Deep neural network processing for user equipment coordination set

Abstract

This application relates to deep neural network processing for User Equipment Coordination Sets (UECS). Techniques and apparatus for deep neural network (DNN) processing for UECS UECS are described. A network entity selects (910) to form an E2E machine learning (ML) configuration for an end-to-end E2E DNN used to process UECS communications. The network entity instructs (915) each of a plurality of devices participating in the UECS to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN of the E2E DNN that transmits UECS communications over an E2E communication link, wherein the plurality of devices includes at least one base station, a coordinating user equipment (UE), and at least one additional UE. The network entity receives (940) feedback associated with the UECS communications and identifies (945) adjustments to the E2E ML configuration. The network entity then instructs at least some of the plurality of devices participating in the UECS to update the corresponding sub-DNN of the E2E DNN based on this adjustment.

Description

[0001] Case Analysis

[0002] This application is a divisional application of Chinese invention patent application 202180046564.7, filed on June 8, 2021. Technical Field

[0003] This application relates to deep neural network processing for user equipment coordination sets. Background Technology

[0004] In a wireless network, a base station provides user equipment (UE) with connectivity to various services (such as data and / or voice services) within the cell coverage area. The base station typically determines the configuration of the wireless connection used by the UE to access the service. For example, the base station determines the bandwidth and timing configuration of the wireless connection.

[0005] The quality of the wireless connection between the base station and the UE typically varies based on multiple factors, such as signal strength, bandwidth limitations, and interference. For example, a first UE operating at the edge of the cell coverage area typically receives a weaker signal from the base station compared to a second UE operating relatively close to the center of the cell coverage area. Therefore, service quality can sometimes degrade as the UE moves to different areas of the cell coverage area. With recent advancements in wireless communication systems, such as 5G New Radio (5G NR), new methods can be used to improve service quality. Summary of the Invention

[0006] This document describes the techniques and apparatus for deep neural network (DNN) processing for a User Equipment Coordination Set (UECS). In each aspect, a network entity selects an end-to-end (E2E) machine learning (ML) configuration that forms an E2E DNN for processing UECS communications. The network entity instructs each of a plurality of devices participating in the UECS to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN of the E2E DNN that transmits UECS communications over an E2E communication link, wherein the plurality of devices includes at least one base station, a coordinating user equipment (UE), and at least one additional UE. The network entity receives feedback associated with the UECS communications and identifies adjustments to the E2E ML configuration. The network entity then instructs at least some of the plurality of devices participating in the UECS to update the corresponding sub-DNN of the E2E DNN based on the adjustments.

[0007] In each aspect, the coordinating user equipment (UE) and network entities confirm a training schedule that indicates a period of time for maintaining one or more fixed ML architectures of a second part of an end-to-end (E2E) machine learning (ML) configuration, which forms a second set of sub-deep neural networks (sub-DNNs) of the E2E DNN, which transmits Radio-based User Equipment Coordination Set (UECS) communications via E2E communication links. The coordinating UE determines adjustments to a first part of the E2EML configuration based on the training schedule, which forms a first set of sub-DNNs that transmits UECS communications over the local radio network via E2E communication links. In response to determining these adjustments, the coordinating UE instructs one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first set of sub-DNNs using the adjustments to the first part of the E2E ML configuration.

[0008] Details of one or more embodiments of DNN processing for UECS are set forth in the accompanying drawings and the following description. Other features and advantages will be apparent from the specification and drawings, as well as from the claims. This summary is provided to introduce the subject matter further described in the detailed description and drawings. Therefore, this summary should not be construed as describing essential features, nor should it be used to limit the scope of the claimed subject matter. Attached Figure Description

[0009] The following describes in detail one or more aspects of deep neural network (DNN) processing for the User Equipment Coordination Set (UECS). The same reference numerals are used to denote similar elements in different instances of the specification and figures: Figure 1 An exemplary environment is illustrated that enables various aspects of DNN processing for UECS; Figure 2 An exemplary device diagram is shown, illustrating a device capable of implementing various aspects of DNN processing for UECS; Figure 3 An exemplary device diagram is shown, illustrating a device capable of implementing various aspects of DNN processing for UECS; Figure 4 The illustration shows an exemplary operating environment in a wireless communication system utilizing multiple deep neural networks, based on aspects of DNN processing for UECS. Figure 5 The illustration shows an example of generating multiple neural network formation configurations based on aspects of DNN processing used in UECS; Figure 6 The illustration shows an exemplary operating environment that enables DNN processing for UECS from various aspects. Figure 7The illustration shows an exemplary operating environment that enables DNN processing for UECS from various aspects. Figure 8 The illustration shows an exemplary operating environment that enables DNN processing for UECS from various aspects. Figure 9 The diagram illustrates an exemplary transaction graph between various network entities that implement DNN processing for UECS; Figure 10 The diagram illustrates another exemplary transaction graph between various network entities that implement DNN processing for UECS; Figure 11 An exemplary method for DNN processing in UECS is illustrated; and Figure 12 Another exemplary method for DNN processing for UECS is illustrated. Detailed Implementation

[0010] In previous wireless communication systems, various factors affected the quality of service provided by the base station to the user equipment (UE), such as the UE's location affecting signal strength. To improve service quality, various parties configure and / or establish a User Equipment Coordination Set (UECS) to perform joint processing (e.g., joint transmission, joint reception) of communications to the target UE.

[0011] Typically, a UECS includes at least two UEs communicating via a local wireless network connection, such as sharing or distributing signal-related information used for downlink and / or uplink UECS communication. By enabling multiple UEs to form a UECS to jointly transmit and receive data intended for a target UE within that UECS, the UEs in the UECS coordinate in a manner similar to distributed antennas for the target UE to improve the effective signal quality between the target UE and the base station. Downlink data intended for the target UE can be transmitted to multiple UEs within the UECS. Each UE demodulates and samples the downlink data, then forwards the samples to a single UE within the UECS (such as the coordinating UE or the target UE) for joint processing. Additionally, uplink data generated by the target UE can be distributed among multiple UEs in the UECS for joint transmission to the base station. By using multiple UEs to transmit uplink data, coordinating the joint transmission and reception of data intended for the target UE significantly increases the effective transmission power of the target UE, thereby improving the effective signal quality.

[0012] Deep neural networks (DNNs) provide solutions for performing various types of operations, such as transmitting UECS communications via end-to-end (E2E) communication. To illustrate, some aspects of DNN processing for UECS are trained to show how a DNN can jointly process (e.g., jointly receive, jointly transmit) UECS communications transmitted via E2E communication links between one or more base stations and multiple UEs included in the UECS. As an example, an end-to-end (E2E) DNN learns to (a) process communications transmitted between a base station and UEs included in the UECS via a first wireless network, and (b) process communications transmitted between UEs via a second local wireless network.

[0013] Network entities such as base stations and / or core network servers determine the end-to-end machine learning configuration (E2E ML configuration) that forms the E2E DNN trained to handle UECS communications, and instruct various devices to use portions of the E2E ML configuration to form sub-DNNs. In some cases, network entities dynamically reconfigure the E2E DNN based on various factors, such as changes in signal and / or link quality, changes in participating devices within the UECS, or changes in coordinating UEs within the UECS. The ability to dynamically adapt the E2E DNN provides flexible solutions in response to these changing factors, thereby improving the overall performance of data transmission and / or recovery over E2E communication links (e.g., higher processing resolution, faster processing, lower bit error rate, improved signal quality, reduced latency).

[0014] This document describes various aspects of DNN processing for UECS. In these aspects, a network entity selects an end-to-end (E2E) machine learning (ML) configuration that forms an E2E DNN for processing UECS communications. The network entity instructs each of a plurality of devices participating in UECS to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN for the E2E DNN that transmits UECS communications over an E2E communication link, wherein the plurality of devices includes at least one base station, a coordinating user equipment (UE), and at least one additional UE. The network entity receives feedback associated with the UECS communications and identifies adjustments to the E2E ML configuration. The network entity then instructs at least some of the plurality of devices participating in UECS to update the corresponding sub-DNN of the E2E DNN based on these adjustments.

[0015] In each aspect, the coordinating user equipment (UE) and network entities confirm a training schedule that indicates a period of time for maintaining one or more fixed ML architectures of a second part of an end-to-end (E2E) machine learning (ML) configuration, which forms a second set of sub-deep neural networks (sub-DNNs) of the E2E DNN, which transmits Radio-based User Equipment Coordination Set (UECS) communications via E2E communication links. The coordinating UE determines adjustments to a first part of the E2EML configuration based on the training schedule, which forms a first set of sub-DNNs that transmits UECS communications over the local radio network via E2E communication links. In response to determining these adjustments, the coordinating UE instructs one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first set of sub-DNNs using the adjustments to the first part of the E2E ML configuration.

[0016] Exemplary Environment

[0017] Figure 1 An exemplary environment 100 is illustrated, comprising multiple user equipment 110s (UEs 110), illustrated as UE 111, UE 112, and UE 113. Each UE 110 is capable of communicating with one or more base stations 120 (illustrated as base stations 121 and 122) via one or more wireless communication links 130 (wireless links 130) (illustrated as wireless link 131). Each UE 110 in UECS 108 (shown as UE 111, UE 112, and UE 113) is capable of communicating with the coordinating UE of UECS and / or the target UE in UECS via one or more local wireless network connections such as local wireless network connections 133, 134, and 135 (e.g., WLAN, Bluetooth, NFC, Personal Area Network (PAN), WiFi-Direct, IEEE 802.15.4, ZigBee, Threading, millimeter wavelength communication (mmWave), etc.). Although illustrated as a smartphone, UE 110 can be implemented as any suitable computing or electronic device, such as a mobile communication device, modem, cellular phone, gaming device, navigation device, media device, laptop computer, desktop computer, tablet computer, smart home appliance, vehicle-based communication system, Internet of Things (IoT) device (e.g., sensor node, controller / actuator node, or combinations thereof). Base station 120 (e.g., Evolved Universal Terrestrial Radio Access Network Node B, E-UTRAN Node B, Evolved Node B, eNodeB, eNB, Next Generation Node B, gNode B, gNB, ng-eNB, etc.) can be implemented in macrocells, microcells, small cells, picocells, distributed base stations, or any combination thereof.

[0018] Base station 120 communicates with user equipment 110 using radio link 131, and radio link 131 can be implemented as any suitable type of radio link. Radio link 131 includes control and data communications, such as downlinks transmitting data and control information from base station 120 to user equipment 110, uplinks transmitting other data and control information from user equipment 110 to base station 120, or both. Radio link 130 may include one or more radio links (e.g., radio links) or bearers implemented using any suitable communication protocol or standard or a combination of communication protocols or standards (such as 3GPP LTE, 5G NR, etc.). Multiple radio links 130 can be aggregated in carrier aggregation or multi-connectivity to provide higher data rates for UE 110. Multiple radio links 130 from multiple base stations 120 can be configured for Coordinated Multipoint (CoMP) communication with UE 110.

[0019] Base stations 120 collectively form a radio access network 140 (e.g., RAN, Evolved Universal Terrestrial Radio Access Network, E-UTRAN, 5G NR RAN, or NR RAN). Base stations 121 and 122 in RAN 140 are connected to core network 150. Base stations 121 and 122 are connected to core network 150 at 102 and 104 respectively, via an NG2 interface for control plane information and an NG3 interface for user plane data communication when connected to the 5G core network, or via an S1 interface for control plane information and user plane data communication when connected to the evolved packet core (EPC) network. Base stations 121 and 122 are capable of communicating at interface 106 using the Xn Application Protocol (XnAP) via the Xn interface or the X2 Application Protocol (X2AP) via the X2 interface to exchange user plane data and control plane information. User equipment 110 can connect to a public network, such as the Internet 160, via core network 150 to interact with remote service 170.

[0020] Base station 121 can designate a set of UEs (e.g., UE 111, UE 112, and UE 113) to form a UECS (e.g., UECS 108) for joint transmission and joint reception of signals for a target UE (e.g., UE 112). Base station 121 may select UE 111 as the coordinating UE because UE 111 is located between UE 112 and UE 113, or because UE 111 is able to communicate with each of the other UEs 112 and 113 in the UECS. Base station 121 selects UE 111 to coordinate messages and in-phase and quadrature (I / Q) samples transmitted between base station 121 and UEs 111, 112, and 113 for the target UE 112. Communication between the UEs can occur using a local wireless network such as PAN, NFC, Bluetooth, WiFi-Direct, local mmWave link, etc. In this example, all three of UEs 111, 112, and 113 receive RF signals from base station 121. UEs 111, 112, and 113 demodulate the RF signal to generate a baseband I / Q analog signal and sample the baseband I / Q analog signal to generate I / Q samples. UEs 112 and 113 use their local radio network to forward the I / Q samples along with system timing information (e.g., system frame number (SFN)) to the coordinating UE 111, which uses its own local radio network transceiver. The coordinating UE 111 then uses the timing information to synchronize and combine the I / Q samples and processes the combined signal to decode data packets for the target UE 112. The coordinating UE 111 then transmits the data packets to the target UE 112 using its local radio network.

[0021] When target UE 112 has uplink data to transmit to base station 121, the target UE transmits the uplink data to coordinating UE 111. Coordinating UE 111 uses its local radio network to distribute the uplink data as I / Q samples to each UE in UECS 108. Each UE in UECS 108 synchronizes with base station 121 to obtain timing information and its data transmission resource allocation. Then, all three UEs in UECS 108 jointly transmit the uplink data to base station 121. Base station 121 receives the uplink data transmitted from UEs 111, 112, and 113, and jointly processes the combined signals to decode the uplink data from target UE 112.

[0022] Exemplary device

[0023] Figure 2 An exemplary device diagram 200 is illustrated in UE 110 and base station 120, which are capable of implementing various aspects of DNN processing for UECS. Figure 3An exemplary device diagram 300 illustrates a core network server 302 capable of implementing various aspects of DNN processing for UECS. The UE 110, base station 120, and / or core network server 302 may include, for clarity, [details omitted]. Figure 2 or Figure 3 Additional features and interfaces omitted in the text.

[0024] UE 110 includes an antenna 202, a radio frequency front-end 204 (RF front-end 204), and a radio transceiver (e.g., LTE transceiver 206 and / or 5G NR transceiver 208) for communicating with a base station 120 in RAN 140. The RF front-end 204 of UE 110 can couple or connect the LTE transceiver 206 and the 5G NR transceiver 208 to the antenna 202 to facilitate various types of wireless communication. The antenna 202 of UE 110 may include an array of multiple antennas configured similarly or differently from each other. The antenna 202 and RF front-end 204 can be tuned to and / or tunable to one or more frequency bands defined by the 3GPP LTE and 5G NR communication standards and implemented by the LTE transceiver 206 and / or the 5G NR transceiver 208. Additionally, antenna 202, RF front-end 204, LTE transceiver 206, and / or 5G NR transceiver 208 can be configured to support beamforming for transmission and reception of communications with base station 120. By way of example and not limitation, antenna 202 and RF front-end 204 can be implemented for operation in sub-gigahertz, sub-6 GHz, and / or higher frequency bands as defined by the 3GPP LTE and 5G NR communication standards.

[0025] User equipment 110 also includes one or more processors 210 and a computer-readable storage medium 212 (CRM 212). Processor 210 may be a single-core or multi-core processor made of various materials such as silicon, polysilicon, high-k dielectrics, copper, etc. The computer-readable storage medium described herein does not include propagating signals. CRM 212 may include any suitable memory or storage device that can be used to store device data 214 of UE 110, such as random access memory (RAM), static RAM (SRAM), dynamic RAM (DRAM), non-volatile RAM (NVRAM), read-only memory (ROM), or flash memory. Device data 214 includes user data, multimedia data, beamforming codebooks, applications, neural network (NN) tables, and / or the operating system of UE 110 executable by one or more processors 210 to implement user plane data, control plane information, and user interaction with user equipment 110.

[0026] In various aspects, CRM 212 includes a neural network table 216 that stores various architectures and / or parameter configurations for forming neural networks, such as, by example and not limitation, specifying fully connected layer neural network architectures, convolutional layer neural network architectures, recurrent neural network layers, multi-connected hidden neural network layers, input layer architectures, output layer architectures, multiple nodes utilized by the neural network, coefficients (e.g., weights and biases) utilized by the neural network, kernel parameters, multiple filters utilized by the neural network, stride / pooling configurations utilized by the neural network, activation functions for each neural network layer, interconnections between neural network layers, parameters for neural network layers to be skipped, etc. Therefore, neural network table 216 includes any combination of neural network forming configuration elements (NN forming configuration elements) (such as architecture and / or parameter configurations) that can be used to create neural network forming configurations (NN forming configurations) that include combinations of one or more NN forming configuration elements that define and form a DNN. In some aspects, a single index value of neural network table 216 maps to a single NN forming configuration element (e.g., a 1:1 correspondence). Alternatively or additionally, individual index values ​​of neural network table 216 are mapped to NN forming configurations (e.g., combinations of NN forming configuration elements). In some implementations, the neural network table includes input characteristics for each NN forming configuration element and / or NN forming configuration, wherein the input characteristics describe properties related to the training data used to generate the NN forming configuration elements and / or NN forming configurations, as further described.

[0027] CRM 212 may also include a User Equipment Neural Network Manager 218 (UE Neural Network Manager 218). Alternatively or additionally, the UE Neural Network Manager 218 may be implemented wholly or partially as hardware logic or circuitry integrated or separate from other components of the User Equipment 110. The UE Neural Network Manager 218 accesses the Neural Network Table 216, such as via index values, and forms the DNN using NN formation configuration elements specified by the NN formation configuration. This includes updating the DNN using any combination of architectural changes and / or parameter changes to the DNN as further described (such as small changes to the DNN involving updated parameters and / or large changes to the node and / or layer connections of the DNN for reconfiguration). In implementations, the UE Neural Network Manager forms multiple DNNs to handle wireless communications (e.g., downlink communications, uplink communications).

[0028] Figure 2The illustrated device diagram of base station 120 includes a single network node (e.g., gNode B). The functionality of base station 120 can be distributed across multiple network nodes or devices and can be distributed in any manner suitable for performing the functions described herein. Base station 120 includes an antenna 252, a radio frequency front-end 254 (RF front-end 254), and one or more radio transceivers (e.g., one or more LTE transceivers 256 and / or one or more 5G NR transceivers 258) for communicating with UE 110. The RF front-end 254 of base station 120 is capable of coupling or connecting the LTE transceiver 256 and the 5G NR transceiver 258 to the antenna 252 to facilitate various types of wireless communication. The antenna 252 of base station 120 may include an array of multiple antennas configured similarly or differently from each other. The antenna 252 and RF front-end 254 can be tuned to and / or tunable to one or more frequency bands defined by the 3GPP LTE and 5G NR communication standards and implemented by the LTE transceiver 256 and / or the 5G NR transceiver 258. Additionally, antenna 252, RF front end 254, LTE transceiver 256 and / or 5G NR transceiver 258 can be configured to support beamforming (such as massive MIMO) for transmitting and receiving communications with UE 110.

[0029] Base station 120 also includes one or more processors 260 and computer-readable storage medium 262 (CRM 262). Processor 260 may be a single-core or multi-core processor made of various materials such as silicon, polysilicon, high-k dielectric, copper, etc. CRM 262 may include any suitable memory or storage device that can be used to store device data 264 of base station 120, such as random access memory (RAM), static RAM (SRAM), dynamic RAM (DRAM), non-volatile RAM (NVRAM), read-only memory (ROM), or flash memory. Device data 264 includes network scheduling data, radio resource management data, beamforming codebooks, applications, and / or the operating system of base station 120, which can be executed by one or more processors 260 to enable communication with UE 110.

[0030] CRM 262 also includes a base station manager 266. Alternatively or additionally, the base station manager 266 may be implemented, in whole or in part, as hardware logic or circuitry integrated or separate from other components of the base station 120. In at least some aspects, the base station manager 266 configures the LTE transceiver 256 and the 5G NR transceiver 258 for communication with the UE 110 and with a core network such as the core network 150.

[0031] CRM 262 also includes a Base Station Neural Network Manager 268 (BS Neural Network Manager 268). Alternatively or additionally, the BS Neural Network Manager 268 may be implemented, in whole or in part, as hardware logic or circuitry integrated or separate from other components of the base station 120. In at least some aspects, such as by selecting a combination of NN forming configuration elements to form a DNN for processing UECS communications, the BS Neural Network Manager 268 selects an NN forming configuration utilized by the base station 120 and / or UE 110 to configure a deep neural network for processing wireless communications. In some embodiments, the BS Neural Network Manager receives feedback from UE 110 and selects an NN forming configuration based on that feedback. Alternatively or additionally, the BS Neural Network Manager 268 receives a neural network forming configuration instruction from elements of the core network 150 via the core network interface 278 or the inter-base station interface 276 and forwards the NN forming configuration instruction to UE 110. In some aspects, the BS neural network manager 268 selects the NN formation configuration in response to recognizing changes in UECS (such as changes in channel conditions, changes in participating UEs, changes in estimated UE locations, changes in coordinating UEs, etc.).

[0032] CRM 262 includes a training module 270 and a neural network table 272. In an implementation, base station 120 manages the formation configuration of the NN and deploys it to UE 110. Alternatively or additionally, base station 120 maintains the neural network table 272. Training module 270 uses known input data to teach and / or train the DNN. For example, training module 270 trains one or more DNNs for various purposes, such as processing communications transmitted through a wireless communication system (e.g., encoding downlink communications, modulating downlink communications, demodulating downlink communications, decoding downlink communications, encoding uplink communications, modulating uplink communications, demodulating uplink communications, decoding uplink communications, processing UECS communications). This includes training one or more DNNs offline (e.g., when the DNN is not actively involved in processing communications) and / or online (e.g., when the DNN is actively involved in processing communications).

[0033] In this implementation, the training module 270 extracts the learned parameter configuration from the DNN to identify NN formation configuration elements and / or NN formation configurations, and then adds and / or updates the NN formation configuration elements and / or NN formation configurations to the neural network table 272. The extracted parameter configuration includes any combination of information defining the behavior of the neural network, such as node connections, coefficients, active layers, weights, biases, pooling, etc.

[0034] The neural network table 272 stores multiple different NN formation configuration elements and / or NN formation configurations generated using the training module 270. In some embodiments, the neural network table includes input characteristics for each NN formation configuration element and / or NN formation configuration, wherein the input characteristics describe attributes related to the training data used to generate the NN formation configuration elements and / or NN formation configurations. For example, input characteristics include, but are not limited to, the number of UEs participating in the UECS, the estimated location of the target UE in the UECS, the estimated location of the coordinating UE in the UECS, the type of local radio network link used by the UECS, power information, signal-to-interference-plus-noise ratio (SINR) information, channel quality indicator (CQI) information, channel state information (CSI), Doppler feedback, frequency band, block error rate (BLER), quality of service (QoS), hybrid automatic repeat request (HARQ) information (e.g., first transmission error rate, second transmission error rate, maximum retransmission), latency, radio link control (RLC), automatic repeat request (ARQ) metric, received signal strength (RSS), uplink SINR, timing measurement, error metric, UE capability, BS capability, power mode, Internet Protocol (IP) layer throughput, end-to-end latency, end-to-end packet loss rate, etc. Therefore, input characteristics sometimes include layer 1, layer 2, and / or layer 3 metrics. In some implementations, a single index value of the neural network table 272 is mapped to a single NN to form configuration elements (e.g., a 1:1 correspondence). Alternatively or additionally, individual index values ​​of the neural network table 272 are mapped to NN forming configurations (e.g., combinations of NN forming configuration elements).

[0035] In one implementation, base station 120 synchronizes neural network table 272 with neural network table 216 such that the NN forming configuration elements and / or input characteristics stored in one neural network table are copied in the second neural network table. Alternatively or additionally, base station 120 synchronizes neural network table 272 with neural network table 216 such that the NN forming configuration elements and / or input characteristics stored in one neural network table represent complementary functions in the second neural network table (e.g., NN forming configuration elements for transmitter path processing in the first neural network table, and NN forming configuration elements for receiver path processing in the second neural network table).

[0036] In various aspects, CRM 262 also includes an end-to-end machine learning controller 274 (E2E ML controller 274). The E2EML controller 274 determines the end-to-end machine learning configuration (E2E ML configuration) for processing information transmitted over the E2E communication link, such as determining the E2E ML configuration for processing UECS communication over the E2E communication link, as further described. Alternatively or additionally, the E2E ML controller analyzes any combination of the ML capabilities of the devices participating in the E2E communication link (e.g., supported ML architectures, supported layers, available processing power, memory limitations, available power budget, fixed-point processing versus floating-point processing, maximum kernel size capability, compute capability). In some implementations, the E2E ML controller obtains metrics characterizing the current operating environment (e.g., signal quality parameters, link quality parameters) and analyzes the current operating environment to determine the E2E ML configuration. For illustration, the E2E ML controller receives any combination of the following: Receive Signal Strength Indicator (RSSI), power information, Signal-to-Interference-Ratio (SINR) information, Reference Signal Received Power (RSRP), Channel Quality Indicator (CQI) information, Channel State Information (CSI), Doppler feedback, Block Error Rate (BLER), Quality of Service (QoS), Hybrid Automatic Repeat Request (HARQ) information (e.g., First Transmission Error Rate, Second Transmission Error Rate, Maximum Repeat Request), Uplink SINR, Timing Measurement, Error Measurement, etc. This includes determining an E2E ML configuration that includes a combination of architectural configuration and parameter configurations defining the DNN, or determining an E2E ML configuration that simply includes parameter configurations for updating the DNN.

[0037] When determining the E2E ML configuration, the E2E ML controller sometimes defines partitions of the E2E ML configuration, distributing the processing functions associated with the E2E ML configuration across multiple devices. For clarity, Figure 2 The E2E ML controller 274 is illustrated separately from the BS neural network manager 268, but in alternative or additional embodiments, the BS neural network manager 268 includes functions performed by the E2E ML controller 274, or vice versa.

[0038] Base station 120 also includes an inter-base station interface 276, such as an Xn and / or X2 interface. Base station manager 266 configures inter-base station interface 276 to exchange user plane data, control plane information, and other data / information between other base stations to manage communication between base station 120 and UE 110. Base station 120 includes a core network interface 278, which base station manager 266 configures to exchange user plane data, control plane information, and / or other data / information with core network functions and / or entities.

[0039] exist Figure 3 In the core network 150, core network server 302 can provide all or part of the functions, entities, services, and / or gateways. Each function, entity, service, and / or gateway in core network 150 can be provided as a service distributed across multiple servers or embodied on a dedicated server within core network 150. For example, core network server 302 can provide all or part of services or functions such as User Plane Function (UPF), Access and Mobility Management Function (AMF), Serving Gateway (S-GW), Packet Data Network Gateway (P-GW), Mobility Management Entity (MME), Evolved Packet Data Gateway (ePDG), etc. Core network server 302 is shown as being embodied on a single server including one or more processors 304 and computer-readable storage medium 306 (CRM 306). Processor 304 can be a single-core processor or a multi-core processor made of various materials such as silicon, polysilicon, high-k dielectric, copper, etc. CRM 306 may include any suitable memory or storage device that can be used to store device data 308 of the core network server 302, such as random access memory (RAM), static RAM (SRAM), dynamic RAM (DRAM), non-volatile RAM (NVRAM), read-only memory (ROM), hard disk drive, or flash memory. Device data 308 includes data that can be executed by one or more processors 304 to support core network functions or entities, and / or the operating system of the core network server 302.

[0040] CRM 306 also includes one or more core network applications 310, which in one embodiment are embodied on CRM 306 (as shown). The one or more core network applications 310 may implement functions such as UPF, AMF, S-GW, P-GW, MME, ePDG, etc. Alternatively or additionally, the one or more core network applications 310 may be implemented, in whole or in part, as hardware logic or circuitry integrated or separate from other components of the core network server 302.

[0041] CRM 306 also includes a core network neural network manager 312, which manages the NN formation configuration for forming a DNN to handle communications transmitted between UE 110 and base station 120, such as UECS communications involving multiple UEs and / or adding local wireless network connections to E2E communication links. In various aspects, the core network neural network manager 312 analyzes various characteristics of the UECS (e.g., the number of participating devices, the estimated location of the target UE, the estimated location of participating UEs, and the type of local wireless network connection) and selects an end-to-end machine learning configuration (E2E ML configuration) capable of forming an end-to-end deep neural network (E2E DNN) to handle UECS communications transmitted over E2E communication links. In various aspects, the core network neural network manager 312 selects one or more NN formation configurations within neural network table 316 to indicate the determined E2E ML configuration.

[0042] In some implementations, the core network neural network manager 312 analyzes various criteria, such as current signal channel conditions (e.g., reported by base station 120, other radio access points, or UE 110 (via base station or other radio access points)), the capabilities of base station 120 (e.g., antenna configuration, cell configuration, MIMO capability, radio capability, processing capability), and the capabilities of UE 110 (e.g., antenna configuration, MIMO capability, radio capability, processing capability). For example, base station 120 obtains various criteria and / or link quality indications (e.g., RSSI, power information, SINR, RSRP, CQI, CSI, Doppler feedback, BLER, HARQ, timing measurement, error metric, etc.) during communication with the UE and forwards these criteria and / or link quality indications to the core network neural network manager 312. The core network neural network manager selects an E2E ML configuration based on these criteria and / or indications to improve the accuracy of DNN processing communication (e.g., lower bit error rate, higher signal quality). The core network neural network manager 312 then communicates the E2E ML configuration to the base station 120 and / or the UE 110, such as by communicating an index of the neural network table. In one implementation, the core network neural network manager 312 receives feedback from the UE and / or BS from the base station 120 and selects an updated E2E ML configuration based on that feedback.

[0043] CRM 306 includes a training module 314 and a neural network table 316. In one implementation, a core network server 302 manages E2E ML configurations and / or portions of partitionable E2E ML configurations and deploys them to multiple devices (e.g., UE 110, base station 120) in a wireless communication system. Alternatively or additionally, the core network server maintains the neural network table 316 externally to CRM 306. The training module 314 teaches and / or trains the DNN using known input data. For example, the training module 314 trains the DNN to handle different types of pilot communications transmitted through the wireless communication system. This includes offline and / or online DNN training. In one implementation, the training module 314 extracts learned NN forming configurations and / or learned NN forming configuration elements from the DNN and stores the learned NN forming configuration elements in the neural network table 316, such as NN forming configurations that can be selected by the core network neural network manager 312 as E2E ML configurations to form an E2E DNN, as further described. Therefore, the NN formation configuration includes any combination of architectural configurations (e.g., node connections, layer connections) and / or parameter configurations (e.g., weights, biases, pooling) that define or influence the behavior of the DNN. In some implementations, a single index value of the neural network table 316 is mapped to a single NN formation configuration element (e.g., a 1:1 correspondence). Alternatively or additionally, a single index value of the neural network table 316 is mapped to an NN formation configuration (e.g., a combination of NN formation configuration elements).

[0044] In some implementations, the training module 314 of the core network neural network manager 312 generates complementary NN formation configurations and / or NN formation configuration elements to those stored in the neural network table 216 at UE 110 and / or the neural network table 272 at base station 121. As an example, the training module 314 generates the neural network table 316 with NN formation configurations and / or NN formation configuration elements that exhibit high variation in architecture and / or parameter configuration relative to the moderate and / or low variation used to generate the neural network tables 272 and / or 216. For example, the NN formation configurations and / or NN formation configuration elements generated by the training module 314 correspond to fully connected layers, full kernel size, frequent sampling and / or pooling, high weighted accuracy, etc. Therefore, the neural network table 316 sometimes includes high-accuracy neural networks at the cost of increased processing complexity and / or time.

[0045] The neural network table 316 stores multiple different NN formation configuration elements generated using the training module 314. In some embodiments, the neural network table includes input characteristics for each NN formation configuration element and / or NN formation configuration, wherein the input characteristics describe attributes related to the training data used to generate the NN formation configuration. For example, the input characteristics may include the number of UEs participating in the UECS, the estimated location of the target UE in the UECS, the estimated location of the coordinating UE in the UECS, the type of local radio network link used by the UECS, power information, SINR information, CQI, CSI, Doppler feedback, RSS, error metric, minimum end-to-end (E2E) latency, expected E2E latency, E2E QoS, E2E throughput, E2E packet loss rate, service cost, etc.

[0046] CRM 306 also includes an end-to-end machine learning controller 318 (E2E ML controller 318). The E2E ML controller 318 determines an end-to-end machine learning configuration (E2E ML configuration) for processing information transmitted over an E2E communication link, such as, as further described, determining an E2E ML configuration for processing UECS communications over the E2E communication link. Alternatively or additionally, the E2E ML controller analyzes any combination of the ML capabilities of the devices participating in the E2E communication link (e.g., supported ML architectures, supported layers, available processing power, memory limitations, available power budget, fixed-point and floating-point processing, maximum kernel size capability, compute capability). In some implementations, the E2E ML controller obtains metrics characterizing the current operating environment (e.g., signal quality parameters, link quality parameters) and analyzes the current operating environment to determine the E2E ML configuration. This includes determining an E2E ML configuration that includes a combination of architecture configuration and parameter configuration(s) defining a DNN(s), or determining an E2E ML configuration that simply includes parameter configurations for updating the DNN.

[0047] When determining the E2E ML configuration, the E2E ML controller 318 sometimes determines partitions of the E2E ML configuration, which distribute the processing functions associated with the E2E ML configuration across multiple devices. For clarity, Figure 3 The E2E ML controller 318 is illustrated separately from the core network neural network manager 312, but in alternative or additional embodiments, the core network neural network manager 312 includes functions performed by the E2E ML controller 318, or vice versa.

[0048] The core network server 302 also includes a core network interface 320 for communicating user plane data, control plane information, and other data / information with other functions or entities in the core network 150, base station 120, or UE 110. In one embodiment, the core network server 302 uses the core network interface 320 to transmit E2E ML configuration or portions of partitionable E2E ML configuration to the base station 120. Alternatively or additionally, the core network server 302 uses the core network interface 320 to receive feedback from the base station 120 and / or UE 110 via the base station 120.

[0049] Configurable machine learning module

[0050] Figure 4 An exemplary operating environment 400 is illustrated, comprising a UE 110 and a base station 120 capable of implementing various aspects of DNN processing for UECS. In this embodiment, the UE 110 and the base station 120 exchange communication with each other via a wireless communication system by using multiple DNN processing communications.

[0051] The base station neural network manager 268 of base station 120 includes a downlink processing module 402 for processing downlink communications, such as those for generating downlink communications to be transmitted to UE 110. For illustration, the base station neural network manager 268 forms one or more deep neural networks 404 (DNN 404) in the downlink processing module 402 using an E2E ML configuration and / or a portion of the E2E ML configuration, as further described. In various aspects, the DNN 404 performs some or all of the transmitter processing chain functions for generating downlink communications, such as receiving input data, proceeding to the coding stage, followed by the modulation stage, and then the radio frequency (RF) analog transmission (Tx) stage. For illustration, the DNN 404 is capable of performing convolutional coding, serial-to-parallel conversion, cyclic prefix insertion, channel coding, time / frequency interleaving, etc. In some aspects, the DNN 404 processes downlink UECS communications.

[0052] Similarly, the UE neural network manager 218 of UE 110 includes a downlink processing module 406, wherein the downlink processing module 406 includes one or more deep neural networks 408 (DNN 408) for processing (received) downlink communication. In various embodiments, the UE neural network manager 218 uses an E2E ML configuration and / or a portion of an E2E ML configuration to form the DNN 408, as further described. In various aspects, the DNN 408 performs some or all of the receiver chain operations and / or processing functions for (received) downlink communication, such as complementary processing to the processing performed by the DNN 404 (e.g., RF analog reception (Rx) phase, demodulation phase, decoding phase). For illustration, the DNN 408 is capable of performing any combination of functions such as extracting data embedded in the Rx signal, recovering control information, recovering binary data, correcting data errors based on forward error correction applied at the transmitter block, extracting payload data from frames and / or time slots, etc.

[0053] Base station 120 and / or UE 110 also use a DNN to process uplink communication. In environment 400, UE neural network manager 218 includes uplink processing module 410, wherein uplink processing module 410 includes one or more deep neural networks 412 (DNN 412) for generating and / or processing uplink communication (e.g., encoding, modulation). In other words, uplink processing module 410 processes pre-transmission communication as part of processing uplink communication. UE neural network manager 218 uses, for example, E2E ML configuration and / or a portion of E2E ML configuration to form DNN 412 to perform some or all of the transmitter processing functions for generating uplink communication transmitted from UE 110 to base station 120.

[0054] Similarly, the uplink processing module 414 of base station 120 includes one or more deep neural networks 416 (DNN 416) for processing (received) uplink communication, wherein the base station neural network manager 268 uses E2E ML configuration and / or a portion of E2E ML configuration to form the DNN 416 to perform some or all of the receiver processing functions for (received) uplink communication, such as uplink communication received from UE 110. Sometimes, DNN 412 and DNN 416 perform complementary functions to each other.

[0055] Typically, a deep neural network (DNN) corresponds to a group of connected nodes organized into three or more layers. The nodes between layers can be configured in various ways, such as a partially connected configuration where a subset of the first node in the first layer connects to a subset of the second node in the second layer, or a fully connected configuration where every node in the first layer connects to every node in the second layer. Nodes can use various algorithms and / or analyses to generate output information based on adaptive learning, such as unilinear regression, multiple linear regression, logistic regression, stepwise regression, binary classification, multi-class classification, multivariate adaptive regression splines, local estimation scatter plot smoothing, etc. Sometimes, one or more algorithms include weights and / or coefficients that change based on adaptive learning. Therefore, the weights and / or coefficients reflect the information learned by the neural network.

[0056] Neural networks can also employ various architectures that determine which nodes are connected within the network, how data is advanced and / or retained within the network, what weights and coefficients are used to process the input data, and how the data is processed, among other things. These various factors collectively describe the NN formation configuration. For illustration, recurrent neural networks (such as Long Short-Term Memory (LSTM) neural networks) form loops between node connections to retain information from previous portions of the input data sequence. The recurrent neural network then uses the retained information for subsequent portions of the input data sequence. As another example, feedforward neural networks pass information to forward connections without forming loops to retain information. Although described in the context of node connections, it should be recognized that the NN formation configuration can include various parameter configurations that influence how the neural network processes the input data.

[0057] The NN formation configuration of a neural network can be characterized by various architectures and / or parameter configurations. For illustration, consider an example of a DNN implementing a convolutional neural network. Typically, a convolutional neural network corresponds to a type of DNN where layers use convolution operations to process data to filter the input data. Accordingly, the convolutional NN formation configuration can be characterized, for example but not limited to, using pooling parameters (e.g., specifying pooling layers to reduce the dimensionality of the input data), kernel parameters (e.g., the filter size and / or kernel type used in processing the input data), weights (e.g., biases used to classify the input data), and / or layer parameters (e.g., layer connections and / or layer types). While described in the context of pooling parameters, kernel parameters, weight parameters, and layer parameters, other parameter configurations can be used to form a DNN. Therefore, the NN formation configuration can include any other type of parameters that can be applied to the DNN to influence how the DNN processes the input data to generate output data. An E2E ML configuration uses one or more NN formation configurations to form an E2E DNN that handles communication from one endpoint to another. For example, a partitionable E2E ML configuration can be configured using the corresponding NameNode for each partition.

[0058] Figure 5 The illustration depicts an example 500 illustrating aspects of generating multiple neural network (NN) formation configurations based on DNN processing for UECS. Sometimes, the aspects of example 500 are... Figure 2 and Figure 3 It can be implemented by any combination of training module 270, base station neural network manager 268, core network neural network manager 312 and / or training module 314.

[0059] Figure 5The upper part includes DNN 502, which represents any suitable DNN used to implement DNN processing for UECS. In embodiments, the neural network manager determines to generate different NN formation configurations, such as NN formation configurations for processing UECS communications. Alternatively or additionally, the neural network generates NN formation configurations based on different transmission environments and / or transmission channel conditions. Training data 504 represents exemplary inputs to DNN 502, such as data corresponding to downlink and / or uplink communications with specific operating configurations and / or specific transmission environments. For illustration, training data 504 can include digital samples of downlink radio signals, recovered symbols, recovered frame data, binary data, etc. In some embodiments, the training module mathematically generates training data or accesses a file storing training data. At other times, the training module obtains real-world communication data. Therefore, the training module can train DNN 502 using mathematically generated data, static data, and / or real-world data. Some implementations generate input characteristics 506 that describe various qualities of the training data, such as operating configuration, transport channel metrics, UE capabilities, UE speed, number of UEs participating in the UECS, estimated location of the target UE in the UECS, estimated location of the coordinating UE in the UECS, and type of local wireless network link used by the UECS.

[0060] DNN 502 analyzes the training data and generates output 508, represented here as binary data. Some implementations iteratively train DNN 502 using the same training dataset and / or additional training data with the same input characteristics to improve the accuracy of the machine learning module. During training, the machine learning module modifies some or all of the architecture and / or parameter configurations of the neural network included in the machine learning module, such as node connections, coefficients, kernel size, etc. At some point in the training, such as when the training module determines that the accuracy meets or exceeds a desired threshold, or when the training process meets or exceeds the number of iterations, the training module determines to extract the architecture and / or parameter configuration 510 of the neural network (e.g., one or more pooling parameters, one or more kernel parameters, one or more layer parameters, weights). The training module then extracts the architecture and / or parameter configuration from the machine learning module to use as NN forming configurations and / or one or more NN forming configuration elements. The architecture and / or parameter configurations can include any combination of fixed architecture and / or parameter configurations and / or variable architecture and / or parameter configurations.

[0061] Figure 5 The lower part includes a neural network table 512 representing a set of configuration elements for the NN, such as... Figure 2 and Figure 3The neural network tables 216, 272, and / or 316 are used. Neural network table 512 stores various combinations of architecture configurations, parameter configurations, and input characteristics; however, alternative implementations omit input characteristics from the table. As the DNN learns additional information, various implementations update and / or maintain the NN formation configuration elements and / or input characteristics. For example, at index 514, the neural network manager and / or training module updates neural network table 512 to include the architecture and / or parameter configuration 510 generated by DNN 502 when analyzing training data 504. At later points in time, such as when determining the E2E ML configuration for handling UECS communication over E2E communication links, the neural network manager selects one or more NN formation configurations from neural network table 512 by matching input characteristics with the current operating environment and / or configuration, such as by matching input characteristics with current channel conditions, the number of UEs participating in the UECS, the estimated location of the target UE in the UECS, the estimated location of the coordinating UE in the UECS, the type of local wireless network link used by the UECS, UE capabilities, UE characteristics (e.g., speed, location, etc.).

[0062] DNN processing for UECS

[0063] The UECS enhances the ability of a target UE to transmit and receive communications with a base station by typically acting as a distributed antenna for that UE. To illustrate, the base station uses a wireless network to transmit downlink data packets using radio frequency (RF) signals to multiple UEs within the UECS. Some or all of the UEs in the UECS receive the RF signals and demodulate them into analog baseband signals, sampling the baseband signals to generate a set of in-phase and quadrature (I / Q) samples. Each UE transmits its I / Q samples to a coordinating UE via its local wireless network. In some aspects, the UE transmits timing information with the I / Q samples. Using the timing information, the coordinating UE combines the I / Q samples and processes the combined I / Q samples to decode user plane data for the target UE. The coordinating UE then transmits data packets to the target UE via its local wireless network.

[0064] Similarly, when a target UE has uplink data to transmit to a base station, the target UE transmits the uplink data to a coordinating UE, which uses its local radio network to distribute the uplink data to each UE in the UECS. In some aspects, each UE in the UECS synchronizes with the base station for timing information and corresponding data transmission resource allocation. Then, the multiple UEs in the UECS jointly transmit the uplink data to the base station. The base station receives the jointly transmitted uplink data from the multiple UEs and processes the combined signal to decode the uplink data from the target UE. By enabling multiple UEs to form a UECS for jointly transmitting and receiving data intended for the target UE, the UEs in the UECS coordinate in a manner similar to a distributed antenna for the target UE to improve the effective signal quality between the target UE and the base station.

[0065] In each aspect, the network entity determines an end-to-end (E2E) machine learning (ML) configuration that forms an E2E deep neural network (DNN) for processing UECS communications transmitted over E2E communication links. For example, the core network server determines the E2E ML configuration based on any combination of factors such as the number of devices participating in the UECS, signal and / or link quality parameters, the capabilities of one or more devices participating in the UECS, the type of local radio network connection used between UEs, and the estimated location of the target UE and / or other participating UE(one or more). The network entity then instructs each of the one or more devices to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN of the corresponding E2E DNN formed using the E2E ML configuration.

[0066] Figure 6 An exemplary environment 600 for implementing DNN processing for UECS is illustrated, according to various aspects. Environment 600 includes... Figure 1 Base station 120 and UECS 108, among which, Figure 1 UEs 111, UE 112, and UE 113 form UECS 108. While environment 600 shows a single base station 120, alternative or additional aspects of the DNN processing for the UECS can utilize multiple base stations, such as an Active Coordination Set (ACS) of multiple base stations for joint wireless communication with the target user equipment, as referenced in [reference missing]. Figure 8 As stated above.

[0067] In the aspect of DNN processing for UECS, an E2E ML controller, such as E2E ML controller 274 of base station 120 or E2E ML controller 318 of core network server 302 (not shown), determines an E2E ML configuration for processing UECS communications (e.g., joint reception, joint transmission) transmitted over one or more E2E communication links. For illustration, the E2E ML controller determines adjustments to: (a) existing E2E ML configurations, such as minor adjustments using parameter updates (e.g., coefficients, weights) to tune (one or more) existing E2E DNNs based on feedback, and / or (b) ML architecture changes (e.g., number of layers, layer downsampling configuration, addition or removal of fully convolutional layers) to reconfigure (one or more) E2E DNNs. In environment 600, the E2E ML controller determines the E2E ML configuration for forming a unidirectional E2E DNN for handling downlink UECS communication. However, in alternative or additional implementations, the E2E ML controller determines the E2E ML configuration to form one or more bidirectional E2E DNNs. In various aspects, the E2E ML controller partitions the E2E ML configuration and instructs multiple devices to use the E2E ML configuration partitions to form sub-DNNs of the E2E DNN.

[0068] E2E ML controllers (e.g., E2E ML controller 274, E2E ML controller 318) can determine E2E ML configurations based on a combination of factors such as the machine learning (ML) capabilities of one or more devices or network entities participating in the E2E communication link and / or UECS (e.g., supported ML architecture, supported layers, available processing power, memory limitations, available power budget, fixed-point processing versus floating-point processing, maximum kernel size capability, computing power). As another example, E2E ML controllers may analyze the current operating environment, such as by analyzing signals and / or link quality indicators (e.g., RSSI, power information, SINR, RSRP, CQI, CSI, Doppler feedback, BLER, HARQ, timing measurements, error metrics, etc.) received from UE 111, UE 112, UE 113 and / or generated by base station 120.

[0069] The E2E ML controller sometimes determines partitions of the E2E ML configuration (and the resulting E2EDNN formed using the E2E ML configuration) to distribute processing among various devices participating in the E2E communication link. In other words, the E2E ML configuration forms a distributed E2E DNN, where multiple devices implement corresponding portions of the distributed E2E DNN. For example, in response to determining the E2E ML configuration associated with processing downlink UECS communication, the E2E ML controller 318 partitions the E2E ML configuration into multiple portions and instructs devices to form corresponding DNNs based on these portions.

[0070] For illustration, base station 120 uses a first portion of the E2E ML configuration determined by the E2E ML controller to form a first sub-DNN labeled Transmit DNN 602 (TX DNN 602). In various aspects, TX DNN 602 processes downlink communications directed to a target UE in UECS 108, such as by performing any combination of transmitter processing chain operations that result in one or more downlink transmissions via radio link 131 to UECS 108. For example, base station 120 uses the air interface resources of the (cellular) wireless network to transmit downlink radio signals via TX DNN 602. At least some UEs in UECS 108 use corresponding sub-DNNs to receive and process downlink radio signals. For illustration, UE 111 uses the second part of the E2E ML configuration to form a second sub-DNN labeled as Receive DNN 604 (RX DNN 604), UE 112 uses the third part of the E2E ML configuration to form a third sub-DNN labeled as Receive DNN 606 (RX DNN 606), and UE 113 uses the fourth part of the E2E ML configuration to form a fourth sub-DNN labeled as Receive DNN 608 (RX DNN 608). RX DNN 606 and RX DNN 608 process downlink communication to the target UE of UECS 108, such as by performing any combination of transmitter processing chain operations that result in one or more downlink transmissions to UECS 108 via radio link 131.

[0071] RX DNN 604, RX DNN 606, and RX DNN 608 form a first sub-DNN set 610, which relates to processing communications transmitted using the wireless network associated with base station 120. For illustration, RX DNN 604, 606, and 608 process downlink communications received via the wireless network associated with base station 120, such as by performing at least some receiver chain operations. As an example, the first sub-DNN set 610 receives samples of downlink radio signals (or down-converted versions of downlink radio signals) from an analog-to-digital converter (ADC) and generates I / Q samples.

[0072] In each aspect, the UEs in the UECS form a second sub-DNN set 612 for processing communications transmitted using sidelinks (e.g., local wireless network connections). As an example, at least some of the sub-DNNs in the second sub-DNN set 612 correspond to the sidelink TX DNNs for processing I / Q samples transmitted to the coordinating UE via the local wireless network connection. For illustration, assume that in environment 600, base station 120 instructs UE 111 to act as the coordinating UE, and instructs UEs 112 and 113 to act as (non-coordinating UE) participants in the UECS. Figure 6 As shown, UE 112 uses the fifth part of the E2E ML configuration to form a fifth sub-DNN, labeled as Transmission DNN 614 (TX DNN 614), which operates as a sidelink TX DNN by receiving the output generated by RX DNN 606 and processing the output to generate a transmission over the local wireless network to the coordinating UE (e.g., UE 111) using local wireless network connection 134. In an alternative or additional implementation, UE 112 forms a single sub-DNN that includes the functionality of both RX DNN 606 and TX DNN 614. Similarly, UE 113 uses the sixth part of the E2E ML configuration to form a sixth sub-DNN that operates as a sidelink TX DNN, labeled as Transmission DNN 616 (TX DNN 616), which receives the output generated by RX DNN 608 and processes the output to generate a transmission over the local wireless network to the coordinating UE (e.g., UE 111) using local wireless network connection 135. However, in alternative or additional implementations, UE 113 forms a single sub-DNN combining the functions of RX DNN 608 and TX DNN 616. TX DNNs 614, 616 can process transmissions toward the coordinating UE, such as by performing any combination of transmitter processing chain operations.

[0073] UE 111, acting as the coordinating UE, uses the seventh part of its E2E ML configuration to form a seventh sub-DNN as part of the second sub-DNN set 612. This seventh sub-DNN is designated as receive DNN 618 (RX DNN 618), which operates as a side-link RX DNN for receiving local radio network communications from various UEs in the UECS and processing the input using various receiver chain operations. As an example, RX DNN 618 decodes and / or extracts I / Q samples from local radio network connection messages from UE 112 and / or UE 113.

[0074] UE 111, acting as the coordinating UE, also uses an eighth part of the E2E ML configuration to form an eighth sub-DNN, designated as the Joint Receive Processing DNN 620 (Joint RX Processing DNN 620). In each aspect, the Joint RX Processing DNN 620 receives baseband I / Q samples generated by various UEs in the UECS and combines the I / Q samples, as further described. For example, the Joint RX Processing DNN 620 receives a first set of I / Q samples generated by the (co-resident) RX DNN 604, a second set of I / Q samples from UE 112 via local radio network connection 134 and through RX DNN 618, and a third set of I / Q samples from UE 113 via local radio network connection 135 and through RX DNN 618. The Joint RX Processing DNN 620 then combines the I / Q samples and processes the combined I / Q samples to recover user plane data and / or control plane information intended for use by the target UE from downlink communications. Subsequently, if the target UE separates from the coordinating UE 111, the joint RX processing DNN 620 generates local radio network communication that forwards the recovered user plane data and / or control plane information to the target UE. Furthermore, although environment 600 illustrates UE 112 and UE 113 forwarding I / Q samples to the coordinating UE (e.g., UE 111), in some embodiments, UE 112, UE 113, and / or UE 111 forward I / Q samples to the target UE for processing. In other words, the target UE can also be the coordinating UE and include a sub-DNN that receives and processes communications transmitted via the local radio network connection.

[0075] Figure 7 An exemplary environment 700 for implementing DNN processing for UECS is illustrated, according to various aspects. Environment 700 includes... Figure 1 Base station 120 and UECS 108, among which, Figure 1 UEs 111, UE 112, and UE 113 form UECS 108. Similar to exemplary environment 600, an E2E ML controller (e.g., E2E ML controller 274) determines the E2E ML configuration that forms a unidirectional E2E DNN for handling uplink UECS communication. In alternative or additional embodiments, the E2E ML controller 318 of core network server 302 determines the E2E ML configuration.

[0076] In the aspect of DNN processing for UECS, the coordinating UE (e.g., UE 111) uses a portion of the E2E ML configuration to form multiple sub-DNNs. For illustration, the coordinating UE forms a joint transport processing DNN 702 (joint TX processing DNN 702), a first transport DNN 704 (TX DNN 704), and a second transport DNN 706 (TX DNN 706). The joint TX processing DNN 702 receives and processes uplink user plane data and / or control plane information from the target UE. For example, as referenced... Figure 6 The target UE (e.g., UE 112 or UE 113) includes a sidelink TX DNN (e.g., TX DNN 614, TX DNN 616) for transmitting uplink UECS communication (e.g., uplink user plane data, control plane information) to the coordinating UE via a local wireless network connection (e.g., local wireless network connection 134, local wireless network connection 135). The coordinating UE then uses a joint TX processing DNN 702 to receive and process the uplink UECS communication. For illustration, the joint TX processing DNN 702 generates multiple outputs, such as a first output pointing to TX DNN 704 and a second output pointing to TX DNN 706.

[0077] The TX DNN 704 receives a first output from the joint TX processing DNN 702 and generates uplink communication for transmission to base station 120 via the (cellular) wireless network. For example, the TX DNN 704 performs at least some transmitter chain operations associated with transmitting user plane data and / or control plane information of the target UE via wireless link 131. Sometimes, the TX DNN 704 applies timing adjustments to the uplink transmission.

[0078] TX DNN 706 operates as a sidelink TX DNN by receiving a second output from the joint TX processing DNN 702 and generating one or more communications for transmission via local wireless network connections 134 and 135. As an example, TX DNN 706 performs at least some transmitter chain operations associated with transmitting a first communication to UE 112 via local wireless network connection 134 and at least some transmitter chain operations associated with transmitting a second communication to UE 113 via local wireless network connection 135.

[0079] Non-coordinated UEs in UECS 108 (e.g., UE 112, UE 113) each form a receive DNN, respectively labeled Receive DNN 708 (RX DNN 708) and Receive DNN 710 (RX DNN 710), which operate as side-link RX DNNs for processing uplink UECS communication received from the coordinated UE via the local wireless network. The non-coordinated UEs also form corresponding transmit DNNs, labeled Transmit DNN 712 (TX DNN 712) and Transmit DNN 714 (TX DNN 714), for transmitting uplink UECS communication to base station 120 via the wireless network using radio link 131. In all aspects, TX DNN 712 and / or 714 will apply timing adjustments to the uplink transmission. Therefore, similar to reference... Figure 6 The UE in UECS 108 forms a first sub-DNN set 716 for processing communications transmitted with base station 120 via wireless link 131 and a second sub-DNN set 718 for processing communications transmitted via one or more side links (e.g., local wireless network).

[0080] Base station 120 forms receive DNN 720 (RX DNN 720), which receives at least some of the uplink transmissions from various UEs of UECS 108. In various aspects, RX DNN 720 combines and / or aggregates uplink transmissions to extract uplink user plane data and / or control plane information originating from the target UE. Alternatively or additionally, RX DNN 720 performs receiver chain operations, such as referencing... Figure 4 Those described are for receiving uplink transmissions.

[0081] Environment 700 typically refers to the functions performed by base station 120 as base station 722. While exemplary environment 700 illustrates a single base station (e.g., base station 120) performing the operations of base station 722, alternative or additional implementations can utilize multiple base stations.

[0082] Figure 8 An exemplary environment 800 for implementing DNN processing for UECS is illustrated, according to various aspects. Environment 800 shows... Figure 7 An exemplary implementation of base station 722 includes multiple base stations: base station 802, base station 804, and base station 806, each base station representing Figure 1An example of base station 120. In various aspects, base stations 802, 804, and 806 perform Coordinated Multipoint (CoMP) communication with the target UE and via the UECS. As an example, base stations 802, 804, and 806 form an Activity Coordination Set (ACS) for joint communication (joint transmission, joint reception) with the UECS. In various implementations, the ACS may be a component of a user-centric cellless (UCNC) network architecture or used to implement a user-centric cellless (UCNC) network architecture.

[0083] In some aspects, the E2E ML controller (e.g., E2E ML controller 274, E2E ML controller 318) determines the E2E ML configuration based on the UECS communicating with multiple base stations. In environment 800, the E2E ML controller determines the E2E ML configuration that forms a unidirectional E2E DNN for handling uplink UECS communication, and uses portions of the E2E ML configuration to instruct each base station to form one or more sub-DNNs. In other words, a portion of the E2E ML configuration forms a set of sub-DNNs distributed across multiple base stations performing CoMP communication with the UECS, as shown in environment 800. For example, base stations 802, 804, and 806 each form a corresponding receive DNN (e.g., RX DNN 808, RX DNN 810, RX DNN 812) for receiving and processing communication transmitted via radio link 131. Base stations 802, 804, and 806 also form corresponding inter-base station DNNs for forwarding or receiving inter-base station communication. In environment 800, base station 802 acts as the master base station of the ACS formed by base stations 802, 804, and 806. For uplink UECS communication, base station 802 forms an inter-base station DNN 814 (inter-BS DNN 814) for receiving and processing communications from other base stations via interface(s) 106. To transmit communications via interface(s) 106, base stations 804 and 806 each form a sub-DNN labeled as inter-base station DNN 816 (inter-BS DNN 816) and inter-base station DNN 818, respectively.

[0084] Base station 802, serving as the primary base station for ACS, forms a joint RX processing DNN 820 that aggregates communications from other base stations. As shown in environment 800, the joint RX processing DNN 820 receives a first input associated with uplink UECS communication from RX DNN 808 via inter-BS DNN 814, and a second and third input associated with uplink UECS communication from base stations 804 and 806.

[0085] Using E2E DNNs to process UECS communications over E2E communication links allows network entities to dynamically determine and / or adjust the E2E ML configuration of the E2E DNN based on various factors and information, such as the ML capabilities of the devices participating in the UECS, changes in the devices participating in the UECS, signal and / or link quality indicators, performance requirements, and available radio network resources. In some aspects, network entities determine partitionable E2E ML configurations to distribute E2E DNN processing, such as by instructing devices with fewer resources to form DNNs with less processing (e.g., less data, less memory, fewer CPU cycles, fewer nodes, fewer layers) relative to devices with more processing resources and / or memory. Dynamic adaptation and partitioning allow network entities to modify the E2E DNN based on changing factors and improve the performance of the E2E communication link (e.g., higher resolution, faster processing, lower bit error rate, improved signal quality, improved latency).

[0086] Adaptation for E2E DNN for handling UECS communication

[0087] Figure 9 and Figure 10 The diagram illustrates an exemplary signaling and control transaction diagram between a core network server, base station, coordinating user equipment, and at least one other user equipment, based on one or more aspects of DNN processing for UECS (e.g., UECS 108), such as aspects adapted for E2E DNN processing of UECS communications. The operation of the signaling and control transactions can be performed by... Figure 3 Core network server 302 and / or Figure 1 The base station 120, the coordinating UE 111, and at least one other UE (typically shown as) Figure 1 UE 110) uses as referenced Figures 1 to 8 To execute any aspect described herein.

[0088] The first example of signaling and control transactions for DNN processing in UECS is provided by Figure 9 The signaling and control transactions are shown in Figure 900. In Figure 900, base station 120 and / or core network server 302 select and adapt the E2E ML configuration to form the E2E DNN for processing UECS communications.

[0089] As shown, at 905, base station 120 and / or core network server 302 receive information from one or more UEs participating in the UECS (such as coordinating UE 111 and one or more UEs 110). As an example, base station 120 receives UE capabilities from coordinating UE 111 and / or one or more UEs 110, such as in response to transmitting a UE capability query message (not shown). Sometimes, coordinating UE 111 and / or one or more UEs 110 transmit indications of ML capabilities (e.g., supported ML architectures, supported tiers, available processing power, memory limits, available power budget, fixed-point processing versus floating-point processing, maximum kernel size capability, computing power). Alternatively or additionally, coordinating UE 111 and / or one or more UEs 110 transmit signal and / or link quality parameters, estimated UE location (e.g., average estimated location in the UECS, estimated location for each UE included in the UECS), battery level, temperature, etc. In some cases, base station 120 forwards UE capabilities, signal quality parameters, link quality parameters, etc., to core network server 302.

[0090] At 910, base station 120 and / or core network server 302 select an initial E2E ML configuration based on information received at 905 via the E2E ML controller. For example, base station 120 and / or core network server 302 select the initial E2E ML configuration from a neural network table based on UE and / or ML capabilities, the number of UEs participating in the UECS, the estimated location of the target UE and / or additional UEs in the UECS, signal quality, etc. When selecting the initial E2E ML configuration, base station 120 and / or core network server 302 sometimes determine partitions of the E2E ML configuration, such as partitions based on UE and / or ML capabilities.

[0091] In some aspects, base station 120 and / or core network server 302 are based on one or more base stations participating in the E2E communication link (such as reference 120 and / or core network server 302). Figure 8 The initial E2E ML configuration is selected by referring to one or more base stations. In some aspects, the selected E2E ML configuration corresponds to the formation of downlink UECS communication for processing via the E2E communication link (as described in reference). Figure 6 The above) or used for processing uplink UECS communication (as referenced) Figure 7 The E2E DNN is configured as a unidirectional E2E ML configuration. Otherwise, the E2E ML configuration corresponds to a bidirectional E2E ML configuration that forms the E2E DNN for handling bidirectional UECS communication.

[0092] At 915, base station 120 or core network server 302 instructs one or more UEs to form one or more sub-DNNs based on one or more portions of the E2E ML configuration. For example, E2E ML controller 274 or 318 partitions the E2EML configuration across multiple devices and determines a corresponding entry in the neural network table for each partition, wherein each entry indicates the architecture and / or parameter configuration for the sub-DNN. Base station 120 then transmits, for example via a transmission through the control channel, an indication of one or more indices in the neural network table (e.g., neural network table 216) to coordinating UE 111 and / or one or more UEs 110. For example, base station 120 indicates the index to coordinating UE 111 using Layer 1 signaling and / or Layer 2 messaging, and coordinating UE 111 forwards the index to other UEs in the UECS using the local radio network.

[0093] At positions 920, 925, and 930, base station 120, coordinating UE 111, and one or more UEs 110 form one or more corresponding sub-DNNs for processing UECS communication. For illustration, base station 120, coordinating UE 111, and / or one or more UEs 110 access corresponding neural network tables to obtain one or more parameters and / or architectures, as referenced. Figure 5 As described above. In some cases, base station 120 forms a TXDNN (e.g., TX DNN 602) for communicating downlink UECS communication to coordinating UE 111 and one or more UE 110, and / or an RX DNN (e.g., RX DNN 720) for receiving uplink UECS communication from coordinating UE 111 and one or more UE 110. Sometimes, base station 120 forms multiple sub-DNNs, such as sub-DNNs for inter-BS communication (e.g., inter-BS DNN 814, inter-BS DNN 816, inter-BS DNN 818) and / or for processing CoMP communication (e.g., joint RX processing DNN 820).

[0094] The coordinating UE 111 forms at least a first sub-DNN (e.g., RX DNN 604, TX DNN 704) for handling radio network communication with the base station, at least a second sub-DNN (e.g., RX DNN 618, TX DNN 706) for handling local radio network communication with other UEs, and at least a third sub-DNN (e.g., joint RX processing DNN 620, joint TX processing DNN 702) for joint processing. One or more UEs 110 acting as participating UEs in the UECS form at least a first sub-DNN (e.g., RX DNN 606, RX DNN 608, TX DNN 712, TX DNN 714) for handling radio network communication with the base station, and at least a second sub-DNN (e.g., TX DNN 614, TX DNN 616, RX DNN 708, RX DNN 710) for handling local radio network communication with the coordinating UE.

[0095] At 935, base station 120, coordinating UE 111, and / or one or more UEs 110 use an E2E DNN formed according to the E2E ML configuration selected at 910 to process UECS communications transmitted via the E2E communication link. For example, refer to Figures 6 to 8 E2E DNN (via sub-DNN) handles uplink and / or downlink UECS communication.

[0096] At 940, the UE 111 (via base station 120) is coordinated to transmit feedback regarding UECS communication to base station 120 and / or core network server 302. For example, the UE 111 is coordinated to transmit signal and / or link quality parameters (e.g., RSSI, power information, SINR, RSRP, CQI, CSI, Doppler feedback, BLER, HARQ, timing measurement, error measurement, etc.) to base station 120. As another example, the UE 111 is coordinated to transmit UE capabilities and / or ML capabilities.

[0097] At 945, base station 120 and / or core network server 302 (via the E2E ML controller) identify adjustments to the E2E ML configuration based on feedback. The identified adjustments can include any combination of architectural changes and / or parameter changes to the E2E ML configuration as further described, such as minor changes involving updating parameters and / or major changes involving reconfiguring node and / or layer connections of the E2E DNN. For example, based on feedback received at 940, base station 120 determines to add and / or remove a UE from the UECS, and determines the adjustment based on the change in participating UEs in the UECS. As another example, base station 120 determines to change the coordinating UE from coordinating UE 111 to another UE, and determines the adjustment based on the change in the coordinating UE. Alternatively or additionally, base station 120 determines the adjustment based on channel impairments identified by the feedback.

[0098] In response to the identified adjustment, the diagram is executed at 950, and at 915, base station 120 and / or core network server 302 instruct one or more UEs to form one or more sub-DNNs based on one or more portions of the adjusted E2E ML configuration. This allows network entities to dynamically adapt the E2E DNN and how the E2E DNN handles UECS communications to optimize (and re-optimize) processing in the event of changes in the operating environment (e.g., changes in channel conditions, changes in participating UEs, changes in coordinating UEs).

[0099] A second example of signaling and control transactions for DNN processing in UECS is provided by Figure 10 The signaling and control transactions are illustrated in Figure 1000. In Figure 1000, the coordination of the UE adaptation is part of the E2E ML configuration for handling UECS communications. In some aspects, the signaling and control transaction operations shown in Figure 1000 are a continuation of or in combination with the signaling and control transactions shown in Figure 900.

[0100] As shown, Figure 1000 begins at Figure 9 In step 935, base station 120, coordinating UE 111, and one or more UEs 110 use a partitioned E2E DNN to process UECS communications transmitted via E2E communication links. At step 1005, coordinating UE 111 determines at least one sub-DNN for adjusting the transmission of UECS communications based on the local wireless network via the E2E communication link. For example, UEs 111, 112, and 113 form sub-DNNs involving communications exchanged between each other using local wireless network connections 134 and / or 135, such as... Figure 9As described at positions 925 and 930. Alternatively or additionally, the coordinating UE 111 determines to adjust the joint processing DNN, such as the joint RX processing DNN 620 and / or the joint TX processing DNN 702. For illustration, in response to identifying that the packet error rate of at least one of the participating UEs has exceeded an acceptable performance threshold, the coordinating UE 111 determines to adjust at least one sub-DNN used for processing UECS communications based on the local radio network. At other times, the UE 111 determines to periodically adjust at least one sub-DNN.

[0101] At points 1010 and 1015, the coordinating UE 111 and base station 120 confirm the sub-DNN training schedule. In some aspects, the coordinating UE 111 requests the training schedule from base station 120, and base station 120 confirms the start time and / or duration of the sub-DNN training to the coordinating UE 111. Alternatively or additionally, the coordinating UE 111 indicates a proposed training schedule to base station 120, and base station 120 confirms the proposed training schedule and / or provides an alternative training schedule to the coordinating UE 111. Therefore, at points 1010 and 1015, the coordinating UE 111 and base station 120 negotiate the sub-DNN training schedule.

[0102] At 1020, base station 120 instructs one or more UEs included in the UECS to maintain a fixed architecture ML configuration for one or more sub-DNNs. For example, base station 120 instructs one or more UEs to maintain a fixed ML configuration (e.g., architecture and / or parameters) for one or more sub-DNNs (e.g., RX DNN 604, RX DNN 606, RX DNN 608, TX DNN 704, TX DNN 712, TX DNN 714) used to process communication with the wireless network-based UECS of base station 120. Base station 120 can explicitly instruct one or more UEs to maintain the fixed ML configuration, as shown at 1020, or implicitly instruct one or more UEs to maintain the fixed ML configuration by confirming the training schedule at 1015. In some aspects, the base station explicitly instructs coordinating UE 111 to maintain the fixed ML architecture, and at 1025, coordinating UE 111 forwards the instruction to maintain the fixed ML architecture to one or more UEs 110. However, at other times, base station 120 explicitly (and individually) instructs each UE in the UECS to maintain a fixed ML architecture. Alternatively or additionally, base station 120 suppresses the transmission of E2E ML configuration adjustments to coordinating UE 111 and / or one or more UE 110 based on an established training schedule.

[0103] At 1030, the coordinating UE 111 initiates a training process with one or more UEs 110, such as a training process to measure reference signals transmitted via local radio network connections 134 and 135. Alternatively or additionally, the coordinating UE 111 requests one or more UEs 110 to forward signal and / or link quality parameters associated with the local radio network connection, such as by sending a request message to each of the one or more UEs 110 via the respective local radio network connection or by broadcasting a request message via the local radio network. For example, the coordinating UE 111 requests one or more UEs 110 to forward RSSI, Link Quality (LQ), Transmit Power Link (TPL), and / or Received Power (RX) parameters based on the one or more local radio network connections. In some aspects, the coordinating UE 111 requests estimated distance and / or location information from each of the one or more UEs 110. Accordingly, at 1035, one or more UEs 110 forward feedback to the coordinating UE 111 via the local radio network.

[0104] At 1040, the coordinating UE 111 determines adjustments to at least one sub-DNN associated with the local wireless network. For example, similar to... Figure 9 As described at 945, the coordinating UE 111 determines architectural and / or parameter changes to one or more sub-DNNs associated with the local wireless network, such as minor changes involving updating parameters and / or major changes involving reconfiguring the nodes and / or layer connections of one or more sub-DNNs. In various aspects, the coordinating UE 111 analyzes the neural network table based on the feedback received at 1035 and identifies adjustments from that neural network table. As another example, the coordinating UE 111 forwards the feedback to the base station 120, such as in Figure 9 As stated at point 940, and an adjustment was requested from base station 120.

[0105] At 1045, coordinating UE 111 instructs one or more UEs 110 to update one or more sub-DNNs for transmitting UECS communications based on the local wireless network via an E2E communication link, such as via an index transmitted to a neural network table through the local wireless network connection. Alternatively or additionally, base station 120 instructs the UE to use, as in Figure 9A similar technique described at 915 is used to update the sub-DNN (not shown). In response to receiving an instruction at 1045 (e.g., from coordinating UE 111 or base station 120), coordinating UE 111 and / or one or more UEs 110 update one or more sub-DNNs at 1050, and at 1055, base station 120, coordinating UE 111, and one or more UEs 110 use the updated E2E DNN to process UECS communications transmitted over the E2E communication link, as described at 935. This allows coordinating UE 111 and / or base station 120 to dynamically adapt the processing of the E2E DNN to a portion of the UECS communications based on the local wireless network, to optimize (and re-optimize) processing in the event of changes in the operating environment (e.g., changes in channel conditions, changes in participating UEs, changes in coordinating UEs). Optionally, at 1060, coordinating UE 111 indicates to base station 120 that training has been completed.

[0106] Exemplary methods

[0107] Based on one or more aspects of DNN processing used in UECS, refer to Figure 11 and Figure 12 Exemplary methods 1100 and 1200 are described.

[0108] Figure 11 An exemplary method 1100 for performing aspects of DNN processing for UECS is illustrated. In some embodiments, the operation of method 1100 is performed by network entities such as base station 120 and / or core network server 302.

[0109] At 1105, the network entity identifies multiple devices participating in UECS communications transmitted via the E2E communication link. For example, base station 120 identifies UE 111, UE 112, and UE 113 as UEs to be included in the UECS. Alternatively or additionally, base station 120 identifies and instructs UE 111 to act as the coordinating UE of the UECS. In various aspects, the multiple devices participating in UECS communications transmitted via the E2E communication link include at least one base station, a coordinating user equipment (UE), and at least one additional UE, as referenced. Figures 6 to 8 As stated above.

[0110] At 1110, the network entity selects the E2E ML configuration for the E2E DNN used to process UECS communications. For example, the core network server 302 or base station 120 selects the configuration via the E2E ML controller, as shown in... Figure 9The E2E ML configuration is described at 910. In some aspects, the core network server and / or base station 120 receive information from one or more UEs (e.g., UE 111, UE 112, UE 113) forming the UECS, such as ML capabilities and / or signal and / or link quality parameters, as shown in... Figure 9 As described at 905. This can include selecting an E2E ML configuration for unidirectional UECS communication (e.g., downlink only, uplink only) or bidirectional UECS communication, as referenced. Figures 6 to 8 As stated above.

[0111] At 1115, the network entity instructs each of the multiple devices participating in the UECS to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN for transmitting UECS communication via the E2E communication link in the E2E DNN. For example, core network server 302 or base station 120 instructs UE 111, UE 112, and / or UE 113 to form multiple sub-DNNs, as shown in... Figure 9 As described at 925 and 930. Alternatively or additionally, core network server 302 and / or another base station instruct base station 120 to form one or more sub-DNNs, as in Figure 9 As stated at point 920.

[0112] At 1120, the network entity receives feedback associated with UECS communication from at least one of a plurality of devices. For illustration, base station 120 receives signal and / or link quality parameters from UE 111, such as in... Figure 9 As described at point 940. Alternatively or additionally, the core network server 302 receives signals and / or link quality parameters from the base station 120.

[0113] At 1125, network entities identify adjustments to the E2E ML configuration based on feedback. For example, the core network server or base station 120 analyzes neural network tables (e.g., neural network table 316, neural network table 272) based on feedback and identifies adjustments to the E2E ML configuration, such as architecture adjustments and / or parameter changes.

[0114] At 1130, the network entity instructs at least one of the multiple devices participating in the UECS to update the corresponding sub-DNN of the E2E DNN based on adjustments. For illustration, base station 120 uses a control channel to transmit instructions for adjustments to UE 111, UE 112, and / or UE 113, as shown in... Figure 9 As described at 915, instructions for indexes in a neural network table are transmitted, for example, by using layer 1 signaling and / or layer 2 messaging.

[0115] In some aspects, method 1100 is repeated iteratively, as indicated at 1135, such as when a network entity receives new feedback and / or information as described at 1120, wherein the feedback and / or information indicates a change in the E2E ML configuration used for the E2E DNN. This allows the network entity to dynamically adapt the DNN and how the DNN handles communication over the E2E communication link to optimize (and re-optimize) processing when UECS communication changes.

[0116] Figure 12 An exemplary method 1200 for performing aspects of DNN processing for UECS is illustrated. In some embodiments, the operation of method 1200 is performed by a coordinating user equipment of UECS, such as UE 111.

[0117] At point 1205, the UE confirms the training schedule with the network entity. This training schedule indicates a fixed second portion of the time period for maintaining the E2E ML configuration. This fixed second portion forms a second set of sub-DNNs (e.g., sub-DNN set 610, sub-DNN set 716) for transmitting UECS communication based on the wireless network via the E2E communication link. For example, UE111 coordinates with base station 120 to confirm the training schedule, as in... Figure 10 As mentioned at positions 1010 and 1020.

[0118] At 1210, the UE determines adjustments to a first portion of the E2E ML configuration based on a training schedule. This first portion forms a first set of sub-DNNs (e.g., sub-DNN set 612, sub-DNN set 718) for transmitting UECS communication based on the local wireless network via the E2E communication link. For example, the coordinating UE 111 determines the adjustments by initiating a training process and analyzing the neural network table, as described at 1040. Alternatively or additionally, the coordinating UE 111 initiates a training process and forwards signal and / or link quality metrics to base station 120, similar to... Figure 9 As described at point 940, and receiving adjustments from the base station, similar to in Figure 9 As stated at point 915.

[0119] At 1215, the UE instructs one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first sub-DNN set using adjustments to the first part of the E2E ML configuration. For example, coordinating UE 111 instructs one or more UEs participating in the UECS (e.g., UE 112, UE 113) to update one or more sub-DNNs, as in Figure 10 As stated at point 1050.

[0120] In some aspects, method 1200 is repeated iteratively, as indicated at 1220. For example, coordinating UE 111 determines to adjust the sub-DNN based on signal and / or link quality parameters failing to meet acceptable performance thresholds or based on a periodic schedule indicating when to evaluate and / or adjust the sub-DNN. This iterative process allows network entities to dynamically adapt the DNN and how the DNN handles UECS communications to optimize (and re-optimize) processing as conditions change.

[0121] The order of the method blocks describing methods 1100 and 1200 is not intended to be construed as limiting, and any number of the described method blocks can be skipped or combined in any order to implement the method or an alternative method. Generally, any of the components, modules, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Some operations of the exemplary methods can be described in the general context of executable instructions stored on computer-readable storage memory local and / or remote on a computer processing system, and implementations can include software applications, programs, functions, etc. Alternatively or additionally, any of the functions described herein can be executed, at least in part, by one or more hardware logic components such as, but not limited to, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chips (SOCs), complex programmable logic devices (CPLDs), etc.

[0122] Several examples are described below: Example 1: A method performed by a network entity for determining an end-to-end (E2E) machine learning (ML) configuration, the E2E ML configuration forming an E2E deep neural network (DNN) for processing User Equipment Coordination Set (UECS) communications transmitted over an E2E communication link in a wireless network, the method comprising: the network entity selecting the E2E ML configuration for forming the E2E DNN for processing the UECS communications; instructing each of a plurality of devices participating in the UECS to use at least a portion of the E2E ML configuration to form a corresponding sub-DNN of the E2E DNN for transmitting the UECS communications over the E2E communication link, the plurality of devices including at least one base station, a coordinating user equipment (UE), and at least one additional UE; receiving feedback associated with the UECS communications from at least one of the plurality of devices; identifying adjustments to the E2E ML configuration based on the feedback; and instructing at least one of the plurality of devices participating in the UECS to update its corresponding sub-DNN of the E2E DNN based on the adjustments.

[0123] Example 2: The method described in Example 1, wherein the feedback includes one or more of the following: Received Signal Strength Indicator (RSSI), power information, Signal-to-Interference-Ratio (SINR) information, Reference Signal Received Power (RSRP), Channel Quality Indicator (CQI) information, Channel State Information (CSI), Doppler feedback, Block Error Rate (BLER), Quality of Service (QoS), Hybrid Automatic Repeat Request (HARQ) information, Uplink SINR, Timing Measurement, Error Measurement, UE Capability, or ML Capability.

[0124] Example 3: The method as described in Example 1, wherein instructing each of the plurality of devices to form a corresponding sub-DNN of the E2E DNN further includes at least one of the following: instructing the at least one base station to form a first sub-DNN of the E2E DNN, the first sub-DNN processing downlink UECS communication to the UECS via the wireless network; instructing the coordinating UE to form a second sub-DNN and a third sub-DNN of the E2E DNN, the second sub-DNN processing downlink UECS communication received from the at least one base station via the wireless network, the third sub-DNN processing local wireless network communication received from the at least one additional UE via the local wireless network; and instructing the at least one additional UE to form a fourth sub-DNN and a fifth sub-DNN, the fourth sub-DNN processing downlink UECS communication received from the at least one base station via the wireless network, the fifth sub-DNN processing inputs from the fourth sub-DNN and generating the local wireless network communication transmitted to the coordinating UE.

[0125] Example 4: The method as described in Example 3, wherein instructing each of the plurality of devices to form a corresponding sub-DNN of the E2E DNN further includes: instructing the coordinating UE to form a sixth sub-DNN, the sixth sub-DNN jointly processing the local wireless network communications from the at least one additional UE and the downlink UECS communications received from the at least one base station through the wireless network.

[0126] Example 5: The method as described in Example 3, wherein the fifth sub-DNN receives in-phase and quadrature (IQ) samples of radio signals carrying downlink UECS communication from the at least one base station from the fourth sub-DNN via the wireless network, and forwards the IQ samples to the coordinating UE using the local wireless network communication.

[0127] Example 6: The method as described in Example 1, wherein selecting the E2E ML configuration for forming the E2EDNN for processing the UECS communication further includes: receiving device capabilities from at least one of the plurality of devices; and determining the E2E ML configuration based on the device capabilities.

[0128] Example 7: The method as described in Example 6, wherein the device capabilities include one or more of the following: supported ML architecture, supported number of layers, available processing power, memory limit, available power budget, fixed-point processing and floating-point processing, maximum kernel size capability, computing power, battery power, temperature, or estimated UE location.

[0129] Example 8: The method as described in Example 1, wherein identifying the adjustment to the E2E ML configuration further comprises: determining, based on the feedback, to remove at least one UE participating in the UE; and identifying the adjustment based on determining that at least one UE participating in the UE is to be removed.

[0130] Example 9: The method as described in Example 1, wherein instructing each of the plurality of devices to form a corresponding sub-DNN of the E2E DNN further comprises: instructing each of the plurality of devices to indicate at least the said portion of the E2E ML configuration by instructing at least one of the following: an ML architecture for the corresponding sub-DNN; or one or more ML parameters for the corresponding sub-DNN. Examples of ML architectures include multiple layers, layer downsampling configurations, adding or removing fully convolutional layers, adding or removing recurrent neural network layers, interconnections between neural network layers, the number and / or configuration of hidden layers, the number of nodes, pooling configurations, input layer architecture, and output layer architecture. Examples of ML parameters include coefficients, weights, kernel parameters, and biases.

[0131] Example 10: The method as described in Example 9, wherein instructing at least the portion of the E2E ML configuration further comprises: transmitting an instruction to at least one of the plurality of devices to update at least the portion of the E2E ML configuration via a control channel.

[0132] Example 11: The method as described in Example 9, wherein indicating at least the portion of the E2E ML configuration further comprises: indicating a first ML architecture to the first additional UE based on one or more ML capabilities of the first additional UE among the at least one additional UE; and indicating a second ML architecture to the second additional UE based on one or more ML capabilities of the second additional UE among the at least one additional UE, the second ML architecture being different from the first ML architecture.

[0133] Example 12: The method as described in Example 1, wherein selecting the E2E ML configuration for forming the E2EDNN for processing the UECS communication further includes: selecting the E2E ML configuration to form the E2E DNN for processing uplink UECS communication.

[0134] Example 13: The method as described in Example 1, wherein instructing each of the plurality of devices to form a corresponding sub-DNN of the E2E DNN further includes at least one of the following: instructing the at least one base station to form a first sub-DNN of the E2E DNN, the first sub-DNN processing uplink UECS communication from the UECS via the wireless network; instructing the coordinating UE to form a second sub-DNN and / or a third sub-DNN of the E2E DNN, the second sub-DNN processing uplink UECS communication directed to the at least one base station and used for transmission via the wireless network, the third sub-DNN processing local wireless network communication directed to the at least one additional UE and used for transmission via the local wireless network; and instructing the at least one additional UE to form a fourth sub-DNN and / or a fifth sub-DNN, the fourth sub-DNN processing uplink UECS communication directed to the at least one base station and used for transmission via the wireless network, the fifth sub-DNN processing input received from the coordinating UE via the local wireless communication network and generating input to the fourth sub-DNN.

[0135] Example 14: The method as described in any of Examples 1 to 13, wherein the network entity is the at least one base station or core network server.

[0136] Example 15: A network entity comprising: at least one processor; and a computer-readable storage medium including instructions that, in response to execution by the at least one processor, instruct the network entity to perform a method for determining an end-to-end (E2E) machine learning (ML) configuration, the E2E ML configuration forming an E2E deep neural network (DNN) for processing User Equipment Coordination Set (UECS) communications transmitted over an E2E communication link in a wireless network, the method comprising: identifying by the network entity a plurality of devices participating in the UECS communications transmitted over the E2E communication link, the plurality of devices including at least one base station, a coordinating user equipment (UE), and at least one additional UE, the coordinating UE and the at least one additional UE forming a UECS; selecting, based on the plurality of devices, the E2E ML configuration for forming the E2E DNN for processing the UECS communications; and instructing each of the plurality of devices to use at least a portion of the E2E ML configuration to form at least a portion of the E2E DNN for transmitting the UECS communications over the E2E communication link.

[0137] Example 16: A network entity as described in Example 15, wherein the computer-readable storage medium includes additional instructions that, in response to execution by the at least one processor, instruct the network entity to perform at least the portion of instructing each of the plurality of devices to form the E2E DNN by at least one of the following: instructing the at least one base station to form a first sub-DNN of the E2E DNN, the first sub-DNN processing downlink communications transmitted to the UECS via the wireless network; instructing the coordinating UE to form a second sub-DNN of the E2E DNN, the second sub-DNN processing downlink UECS communications received from the at least one base station via the wireless network; and instructing the at least one additional UE to form a third sub-DNN, the third sub-DNN processing downlink UECS communications received from the at least one base station via the wireless network.

[0138] Example 17: A network entity as described in Example 16, wherein the computer-readable storage medium includes additional instructions that, in response to execution by the at least one processor, instruct the network entity to perform at least the portion of instructing each of the plurality of devices to form the E2E DNN by at least one of the following: instructing the at least one additional UE to form a fourth sub-DNN, the fourth sub-DNN processing input from the third sub-DNN and transmitting local radio network communications to the coordinating UE; instructing the coordinating UE to form a fifth sub-DNN, the fifth sub-DNN processing the local radio network communications received from the at least one additional UE; and instructing the coordinating UE to form a sixth sub-DNN, the sixth sub-DNN jointly processing the local radio network communications from the at least one additional UE and the downlink UECS communications received from the at least one base station via the radio network.

[0139] Example 18: A network entity as described in Example 17, wherein the fourth sub-DNN receives, via the wireless network, in-phase and quadrature (IQ) samples of radio signals carrying downlink UECS communications from the at least one base station from the third sub-DNN, and transmits the IQ samples to the coordinating UE using the local wireless network communications.

[0140] Example 19: A network entity as described in Example 15, wherein the computer-readable storage medium includes additional instructions that, in response to execution by the at least one processor, instruct the network entity to perform the following operation: selecting the E2E ML configuration for forming the E2E DNN for processing the UECS communication; selecting the E2EML configuration to form the E2E DNN for processing uplink UECS communication.

[0141] Example 20: A network entity as described in Example 19, wherein the at least one base station includes a plurality of base stations configured to perform Coordinated Multipoint (CoMP) communication with the UECS, and wherein the computer-readable storage medium includes additional instructions that, in response to execution by the at least one processor, instruct the network entity to perform the following operation: selecting the E2E ML configuration for forming the E2E DNN for processing the uplink UECS communication; selecting the E2E ML configuration to use a first portion of the E2E ML configuration to form a set of sub-DNNs distributed across the plurality of base stations.

[0142] Example 21: A network entity as described in Example 20, wherein the computer-readable storage medium includes additional instructions that, in response to execution by the at least one processor, instruct the network entity to perform the following operations: instruct each of the plurality of devices to form a corresponding sub-DNN: instruct the master base station among the plurality of base stations to form: a receiving sub-DNN (RX sub-DNN) that processes at least a portion of the uplink UECS communication received by the master base station through the wireless network and generates a first output; an inter-base station sub-DNN (BS sub-DNN) that receives and processes input from a first base station among the plurality of base stations and generates a second output; and a joint receiving processing sub-DNN (joint RX processing sub-DNN) that jointly processes the first output and the second output to recover the uplink UECS communication.

[0143] Example 22: A network entity as described in Example 15, wherein the computer-readable storage medium includes instructions that, in response to execution by the at least one processor, instruct the network entity to perform operations as part of the method, the operations including: receiving feedback from at least one of the plurality of devices; determining an adjustment to the E2E ML configuration based on the feedback; instructing at least one of the plurality of devices to the adjustment to the E2E ML configuration; and instructing at least one of the plurality of devices to use the adjustment to update its corresponding sub-DNN.

[0144] Example 23: A network entity as described in any one of Examples 15 to 22, wherein the network entity is the at least one base station or core network server.

[0145] Example 24: A method performed by a coordinating user equipment (UE) in a user equipment coordination set (UECS) for determining at least a first portion of an end-to-end (E2E) machine learning (ML) configuration, the first portion forming a first set of sub-deep neural networks (sub-DNNs) of an E2E DNN, the first set of sub-DNNs transmitting UECS communications over a local radio network via an E2E communication link, the method comprising: confirming a training schedule with a network entity, the training schedule indicating a period of time for maintaining a second portion of the E2E ML configuration at a fixed time, the second portion forming a second set of sub-DNNs of the E2E DNN, the second set of sub-DNNs transmitting UECS communications over a cellular network via the E2E communication link; determining adjustments to the first portion of the E2E ML configuration based on the training schedule; and instructing one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first set of sub-DNNs using the adjustments to the first portion of the E2E ML configuration.

[0146] Example 25: The method as described in Example 24, wherein determining the adjustment includes determining at least one of the following: updating one or more parameters of at least a portion of the E2E ML configuration; or changing one or more ML architectures of at least said portion of the E2E ML configuration.

[0147] Example 26: The method as described in Example 24, wherein using the adjustment to the first portion of the E2E ML configuration to instruct the one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first sub-DNN set further comprises: instructing a first additional UE among the one or more additional UEs to update a first sub-DNN in the first sub-DNN set, the first sub-DNN generating local radio network communication directed to the coordinating UE; or instructing the first additional UE to update a second sub-DNN in the first sub-DNN set, the second sub-DNN receiving local radio network communication from the coordinating UE.

[0148] Example 27: The method as described in Example 24, wherein determining the adjustment of the first portion of the E2E ML configuration further comprises: determining the adjustment of the first portion of the E2E ML configuration, the adjustment forming a joint processing DNN at the coordinating UE.

[0149] Example 28: A computer-readable storage medium including processor-executable instructions that, in response to execution by at least one processor, instruct a device to perform a method as described in any one of Examples 1 to 14 and 24-227.

[0150] Although techniques and apparatus for DNN processing for UECS have been described in feature- and / or method-specific language, it should be understood that the subject matter of the appended claims is not necessarily limited to the specific features or methods described. Rather, the specific features and methods are disclosed as exemplary embodiments of DNN processing for UECS.

Claims

1. A network entity, comprising: At least one processor; and A computer-readable storage medium including instructions that, in response to execution by the at least one processor, cause the network entity to: Identify multiple devices participating in user equipment coordination set (UECS) communication transmitted via end-to-end E2E communication, the multiple devices including at least one base station, a coordinating user equipment (UE) and at least one additional UE, the coordinating UE and the at least one additional UE forming the UECS; Based on the plurality of devices, select an E2E machine learning (ML) configuration for forming an E2E deep neural network (DNN) to process the UECS communication; The training schedule is confirmed with the coordinating UE, the training schedule indicating that at least a first portion of the E2E ML configuration is maintained for a first time period, the at least first portion of the E2E ML configuration forming a first set of sub-DNNs of the E2E DNN; as well as Each of the plurality of devices is instructed to use at least the first portion of the E2E ML configuration to form a corresponding non-overlapping sub-DNN of the E2E DNN that transmits the UECS communication via the E2E communication in the wireless network.

2. The network entity according to claim 1, wherein, In order to instruct each of the plurality of devices to form a corresponding non-overlapping sub-DNN of the E2E DNN, the network entity is further configured as follows: The at least one base station is instructed to form a first sub-DNN of the E2E DNN, the first sub-DNN processing downlink communication transmitted to the UECS through the wireless network; The coordinating UE is instructed to form a second sub-DNN of the E2E DNN, the second sub-DNN processing downlink UECS communications received from the at least one base station via the wireless network; as well as The at least one additional UE is instructed to form a third sub-DNN, which processes downlink UECS communications received from the at least one base station via the wireless network.

3. The network entity according to claim 2, wherein, In order to instruct each of the plurality of devices to form a corresponding non-overlapping sub-DNN of the E2E DNN, the network entity is further configured as follows: The at least one additional UE is instructed to form a fourth sub-DNN, which processes the input from the third sub-DNN and transmits local wireless network communication to the coordinating UE; The coordinating UE is instructed to form a fifth sub-DNN, which processes the local wireless network communications received from the at least one additional UE; and The coordinating UE is instructed to form a sixth sub-DNN, which processes the outputs from the fifth sub-DNN and the outputs from the second sub-DNN.

4. The network entity according to claim 3, wherein, The fourth sub-DNN is configured as follows: Receive in-phase and quadrature IQ samples from the third sub-DNN, carrying wireless signals from the at least one base station via the wireless network for the downlink UECS communication. as well as The IQ sample is transmitted to the coordinating UE using the local wireless network communication.

5. The network entity according to claim 1, wherein, In order to select the E2E ML configuration for forming the E2EDNN used to process the UECS communications, the network entity is further configured as follows: Select the E2E ML configuration to form the E2E DNN for handling uplink UECS communication.

6. The network entity according to claim 5, wherein, The at least one base station includes multiple base stations configured to perform coordinated multipoint CoMP communication with the UECS, and In order to select the E2E ML configuration to form the E2EDNN for processing the uplink UECS communication, the network entity is further configured as follows: Select the E2E ML configuration to use the first part of the E2E ML configuration to form a set of sub-DNNs distributed across the multiple base stations.

7. The network entity according to claim 6, wherein, To instruct each of the plurality of devices to form the corresponding non-overlapping sub-DNN, the network entity is further configured to instruct the primary base station among the plurality of base stations to form a receive sub-DNN—an RX sub-DNN, the RX sub-DNN being configured to process at least a portion of the uplink UECS communication received by the primary base station through the wireless network and generate a first output, and further including: Inter-base station sub-DNN—BS sub-DNN, wherein the BS sub-DNN receives and processes input from a first base station among the plurality of base stations and generates a second output; and Joint Receive Processing Sub-NN – Joint RX Processing Sub-NN, which jointly processes the first output and the second output to restore the uplink UECS communication.

8. The network entity according to claim 1, wherein, The network entity is further configured as follows: Receive feedback from at least a first device among the plurality of devices; The adjustments to the E2E ML configuration are determined based on the feedback. Instruct at least a second device among the plurality of devices to adjust the E2E ML configuration; as well as The device is instructed to use the adjustment to update its corresponding sub-DNN, wherein the first device and the second device are of the same type.

9. The network entity according to any one of claims 1 to 8, wherein, The training schedule indicates a second time period for maintaining a second portion of the E2E ML configuration, the second portion forming a second set of sub-DNNs of the E2E DNN, wherein the network entities are further configured as follows: The system instructs one or more additional UEs participating in the UECS to update one or more sub-DNNs in the first sub-DNN set based on the training schedule using adjustments to the first part of the E2E ML configuration.

10. The network entity according to claim 9, wherein: The network entity is further configured to instruct a first additional UE among the one or more additional UEs to update a first sub-DNN in the first sub-DNN set, the first sub-DNN generating local wireless network communication directed to the coordinating UE; or The network entity is further configured to instruct the first additional UE to update a second sub-DNN in the first sub-DNN set, wherein the second sub-DNN receives the local wireless network communication from the coordinating UE. or The adjustments to the first part of the E2E ML configuration form a joint processing DNN at the coordinating UE.

11. A method performed by a coordinating user equipment (UE), the method comprising: Confirm the training schedule with the network entity, the training schedule indicating a first time period for maintaining at least a first portion of the end-to-end E2E machine learning ML configuration, the at least first portion of the E2E ML configuration forming a first set of sub-DNNs of the E2E deep neural network DNN; The adjustments to the first portion of the E2E ML configuration are determined based on the training schedule; as well as Instruct one or more additional UEs participating in the User Equipment Coordination Set (UECS) to update one or more non-overlapping sub-DNNs in the first sub-DNN set using adjustments to the first portion of the E2E ML configuration.

12. The method according to claim 11, wherein, The one or more additional UEs indicated include: The first additional UE among the one or more additional UEs is instructed to update the first sub-DNN in the first sub-DNN set, the first sub-DNN generating local wireless network communication directed to the coordinating UE; or The first additional UE is instructed to update the second sub-DNN in the first sub-DNN set, and the second sub-DNN receives the local wireless network communication from the coordinating UE.

13. The method according to any one of claims 11 to 12, wherein, Determining the adjustments to the first portion of the E2E ML configuration includes: The adjustment of the first part of the E2E ML configuration is determined at the coordinating UE to form a joint processing DNN.

14. The method of claim 13, further comprising: Form the first sub-DNN for processing wireless network communication with the base station; Form at least a second sub-DNN for handling local wireless network communications with one or more attached UEs; as well as Form at least a third sub-DNN for the joint processing.

15. A coordinating user equipment, comprising: At least one processor; and A computer-readable storage medium including instructions that, in response to execution by the at least one processor, cause the coordinated user equipment to perform the method according to any one of claims 11 to 14.

Images