, CQI, etc.) from a controller / processor 280. The transmit processor 264 may also generate reference symbols for one or more reference signals. The symbols from the transmit processor 264 may be precoded by a TX MIMO processor 266 if applicable, further processed by modulators 254a-254r (e.g., for DFT-s-OFDM, CP-OFDM, etc.), and transmitted to the base station 110. At the base station 110, uplink signals from the UE 120 and other UEs may be received by antennas 234, processed by demodulator 254, detected by a MIMO detector 236 if applicable, and further processed by receive processor 238 to obtain decoded data and control information transmitted by the UE 120. The receive processor 238 may provide the decoded data to a data sink 239 and the decoded control information to the controller / processor 240. The base station 110 may include a communication unit 244 and may 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.

[0042]

[0051] The controller / processor 240 of the base station 110, the controller / processor 280 of the UE 120, and / or any other component(s) of FIG. 2 may perform one or more techniques associated with machine learning to optimize an iterative process using an artificial neural network, 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 any other component(s) of FIG. 2 may perform or direct the operation of, for example, the processes of FIGS. 6-8 and / or other processes as described. The memories 242 and 282 may store data and program codes for the base station 110 and the UE 120, respectively. The scheduler 246 may schedule UEs for data transmission on the downlink and / or uplink.

[0043]

[0052] In some aspects, the UE 120, base station 110, or other disclosed network may include a means for receiving, a means for capturing, a means for training, and a means for applying. Such means may include one or more components of the UE 120 or base station 110 described with respect to FIG.

[0044]

[0053] As noted above, Figure 2 is provided as one example only, other examples may differ from those described with respect to Figure 2.

[0045]

[0054] 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 UE handsets, customer premises equipment (CPE), vehicles, Internet of Things (IoT) devices, etc. Examples of different types of applications include ultra-reliable low-latency communications (URLLC) applications, massive machine-type communications (mMTC) applications, enhanced mobile broadband (eMBB) applications, vehicle-to-anything (V2X) applications, etc. Furthermore, in some cases, a single device can support different applications or services simultaneously.

[0046]

[0055] 3 illustrates an exemplary implementation of a system-on-chip (SOC) 300 that may include a central processing unit (CPU) 302 or a multi-core CPU configured to generate a reconstruction of a channel sequence for a wireless propagation channel according to some aspects of the disclosure. The SOC 300 may be included in a base station 110 or a UE 120. Variables (e.g., neural signals and synaptic weights), system parameters associated with a computation device (e.g., a neural network with weights), delays, frequency bin information, and task information may be stored in a memory block associated with a neural processing unit (NPU) 308, a memory block associated with the CPU 302, a memory block associated with a graphics processing unit (GPU) 304, a memory block associated with a digital signal processor (DSP) 306, a memory block 318, or may be distributed across multiple blocks. Instructions executed in the CPU 302 may be loaded from a program memory associated with the CPU 302 or may be loaded from the memory block 318.

[0047]

[0056] The SOC 300 may also include a connectivity block 310, which may include a GPU 304, a DSP 306, fifth generation (5G) connectivity, fourth generation long term evolution (4G LTE) connectivity, Wi-Fi connectivity, USB connectivity, Bluetooth connectivity, etc., as well as additional processing blocks adapted to specific functions, such as a multimedia processor 312, which may detect and recognize gestures. In one implementation, the NPU is implemented in the CPU, DSP, and / or GPU. The SOC 300 may also include a navigation module 320, which may include a sensor processor 314, image signal processors (ISP) 316, and / or a global positioning system.

[0048]

[0057] The SOC 300 may be based on an ARM instruction set. In one aspect of the disclosure, the instructions loaded into the general-purpose processor 302 may comprise code for receiving, from a transmitting device, a latent representation of a channel sequence for a wireless signal. The instructions loaded into the general-purpose processor 302 may also comprise code for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal.

[0049]

[0058] Deep learning architectures may perform object recognition tasks by learning to represent inputs at successively higher levels of abstraction within each layer, thereby building useful feature representations of the input data. In this way, deep learning addresses a major bottleneck of traditional machine learning. Prior to the advent of deep learning, machine learning approaches to object recognition problems may have relied heavily on human-designed features, possibly in combination with shallow classifiers. A shallow classifier may be, for example, a two-class linear classifier that can compare a weighted sum of feature vector components to a threshold to predict which class an input belongs to. Human-designed features may be templates or kernels that are adapted to a particular problem domain by an engineer with domain expertise. In contrast, deep learning architectures may learn, but through training, to represent features that are similar to those that a human engineer could design. Furthermore, deep networks may learn to represent and recognize new types of features that humans may not have thought of.

[0050]

[0059] A deep learning architecture may learn a hierarchy of features. When presented with visual data, for example, a first layer may learn to recognize relatively simple features such as edges in the input stream. In another example, when presented with auditory data, the first layer may learn to recognize spectral power at specific frequencies. A second layer, taking as input the output of the first layer, may learn to recognize combinations of features such as simple shapes in the case of visual data, or combinations of sounds in the case of auditory data. For example, higher layers may learn to represent complex shapes in visual data or words in auditory data. Even higher layers may learn to recognize common visual objects or spoken phrases.

[0051]

[0060] Deep learning architectures may work particularly well when applied to problems that have a natural hierarchical structure. For example, classification of electric vehicles may benefit from first learning to recognize wheels, windshields, and other features. These features may be combined in different ways at higher layers to recognize cars, trucks, and planes.

[0052]

[0061] Neural networks may be designed with various connectivity patterns. In feedforward networks, each neuron in a given layer communicates with neurons in a higher layer, so that information is passed from lower layers to higher layers. As described above, hierarchical representations may be constructed in successive layers of a feedforward network. Neural networks may also have recurrent or feedback (also called top-down) connections. In recurrent connections, the output from a neuron in a given layer may be transmitted to another neuron in the same layer. Recurrent architectures may be useful in recognizing patterns across two or more of the input data chunks delivered in a sequence to the neural network. Connections from neurons in a given layer to neurons in a lower layer are called feedback (or top-down) connections. Networks with many feedback connections may be useful when recognition of high-level concepts can help distinguish certain low-level features of the input.

[0053]

[0062] The connections between layers of a neural network may be fully connected or locally connected. FIG. 4A illustrates an example of a fully connected neural network 402. In the fully connected neural network 402, a neuron in a first layer may transmit its output to every neuron in a second layer, so that each neuron in the second layer receives input from every neuron in the first layer. FIG. 4B illustrates an example of a locally connected neural network 404. In the locally connected neural network 404, a neuron in a first layer may be connected to a limited number of neurons in the second layer. More generally, the locally connected layers of the locally connected neural network 404 may be configured such that each neuron in a layer has the same or similar connectivity pattern, but with connection strengths that may have different values ​​(e.g., 410, 412, 414, and 416). Because higher layer neurons in a given region may receive inputs that are tuned through training to the properties of a constrained subset of all inputs to the network, the connectivity patterns of local connections may give rise to spatially distinct receptive fields within the higher layers.

[0054]

[0063] One example of a locally connected neural network is a convolutional neural network. Figure 4C shows an example of a convolutional neural network 406. The convolutional neural network 406 can be configured (e.g., 408) such that the connection strengths associated with the inputs to each neuron in the second layer are shared. Convolutional neural networks may be suitable for problems where the spatial location of the inputs is meaningful.

[0055]

[0064] One type of convolutional neural network is the deep convolutional network (DCN). Figure 4D shows a detailed embodiment of a DCN 400 designed to recognize visual features from an image 426 input from an image capture device 430, such as an on-board camera. The DCN 400 of this embodiment can be trained to identify traffic signs and numbers on traffic signs. Of course, the DCN 400 can be trained for other tasks, such as recognizing lane markings or identifying traffic signals.

[0056]

[0065] The DCN 400 can be trained using supervised learning. During training, the DCN 400 can be presented with an image, such as an image 426 of a speed limit sign, and can then compute a forward pass to generate the output 422. The DCN 400 can include a feature extraction section and a classification section. Upon receiving the image 426, the convolution layer 432 can apply a convolution kernel (not shown) to the image 426 to generate the first set of feature maps 418. As an example, the convolution kernel for the convolution layer 432 can be a 5×5 kernel that generates a 28×28 feature map. In this example, four different feature maps are generated in the first set of feature maps 418, so four different convolution kernels were applied to the image 426 in the convolution layer 432. The convolution kernel may also be referred to as a filter or a convolution filter.

[0057]

[0066] The first set of feature maps 418 may be subsampled by a max pooling layer (not shown) to generate a second set of feature maps 420. The max pooling layer reduces the size of the first set of feature maps 418. That is, the size of the second set of feature maps 420, such as 14×14, is smaller than the size of the first set of feature maps 418, such as 28×28. The reduced size provides similar information to subsequent layers while reducing memory consumption. The second set of feature maps 420 may be further convolved through one or more subsequent convolutional layers (not shown) to generate one or more subsequent sets of feature maps (not shown).

[0058]

[0067] In the example of Figure 4D, the second set of feature maps 420 are convolved to generate a first feature vector 424. Furthermore, the first feature vector 424 is further convolved to generate a second feature vector 428. Each feature in the second feature vector 428 may include a number corresponding to a possible feature of the image 426, such as "sign", "60", and "100". A softmax function (not shown) may convert the numbers in the second feature vector 428 into probabilities. Thus, the output 422 of the DCN 400 is the probability that the image 426 contains one or more features.

[0059]

[0068] In this example, the probability in output 422 for "sign" and "60" is higher than the probability for other of the outputs 422, such as "30", "40", "50", "70", "80", "90", and "100". Prior to training, the output 422 generated by DCN 400 may be inaccurate. Thus, an error may be calculated between the output 422 and a target output. The target output is the ground truth of image 426 (e.g., "sign" and "60"). The weights of DCN 400 may then be adjusted so that the output 422 of DCN 400 is more closely aligned with the target output.

[0060]

[0069] To adjust the weights, the learning algorithm or process may calculate a gradient vector for the weights. The gradient may indicate the amount by which the error would increase or decrease if the weights were adjusted. In the top layer, the gradient may correspond directly to the values ​​of the weights connecting the activated neurons in the penultimate layer to the neurons in the output layer. In lower layers, the gradient may depend on the values ​​of the weights and the calculated error gradients of the upper layers. The weights may then be adjusted to reduce the error. This method of adjusting weights is sometimes called "backpropagation" because it involves a "backward pass" through the neural network.

[0061]

[0070] In practice, the error gradient of the weights may be calculated over a small number of examples such that the calculated gradient approximates the true error gradient. This approximation method is sometimes called stochastic gradient descent. Stochastic gradient descent may be iterated until the achievable error rate of the entire system stops decreasing or until the error rate reaches a target level. After training, the DCN may be presented with a new image (e.g., a speed limit sign in image 426) and a forward pass through the network may result in an output 422, which may be considered the inference or prediction of the DCN.

[0062]

[0071] Deep belief networks (DBNs) are probabilistic models with multiple layers of hidden nodes. DBNs may be used to extract hierarchical representations of a training dataset. DBNs may be obtained by stacking layers of Restricted Boltzmann Machines (RBMs). RBMs are a type of artificial neural network that can learn probability distributions over a set of inputs. RBMs are frequently used in unsupervised learning because they can learn probability distributions without information about the class into which each input should be categorized. Using a hybrid unsupervised-supervised paradigm, the lower RBM of the DBN can be trained in an unsupervised manner and can function as a feature extractor, and the upper RBM can be trained in a supervised manner (on the joint distribution of inputs from previous layers and the target class) and can function as a classifier.

[0063]

[0072] Deep convolutional networks (DCNs) are networks of convolutional networks, composed of additional pooling and normalization layers. DCNs have achieved state-of-the-art performance for many tasks. DCNs can be trained using supervised learning, where both the input and output targets are known for a large number of examples and are used to modify the network weights by using gradient descent.

[0064]

[0073] The DCN may be a feedforward network. In addition, as described above, connections from neurons in a first layer of the DCN to groups of neurons in the next higher layer are shared across neurons in the first layer. The feedforward and shared connections of the DCN can be exploited for high speed processing. The computational burden of the DCN may be much smaller than that of a similarly sized neural network with, for example, recurrent or feedback connections.

[0065]

[0074] The processing of each layer of the convolutional network may be considered as a spatially invariant template or basis projection. If the input is initially decomposed into multiple channels, such as the red, green, and blue channels of a color image, the convolutional network trained on that input may be considered as three-dimensional, with two spatial dimensions along the axes of the image and a third dimension capturing color information. The output of the convolutional connections may be considered to form a feature map in the subsequent layer, with each element of the feature map (e.g., 220) receiving input from a range of neurons in the previous layer (e.g., feature map 218) and from each of multiple channels. The values ​​in the feature map may be further processed using nonlinearities such as rectification, max(0,x), etc. Values ​​from neighboring neurons may be further pooled, corresponding to downsampling, to provide additional local invariance and dimensionality reduction. Normalization, corresponding to whitening, may also be applied through lateral inhibition between neurons in the feature map.

[0066]

[0075] The performance of deep learning architectures can improve as more labeled data points become available or as computational power increases. Modern deep neural networks are routinely trained with computational resources thousands of times greater than those available to a typical researcher only 15 years ago. New architectures and training paradigms can further improve deep learning performance. Rectified linear units may reduce the training problem known as vanishing gradients. New training techniques may reduce overfitting and therefore allow larger models to achieve better generalization. Encapsulation techniques may extract data within a given receptive field, further improving overall performance.

[0067]

[0076] 5 is a block diagram illustrating a deep convolutional network 550. The deep convolutional network 550 can include multiple different types of layers based on connectivity and weight sharing. As shown in FIG. 5, the deep convolutional network 550 includes convolution blocks 554A, 554B. Each of the convolution blocks 554A, 554B can be configured with a convolution layer (CONV) 356, a normalization layer (LNorm) 558, and a max pooling layer (MAX POOL) 560.

[0068]

[0077] The convolution layer 556 may include one or more convolution filters that may be applied to the input data to generate feature maps. Although only two of the convolution blocks 554A, 554B are shown, the disclosure is not so limited and instead any number of convolution blocks 554A, 554B may be included in the deep convolution network 550 according to design preferences. The normalization layer 558 may normalize the output of the convolution filters. For example, the normalization layer 558 may provide whitening or lateral inhibition. The max pooling layer 560 may provide downsampling aggregation across the space for local invariance and dimensionality reduction.

[0069]

[0078] For example, a parallel filter bank of a deep convolutional network can be loaded onto the CPU 302 or GPU 304 of the SOC 300 to achieve high performance and low power consumption. In alternative embodiments, the parallel filter bank can be loaded onto the DSP 306 or ISP 316 of the SOC 300. In addition, the deep convolutional network 550 can access other processing blocks that may be present on the SOC 300, such as the sensor processor 314 and the navigation module 320, which are dedicated to sensors and navigation, respectively.

[0070]

[0079] The deep convolutional network 550 may also include one or more fully connected layers 562 (FC1 and FC2). The deep convolutional network 550 may further include a logistic regression (LR) layer 564. Between each layer 556, 558, 560, 562, 564 of the deep convolutional network 550, there are weights (not shown) that are to be updated. The output of each of the layers (e.g., 556, 558, 560, 562, 564) can serve as an input of a subsequent one of the layers (e.g., 556, 558, 560, 562, 564) in the deep convolutional network 550 to learn a hierarchical feature representation from the input data 552 (e.g., image, audio, video, sensor data, and / or other input data) provided at the first one of the convolution blocks 554A. The output of the deep convolutional network 550 is a classification score 566 for the input data 552. The classification score 566 may be a set of probabilities, each probability being a probability for the input data including features from the set of features.

[0071]

[0080] As noted above, Figures 3-5 are provided as examples, and other embodiments may differ from those described with respect to Figures 3-5.

[0072]

[0081] 6 is a block diagram illustrating a conventional architecture 600 for channel feature extraction. Referring to FIG 6, the exemplary architecture 600 may be an autoencoder including an encoder 602 (e.g., an inference neural network) and a decoder 604 (e.g., a generative neural network).

[0073]

[0082] A channel sequence (e.g., a high-dimensional representation) of a wireless propagation channel may be received as an input to an encoder 602. The encoder 602 may process the channel sequence to generate a latent channel representation (e.g., a low-dimensional representation) of the wireless propagation channel. The latent channel representation may be provided to a decoder 604. The decoder 604 processes the latent channel representation to generate a reconstruction (high-dimensional representation) of the channel sequence. This conventional architecture 600 may beneficially utilize large amounts of unlabeled channel data. However, as described, the conventional architecture 600 lacks interpretability and control of the latent vectors.

[0074]

[0083] 7 illustrates an example architecture 700 for model-based channel feature extraction, according to an aspect of the present disclosure. As shown in FIG. 7, the architecture 700 includes an encoder 702 (e.g., similar to the encoder 602) and a decoder 704. The encoder 702 may include multiple layers of convolutional filters (e.g., the convolutional layers 556 shown in FIG. 5), fully connected layers, or any other type of neural network layers. The encoder 702 receives as input at a first layer a channel sequence (high-dimensional representation) H t The channel sequence H t can be processed through successive layers of convolutional filters in the encoder 702. Each layer of the encoder 702 processes the channel sequence H t The features of the latent channel representation z t can be generated.

[0075]

[0084] The exemplary architecture 700 includes a decoder 704. Rather than using a neural network as the encoder 702, a physical propagation channel model between the transmitter (e.g., base station 110) and the receiver (e.g., UE 120) can decode the latent representation. That is, a physics-based relationship between the latent representation and the channel can be applied to generate a reconstruction of the input channel sequence. In one embodiment, the physical propagation channel model is a non-line of sight (NLOS) impulse response, as presented in 3GPP TR 38.901 and given by:

[0076]

number

[0077]

number

[0078]

number

[0079]

number

[0080]

number

[0081]

number

[0082]

number

[0083]

number

[0084]

number

[0085]

number

[0086]

number

[0087]

number

[0088]

number

[0089]

[0085] By using a physical propagation channel model-based decoder (e.g., 704) to decode the latent representation and generate a reconstruction of the input, the latent representation can be made to conform to physical parameters, such as delay and angle of arrival. This can also allow for more efficient training, since only the encoder 702 is trained as opposed to also training the decoder 704. In addition, the decoder 704 can decode the latent representation z t The physical propagation channel model-based decoder 704 enables interpretability since parts of z can be mapped to aspects of the physical propagation channel. For example, t The first element of may be defined to correspond to the angle of arrival. Furthermore, the example architecture 700 may enable dynamic modeling (e.g., Kalman filtering, ray tracing) based on a physical propagation channel model. t is interpretable, which can improve predictability.

[0090] In some aspects, the known physical model is not completely known. That is, some parameter values ​​of the model, such as the receiver (Rx) array response, may be unknown. In these aspects, the network can estimate the unknown features. This approach is sometimes called a hybrid approach.

[0091]

[0087] In some aspects, the physical propagation channel model may be represented as a sum of multipath components (MPC). Each multipath component may have, for example, a delay and a complex magnitude. In addition, each multipath component may have an associated transmitter (Tx) array response and a receiver (Rx) array response. In some aspects, the Tx array response and the Rx array response may be represented as a learnable function of the angle response (e.g., as a neural network layer). Furthermore, the physical propagation channel model may also include the impulse response of the Tx chain and the Rx chain. In some embodiments, the physical propagation channel model may be given by the following equation: H[k]=H RX [k]H OTA [k]H TX [k] (3)

[0092]

number

[0093]

number

[0094]

number

[0095] In some aspects, the exemplary architecture 700 can be trained via unsupervised learning based on measured channel data. Once trained, the encoder of the NN can be used to extract latent features from a given channel. The decoder of the NN can generate new channel samples and / or channel sample sequences from the latent channel features. The trained encoder / decoder neural network pair can also be configured and trained for channel state feedback between a UE and a base station (e.g., a gNodeB (gNB)).

[0096] In some aspects, the example architecture 700 can perform bandwidth stitching. For example, given two frequency bands separated by a guard band, the unknown channel in the guard band can be estimated to form a continuous channel response.

[0097] In some aspects, a transmitting device (e.g., base station 110) jointly trains an encoder and decoder pair such that a cascade of the encoder and decoder reconstructs the original channel sequence given as input to the encoder. However, only the encoder may be used and the decoder discarded. That is, the transmitting device may train a decoder only for the purpose of training the encoder, not for use in reconstructing the encoder input. Instead, the receiving device may independently train a decoder (e.g., a decoder 704 that may use a physical propagation channel model) that may be different from the decoder that the transmitting device has internally. In this way, the transmitting device (e.g., base station 110) may obscure its encoder design and, in some aspects, avoid revealing its encoder design to the receiving device (e.g., UE 120). Similarly, the receiving device (e.g., UE 120) may obscure its decoder design and, in some aspects, avoid disclosing its decoder design to the transmitting device (e.g., base station 110). Furthermore, in some aspects, neither a transmitting device (eg, a base station 110) nor a receiving device (eg, a UE 120) forces the other device to use a particular encoder or decoder.

[0098]

[0091] Thus, aspects of the present disclosure provide more efficient and interpretable learning compared to conventional non-physical propagation channel based approaches. Furthermore, the improved interpretability allows for less complex specification of neural network pairs across gNBs and UEs. Because the latent representation has interpretable meaning, the encoder (e.g., 702) may be left as a standalone implementation. This is in contrast to conventional autoencoder-based neural networks that use joint training and specification of compatible encoder-decoder pairs.

[0099]

[0092] Figure 8 is a flow diagram illustrating an example process 800 performed, for example, by a processor, according to various aspects of the disclosure. At block 802, the process 800 receives a latent representation of channel characteristics for a wireless signal from a transmitting device. The transmitting device can convert the channel sequence into the latent representation using an encoder (e.g., encoder 702 of Figure 7). The latent representation can include, for example, one or more of a set of delay, complex amplitude, angle of departure, and angle of arrival. In some aspects, the latent representation can be transmitted to a receiving device as channel condition feedback.

[0100] At block 804, the process 800 applies the physical propagation channel model to the latent representation via a decoder to reconstruct channel characteristics for the wireless signal. The receiving device can use a decoder (e.g., decoder 704 of FIG. 7) to reconstruct the channel characteristics from the latent representation.

[0101] In some aspects, the physical propagation channel model may include a sum of multipath components. Each of the multipath components may include, for example, a delay and a complex amplitude. In other embodiments, each multipath component may include a transmitter (Tx) array response and a receiver (Rx) array response. Furthermore, the Tx array response and the Rx array response may be represented as learnable functions of the azimuth and zenith angles of departure and the azimuth and zenith angles of arrival, respectively.

[0102] In some aspects, additional effects on the wireless signal not captured by the physical channel model can be captured via additional learnable neural network layers. For example, the additional effects can include receiver (Rx) chain impulse response and transmitter (Tx) chain impulse response.

[0103] In some aspects, the encoder may be a first neural network (e.g., encoder 702 in FIG. 7) and the decoder is a second neural network (e.g., decoder 704 in FIG. 7) based on a physical propagation channel model. The decoder and encoder can be trained jointly or independently.

[0104] 9 is a flow diagram illustrating an example process 900, performed, for example, by a processor, according to various aspects of the present disclosure. At block 902, the process 900 receives an input comprising channel characteristics for a wireless signal. As shown in FIG. 7, the encoder 702 receives as input a channel sequence (high-dimensional representation) H t Receive.

[0105] At block 904, the process 900 processes the channel characteristics via an encoder to generate a latent representation of the channel characteristics for the wireless signal, the latent representation including at least one element that maps to a parameter of the physical propagation channel. The transmitting device can convert the channel characteristics to the latent representation using an encoder (e.g., encoder 702 of FIG. 7). The latent representation can include, for example, one or more of a set of delay, complex amplitude, angle of departure, and angle of arrival. In some aspects, the latent representation can be transmitted to the receiving device as channel condition feedback.

[0106]

[0099] At block 904, the process 900 transmits the latent representation to a receiving device. As shown in Figure 7, the encoder 702 transmits the latent representation of the channel features to a decoder 704.

[0107]

[0100] Implementation examples are provided in the following numbered clauses. 1. A method of wireless communication by a receiving device, comprising: receiving from a transmitting device a latent representation of a channel sequence for a wireless signal; applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal; A method comprising: 2. The method of claim 1, wherein the physical propagation channel model comprises a sum of multipath components. 3. The method of claim 1 or 2, wherein each of the multipath components has a delay and a complex amplitude. 4. The method of any one of clauses 1 to 3, wherein each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response. 5. The method of any one of clauses 1 to 4, wherein the Tx array response and the Rx array response are represented as learnable functions of azimuth and zenith departure angles and azimuth and zenith arrival angles, respectively. 6. The method of any one of clauses 1 to 5, further comprising capturing additional effects on the wireless signal not captured by the physical propagation channel model via additional trainable neural network layers. 7. The method of any one of clauses 1 to 6, wherein the additional effects include a receiver (Rx) chain impulse response and a transmitter (Tx) chain impulse response. 8. The method of any one of clauses 1 to 7, wherein the latent representation is transmitted to a receiving device as channel state feedback. 9. The method of any one of clauses 1 to 8, wherein the reconstructed channel sequence comprises a continuous channel response across multiple frequency bands separated by guard bands. 10. The method of any one of clauses 1 to 9, wherein the latent representation comprises one or more of the following sets: delay, complex amplitude, launch angle, and arrival angle. 11. The method of any one of clauses 1 to 10, wherein the latent representation comprises an encoded channel sequence, and the receiving device reconstructs the channel sequence from the latent representation using a decoder. 12. The method of any one of clauses 1 to 11, wherein the decoder is a neural network based on a physical propagation channel model. 13. The method of any one of clauses 1 to 12, further comprising training a decoder at the receiving device independent of the encoder at the transmitting device. 14. A method of wireless communication by a transmitting device, comprising: receiving an input comprising a channel sequence for a wireless signal via an encoder; processing, via an encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal, the latent representation including at least one element that maps to parameters of a physical propagation channel model; transmitting the latent representation to a receiving device; A method comprising: 15. The method of claim 14, wherein the physical propagation channel model comprises a sum of multipath components. 16. The method of claim 14 or 15, wherein each of the multipath components has a delay and a complex amplitude. 17. The method of any one of clauses 14 to 16, wherein each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response. 18. A method according to any one of clauses 14 to 17, wherein the Tx array response and the Rx array response are represented as learnable functions of azimuth and zenith departure angles and azimuth and zenith arrival angles, respectively. 19. The method of any one of clauses 14 to 18, wherein the latent representation is transmitted to a receiving device as channel state feedback. 20. A method according to any one of clauses 14 to 19, wherein the encoder comprises a neural network. 21. The method of any one of clauses 14 to 20, wherein the transmitting device jointly trains an encoder-decoder pair such that a cascade of an encoder and a second decoder of the encoder-decoder pair reconstructs the channel sequence input at the encoder. 22. An apparatus for wireless communication by a receiving device, comprising: Memory, at least one processor coupled to the memory; at least one processor receiving from a transmitting device a latent representation of a channel sequence for the wireless signal; applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal; It is configured as follows: Device. 23. The apparatus of clause 22, wherein the physical propagation channel model comprises a sum of multipath components. 24. The apparatus of claim 22 or 23, wherein each of the multipath components has a delay and a complex amplitude. 25. The apparatus of any one of clauses 22 to 24, wherein each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response. 26. The apparatus of any one of clauses 22 to 25, wherein the at least one processor is further configured to represent the Tx array response and the Rx array response as learnable functions of azimuth and zenith departure angles and azimuth and zenith arrival angles, respectively. 27. The apparatus of any one of clauses 22 to 26, wherein at least one processor is further configured to capture additional effects on the wireless signal not captured by the physical propagation channel model via additional trainable neural network layers. 28. The apparatus of any one of clauses 22 to 27, wherein the additional effects include a receiver (Rx) chain impulse response and a transmitter (Tx) chain impulse response. 29. The apparatus of any one of clauses 22 to 28, wherein at least one processor is further configured to receive, via a receiving device, the latent representation as channel condition feedback. 30. The apparatus of any one of clauses 22 to 29, wherein the reconstructed channel sequence comprises a continuous channel response across multiple frequency bands separated by guard bands. 31. The apparatus of any one of clauses 22 to 30, wherein the latent representation comprises one or more of the following sets: delay, complex amplitude, angle of departure, and angle of arrival. 32. The apparatus of any one of clauses 22 to 31, wherein the latent representation comprises an encoded channel sequence, and the receiving device uses a decoder to reconstruct the channel sequence from the latent representation. 33. An apparatus according to any one of clauses 22 to 32, wherein the decoder is a neural network based on a physical propagation channel model. 34. The apparatus of any one of clauses 22 to 33, wherein the at least one processor is further configured to train a decoder at the receiving device independent of the encoder at the transmitting device. 35. An apparatus for wireless communication by a transmitting device, comprising: Memory, at least one processor coupled to the memory; at least one processor receiving an input comprising a channel sequence for a wireless signal via an encoder; processing, via an encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal, the latent representation including at least one element that maps to a parameter of a physical propagation channel model; transmitting the latent representation to a receiving device; It is configured as follows: Device. 36. The apparatus of clause 35, wherein the physical propagation channel model comprises a sum of multipath components. 37. The apparatus of claim 35 or 36, wherein each of the multipath components has a delay and a complex amplitude. 38. The apparatus of any one of clauses 35 to 37, wherein each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response. 39. The apparatus of any one of clauses 35 to 38, wherein the at least one processor is further configured to represent the Tx array response and the Rx array response as learnable functions of azimuth and zenith departure angles and azimuth and zenith arrival angles, respectively. 40. The apparatus of any one of clauses 35 to 39, wherein at least one processor is further configured to transmit the latent representation to a receiving device as channel condition feedback. 41. The apparatus of any one of clauses 35 to 40, wherein the encoder comprises a neural network. 42. The apparatus of any one of clauses 35 to 41, wherein at least one processor is further configured to jointly train, via the transmitting device, an encoder-decoder pair such that a cascade of an encoder and a second decoder of the encoder-decoder pair reconstructs the channel sequence input at the encoder. 43. An apparatus for wireless communication by a receiving device, comprising: means for receiving, from a transmitting device, a latent representation of a channel sequence for a wireless signal; means for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal; An apparatus comprising: 44. A method of wireless communication by a transmitting device, comprising: means for receiving an input comprising a channel sequence for a wireless signal; means for processing, via an encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal, the latent representation including at least one element that maps to a parameter of the physical propagation channel; means for transmitting the latent representation to a receiving device; A method comprising: 45. A non-transitory computer-readable medium having encoded thereon program code for wireless communication by a receiving device, the program code being executed by a processor; program code for receiving, from a transmitting device, a latent representation of a channel sequence for a wireless signal; program code for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct a channel sequence for the wireless signal; Equipped with Non-transitory computer-readable medium. 46. ​​A non-transitory computer-readable medium having encoded thereon program code for wireless communication by a transmitting device, the program code being executed by a processor; program code for receiving an input comprising a channel sequence for a wireless signal; program code for processing, via an encoder, the channel sequence to generate a latent representation of the channel sequence for the wireless signal, the latent representation including at least one element that maps to a parameter of the physical propagation channel; program code for transmitting the latent representation to a receiving device; Equipped with Non-transitory computer-readable medium.

[0108]

[0101] The above disclosure provides illustration and description, but is not intended to be exhaustive or to limit the embodiments to the precise form disclosed. Modifications and variations may be made in light of the above disclosure or may be acquired from practice of the embodiments.

[0109]

[0102] When used, the term "component" is intended to be interpreted broadly as hardware, firmware, and / or a combination of hardware and software. When used, a processor is implemented in hardware, firmware, and / or a combination of hardware and software.

[0110]

[0103] Some aspects are described with respect to thresholds. As used, meeting a threshold can refer to a value being greater than the threshold, greater than or equal to the threshold, less than the threshold, less than or equal to the threshold, equal to the threshold, not equal to the threshold, etc., depending on the context.

[0111]

[0104] It will be apparent that the described systems and / or methods can be implemented in different 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 is not a limiting aspect. Thus, the operation and behavior of the systems and / or methods have been described without reference to specific software code. It should be understood that software and hardware can be designed to implement the systems and / or methods based at least in part on the description.

[0112]

[0105] Although certain combinations of features are recited in the claims and / or disclosed herein, such combinations do not limit the disclosure of the various aspects. In fact, many of these features may be combined in ways not specifically recited in the claims and / or disclosed herein. Although each dependent claim recited below may depend directly on only one claim, the disclosure of the various aspects includes each dependent claim in combination with any other claim in the set of claims. A phrase referring to a listing of "at least one of" items refers to any combination of those items, including a single element. By way of example, "at least one of a, b, or c" is intended to include a, b, c, ab, ac, bc, and abc, as well as any combination having multiple identical elements (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other order of a, b, and c).

[0113]

[0106] No element, act, or instruction used should be construed as critical or essential unless expressly described as such. Also, when used, the articles "a" and "an" are intended to include one or more items and may be used interchangeably with "one or more." Furthermore, when 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 may be used interchangeably with "one or more." When only one item is intended, the phrase "only one" or similar language is used. Also, when used, terms such as "has," "have," "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 otherwise specified.

Claims

1. 1. A method of wireless communication by a receiving device, comprising: receiving from a transmitting device a latent representation of a channel sequence for a wireless signal; applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal, where the physical propagation channel model is represented by a sum of multipath components. A method comprising:

2. The method of claim 1 , wherein each of the multipath components has a delay and a complex amplitude.

3. 3. The method of claim 1, wherein each of the multipath components has a transmitter (Tx) array response and a receiver (Rx) array response, preferably the Tx array response and the Rx array response are represented as learnable functions of azimuth and zenith departure angles and azimuth and zenith arrival angles, respectively.

4. The method further comprises capturing, via a trainable neural network layer, additional effects on the wireless signal that are not captured by the physical propagation channel model; The method of claim 1, wherein the additional effects preferably include a receiver (Rx) chain impulse response and a transmitter (Tx) chain impulse response.

5. The method of claim 1 , wherein the latent representation is transmitted to the receiving device as channel condition feedback.

6. The method of claim 1 , wherein the reconstructed channel sequence comprises a continuous channel response across multiple frequency bands separated by guard bands.

7. The method of claim 1 , wherein the latent representation comprises one or more of the following sets: delay, complex amplitude, angle of departure, and angle of arrival.

8. The method of claim 1 , wherein the latent representation comprises an encoded channel sequence, and the receiving device uses the decoder to reconstruct the channel sequence from the latent representation.

9. The method of claim 8 , wherein the decoder is a neural network based on the physical propagation channel model.

10. The method of claim 8 , further comprising training the decoder at the receiving device independently of an encoder at the transmitting device.

11. 1. A method of wireless communication by a transmitting device, comprising: receiving an input comprising a channel sequence for a wireless signal via an encoder; processing the channel sequence via the encoder to generate a latent representation of the channel sequence for the wireless signal, the latent representation including at least one element that maps to parameters of a physical propagation channel model, where the physical propagation channel model is represented by a sum of multipath components; transmitting the latent representation to a receiving device; A method comprising:

12. The method of claim 11 , wherein the encoder comprises a neural network.

13. The method of claim 11 , wherein the transmitting device jointly trains an encoder-decoder pair such that a cascade of the encoder and decoder of the pair reconstructs the channel sequence input at the encoder.

14. 1. An apparatus for wireless communication by a receiving device, comprising: means for receiving, from a transmitting device, a latent representation of a channel sequence for a wireless signal; means for applying, via a decoder, a physical propagation channel model to the latent representation to reconstruct the channel sequence for the wireless signal, wherein the physical propagation channel model is represented by a sum of multipath components; An apparatus comprising:

15. 1. An apparatus for wireless communication by a transmitting device, comprising: means for receiving an input including a channel sequence for a wireless signal; means for processing, via an encoder, the channel sequence for the wireless signal to generate a latent representation of the channel sequence, the latent representation including at least one element that maps to parameters of a physical propagation channel, wherein the physical propagation channel model is represented by a sum of multipath components; means for transmitting the latent representation to a receiving device; A method comprising:

Images