Method for online training device performing AI / ML-based CSI feedback

Abstract

As described herein, there are one or more techniques relating to training performance of one or more wireless transmit receive units (WTRUs) configured to generate artificial intelligence and / or machine learning based channel state information feedback.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 394,167, filed on August 1, 2022, entitled “Methods for Online Training for Devices Performing AI / ML Based CSI Feedback,” the entire contents of which are incorporated herein by reference. Summary of the invention

[0003] As described herein, there are one or more techniques related to training performance of one or more wireless transmit receive units (WTRUs) configured to generate artificial intelligence and / or machine learning based channel state information feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] A more detailed understanding may be obtained from the following description given by way of example in conjunction with the accompanying drawings, in which like reference numerals indicate like elements, and in which:

[0005] Figure 1A is a system diagram illustrating an example communication system in which one or more disclosed embodiments may be implemented;

[0006] Figure 1B is an example of an embodiment in which Figure 1A A system diagram of an example wireless transmit / receive unit (WTRU) for use within an illustrated communication system;

[0007] Figure 1C is an example of an embodiment in which Figure 1A A system diagram of an example Radio Access Network (RAN) and an example Core Network (CN) used within the illustrated communication system;

[0008] Figure 1D is an example of an embodiment in which Figure 1A A system diagram of another example RAN and another example CN used within the illustrated communication system;

[0009] Figure 2 Examples of configuration for CSI report settings, resource settings, and links are illustrated.

[0010] Figure 3 An example of codebook-based pre-coding using feedback information is illustrated;

[0011] Figure 4An example table illustrating a downlink codebook for 2Tx is shown;

[0012] Figure 5 Illustrate a sample process for online update of AI / ML encoder models;

[0013] Figure 6 Another example process for online updating of AI / ML encoder models is illustrated; and

[0014] Figure 7 An example process for online updating of an AI / ML decoder model is illustrated. DETAILED DESCRIPTION

[0015] Table 1 is a non-exhaustive list of acronyms that may be used in this article.

[0016]

[0017]

[0018]

[0019] Table 1

[0020] Figure 1A 1 is a diagram illustrating an example communication system 100 in which one or more disclosed embodiments may be implemented. The communication system 100 may be a multiple access system that provides content such as voice, data, video, messaging, broadcast, etc. to multiple wireless users. The communication system 100 may enable multiple wireless users to access such content through sharing of system resources (including wireless bandwidth). For example, the communication system 100 may employ one or more channel access methods such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal FDMA (OFDMA), single carrier FDMA (SC-FDMA), zero tail unique word discrete Fourier transform spread OFDM (ZT-UW-DFT-S-OFDM), unique word OFDM (UW-OFDM), resource block filtered OFDM, and filter bank multi-carrier (FBMC), etc.

[0021] like Figure 1AAs shown, the communication system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, a radio access network (RAN) 104, a core network (CN) 106, a public switched telephone network (PSTN) 108, the Internet 110 and other networks 112, but it should be understood that the disclosed embodiments contemplate any number of WTRUs, base stations, networks, and / or network elements. Each of the WTRUs 102a, 102b, 102c, 102d may be any type of device configured to operate and / or communicate in a wireless environment. By way of example, the WTRUs 102a, 102b, 102c, 102d (any of which may be referred to as a station (STA)) may be configured to send and / or receive wireless signals and may include user equipment (UE), a mobile station, a fixed or mobile subscriber unit, a subscription-based unit, a pager, a cellular phone, a personal digital assistant (PDA), a smart phone, a laptop, a netbook, a personal computer, a wireless sensor, a hotspot or Mi-Fi device, an Internet of Things (IoT) device, a watch or other wearable device, a head-mounted display (HMD), a vehicle, a drone, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in an industrial and / or automated process chain environment), consumer electronic devices, and devices operating on a commercial and / or industrial wireless network, etc. Any of the WTRUs 102a, 102b, 102c, and 102d may be interchangeably referred to as a UE.

[0022] The communication system 100 may also include a base station 114a and / or a base station 114b. Each of the base stations 114a, 114b may be any type of device configured to wirelessly interface with at least one of the WTRUs 102a, 102b, 102c, 102d to facilitate access to one or more communication networks, such as the CN 106, the Internet 110, and / or other networks 112. By way of example, the base stations 114a, 114b may be a base transceiver station (BTS), a Node B, an evolved Node B (eNB), a Home Node B, a Home evolved Node B, a Next Generation Node B (such as a gNode B (gNB)), a New Radio (NR) Node B, a site controller, an access point (AP), a wireless router, and the like. Although the base stations 114a, 114b are each depicted as a single element, it should be appreciated that the base stations 114a, 114b may include any number of interconnected base stations and / or network elements.

[0023] Base station 114a may be part of RAN 104, which may also include other base stations and / or network elements (not shown), such as base station controllers (BSC), radio network controllers (RNC), and relay nodes. Base station 114a and / or base station 114b may be configured to send and / or receive wireless signals on one or more carrier frequencies, which may be referred to as cells (not shown). These frequencies may be in a licensed spectrum, an unlicensed spectrum, or a combination of a licensed spectrum and an unlicensed spectrum. A cell may provide coverage of wireless services to a specific geographic area, which may be relatively fixed or may change over time. The cell may be further divided into cell sectors. For example, the cell associated with base station 114a may be divided into three sectors. Therefore, in one embodiment, base station 114a may include three transceivers, i.e., one transceiver for each sector of the cell. In an embodiment, base station 114a may employ multiple-input multiple-output (MIMO) technology and may utilize multiple transceivers for each sector of the cell. For example, beamforming may be used to send and / or receive signals in a desired spatial direction.

[0024] The base stations 114a, 114b may communicate with one or more of the WTRUs 102a, 102b, 102c, 102d over an air interface 116, which may be any suitable wireless communication link (e.g., radio frequency (RF), microwave, centimeter wave, micrometer wave, infrared (IR), ultraviolet (UV), visible light, etc.). The air interface 116 may be established using any suitable radio access technology (RAT).

[0025] More specifically, as noted above, the communication system 100 may be a multiple access system and may employ one or more channel access schemes, such as CDMA, TDMA, FDMA, OFDMA, and SC-FDMA, among others. For example, the base station 114a in the RAN 104 and the WTRUs 102a, 102b, 102c may implement a radio technology, such as Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access (UTRA), that may use Wideband CDMA (WCDMA) to establish the air interface 116. WCDMA may include communication protocols, such as High Speed ​​Packet Access (HSPA) and / or Evolved HSPA (HSPA+). HSPA may include High Speed ​​Downlink (DL) Packet Access (HSDPA) and / or High Speed ​​Uplink (UL) Packet Access (HSUPA).

[0026] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology such as Evolved UMTS Terrestrial Radio Access (E-UTRA) that may establish the air interface 116 using Long Term Evolution (LTE) and / or Advanced LTE (LTE-A) and / or Advanced LTE Pro (LTE-APro).

[0027] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology (such as NR radio access) that may establish the air interface 116 using NR.

[0028] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement multiple radio access technologies. For example, the base station 114a and the WTRUs 102a, 102b, 102c may together implement LTE radio access and NR radio access, for example using the dual connectivity (DC) principle. Thus, the air interface utilized by the WTRUs 102a, 102b, 102c may be characterized by multiple types of radio access technologies and / or transmissions transmitted to / from multiple types of base stations (e.g., eNBs and gNBs).

[0029] In other embodiments, the base station 114a and the WTRUs 102a, 102b, 102c may implement radio technologies such as IEEE 802.11 (i.e., Wireless Fidelity (WiFi)), IEEE 802.16 (i.e., Worldwide Interoperability for Microwave Access (WiMAX)), CDMA2000, CDMA2000 1X, CDMA2000 EV-DO, Interim Standard 2000 (IS-2000), Interim Standard 95 (IS-95), Interim Standard 856 (IS-856), Global System for Mobile Communications (GSM), Enhanced Data rates for GSM Evolution (EDGE), and GSM EDGE (GERAN).

[0030] Figure 1AThe base station 114b in may be, for example, a wireless router, a Home NodeB, a Home eNodeB, or an access point, and may utilize any suitable RAT to facilitate wireless connectivity in local areas such as commercial venues, homes, vehicles, campuses, industrial facilities, sky corridors (e.g., for use by drones), and roads. In one embodiment, the base station 114b and the WTRUs 102c, 102d may implement a radio technology (such as IEEE 802.11) to establish a wireless local area network (WLAN). In an embodiment, the base station 114b and the WTRUs 102c, 102d may implement a radio technology (such as IEEE 802.15) to establish a wireless personal area network (WPAN). In yet another embodiment, the base station 114b and the WTRUs 102c, 102d may utilize a cellular-based RAT (e.g., WCDMA, CDMA2000, GSM, LTE, LTE-A, LTE-A Pro, NR, etc.) to establish a picocell or a femtocell. As Figure 1A As shown, the base station 114 b may have a direct connection to the Internet 110. Therefore, the base station 114 b may not need to access the Internet 110 via the CN 106.

[0031] The RAN 104 may be in communication with the CN 106, which may be any type of network configured to provide voice, data, applications, and / or Voice over Internet Protocol (VoIP) services to one or more of the WTRUs 102a, 102b, 102c, 102d. The data may have different quality of service (QoS) requirements, such as different throughput requirements, latency requirements, error tolerance requirements, reliability requirements, data throughput requirements, mobility requirements, etc. The CN 106 may provide call control, billing services, mobile location-based services, prepaid calling, Internet connectivity, video distribution, etc., and / or perform high-level security functions, such as user authentication. Although not described in detail in the specification, the CN 106 may be configured to provide voice, data, applications, and / or Voice over Internet Protocol (VoIP) services to one or more of the WTRUs 102a, 102b, 102c, 102d. Figure 1A Although not shown in the figure, it will be appreciated that the RAN 104 and / or the CN 106 may be in direct or indirect communication with other RANs that employ the same RAT as the RAN 104 or a different RAT. For example, in addition to being connected to the RAN 104, which may utilize NR radio technology, the CN 106 may also be in communication with another RAN (not shown) that employs GSM, UMTS, CDMA 2000, WiMAX, E-UTRA, or WiFi radio technology.

[0032] The CN 106 may also act as a gateway for the WTRUs 102a, 102b, 102c, 102d to access the PSTN 108, the Internet 110, and / or other networks 112. The PSTN 108 may include a circuit-switched telephone network that provides plain old telephone service (POTS). The Internet 110 may include a global system of interconnected computer networks and devices that use common communication protocols, such as the Transmission Control Protocol (TCP), the User Datagram Protocol (UDP), and / or the Internet Protocol (IP) in the TCP / IP Internet protocol suite. The networks 112 may include wired communication networks and / or wireless communication networks owned and / or operated by other service providers. For example, the networks 112 may include another CN connected to one or more RANs, which may employ the same RAT as the RAN 104 or a different RAT.

[0033] Some or all of the WTRUs 102a, 102b, 102c, 102d in the communication system 100 may include multi-mode capabilities (e.g., the WTRUs 102a, 102b, 102c, 102d may include multiple transceivers for communicating with different wireless networks via different wireless links). Figure 1A The illustrated WTRU 102c may be configured to communicate with the base station 114a, which may employ a cellular-based radio technology, and with the base station 114b, which may employ an IEEE 802 radio technology.

[0034] Figure 1B is a system diagram illustrating an example WTRU 102. Figure 1B As shown, the WTRU 102 may include, among other things, a processor 118, a transceiver 120, a transmit / receive element 122, a speaker / microphone 124, a keypad 126, a display / touchpad 128, non-removable memory 130, removable memory 132, a power source 134, a global positioning system (GPS) chipset 136, and / or other peripherals 138. It will be appreciated that the WTRU 102 may include any sub-combination of the foregoing elements while remaining consistent with an embodiment.

[0035] The processor 118 may be a general purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), a plurality of microprocessors, one or more microprocessors associated with a DSP core, a controller, a microcontroller, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), any other type of integrated circuit (IC), state machine, etc. The processor 118 may perform signal decoding, data processing, power control, input / output processing, and / or any other functionality that enables the WTRU 102 to operate in a wireless environment. The processor 118 may be coupled to the transceiver 120, which may be coupled to the transmit / receive element 122. Although Figure 1B The processor 118 and the transceiver 120 are depicted as separate components, but it is understood that the processor 118 and the transceiver 120 may be integrated together in an electronic package or chip.

[0036] The send / receive element 122 may be configured to send a signal to a base station (e.g., base station 114a) or receive a signal from a base station via an air interface 116. For example, in one embodiment, the send / receive element 122 may be an antenna configured to send and / or receive an RF signal. In an embodiment, the send / receive element 122 may be a transmitter / detector configured to send and / or receive, for example, an IR, UV, or visible light signal. In another embodiment, the send / receive element 122 may be configured to send and / or receive both an RF signal and an optical signal. It should be understood that the send / receive element 122 may be configured to send and / or receive any combination of wireless signals.

[0037] Although the transmit / receive element 122 Figure 1B Although depicted as a single element in the figure, the WTRU 102 may include any number of transmit / receive elements 122. More specifically, the WTRU 102 may employ MIMO technology. Thus, in one embodiment, the WTRU 102 may include two or more transmit / receive elements 122 (e.g., multiple antennas) for transmitting and receiving wireless signals over the air interface 116.

[0038] The transceiver 120 may be configured to modulate signals to be transmitted by the transmit / receive element 122 and to demodulate signals received by the transmit / receive element 122. As noted above, the WTRU 102 may have multi-mode capabilities. For example, the transceiver 120 may include multiple transceivers for enabling the WTRU 102 to communicate via multiple RATs, such as NR and IEEE 802.11.

[0039] The processor 118 of the WTRU 102 may be coupled to and may receive user input data from a speaker / microphone 124, a keypad 126, and / or a display / touchpad 128 (e.g., a liquid crystal display (LCD) display unit or an organic light emitting diode (OLED) display unit). The processor 118 may also output user data to the speaker / microphone 124, the keypad 126, and / or the display / touchpad 128. In addition, the processor 118 may access information from and store data in any type of suitable memory, such as a non-removable memory 130 and / or a removable memory 132. The non-removable memory 130 may include a random access memory (RAM), a read-only memory (ROM), a hard disk, or any other type of memory storage device. The removable memory 132 may include a subscriber identity module (SIM) card, a memory stick, and a secure digital (SD) memory card, among others. In other embodiments, the processor 118 may access information from and store data in a memory that is not physically located on the WTRU 102, such as on a server or a home computer (not shown).

[0040] The processor 118 may receive power from the power source 134, and may be configured to distribute and / or control power to the other components in the WTRU 102. The power source 134 may be any suitable device for powering the WTRU 102. For example, the power source 134 may include one or more dry cell batteries (e.g., nickel cadmium (NiCd), nickel zinc (NiZn), nickel metal hydride (NiMH), lithium ion (Li-ion), etc.), solar cells, fuel cells, etc.

[0041] The processor 118 may also be coupled to the GPS chipset 136, which may be configured to provide location information (e.g., longitude and latitude) regarding the current location of the WTRU 102. In addition to or in lieu of the information from the GPS chipset 136, the WTRU 102 may receive location information from a base station (e.g., base stations 114a, 114b) over the air interface 116 and / or determine its location based on the timing of signals received from two or more nearby base stations. It will be appreciated that the WTRU 102 may acquire location information by any suitable location-determination method while remaining consistent with an embodiment.

[0042] The processor 118 may also be coupled to other peripherals 138, which may include one or more software modules and / or hardware modules that provide additional features, functionality, and / or wired or wireless connectivity. For example, the peripherals 138 may include an accelerometer, an electronic compass, a satellite transceiver, a digital camera (for photos and / or video), a universal serial bus (USB) port, a vibration device, a television transceiver, a hands-free headset, Modules, frequency modulation (FM) radio units, digital music players, media players, video game player modules, Internet browsers, virtual reality and / or augmented reality (VR / AR) devices and activity trackers, etc. The peripheral device 138 may include one or more sensors. The sensor may be one or more of the following: a gyroscope, an accelerometer, a Hall effect sensor, a magnetometer, an orientation sensor, a proximity sensor, a temperature sensor, a time sensor; a geographic location sensor, an altimeter, a light sensor, a touch sensor, a magnetometer, a barometer, a gesture sensor, a biometric sensor, and a humidity sensor, etc.

[0043] The WTRU 102 may include a full-duplex radio for which transmission and reception of some or all of the signals (e.g., associated with specific subframes for both UL (e.g., for transmission) and DL (e.g., for reception)) may be concurrent and / or simultaneous. The full-duplex radio may include an interference management unit for reducing and / or substantially eliminating self-interference via signal processing performed by hardware (e.g., a choke) or via a processor (e.g., a separate processor (not shown) or via the processor 118). In an embodiment, the WTRU 102 may include a half-duplex radio for which transmission and reception of some or all of the signals (e.g., associated with specific subframes for both UL (e.g., for transmission) or DL ​​(e.g., for reception)) may be concurrent and / or simultaneous.

[0044] Figure 1C 1 is a system diagram illustrating the RAN 104 and the CN 106 in accordance with an embodiment. As noted above, the RAN 104 may employ an E-UTRA radio technology to communicate with the WTRUs 102a, 102b, 102c over the air interface 116. The RAN 104 may also be in communication with the CN 106.

[0045] The RAN 104 may include evolved Node-Bs 160a, 160b, 160c, though it will be appreciated that the RAN 104 may include any number of evolved Node-Bs while remaining consistent with an embodiment. The evolved Node-Bs 160a, 160b, 160c may each include one or more transceivers for communicating with the WTRUs 102a, 102b, 102c over the air interface 116. In one embodiment, the evolved Node-Bs 160a, 160b, 160c may implement MIMO technology. Thus, the evolved Node-B 160a, for example, may use multiple antennas to transmit wireless signals to and / or receive wireless signals from the WTRU 102a.

[0046] Each of the evolved Node Bs 160a, 160b, 160c may be associated with a particular cell (not shown) and may be configured to handle radio resource management decisions, handover decisions, and scheduling of users in the UL and / or DL, among other things. Figure 1C As shown, the eNode-Bs 160a, 160b, 160c may communicate with one another via an X2 interface.

[0047] Figure 1C The illustrated CN 106 may include a mobility management entity (MME) 162, a serving gateway (SGW) 164, and a packet data network (PDN) gateway (PGW) 166. Although the foregoing elements are depicted as part of the CN 106, it should be understood that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0048] The MME 162 may be connected to each of the evolved Node-Bs 162a, 162b, 162c in the RAN 104 via an S1 interface and may serve as a control node. For example, the MME 162 may be responsible for authenticating users of the WTRUs 102a, 102b, 102c, bearer activation / deactivation, and selecting a particular serving gateway during an initial attach of the WTRUs 102a, 102b, 102c. The MME 162 may provide a control plane function for switching between the RAN 104 and other RANs (not shown) that employ other radio technologies, such as GSM and / or WCDMA.

[0049] The SGW 164 may be connected to each of the evolved Node-Bs 160a, 160b, 160c in the RAN 104 via an S1 interface. The SGW 164 may generally route and forward user data packets to / from the WTRUs 102a, 102b, 102c. The SGW 164 may perform other functions, such as anchoring the user plane during an inter-evolved Node-B handover, triggering paging when DL data is available for the WTRUs 102a, 102b, 102c, and managing and storing the context of the WTRUs 102a, 102b, 102c.

[0050] The SGW 164 may be connected to the PGW 166, which may provide the WTRUs 102a, 102b, 102c with access to packet-switched networks, such as the Internet 110, to facilitate communications between the WTRUs 102a, 102b, 102c and IP-enabled devices.

[0051] The CN 106 may facilitate communications with other networks. For example, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to circuit-switched networks, such as the PSTN 108, to facilitate communications between the WTRUs 102a, 102b, 102c and traditional land-line communications devices. For example, the CN 106 may include, or may be in communication with, an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server) that acts as an interface between the CN 106 and the PSTN 108. In addition, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to other networks 112, which may include other wired networks and / or wireless networks owned and / or operated by other service providers.

[0052] Although the WTRU Figures 1A to 1D Although described as wireless terminals, it is contemplated that in certain representative embodiments, such terminals may (eg, temporarily or permanently) use a wired communications interface with a communications network.

[0053] In a representative embodiment, the other network 112 may be a WLAN.

[0054] A WLAN in infrastructure basic service set (BSS) mode may have an access point (AP) for a BSS and one or more stations (STAs) associated with the AP. The AP may have access or an interface to a distribution system (DS) or another type of wired / wireless network that carries traffic to and / or carries traffic away from the BSS. Traffic originating from outside the BSS and going to the STA may be reached by the AP and may be delivered to the STA. Traffic originating from the STA and going to a target outside the BSS may be transmitted to the AP to be delivered to the corresponding target. Traffic between STAs within the BSS may be transmitted by the AP, for example, wherein the source STA may transmit traffic to the AP, and the AP may deliver traffic to the target STA. Traffic between STAs within the BSS may be considered and / or referred to as point-to-point traffic. Point-to-point traffic may be transmitted between the source STA and the target STA (e.g., directly between them) using a direct link setup (DLS). In certain representative embodiments, the DLS may use 802.11e DLS or 802.11z tunnel DLS (TDLS). A WLAN using an independent BSS (IBSS) mode may not have an AP, and STAs within or using the IBSS (eg, all STAs in the STA) may communicate directly with each other. The IBSS communication mode may sometimes be referred to herein as an "ad hoc" communication mode.

[0055] When using the 802.11ac infrastructure operating mode or a similar operating mode, the AP may send beacons on a fixed channel, such as a primary channel. The primary channel may be a fixed width (e.g., a 20 MHz wide bandwidth) or a dynamically set width. The primary channel may be an operating channel of the BSS and may be used by the STA to establish a connection with the AP. In certain representative embodiments, carrier sense multiple access / collision avoidance (CSMA / CA) may be implemented, for example, in an 802.11 system. For CSMA / CA, a STA (e.g., each STA) (including the AP) may sense the primary channel. If the primary channel is sensed / detected and / or determined to be busy by a particular STA, the particular STA may back off. One STA (e.g., only one station) may transmit at any given time in a given BSS.

[0056] High throughput (HT) STAs may communicate using a 40 MHz wide channel (eg, via a combination of a primary 20 MHz channel and adjacent or non-adjacent 20 MHz channels) to form a 40 MHz wide channel.

[0057] Very high throughput (VHT) STA can support 20MHz, 40MHz, 80MHz and / or 160MHz wide channels. 40MHz channels and / or 80MHz channels can be formed by combining continuous 20MHz channels. 160MHz channels can be formed by combining 8 continuous 20MHz channels, or by combining two non-continuous 80MHz channels (this can be called 80+80 configuration). For 80+80 configuration, after channel coding, the data can pass through a segment parser that can divide the data into two streams. Each stream can be processed by inverse fast Fourier transform (IFFT) and time domain processing separately. These streams can be mapped to two 80MHz channels, and the data can be sent by sending STA. At the receiver of the receiving STA, the above-mentioned operation for 80+80 configuration can be reversed, and the combined data can be transmitted to the medium access control (MAC).

[0058] 802.11af and 802.11ah support operating modes below 1GHz. Channel operating bandwidths and carriers are reduced in 802.11af and 802.11ah relative to those used in 802.11n and 802.11ac. 802.11af supports 5MHz, 10MHz, and 20MHz bandwidths in the TV White Space (TVWS) spectrum, and 802.11ah supports 1MHz, 2MHz, 4MHz, 8MHz, and 16MHz bandwidths using non-TVWS spectrum. According to a representative embodiment, 802.11ah may support meter type control / machine type communication (MTC), such as MTC devices in macro coverage areas. MTC devices may have certain capabilities, such as limited capabilities, including support for (e.g., only support for) certain bandwidths and / or limited bandwidths. MTC devices may include batteries with battery life above a threshold (e.g., to maintain very long battery life).

[0059] WLAN systems that can support multiple channels and channel bandwidths (such as 802.11n, 802.11ac, 802.11af, and 802.11ah) include channels that can be designated as primary channels. The primary channel may have a bandwidth equal to the maximum common operating bandwidth supported by all STAs in the BSS. The bandwidth of the primary channel may be set and / or limited by a STA (which supports the minimum bandwidth operating mode) from all STAs operating in the BSS. In the example of 802.11ah, for STAs (e.g., MTC-type devices) that support (e.g., only support) a 1MHz mode, the primary channel may be 1MHz wide, even if the AP and other STAs in the BSS support 2MHz, 4MHz, 8MHz, 16MHz, and / or other channel bandwidth operating modes. Carrier sensing and / or network allocation vector (NAV) settings may depend on the state of the primary channel. If the primary channel is busy, for example, because a STA (which only supports a 1MHz operating mode) is transmitting to the AP, all available bands may be considered busy even if most of the available bands remain idle.

[0060] In the United States, the available frequency band for 802.11ah is 902MHz to 928MHz. In South Korea, the available frequency band is 917.5MHz to 923.5MHz. In Japan, the available frequency band is 916.5MHz to 927.5MHz. The total bandwidth available for 802.11ah is 6MHz to 26MHz, depending on the country code.

[0061] Figure 1D1 is a system diagram illustrating the RAN 104 and the CN 106 in accordance with an embodiment. As noted above, the RAN 104 may employ NR radio technology to communicate with the WTRUs 102a, 102b, 102c over the air interface 116. The RAN 104 may also be in communication with the CN 106.

[0062] The RAN 104 may include gNBs 180a, 180b, 180c, though it will be appreciated that the RAN 104 may include any number of gNBs while remaining consistent with an embodiment. The gNBs 180a, 180b, 180c may each include one or more transceivers for communicating with the WTRUs 102a, 102b, 102c over the air interface 116. In one embodiment, the gNBs 180a, 180b, 180c may implement MIMO technology. For example, the gNBs 180a, 180b may utilize beamforming to send signals to and / or receive signals from the gNBs 180a, 180b, 180c. Thus, the gNB 180a may, for example, use multiple antennas to send wireless signals to and / or receive wireless signals from the WTRU 102a. In an embodiment, the gNBs 180a, 180b, 180c may implement carrier aggregation technology. For example, gNB 180a may transmit multiple component carriers to WTRU 102a (not shown). A subset of these component carriers may be on unlicensed spectrum, while the remaining component carriers may be on licensed spectrum. In an embodiment, gNBs 180a, 180b, 180c may implement coordinated multi-point (CoMP) technology. For example, WTRU 102a may receive coordinated transmissions from gNB 180a and gNB 180b (and / or gNB 180c).

[0063] The WTRUs 102a, 102b, 102c may communicate with the gNBs 180a, 180b, 180c using transmissions associated with parameter sets that may be scalable. For example, OFDM symbol spacing and / or OFDM subcarrier spacing may vary for different transmissions, different cells, and / or different portions of the wireless transmission spectrum. The WTRUs 102a, 102b, 102c may communicate with the gNBs 180a, 180b, 180c using subframes or Transmission Time Intervals (TTIs) of varying or scalable lengths (e.g., containing different numbers of OFDM symbols and / or varying absolute time lengths over time).

[0064] The gNBs 180a, 180b, 180c may be configured to communicate with the WTRUs 102a, 102b, 102c in a standalone configuration and / or a non-standalone configuration. In a standalone configuration, the WTRUs 102a, 102b, 102c may communicate with the gNBs 180a, 180b, 180c while not accessing other RANs (e.g., such as the eNodeBs 160a, 160b, 160c). In a standalone configuration, the WTRUs 102a, 102b, 102c may use one or more of the gNBs 180a, 180b, 180c as mobility anchor points. In a standalone configuration, the WTRUs 102a, 102b, 102c may communicate with the gNBs 180a, 180b, 180c using signals in an unlicensed band. In a non-standalone configuration, the WTRUs 102a, 102b, 102c may communicate / connect with the gNB 180a, 180b, 180c while also communicating / connecting with another RAN, such as the eNodeB 160a, 160b, 160c. For example, the WTRUs 102a, 102b, 102c may implement the DC principle to communicate with one or more gNBs 180a, 180b, 180c and one or more eNodeBs 160a, 160b, 160c substantially simultaneously. In a non-standalone configuration, the eNodeB 160a, 160b, 160c may act as a mobility anchor for the WTRUs 102a, 102b, 102c, and the gNB 180a, 180b, 180c may provide additional coverage and / or throughput for serving the WTRUs 102a, 102b, 102c.

[0065] Each of the gNBs 180a, 180b, 180c may be associated with a specific cell (not shown) and may be configured to handle radio resource management decisions, handover decisions, scheduling of users in UL and / or DL, support of network slicing, interworking between DC, NR and E-UTRA, routing of user plane data towards a user plane function (UPF) 184a, 184b and routing of control plane information towards an access and mobility management function (AMF) 182a, 182b, etc. Figure 1D As shown, gNBs 180a, 180b, and 180c may communicate with each other via an Xn interface.

[0066] Figure 1DThe illustrated CN 106 may include at least one AMF 182a, 182b, at least one UPF 184a, 184b, at least one session management function (SMF) 183a, 183b, and possible data networks (DNs) 185a, 185b. Although the aforementioned elements are depicted as part of the CN 106, it should be understood that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0067] The AMF 182a, 182b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 104 via the N2 interface and may act as a control node. For example, the AMF 182a, 182b may be responsible for authenticating users of the WTRU 102a, 102b, 102c, support of network slicing (e.g., handling of different protocol data unit (PDU) sessions with different requirements), selecting a specific SMF 183a, 183b, management of registration areas, termination of non-access stratum (NAS) signaling, and mobility management, etc. The AMF 182a, 182b may use network slicing to customize CN support for the WTRU 102a, 102b, 102c based on the type of services utilized by the WTRU 102a, 102b, 102c. For example, different network slices may be established for different use cases, such as services relying on ultra-reliable low latency (URLLC) access, services relying on enhanced massive mobile broadband (eMBB) access, and services for MTC access. The AMF 182a, 182b may provide a control plane function for switching between the RAN 104 and other RANs (not shown) that employ other radio technologies, such as LTE, LTE-A, LTE-A Pro, and / or non-3GPP access technologies such as WiFi.

[0068] The SMF 183a, 183b may be connected to the AMF 182a, 182b in the CN 106 via the N11 interface. The SMF 183a, 183b may also be connected to the UPF 184a, 184b in the CN 106 via the N4 interface. The SMF 183a, 183b may select and control the UPF 184a, 184b, and configure the service routing performed by the UPF 184a, 184b. The SMF 183a, 183b may perform other functions, such as managing and allocating UE IP addresses, managing PDU sessions, controlling policy implementation and QoS, and providing DL data notifications. The PDU session type may be IP-based, non-IP-based, Ethernet-based, etc.

[0069] The UPF 184a, 184b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 104 via the N3 interface, which may provide the WTRUs 102a, 102b, 102c with access to packet-switched networks, such as the Internet 110, to facilitate communications between the WTRUs 102a, 102b, 102c and IP-enabled devices. The UPF 184, 184b may perform other functions, such as routing and forwarding packets, enforcing user plane policies, supporting multi-homed PDU sessions, handling user plane QoS, buffering DL packets, and providing mobility anchoring, etc.

[0070] The CN 106 may facilitate communications with other networks. For example, the CN 106 may include or may communicate with an IP gateway (e.g., an IP Multimedia Subsystem (IMS) server) that acts as an interface between the CN 106 and the PSTN 108. In addition, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to other networks 112, which may include other wired networks and / or wireless networks owned and / or operated by other service providers. In one embodiment, the WTRUs 102a, 102b, 102c may be connected to the DNs 185a, 185b via the UPFs 184a, 184b via the N3 interfaces to the UPFs 184a, 184b and the N6 interfaces between the UPFs 184a, 184b and the local DNs 185a, 185b.

[0071] Usually, in Figures 1A to 1D Any network-side devices / nodes / functions / base stations described in and / or anywhere herein may be interchangeable, and references to the network may refer to references to specific entities, such as devices, nodes, functions, base stations, clouds, etc., as disclosed herein.

[0072] Given that Figures 1A to 1D as well as Figures 1A to 1D Corresponding to the description of the present invention, one or more or all of the functions described herein with reference to one or more of the following may be performed by one or more simulation devices (not shown): WTRU102a to 102d, base station 114a to 114b, evolved Node B 160a to 160c, MME 162, SGW 164, PGW 166, gNB180a to 180c, AMF 182a to 182b, UPF 184a to 184b, SMF 183a to 183b, DN 185a to 185b and / or any other device described herein. The simulation device may be one or more devices configured to mimic one or more or all of the functions described herein. For example, the simulation device may be used to test other devices and / or simulate network and / or WTRU functions.

[0073] The simulation device may be designed to implement one or more tests of other devices in a laboratory environment and / or in an operator network environment. For example, one or more simulation devices may perform one or more functions or all functions while being fully or partially implemented and / or deployed as part of a wired and / or wireless communication network in order to test other devices within the communication network. One or more simulation devices may perform one or more functions or all functions while being temporarily implemented / deployed as part of a wired and / or wireless communication network. A simulation device may be directly coupled to another device for the purpose of testing and / or performing tests using over-the-air wireless communications.

[0074] One or more simulation devices can perform one or more (including all) functions without being implemented / deployed as part of a wired and / or wireless communication network. For example, the simulation device can be used in a test scenario in a test laboratory and / or a non-deployed (e.g., testing) wired and / or wireless communication network to implement testing of one or more components. One or more simulation devices can be test equipment. Direct RF coupling and / or wireless communication via RF circuits (e.g., which can include one or more antennas) can be used by the simulation device to send and / or receive data.

[0075] Channel state information (CSI) feedback enhancements (such as reduced overhead, improved accuracy, prediction) may be advantages gained from the use of artificial intelligence and / or machine learning (AI / ML) systems and algorithms as they relate to the NR air interface. For example, typically in NR, there may be a large overhead associated with the downlink (DL) CSI reference signal (CSI-RS) and the corresponding uplink (UL) CSI reports. This overhead may increase as the system bandwidth and the number of antennas for massive MIMO systems for Advanced NR and beyond increase.

[0076] The ML techniques may enable the WTRU to compress the CSI measurements that it transmits to the network (e.g., a network node, network function, base station, gNB, etc., or any other type of network device). The AI / ML encoder at the WTRU may encode (e.g., compress) measurements of channel conditions, which are then transmitted to the network where they may be decompressed by the AI / ML decoder. The AI / ML encoder-decoder pair may be trained together as part of an autoencoder (AE) structure, or separately.

[0077] To reduce WTRU complexity and / or reduce training latency, in some embodiments, the WTRU may be configured with a relatively small AI / ML encoder model containing a small number of weights. However, this means that the AI / ML encoder may no longer be robust to all possible channel condition measurements. Therefore, in some other embodiments, the encoder may need to be retrained when the observed channel conditions change.

[0078] The channel state information may include at least one of the following: a channel quality index (CQI), a rank indicator (RI), a precoding matrix index (PMI), an L1 channel measurement (e.g., RSRP, such as L1-RSRP or SINR), a CSI-RS resource indicator (CRI), a SS / PBCH block resource indicator (SSBRI), a layer indicator (LI), and / or any other measurement quantity measured by the WTRU from a configured reference signal (e.g., a CSI-RS or a SS / PBCH block or any other reference signal).

[0079] Figure 2 An example of a configuration 200 of CSI reporting settings (eg, CSI reporting settings 202A-202B), resource settings (eg, resource settings 204A-204C), and links 206 (eg, link 0 to link 3) is shown.

[0080] The WTRU may be configured to report CSI via the uplink control channel on the PUCCH, or upon request by the gNB for UL PUSCH grant. Depending on the configuration, the CSI-RS may cover the entire bandwidth of the bandwidth part (BWP) or only a small portion thereof. Within the CSI-RS bandwidth, the CSI-RS may be configured in every PRB or every other PRB. In the time domain, the CSI-RS resources may be configured as periodic, semi-persistent, or aperiodic. Semi-persistent CSI-RS may be similar to periodic CSI-RS, except that the resource may be activated (deactivated) by a MAC CE; and the WTRU reports relevant measurements only when the resource is activated. For aperiodic CSI-RS, the WTRU may be triggered to report the CSI-RS measured on the PUSCH via a request in the DCI. Periodic reports may be carried via the PUCCH, while semi-persistent reports may be carried on either the PUCCH or the PUSCH. The reported CSI may be used by the scheduler in allocating the best resource blocks, possibly based on the time-frequency selectivity of the channel, determining the precoding matrix, beams, transmit mode, and selecting the appropriate MCS. The reliability, accuracy, and timeliness of WTRU CSI reporting may be critical to meeting URLLC service requirements.

[0081] The WTRU may be configured with a CSI measurement setup 208, which may include one or more CSI report setups 202, a resource setup 204, and / or a link 206 between the one or more CSI report setups 202 and the one or more resource setups 204. In some embodiments, the CSI measurement setup 208 may include at least one of: a CSI report setup 202; a resource setup 204; and / or a reference transmission scheme setup for CQI.

[0082] As described herein, CSI measurement settings 208 may include one or more configuration parameters.

[0083] For the CSI measurement setup 208 , examples of configuration parameters may be N≥1 CSI report setups 202 , M≥1 resource setups 204 , and one or more links 206 linking the N CSI report setups 202 with the M resource setups 204 .

[0084] For the CSI measurement setting 208, another example of a configuration parameter may be a CSI reporting setting 202, which includes at least one of the following: time domain behavior, such as aperiodic or periodic / semi-persistent; frequency granularity for at least PMI and CQI; CSI report type (e.g., PMI, CQI, RI, CRI, etc.); and / or, if PMI is reported, PMI type (Type I or II) and codebook configuration.

[0085] For CSI measurement setting 208, another example of configuration parameters may be resource setting 202, which includes at least one of the following: time domain behavior, such as aperiodic or periodic / semi-persistent; RS type (e.g., for channel measurement or interference measurement); and / or, S≥1 resource sets, each of which may contain K s resources.

[0086] For the CSI measurement setup 208, another example of a configuration parameter may be one or more frequency granularities supported for CSI reporting for a component carrier, such as wideband CSI; partial-band CSI; and / or sub-band CSI.

[0087] Figure 3 An example of codebook based precoding 300 using feedback information is illustrated. As shown, the feedback information 308 may include a precoding matrix index (PMI), which may be referred to as a codeword index in the codebook.

[0088] As shown, the codebook may include a set of precoding vectors / matrices for each rank and multiple antenna ports, and each of the precoding vectors / matrices may have its own index so that a first device or WTRU (e.g., receiver 306) may inform a second device or WTRU (e.g., transmitter 302) of a preferred precoding vector / matrix index via one or more MIMO channels 304. Codebook-based precoding may have degraded performance compared to non-codebook-based precoding due to its limited number of precoding vectors / matrices. However, an advantage of codebook-based precoding may be lower control signaling / feedback overhead.

[0089] Figure 4 An example of a downlink codebook 400 for 2Tx is illustrated. A CSI processing unit (CPU) may be referred to as a minimum CSI processing unit, and a WTRU may support (e.g., run) one or more CPUs (e.g., N CPUs). A WTRU with N CPUs may estimate N CSI feedback calculations in parallel, where N may be the WTRU capability. If a WTRU is requested to estimate more than N CSI feedbacks simultaneously, in some embodiments, the WTRU may only perform the high priority N CSI feedbacks and may not estimate the remaining CSI feedbacks.

[0090] The start and end of CPU processing can be determined based on the CSI report type (e.g., aperiodic, periodic, semi-persistent). For example, for aperiodic CSI reporting, the CPU is occupied from the first OFDM symbol after the PDCCH trigger until the last OFDM symbol of the PUSCH carrying the CSI report. For periodic and semi-persistent CSI reporting, the CPU is occupied from the first OFDM symbol of one or more associated measurement resources (no earlier than the CSI reference resource) until the last OFDM symbol of the CSI report.

[0091] The number of CPUs occupied may vary based on the CSI measurement type (e.g., beam-based or non-beam-based). For example, for non-beam-related reporting, when K in the CSI-RS resource set is s When CSI-RS resources are used for channel measurement, K sFor beam-related reports (e.g., "cri-RSRP", "ssb-Index-RSRP", or "None"), 1 CPU may be occupied due to the relatively low CSI computation complexity, regardless of the number of CSI-RS resources in the CSI-RS resource set used for channel measurement, and "None" is used for P3 operation or aperiodic TRS transmission. For aperiodic CSI reports with a single CSI-RS resource, 1 CPU may be occupied. For CSI reporting Ks CSI-RS resources, K s The CPU may be occupied because the WTRU needs to perform CSI measurements for each CSI-RS resource.

[0092] When the number of unused CPUs (N u ) is less than the number of CPUs required for CSI reporting (N r ), one or more example WTRU behaviors may be used: In case of UCI / HARQ on PUSCH without data, the WTRU may drop N based on priority. r -N u CSI report; and / or, in other cases, the WTRU may prioritize r -N u Dummy information is reported in the CSI report to avoid rate matching processing for PUSCH.

[0093] In general, artificial intelligence (AI) can be broadly defined as behaviors exhibited by machines. Such behaviors can, for example, mimic cognitive functions for sensing, inference, adaptation, and action.

[0094] In general, machine learning (ML) may refer to a type of algorithm that is based on learning to solve problems through experience ("data") without explicit programming ("configuration rule sets"). Machine learning may be considered a subset of AI. Different machine learning paradigms can be envisioned based on the nature or feedback of the data available for learning the algorithm. In an example paradigm, a supervised learning method may involve learning a function that maps inputs to outputs based on labeled training examples, where each training example may be a pair consisting of an input and a corresponding output. In an example paradigm, an unsupervised learning method may involve detecting patterns in data without pre-existing labels. In an example paradigm, a reinforcement learning method may involve performing a series of actions in an environment to maximize cumulative rewards. In some of the methods discussed herein, it is possible to apply a machine learning algorithm using a combination or interpolation of the above-mentioned paradigms. For example, a semi-supervised learning method may use a combination of a small amount of labeled data and a large amount of unlabeled data during training. In this regard, semi-supervised learning falls between unsupervised learning (no labeled training data) and supervised learning (only labeled training data).

[0095] In general, deep learning refers to a class of machine learning algorithms that employ artificial neural networks, specifically, deep neural networks (DNNs). Deep neural networks (DNNs) are a special class of machine learning models inspired by the human brain, in which the input is linearly transformed and passed through a nonlinear activation function multiple times. DNNs typically consist of multiple layers, each of which consists of a linear transformation and a given nonlinear activation function. DNNs can be trained using training data via a back-propagation algorithm. Recently, DNNs have demonstrated state-of-the-art performance in various fields such as speech, vision, natural language, etc. and for various machine learning settings including supervised, unsupervised, and semi-supervised.

[0096] The term "AI / ML" may refer to techniques involving methods / processes such as those described herein, and may refer to achieving behavior and / or compliance with requirements through data-based learning without explicit configuration of a sequence of action steps. Such methods may enable learning of complex behaviors that may be difficult to specify and / or achieve using traditional methods.

[0097] Normalized mean square error (NMSE) can be used to evaluate the quality of CSI compression and reconstruction. NMSE is defined as:

[0098]

[0099] Where H represents the CSI matrix (e.g., channel response) at the input of the AI ​​NN encoder, represents the reconstructed matrix at the output of the AI ​​NN decoder, and the operator |||| F Indicates the Frobenius (Euclidean) norm.

[0100] Another measure of the quality of CSI compression and reconstruction is the cosine similarity ρ, which is defined as:

[0101]

[0102] in The vector on subcarrier 'n' representing the reconstructed channel matrix (at the output of the AI ​​NN decoder).

[0103] AI encoder, AI / ML encoder, AI NN encoder, or AI / ML NN encoder are used interchangeably herein. AI decoder, AI / ML decoder, AI NN decoder, or AI / ML NN decoder are used interchangeably herein. Unless otherwise specified, AI / ML training and retraining can be used interchangeably as discussed herein.

[0104] ML-based approaches for CSI compression may require data sets to train for a wide variety of channel conditions. Performing this process online not only requires a very large amount of training data, but also results in significant latency, power, and memory requirements at the WTRU. Training the encoder for a subset of channel conditions prior to deployment, and then updating the encoder as additional data (e.g., measurements of channel conditions) becomes available, may improve some of these issues. Specifically, in some embodiments, the AI / ML encoder model at the WTRU may be trained independently of the autoencoder-decoder at the network (e.g., gNB).

[0105] Certain updates to the WTRU AI / ML encoder model based on retraining using actual measurements may also require corresponding updates to the network AI / ML decoder model. In some embodiments, the desired decoder updates can be efficiently communicated to the network.

[0106] In some embodiments, the WTRU may initially use an encoder model trained for a subset of channel conditions. When the WTRU encounters channel conditions outside the original training set, it may request additional training data from the network. The WTRU AI / ML encoder may request the appropriate additional training set and determine when retraining is successful.

[0107] In some embodiments, the WTRU may be configured with one or more AI / ML models. The WTRU may use one or more AI / ML models to obtain CSI feedback reports for transmission to the network. The WTRU may input CSI measurements (e.g., obtained from measurements performed on reference signals) into the AI / ML encoder and may feed back the output of the AI / ML encoder to the gNB.

[0108] As discussed herein, an AI / ML model may be configured with one or more pieces of information.

[0109] For example, the AI / ML model configuration may include parameters related to the original training of the AI / ML model. For example, the model configuration may indicate the channel type for which the AI / ML model is applicable. For example, the model configuration may indicate the source of the AI / ML model. The source may include a node identifier or parameters associated with conditions at the training node.

[0110] For example, the AI / ML model may be configured to include a data set for training the AI / ML model. The data set for training the AI / ML model may include at least one of: a synthetic data set; a field data set; a measurement-based data set obtained at least in part through a reference signal configuration, where the reference signal configuration may be provided to the WTRU; and / or a data set from other nodes, where the WTRU may be provided with resources on which to receive data from the other nodes.

[0111] For example, the AI / ML model can be configured to include an AI / ML model performance threshold. Such a threshold can be compared to an error output by the AI / ML encoder. Such a threshold can be compared to the performance of a function associated with the AI / ML model. For example, the threshold can be compared to a compression rate of a feedback report AI / ML function.

[0112] For example, the AI / ML model may be configured to include an AI / ML model performance reporting resource. For example, the WTRU may be configured with a reporting resource to report the performance of the AI / ML model.

[0113] For example, the AI / ML model may be configured to include training request resources. For example, the WTRU may be configured with resources on which it may send a request to train the AI / ML model (e.g., a request for a dataset for training, or a request for the transmission of signals such as reference signals that enable the WTRU to train the AI / ML model online) to another node (e.g., a gNB).

[0114] If the AI / ML model is determined to be suitable based on one or more criteria, the WTRU may use the model. If the AI / ML model satisfies at least one reporting criterion, suitability may be determined. For example, if the AI / ML model satisfies the compression rate criterion, it may be considered suitable. In this example, the WTRU may consider the AI / ML model to be suitable if its compression rate is greater than or less than a configurable threshold. In another example, if the AI / ML model satisfies the error (e.g., MMSE) criterion, the AI / ML model may be considered suitable. In this example, if the error or MMSE of the AI / ML model is less than a configurable threshold, the WTRU may consider the AI / ML model to be suitable.

[0115] Depending on the compression rate obtained for the appropriate AI / ML model, the WTRU may require different payload sizes for the CSI feedback report. The WTRU may be configured with one or more CSI feedback reporting resources. For example, the WTRU may have a set of CSI feedback reporting resources, each resource being suitable for a different compression rate (or feedback payload). The WTRU may be configured with CSI feedback reporting resources for fallback CSI reporting. Wherein conventional methods (e.g., without using AI / ML) may be used to obtain the fallback CSI report. Such fallback CSI report may be sent by the WTRU when no AI / ML model is considered suitable.

[0116] The WTRU may be configured with training opportunities and / or resources. The training opportunities may occur periodically, aperiodically, or semi-periodically. The training opportunities may be configured with at least one of: a period, an offset, a start time, and / or an end time.

[0117] One or more factors / events may trigger the WTRU to perform AI / ML training.

[0118] The WTRU may be triggered to perform AI / ML training by timing. For example, for periodically configured training opportunities, the WTRU may train the model at configurable time intervals.

[0119] The WTRU may be triggered to perform AI / ML training by receiving an indication from the network (e.g., gNB). For example, for non-periodic or semi-persistent training opportunities, the WTRU may start training upon receiving an indication from the gNB. The training indication may be a dynamic indication (e.g., in a DCI or MAC CE) or a non-dynamic indication (e.g., an RRC indication). The training indication may be received implicitly by the WTRU as part of a reconfiguration of functionality associated with the AI / ML model. For example, if the AI / ML model is configured for CSI feedback reporting, the WTRU may be triggered to start AI / ML model training when the WTRU receives a reconfiguration of one or more CSI feedback reports.

[0120] The WTRU may be triggered to perform AI / ML training based on the performance of the WTRU. For example, the WTRU may be triggered to perform AI / ML training based on at least one of the following: BLER performance (e.g., BLER performance below or above a threshold); HARQ-ACK or HARQ-NACK ratio; measurements, such as if L3 (e.g., RSRP, RSSI, RSRQ, CO) or L1 measurements (e.g., RI, PMI, CQI, LI, CRI, RSRP) are above or below a threshold, the WTRU may be triggered to start training the AI / ML model; changes in BWP; changes, additions, removals, activations, or deactivations of cells; changes in TRP; and / or feedback payload size.

[0121] The WTRU may be triggered to perform AI / ML training based on the performance of the model. For example, the WTRU may be triggered to perform AI / ML model training based on at least one of the AI / ML output error (e.g., MMSE error) and / or compression rate.

[0122] For example, for AI / ML output errors, the WTRU may be configured with an AI / ML encoder and decoder (e.g., a decoder used by a gNB). The WTRU may test the output of the decoder using actual channel measurements and may determine an error value. If the error value is greater than, less than, and / or equal to a threshold, the WTRU may be triggered to start training the AI / ML model. When the error is greater than a threshold, the WTRU may be triggered to train the AI / ML model because such an error may result in degraded performance. When the error is less than a threshold, the WTRU may be triggered to train the AI / ML model because this may mean that higher feedback compression is possible.

[0123] For example, for compression ratio, when the compression ratio is greater than a threshold, the WTRU may be triggered to start training the AI / ML model. In this case, for lower compression ratios that are still suitable, lower AI / ML errors may be achieved. In another example, when the compression ratio is less than a threshold, the WTRU may be triggered to start training the AI / ML model. In this case, greater compression ratios and acceptable error rates are possible.

[0124] The WTRU may be triggered to perform AI / ML training on a time basis starting from the last training opportunity. For example, the WTRU may be configured with a time period. The WTRU may start or restart the timer when first configured with the AI / ML model or when training of the AI / ML model is completed. At the expiration of the time period, the WTRU may start training the AI / ML model.

[0125] When triggered to perform AI / ML model training, the WTRU may receive RS configuration. The WTRU may use such resources to obtain channel conditions to train the AI / ML model.

[0126] When the WTRU is triggered to train the AI / ML model, the WTRU may indicate to the gNB that it is starting AI / ML model training. The indication may include the trigger or reason for the training and may also include the functionality associated with the AI / ML model.

[0127] The WTRU may be triggered to request an AI / ML model training opportunity from the gNB. The trigger may include any of the triggers described above. When the WTRU is triggered to request an AI / ML model training opportunity, the WTRU may report such a request, for example, on resources configured for such a request.

[0128] The training occasion may last for multiple time slots or subframes. During this period, the WTRU may have CSI feedback reporting instances. In such cases, the WTRU may send feedback using the AI / ML model built before retraining, or a new model being trained, or traditional feedback (e.g., not based on an AI / ML model).

[0129] The WTRU may test the AI / ML model to determine whether it is suitable. The testing of the AI / ML model may be performed after the training opportunity ends. The testing of the AI / ML model may be performed simultaneously with the AI / ML model training. During the training opportunity, the WTRU may test the AI / ML model after every x updates of the model, where x>=1. The testing of the AI / ML model may be implemented similarly to the model performance described herein. The WTRU may be configured with an AI / ML decoder (e.g., an AI / ML decoder used at a gNB). The WTRU may perform one or more CSI measurements on one or more RSs. The WTRU may input the CSI measurements into the AI / ML encoder model. The WTRU may determine the compression rate of the output of the encoder model. The WTRU may then input the output of the encoder model into the decoder model. The WTRU may compare the output of the decoder model with the CSI measurements initially obtained. Based on the comparison, the WTRU may adjust some parameters of the encoder or decoder AI / ML model.

[0130] The WTRU may stop training the AI / ML model when one or more termination conditions are met.

[0131] For example, the AI / ML training end condition may include the expiration of a training opportunity time or duration. For example, a training opportunity may be configured with a training opportunity time or duration. In some embodiments, the training time may be a number of time slots, subframes, or RS instances, or may be based on a time period, such as seconds, minutes, or any other time.

[0132] For example, the AI / ML training end condition may include when the compression rate reaches a desired value. For example, when the compression rate becomes greater than a threshold, the training opportunity may end. This value may be predetermined or may be provided to the WTRU when the training is triggered.

[0133] For example, the AI / ML training end condition may include when the error performance reaches a desired value. For example, when the error performance (e.g., MMSE) becomes less than a threshold, the training opportunity may end. This value may be predetermined or may be provided to the WTRU when the training is triggered.

[0134] For example, AI / ML training end conditions may include when the model performance does not converge. For example, if the error performance deteriorates or the compression rate decreases, the WTRU may end AI / ML model training. The WTRU may compare the error performance to the performance (or its offset) achieved before training or a threshold. When the performance becomes worse than the offset or threshold, the WTRU may end AI / ML model training.

[0135] The WTRU may report to the gNB when AI / ML model training has ended. The WTRU may also report the reason why AI / ML model training has ended (e.g., whether the compression rate exceeds a threshold, whether the error rate is less than a threshold, etc.). In the case where AI / ML model training has ended because a new suitable AI / ML model has been determined, the WTRU may report the model (e.g., encoder or decoder AI / ML model) or its parameters. In addition, the WTRU may start using the new model to obtain CSI feedback. In another example, the WTRU may wait until the gNB confirms the applicability of the new AI / ML model before using the new AI / ML model (e.g., the gNB transmits an approval message about the new AI / ML model).

[0136] In the case where AI / ML model training has ended without determining a new suitable AI / ML model, the WTRU may report AI / ML model training failure to the gNB. The WTRU may use the previous AI / ML model or revert to fallback CSI feedback reporting (e.g., no AI / ML). The WTRU may request a new AI / ML model or a new AI / ML training opportunity.

[0137] When the WTRU stops AI / ML training, it may stop monitoring the RS configured with AI / ML model training.

[0138] If the WTRU does not determine a suitable new AI / ML model, the WTRU may test the original AI / ML model. If the original AI / ML model is determined to be unsuitable or has failed, the WTRU may indicate to the gNB that no AI / ML model is suitable. The suitability of the original model may be determined by comparing the performance of the model to a suitability threshold (e.g., a compression rate threshold or an error / MMSE threshold), which may be different from the threshold used to trigger AI / ML training. For example, training may be triggered if the MMSE error is above a first threshold. The threshold may be used to try to find a new model before the original model actually fails. If no new model is found and the original model MMSE error exceeds a second threshold, the WTRU may determine that the original AI / ML model has failed. If no new AI / ML model is found and the original model fails or is deemed unsuitable, the WTRU may fall back to traditional functionality.

[0139] The WTRU may be configured with resources for feedback CSI reporting. Some resources may be configured or reserved for reporting CSI feedback obtained via AI / ML models (e.g., compressed feedback reports). Some resources may be configured or reserved for reporting CSI feedback obtained via traditional methods (e.g., RI, PMI, CQI, LI, CRI, RSSP, RSSI, RSRQ, CO).

[0140] The WTRU may be configured with multiple feedback resources for reporting compressed feedback reports. Each resource may enable a different payload. Each resource may enable reporting of feedback reports with different compression rates. For example, a first feedback resource may enable reporting of a large payload for a low compression rate feedback report, and a second feedback resource may enable reporting of a smaller payload for a higher compression rate feedback report.

[0141] The WTRU may select feedback resources based on the compression rate of the feedback reports.

[0142] The WTRU may be provided (e.g., dynamically) or configured (e.g., semi-statically) with an identification or indication of feedback resources that occur during or after a training opportunity. The identification or indication may be included in a signal received by the WTRU that triggers training of the AI / ML model.

[0143] As disclosed herein, the WTRU may select an appropriate feedback resource based on one or more factors.

[0144] For example, one factor for determining the feedback resource may be the compression rate. For example, the WTRU may select the feedback resource based on the compression rate of the feedback report. For example, the WTRU may select the feedback resource with the smallest payload size that supports the compression rate of the feedback report.

[0145] For example, one factor for determining feedback resources may be the use of a new model. For example, when the WTRU has trained a new suitable model, the WTRU may select resources configured for use with the new AI / ML model.

[0146] For example, one factor used to determine the feedback resource may be the use of an old model. For example, if the WTRU determines that training has not resulted in a better model and the WTRU continues to use an old or original AI / ML model, the WTRU may select a specific feedback resource.

[0147] For example, one factor for determining feedback resources may be the use of fallback or legacy CSI feedback. For example, if no AI / ML is deemed appropriate, the WTRU may use a specific feedback resource. In another example, the WTRU may be configured to report legacy feedback at a specific time instance. The WTRU may use feedback resources configured for legacy feedback.

[0148] For example, one factor for determining feedback resources may be the type of report. For example, some resources may be associated with CSI feedback reporting and other resources may be associated with reporting parameters associated with AI / ML model training. For example, the WTRU may be configured with feedback resources to report information about ongoing or completed AI / ML model training. For example, the WTRU may be configured with resources to request new AI / ML model training opportunities.

[0149] The WTRU may perform AI / ML model training and may obtain a new AI / ML model (e.g., for generation of CSI feedback reports). The WTRU may select a first feedback resource based on a first AI / ML model and select a second feedback resource based on a second AI / ML model after the AI / ML model training. In some cases, the two feedback resources may be the same (e.g., the same configuration may be used at different time instances). For example, if the first and second AI / ML models produce feedback with the same compression rate, the feedback resources may be the same. In this case, the WTRU may explicitly indicate to the gNB that the AI / ML model has changed. The indication may be included in a feedback report (e.g., the first feedback report) using the second AI / ML model.

[0150] The WTRU may be configured with resources to report that a suitable AI / ML model does not exist. For example, the WTRU may be triggered to train a new AI / ML model because the old model is no longer suitable. If the training does not produce a suitable new AI / ML model, the WTRU may use specific resources to indicate that a suitable model does not exist.

[0151] The WTRU may request one or more feedback resources from the gNB. In the request for new feedback resources, the WTRU may include desired feedback resource parameters including at least one of: payload size (or associated compression ratio), periodicity of the feedback resource, remaining duration until a future training opportunity, and / or AI / ML model parameters (e.g., error performance).

[0152] In some cases, the WTRU may be configured with a first feedback resource having a first payload. The first payload may be suitable for a first compression rate of a first AI / ML model. After training the model, the WTRU may obtain a second AI / ML model. The second compression rate of the second AI / ML model may be different from the first compression rate. The WTRU may use an instance of the first feedback resource to report a feedback report (e.g., a first feedback report) obtained from a second AI / ML model (e.g., a retrained AI / ML model).

[0153] If the payload of the new feedback report with the second compression rate is smaller than the available payload of the first feedback resource, then the WTRU may include additional information in the feedback report in addition to the CSI feedback report obtained from the second AI / ML model. The additional information may include at least one of the following: information in the feedback report indicating the second compression rate or CSI feedback report payload size, which may enable the gNB to reconfigure the WTRU with feedback resources that better match the new compression rate or feedback report payload; including legacy CSI feedback report types (e.g., RI, CQI, PMI, CRI, LI, RSSI, RSRP, RSRQ, CO); and / or AI / ML model parameters of the first or second encoder or decoder AI / ML model.

[0154] The additional information may also depend on the feedback resource type. For example, if the feedback resource type is a control channel (e.g., PUCCH resource), the WTRU may multiplex the additional information in the feedback report. If the feedback resource type is a data channel (e.g., PUSCH resource), the WTRU may multiplex the additional information or may include the additional information in the MAC CE.

[0155] If the payload of the new feedback report with the second compression rate is larger than the available payload of the first feedback resource, the WTRU may modify the feedback report in at least one of the following ways: splitting the CSI feedback report into multiple CSI feedback sub-reports and sending each sub-report in a different feedback resource; including a request for a larger feedback resource payload; sending traditional CSI feedback (e.g., the WTRU may fall back to traditional CSI feedback and / or the WTRU may replace the discarded channel values ​​with traditional CSI feedback reports); and / or sending a partial CSI feedback report.

[0156] For the example of sending a partial CSI feedback report, the WTRU may drop some bits from the report. The dropped bits may be the LSBs of the channel coefficients. The dropped bits may be associated with a specific channel coefficient. The dropped bits may correspond to a specific channel region (e.g., a subband, beam, or time period). The dropped bits may correspond to a specific reference signal. The dropped bits may correspond to a specific antenna port.

[0157] In some embodiments as discussed above, the WTRU or UE may send CSI feedback in a legacy mode. For example, the UE may fall back to legacy CSI feedback. In some embodiments, the UE or WTRU may send an identifier of the discarded channel value or send the discarded bits via a legacy CSI feedback report (e.g., an uncompressed CSI feedback result).

[0158] In some embodiments, the UE or WTRU may send a request for a larger feedback resource payload. For example, if a high compression rate is not available (e.g., conventional CSI feedback is being used), the UE or WTRU may first send a request to schedule a larger feedback resource payload (e.g., to extend the amount of time and bandwidth available for sending such feedback reports).

[0159] In some embodiments, the UE or WTRU may segment or split the CSI feedback report into multiple CSI feedback sub-reports and send each sub-report or part of the report in different feedback resources. In some such embodiments, each segment or part of the report may contain a sub-field or identifier to indicate that it is a part of the report, to indicate its order relative to other parts (e.g., "segment 1", "segment 2", or "segment 1 of 5 segments", etc.), or to contain any other such identifier. For example, one or more bits may be allocated to identify the segment ordering.

[0160] For example, at 602, the WTRU may be configured with a decoder model and one or more encoder models. At 604, the WTRU may compress CSI measurements using the identified encoder model. At 606, upon determining to update the model via any of the triggers described above, the WTRU may perform channel measurements or measure channel conditions and use the measurements of channel conditions to train the internal encoder and decoder. At 608, the WTRU may determine whether the new encoder model performs better than the previous model (e.g., based on compression rate, error rate, throughput, or any other such condition or combination of conditions or features, as discussed above). If so, then in some embodiments, at 610, the WTRU may determine whether the performance improvement is greater than a predetermined value. If so, at 612, the WTRU may notify the gNB or other device of the new model and switch to using the newly trained encoder module. If not, then in some embodiments, at 614, the WTRU may continue to use the original model. This may be done because the performance improvement is small or negligible and not worth the delay or resources required to transmit new model parameters and switch models. If the new encoder model does not perform better than the previous model, then in some embodiments, the WTRU may determine if the performance degradation of the new model is greater than a configured value at 616. If yes, at 618, the WTRU may notify the gNB or other device that it is switching to a legacy encoder (e.g., an encoder without compression, or an encoder with a predetermined configuration) and may switch to using the legacy encoder. If no, at 620, the WTRU may continue to use the original model. This may be done based on whether the performance is degrading rapidly (e.g., due to interference).

[0161] Figure 5An example process 500 for online update of an AI / ML encoder model is illustrated. In some embodiments, the UE or WTRU may be configured with one or more triggers for AI / ML model retraining; thresholds for AI / ML encoder selection; a set of one or more reference signals (RS) for retraining; and / or a set of resources for CSI feedback reporting. In some embodiments, when the trigger conditions are met, the UE may retrain the AI / ML encoder using the retraining RS. The UE may measure the reconstruction error and determine the AI / ML encoder model to be used (e.g., a retrained, previous, or fallback or traditional model or mode) based on the measured reconstruction error and the threshold. For example, in some embodiments, if the retrained model is better than the old model by more than a first threshold, the retrained model may be selected. If the retrained model is worse than the old model by more than a second threshold (which may be equal to or different from the first threshold), a fallback or traditional model may be used (which may include not utilizing compression in some embodiments). Otherwise, the old (pre-retrained) model may be used. The UE may select CSI feedback resources based on the selected or determined AI / ML encoder, and may send a report including compressed (or in some cases uncompressed) CSI feedback and updated encoder parameters.

[0162] Thus, in some implementations, the systems and methods discussed herein may be implemented by a wireless transmit receive unit (WTRU) and may include receiving configuration information, wherein the configuration information includes one or more of: a first artificial intelligence or machine learning (AI / ML) model and a second AI / ML model; trigger information related to performing AI / ML model retraining; a set of reconstruction error thresholds for AI / ML model selection; a set of measured reference signals for AI / ML model retraining; and / or a set of CSI feedback resources. The WTRU may measure one or more reference signals from the set of reference signals, and may generate a third AI / ML model by retraining the first AI / ML model based on the measurement of the one or more reference signals and the trigger information. The WTRU may measure a reconstruction error of a third AI / ML encoder, and may select an AI / ML model from the first AI / ML model, the second AI / ML model, or a third AI / ML model (one of which may be a legacy model or even an uncompressed coded model) based on the measured reconstruction error and the set of AI / ML model selection thresholds. The WTRU may compress a channel state information (CSI) report using a selected one of the AI / ML models to generate compressed CSI; and may transmit a message including one or more of the compressed CSI or the AI / ML encoder in the CSI feedback resource in the CSI feedback resource set. In some embodiments, the CSI feedback resource is determined based on the selected AI / ML encoder. In some embodiments, the second AI / ML encoder is a fallback encoder.

[0163] Figure 6 An example process 600 for performing an online update of an AI / ML encoder model is illustrated. The WTRU process and decision making when deciding to use the original AI / ML encoder model or a newly retrained AI / ML encoder model or even fall back to a legacy non-AI / ML encoder is shown.

[0164] In some cases, due to one or more reasons, the WTRU may determine to update the internally configured AI / ML decoder model.

[0165] For example, the WTRU may update the AI / ML model based on a periodicity. The WTRU may perform online training to periodically update the internal decoder model. The period of retraining may be configured by the gNB. The WTRU may be configured to retrain a subset of the AI / ML decoder model weights or layers through the gNB. The membership of the subset of model weights or layers that undergo online retraining may change over time.

[0166] For example, the WTRU may update the AI / ML model based on compression quality. When the quality of the encoding-decoding process degrades below a pre-configured threshold, the WTRU may determine to perform online retraining. The WTRU may first try a different encoder model paired with the existing AI / ML decoder model. If the quality of the encoding-decoding process is worse than a pre-configured threshold for all tested encoder models, the WTRU may determine to retrain the decoder model.

[0167] For example, the WTRU may update the AI / ML model based on the gNB configuration. The gNB may configure the WTRU to perform online retraining of its internal AI / ML decoder model. The WTRU may be configured to perform retraining of the decoder model only once per trigger event, or may be configured to perform periodic online retraining of the internal decoder model. The gNB may provide the WTRU with training data, such as labeled data only for training the decoder model, or some input data (e.g., uncompressed channel matrices for combined retraining of the encoder and decoder).

[0168] When the WTRU successfully retrains its internal AI / ML encoder model, it may notify the gNB. The WTRU may additionally notify the gNB of specific AI / ML model weights or layers that have been updated after online training.

[0169] The WTRU may also specify the AI / ML encoder model to be used for retraining when notifying the gNB that its AI / ML decoder model was successfully retrained. This may be indicated using an identifier, or the AI / ML encoder model weights may be included in the message to the gNB.

[0170] Upon notifying the successful update of the AI / ML decoder model within the gNB, the WTRU may wait to receive a trigger signal from the gNB to start using the retrained decoder model to test compression performance. The amount of time the WTRU should wait after receiving the trigger signal from the gNB before starting to use the newly retrained AI / ML decoder model may be preconfigured.

[0171] The trigger signal from the gNB can be included in the MAC Control Element (MAC-CE) message or can be part of the Downlink Control Information (DCI).

[0172] Upon notification of a successful update of the AI / ML decoder model within the gNB, the WTRU may receive a set of weights for the decoder model. These weights may include updated weights for the entire AI / ML decoder model or for portions of the model. The WTRU may also receive a trigger signal from the gNB simultaneously with the updated model weights or separately. Upon receipt of the trigger signal, the WTRU may begin using the newly configured AI / ML decoder weights (e.g., allowing the gNB to begin using the weights with the corresponding encoder) after a previously configured delay.

[0173] After notifying the gNB of the successful retraining of its internal AI / ML decoder model, the WTRU may switch to the fallback non-AI / ML decoder model if it fails to receive a trigger signal from the gNB within a specified duration.

[0174] Figure 7 Another example process 700 for online updating of AI / ML decoder models is illustrated. In one embodiment, at 702, the WTRU may be configured with one or more AI / ML models for CSI feedback reporting (e.g., compression), and at 704, the CSI measurements may be compressed using the identified encoder model. At 706, the WTRU may be configured to perform training (e.g., retraining or online training) of the AI / ML encoder. The WTRU processes and decisions in determining whether to use the original AI / ML decoder model or the newly retrained AI / ML decoder model or even fall back to the traditional non-AI / ML decoder may be similar to those of process 600 discussed above. However, in another embodiment, peripheral uncompressed measurements of channel conditions (e.g., peripheral CSI values) may be used for training. For example, at 708, the WTRU may determine whether there are specific CSI samples that cause large performance degradation, which may be referred to as peripheral samples. If not, then at 710, the WTRU may select a new or original decoder model based on performance improvement, as discussed above in conjunction with process 600. If yes, at 712 in some embodiments, the WTRU may use the peripheral CSI samples to retrain the internal decoder model.

[0175] The ML decoder located at the gNB may also be updated periodically or aperiodically. In one embodiment, the WTRU may transmit one or more data samples (e.g., channel coefficients) to the gNB to retrain the decoder model. In some embodiments, at 714 (in addition to or in lieu of 702-712), the WTRU may transmit samples that result in large errors (e.g., the WTRU may transmit the N worst samples). This may include both uncompressed channel coefficients and compressed channel coefficients (e.g., labeled data). This will allow the gNB to ultimately decide whether to update the decoder model.

[0176] When the WTRU encounters samples, it may transmit the samples individually, or when the WTRU collects enough samples to satisfy a preconfigured batch size, it may transmit the samples in batches. In addition, the WTRU may transmit the encoder model that the WTRU used to make the failure determination. To this end, the WTRU may transmit an encoder identifier or a model weight to the gNB.

[0177] When the requesting WTRU determines to update its decoder model, the requesting WTRU may receive an indication from the gNB.

[0178] If the gNB updates its decoder model using only the uncompressed channel samples and the compressed channel samples provided by the requesting WTRU, it may be sufficient to indicate when the gNB decides to update the decoder model at 716. Based on this indication, the WTRU may continue to use the original decoder weights at 718 or may update the model with internally learned weights at 722.

[0179] If the gNB determines to use uncompressed channel samples and compressed channel samples from more than one WTRU to retrain the decoder model, then in addition to transmitting an indication at 716 when deciding to update the decoder model, the gNB may also transmit additional uncompressed channel samples and compressed channel samples (e.g., marker data) from the other WTRUs to the requesting WTRU to cause the WTRU to retrain its internal decoder model in a similar manner. In other words, these additional samples are from the other WTRUs. The gNB may use these samples to retrain its decoder model and then transmit these samples to the requesting WTRU so that the requesting WTRU retrains the decoder (e.g., which is a copy of the decoder used by the gNB) at 724. In some cases, if the gNB retrains its decoder model, the requesting WTRU also needs to retrain its copy of the decoder so they are synchronized.

[0180] The WTRU may be pre-configured with decoder parameters that the gNB uses to update its model, such as batch size, other parameters disclosed herein, etc.

[0181] In another embodiment, at 714, the WTRU may transmit only peripheral uncompressed measurements of channel conditions (e.g., peripheral CSI values) to the gNB (after or in place of 702-712). The gNB may then train the autoencoder with the newly reported uncompressed measurements, and then transmit updated decoder weights and new or updated encoder weights to the WTRU, at 716. The peripheral CSI values ​​may be values ​​that result in large errors after decompression. The gNB may select an appropriate synthetic data set to retrain the autoencoder structure upon receiving the uncompressed CSI values ​​from the WTRU.

[0182] A WTRU belonging to a system in which there are multiple AI / ML encoder models with a small number of weights and an AI / ML decoder model with a large number of weights may periodically update its encoder model to obtain better compression performance. In addition, the WTRU may also determine to update its internal AI / ML decoder model, for example, when the compression quality degrades below a predetermined threshold, such as when the error between the actual uncompressed channel matrix and the channel matrix determined by the encoder and decoder combination is greater than a preconfigured value.

[0183] The degradation of compression quality of the WTRU may be caused by, for example, uncompressed channel matrix values ​​of the input that are not part of the value set used for the initial training of the encoder-decoder pair. These peripheral channel matrix values ​​may cause the error at the output of the decoder to be larger than a predetermined threshold.

[0184] The WTRU may transmit one or more channel matrix values ​​to the gNB for online retraining of the AI / ML decoder model. The data may include uncompressed channel matrix values ​​as well as compressed output of the encoder, such as labeled data. The gNB may then retrain the AI / ML decoder model using the labeled data provided by the WTRU.

[0185] The WTRU, when transmitting online training data to the gNB, may limit the number of sample values ​​it transmits in order to limit the feedback size. The gNB may then determine a subset of previously internally stored channel matrix values ​​to which the values ​​reported by the WTRU belong, and may use a larger number of values ​​to retrain the AI / ML decoder model. The data set used by the gNB for online training of the AI / ML decoder model may be selected from several previously stored data subsets, or may be dynamically determined by the gNB when receiving data samples from the WTRU. The gNB may determine the appropriate data subset, for example, using a similarity metric between the reported channel matrix values ​​and previously stored values, such as cosine similarity, normalized mean square error (NMSE), etc.

[0186] The WTRU may determine to transmit a portion of the samples of channel matrix values ​​that are determined to result in compression failure. For example, the WTRU may determine to transmit the N values ​​that result in the largest compression error. The number of values ​​to report and the threshold for determining compression failure may be preconfigured to the WTRU.

[0187] The WTRU may inform the gNB of the encoder model that the WTRU used to make the compression failure determination. The WTRU may transmit the encoder identifier to the gNB, or may transmit the model weights used by the WTRU in the current AI / ML decoder model to the gNB.

[0188] The WTRU may transmit the channel matrix values ​​to the gNB individually, as they are generated after measuring the channel conditions based on the reference signal, or in a batch (containing multiple such channel matrix values). The batch size and / or periodicity of the reports may be pre-configured for the WTRU.

[0189] While transmitting peripheral channel matrix values ​​that cause large compression errors to the gNB, the WTRU may wait to receive updated AI / ML decoder model weights from the gNB. The WTRU may receive updated weights for the entire decoder model or for some parts of the decoder model. The WTRU may additionally be configured to receive a trigger signal from the gNB to activate the new AI / ML decoder model.

[0190] Alternatively, when transmitting peripheral channel matrix values ​​that cause large compression errors to the gNB, the WTRU may receive an additional data set from the gNB for online retraining of the AI / ML decoder model. The WTRU may also be provided with an identifier of the encoder model and / or model weights that the gNB uses to update the AI / ML decoder model weights.

[0191] The WTRU may then retrain the AI / ML decoder model using the data samples that it has determined to have caused compression failures along with the additional data provided by the gNB. The WTRU may additionally receive a trigger signal from the gNB to indicate when it may start using the newly retrained AI / ML decoder model.

[0192] In an alternative, when outlying channel matrix values ​​that cause large compression errors are transmitted to the gNB, the WTRU may receive an indication from the gNB when a newly retrained AI / ML decoder model may be activated. Here, the WTRU may retrain the entire decoder model or certain portions of the decoder model (e.g., certain weights and / or layers) using locally generated data (such as the uncompressed channel matrix values ​​that caused the compression failure).

[0193] The data samples reported by the WTRU that result in large compression errors and the additional data samples provided by the gNB for performing AI / ML decoder model retraining may include uncompressed channel matrix values ​​or a combination of uncompressed channel matrix values ​​and compressed output from the encoder model, such as labeled data. If only uncompressed channel matrix values ​​are provided, the encoder-decoder pair must be retrained jointly. However, if labeled data is provided, the AI / ML decoder model weights may be trained independently.

[0194] The WTRU may first request channel resources from the gNB to transmit peripheral channel matrix values ​​for retraining the AI / ML decoder model. The channel resource request may include an indication of the amount of resources required to send the data.

[0195] The WTRU may be configured with one or more AI / ML encoder models (e.g., with a small number of weights) and a mechanism for determining whether the encoded output can be successfully decoded by the AI / ML decoder. This may be used, for example, when no ML decoder model (e.g., a copy of the ML decoder used by the gNB) is configured based on the CSI feedback for which the WTRU reports AI / ML.

[0196] In one case, the WTRU may determine a metric associated with the ML encoder output and compare the metric to a configured value / range. When the WTRU ML encoder output is outside the configured value / range, a CSI compression error event may occur, indicating that the compressed CSI may not be correctly decoded (decompressed) by the gNB ML decoder.

[0197] In another case, the WTRU may use a dedicated ML model to predict the success of the gNB ML decoder (e.g., no CSI compression error event occurs) or the failure of the decoder (e.g., a CSI compression error event occurs).

[0198] In another case, the WTRU may use a dedicated ML model to predict the CSI reconstruction error; the CSI reconstruction error is a measure of the distance between the decompressed CSI at the gNB ML decoder output and the actual CSI measured by the WTRU and applied at the WTRU ML encoder input (e.g., NMSE or cosine similarity may be used to measure the CSI reconstruction error).

[0199] When the WTRU detects a CSI compression error event, the WTRU may transmit an ML model retraining or ML model update request to the gNB.

[0200] The WTRU may report its ability to detect CSI compression error events. This may include: based on ML encoder output; dedicated ML models to predict decoding success / failure; and / or dedicated ML models to predict CSI reconstruction errors.

[0201] The gNB may configure the parameters of the CSI compression error event based on the reported WTRU capabilities.

[0202] For a configured AI / ML encoder model for CSI compression, the WTRU may receive an indication of a range of encoder outputs that will result in successful decoding by the decoder, e.g., the decoder output differing from the uncompressed channel coefficients by less than a threshold.

[0203] The AI / ML model configuration may include a default value for this range. The gNB may semi-statically update this range via RRC signaling.

[0204] The gNB may configure the WTRU with a metric to measure the ML encoder performance, where the metric may be a measure of the distance between the ML encoder output and a reference output corresponding to a reference uncompressed channel. For example, the metric may be the Frobenius norm of the difference between the ML encoder output and the reference output.

[0205] The WTRU may be provided with reference input data (e.g., uncompressed channel coefficients) and corresponding reference compressed outputs. The WTRU may calculate a metric between the reference compressed output and the compressed outputs of other inputs (e.g., uncompressed channel coefficients). If the calculated metric (e.g., Frobenius norm, etc.) exceeds a configured threshold, the WTRU may determine that a CSI compression error event has occurred.

[0206] The WTRU may be equipped with a dedicated ML model to predict the CSI reconstruction error. The WTRU may be configured with a threshold or a set of thresholds for the CSI reconstruction error based on which it determines whether a CSI compression error event occurs.

[0207] The WTRU applies the estimated channel matrix (e.g., uncompressed CSI) to the input of the dedicated ML model. The output of the dedicated ML model may be the predicted CSI reconstruction error.

[0208] When the indicator of CSI reconstruction error is cosine similarity, the WTRU may compare the prediction error with a configured (first) threshold. If the prediction error (cosine similarity) is less than the configured first threshold, the WTRU may determine that a CSI compression error event occurs.

[0209] When the indicator of CSI reconstruction error is NMSE, the WTRU may compare the prediction error with a configured (second) threshold. If the prediction error (NMSE) is greater than the configured second threshold, the WTRU may determine that a CSI compression error event has occurred.

[0210] When the WTRU detects that a CSI compression error event occurs, the WTRU may report the event to the gNB (e.g., provide a CSI feedback report). The WTRU may also be configured to report the estimated CSI reconstruction error.

[0211] For scheduled CSI feedback reporting opportunities, for example when the WTRU detects a CSI compression error event, the WTRU may fall back to reporting conventional CSI measurements.

[0212] When the WTRU detects a CSI compression error event, the WTRU may transmit an ML model retraining or ML model update request to the gNB. The request may be explicitly signaled (e.g., via RRC signaling) or may be implicitly indicated by the reporting of the CSI compression error event.

[0213] In some cases, there may be partial encoder training. The WTRU may be provided with a mapping between certain encoder weights and reference signal configurations (e.g., time or frequency configurations such as subbands, subcarriers, etc.) such that when the WTRU determines that a limited number of encoder weights need to be updated by freezing other weights, it requests a specific reference signal pattern for online training.

[0214] The encoder can be trained to adapt to the decorrelation between specific channel parameters or channel characteristics and the weights they affect. This decorrelation of individual parameters or characteristics can be achieved in several ways. For example, decorrelation can be performed to provide flexible retraining when specific network / hardware enforced parameters such as subbands, subcarriers, number of transmit / receive antennas, etc. are changed. For example, decorrelation can also be implemented to take into account channel characteristics that are implicitly affected by the environment, such as Doppler, delay spread, number of multipaths, channel rank, etc.

[0215] The encoder can be trained to adapt to the association between specific channel parameters or channel characteristics and the weights they affect. This association of individual parameters or characteristics can be achieved in several ways. When specific network / hardware enforced parameters such as subbands, subcarriers, number of transmit / receive antennas, etc. change, de-association can be performed to provide flexible retraining. Association can also be achieved to take into account channel characteristics that are implicitly affected by the environment, such as Doppler, delay spread, number of multipaths, channel rank, etc. In the absence of association, the WTRU can assume that the model weights are independent of the channel characteristics.

[0216] Therefore, from the perspective of partial model retraining, changes in data dimensionality caused by network / hardware enforced parameters or data distribution caused by environmental changes may only significantly affect specific weights within the encoder neural network model.

[0217] For the purpose of partial model retraining, a list of predefined channel characteristics or target configurations may be needed. The list may include, but is not limited to, one or more of the following parameters: subbands, subcarriers, number of transmit / receive antennas, Doppler, delay spread, number of multipaths, channel rank, etc. For each of these defined parameters, specific weights in the encoder model may be explicitly defined.

[0218] Therefore, to make partial encoder training feasible, one or more configurations can be used, each providing a different level of autonomy.

[0219] For example, one configuration may be one in which the encoder model is preconfigured (eg, in a standard). Additionally, a lookup table may also be specified that associates specific weights of the encoder model with each of the channel characteristics / configurations.

[0220] For example, one configuration may be that the model is not pre-configured, but the gNB indicates the architecture to the WTRU and the gNB may also signal a lookup table that provides weights to the channel characteristic / configuration associations.

[0221] For example, one configuration may be that the model and lookup table are pre-configured at the WTRU (eg, by the vendor), but it should be ensured that the model and lookup table comply with the list of channel parameters / configurations listed in the standard.

[0222] Once the WTRU or gNB has identified the need for encoder retraining, the gNB may further analyze the CSI, or may receive input from other WTRUs, to quantify the degree of variation in channel parameters. Based on this analysis, the gNB may indicate to the WTRU the level of retraining required by the WTRU. Considering the example of a convolutional neural network, the gNB may instruct the WTRU to retrain based on one or more conditions, such as: for large variations in data statistics, or large errors in CSI compression, one or more layers specified by a lookup table; for moderate levels of errors and variations, only a subset of layers may be updated, or specific channels within a layer may be updated; and / or, for small errors and variations, only a (e.g., very small) subset or specific weights within a specific layer may be updated.

[0223] Once the gNB or WTRU has identified the channel characteristics / configuration that impacts the encoder performance, the WTRU or gNB may identify the optimal RS density and configuration that can effectively capture the variations in the channel associated with the channel characteristics / configuration. For example, if the channel has high Doppler, the RS symbols may be spread across the OFDM symbols to capture the variations in the channel over time, whereas if the delay spread is high and the Doppler is low, the RS symbols may be arranged across subcarriers / Rb to ensure that the transitions in the channel across frequencies are effectively captured. This may inherently improve training performance as the data will capture the data statistics more effectively.

[0224] If the calculations for the optimal RS density and configuration are done at the gNB, the RS pattern may need to be signaled to the WTRU. Alternatively, if the RS pattern calculations are done at the WTRU, the pattern information may be signaled to the gNB.

[0225] According to one or more embodiments of the present invention, there may be techniques and methods for training AI / ML models for CSI feedback. These techniques and methods may solve one or more problems arising in wireless systems (e.g., any of the scenarios in FIG. 1 ), and may help answer questions such as: How to perform online training of an AI / ML WTRU encoder model independently of an AI / ML gNB decoder? How does a WTRU determine that an AI / ML network (e.g., gNB) decoder needs to be updated, and how to indicate the decoder update to the network? How to request additional training data for a WTRU using an AI / ML model pre-trained on a subset of channels? These problems and other related problems that may arise from AI / ML techniques for wireless reporting systems may find solutions in methods, devices, and / or systems that address situations where, for example, a WTRU trains an ML encoder model and / or a WTRU determines and helps update ML decoder weights.

[0226] In one embodiment, the WTRU may be configured with one or more AI / ML models for CSI feedback reporting (e.g., compression) and may compress CSI measurements using the identified encoder model. The WTRU may be configured to perform training (e.g., retraining or online training) of the AI / ML encoder. The WTRU may be triggered to perform AI / ML encoder training by: timing; receiving an indicator from the gNB; the performance of the WTRU; and / or the performance of the ML model. The WTRU may test the AI / ML model (e.g., the AI / ML encoder model) to determine if it is appropriate. For example, the WTRU may perform AI / ML encoder training using RS configuration. The WTRU may test the output of the decoder using actual channel measurements and may determine an error value. The WTRU may compare the error value with a set of thresholds to determine whether to use the newly trained model, continue to use the old model, or switch to legacy CSI reporting. The WTRU may be configured with a set of CSI feedback resources. The WTRU may select an appropriate CSI feedback resource based on: compression rate; selected AI / ML encoder model; use of fallback or legacy CSI reporting and report type. The WTRU may request one or more feedback resources from the gNB. The WTRU may include desired feedback parameters (e.g., payload size). The WTRU may be configured with a first feedback resource having a first payload: this may be suitable for a first compression rate of the AI / ML encoder. After training (e.g., retraining, online training), the WTRU may determine a second AI / ML model having a second compression rate. If the payload size associated with the second compression rate is greater than the first payload size, the WTRU may: include information indicating the second compression rate; send a partial CSI feedback report; send traditional CSI feedback; and / or, split the CSI feedback report into multiple sub-reports and send each sub-report in a different feedback resource.

[0227] In one embodiment, the WTRU may update the internal AI / ML decoder model. In one example, the WTRU may calculate updated decoder model weights based on peripheral channel coefficient values ​​and transmit them to the gNB. The WTRU may update only a few decoder parameters (freeze some layers) and it reports only the changed decoder parameters. In one example, the WTRU may transmit only peripheral uncompressed measurements of channel conditions to the gNB. The gNB may then train the autoencoder with the newly reported uncompressed measurements and then transmit the updated decoder weights and the new or updated encoder weights to the WTRU. In one example, the WTRU may transmit one or more samples for the gNB to retrain the decoder model. The WTRU may transmit samples that result in large errors, for example, the WTRU may transmit the N worst samples. This may include uncompressed channel coefficients and compressed channel coefficients, such as marker data. The WTRU may transmit the encoder model it uses to make a failure determination. When the gNB determines to update the decoder model based on the WTRU input, it may transmit an indication to the WTRU. If the updated decoder model only uses samples provided by the WTRU, one indication is sufficient. If the gNB uses samples from multiple WTRUs to update the decoder model, it may transmit additional samples from other WTRUs for the WTRUs to update their AI / ML decoder models.

[0228] As described herein, a higher layer may refer to one or more layers in a protocol stack, or a specific sublayer within a protocol stack. A protocol stack may include one or more layers in a WTRU or a network node (e.g., an eNB, a gNB, other functional entities, etc.), each of which may have one or more sublayers. Each layer / sublayer may be responsible for one or more functions. Each layer / sublayer may communicate directly or indirectly with one or more layers in other layers / sublayers. In some cases, the layers may be numbered, such as layer 1, layer 2, and layer 3. For example, layer 3 may include one or more of the following: non-access stratum (NAS), Internet protocol (IP), and / or radio resource control (RRC). For example, layer 2 may include one or more of the following: packet data convergence control (PDCP), radio link control (RLC), and / or medium access control (MAC). For example, layer 3 may include physical (PHY) layer type operations. The larger the number of a layer, the higher the layer is relative to other layers (e.g., layer 3 is higher than layer 1). In some cases, the foregoing examples may be referred to as layers / sublayers themselves, regardless of the layer number, and may be referred to as higher layers as described herein. For example, from highest to lowest, a higher layer may refer to one or more of the following layers / sublayers: a NAS layer, an RRC layer, a PDCP layer, an RLC layer, a MAC layer, and / or a PHY layer. Any reference to a higher layer in conjunction with a process, device, or system herein will refer to a layer higher than the layer of the process, device, or system. In some cases, references to a higher layer herein may refer to a function or operation performed by one or more layers described herein. In some cases, references to a higher layer herein may refer to information transmitted or received by one or more layers described herein. In some cases, references to a higher layer herein may refer to a configuration transmitted and / or received by one or more layers described herein.

[0229] Although features and elements are described above in specific combinations, it will be understood by those of ordinary skill in the art that each feature or element may be used alone or in any combination with other features and elements. In addition, the methods described herein may be implemented in a computer program, software, or firmware incorporated in a computer-readable medium for execution by a computer or processor. Examples of computer-readable media include electronic signals (sent via a wired or wireless connection) and computer-readable storage media. Examples of computer-readable storage media include, but are not limited to, read-only memory (ROM), random access memory (RAM), registers, cache memory, semiconductor memory devices, magnetic media (such as built-in hard disks and removable disks), magneto-optical media, and optical media (such as CD-ROM disks and digital versatile disks (DVDs)). A processor associated with software may be used to implement a radio frequency transceiver for a WTRU, UE, terminal, base station, RNC, or any host computer.

Claims

1. A method comprising: training, by a first device including a wireless transmit / receive unit (WTRU), a machine learning model for compressing channel state information based on a trigger; testing, by the WTRU, the machine learning model against actual measurements of channel conditions to determine an error value; determining, by the WTRU based on the determined error value, whether to revert to a previous machine learning model or to use the machine learning model; as well as The WTRU transmits a message to a second device including channel state information feedback generated using the machine learning model.

2. The method according to claim 1, wherein: The trigger includes a time period, receipt of an indication, a measure of performance of a device, or a measure of performance of the machine learning model. The method according to claim 1 , further comprising measuring, by the first device, channel state information. 4 . The method of claim 1 , further comprising receiving, by the first device, an indication of a plurality of feedback resources.

5. The method according to claim 4 also includes selecting, by the first device, a feedback resource for transmitting the message including channel state information feedback, and the selection is based on the determined machine learning model.

6. The method of claim 1 further comprises retraining the machine learning model based on events.

7. The method of claim 1, further comprising sending, by the first device to the second device, a second message comprising channel state information based on a payload size exceeding a configured threshold.

8. The method according to claim 1, wherein: The reference signal configuration is used to train the machine learning model.

9. The method of claim 1, further comprising identifying that the determined error value exceeds a threshold, and in response to the identifying, sending a message including uncompressed channel state information.

10. The method of claim 1 further comprises receiving configuration information from a network, wherein the configuration information comprises parameters or weights of the machine learning model.

11. A first device, comprising: one or more transceivers; and One or more processors, wherein the one or more processors are configured to: Based on the trigger, a machine learning model is trained to compress the channel state information. testing the machine learning model against actual measurements of channel conditions to determine an error value, determining whether to revert to a previous machine learning model or to use the machine learning model based on the determined error value, and A message including channel state information generated using the machine learning model is transmitted to a second device via the one or more transceivers.

12. The device according to claim 11, wherein The trigger includes a time period, receipt of an indication, a measure of performance of a device, or a measure of performance of the machine learning model.

13. The device according to claim 11, wherein: The one or more processors are further configured to measure channel state information.

14. The device according to claim 11, wherein: The one or more processors are also configured to receive an indication of a plurality of feedback resources.

15. The device according to claim 14, wherein: The one or more processors are further configured to select a feedback resource for transmitting the message comprising channel state information, the selection being based on the determined machine learning model.

16. The apparatus according to claim 11, wherein: The one or more processors are also configured to retrain the machine learning model based on events.

17. The apparatus according to claim 11, wherein: The one or more processors are further configured to send, via the one or more transceivers, a second message to the second device, the second message comprising channel state information feedback based on a payload size exceeding a configured threshold.

18. The apparatus according to claim 11, wherein: The reference signal configuration is used to train the machine learning model.

19. The apparatus according to claim 11, wherein: The one or more processors are further configured to identify that the determined error value exceeds a threshold, and in response to the identification, send a message including uncompressed channel state information feedback.

20. The apparatus of claim 11, wherein: The one or more processors are also configured to receive configuration information from a network, the configuration information comprising parameters or weights of the machine learning model.

Images