A METHOD FOR CREATING LABELS FOR ONLINE LEARNING: EXPLORING CHANNELS USING SYNONYMOUS DATA

Abstract

The invention relates to a wireless transmit / receive unit (WTRU) capable of receiving pseudo-random data (PRD), one or more channel state information reference signals (CSI-RS), and / or one or more learning reference signals (L-RS). The WTRU can determine CSI measurements based on received CSI-RS and / or determine L-RS measurements based on received L-RS. The WTRU can generate one or more CSI labels based on PRD, CSI-RS measurements, and / or L-RS measurements. The WTRU can generate one or more CSI outputs using an artificial intelligence / machine learning (AI / ML) model based on PRD and / or CSI-RS measurements. The WTRU can train an AI / ML model with one or more CSI labels and one or more CSI outputs.

Description

METHODS FOR LABEL GENERATION FOR ONLINE LEARNING OF CHANNEL SOUNDING USING PSEUDO-RANDOM DATACROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 584,204 filed on September 21 , 2023, the entire contents of which are incorporated herein by reference.BACKGROUND

[0002] Artificial intelligence (Al) may be broadly defined as the behavior exhibited by machines that mimic cognitive functions to sense, reason, adapt and / or act. An Al component may refer to the realization of behaviors and / or conformance to requirements by learning based on data without explicit configuration of sequence of steps of actions. Such Al component may enable learning complex behaviors which may be difficult to specify and / or implement when using legacy methods.

[0003] Machine learning (ML) may refer to the type of algorithms that solve a problem based on learning through experience (e.g. data), without being explicitly programmed (e.g., configuring a set of rules). ML may be considered as a subset of Al. Different ML paradigms may be envisioned based on the nature of data and / or feedback available to the learning algorithm. In an example, a supervised learning approach may involve learning a function that maps input to an output based on labeled training example. Each training example may be, e.g., a pair consisting of an input and its corresponding output. In another example, an unsupervised learning approach may involve detecting patterns in the data with no preexisting labels. In another example, a reinforcement learning approach may involve performing sequence of actions in an environment to maximize the cumulative reward.SUMMARY

[0004] A wireless transmit / receive unit (WTRU) may receive pseudo-random data (PRD), one or more channel state information reference signals (CSI-RSs), and / or one or more learning reference signals (L- RSs). The WTRU may determine CSI measurements based on the received CSI-RSs measurements. The WTRU may determine L-RS measurements based on the received L-RSs. The WTRU may generate one or more CSI labels based on the PRD, the CSI-RS measurements, and / or the L-RS measurements. The WTRU may generate one or more output CSI using an artificial intelligence / machine learning (AI / ML) model and based on the PRD and / or the CSI-RS measurements. The WTRU may train the AI / ML model with the one or more CSI labels and the one or more output CSI.

[0005] The WTRU may determine a change of a loss function of the AI / ML model based on the one or more CSI labels and the one or more output CSI. The WTRU may determine whether the training is complete based on a comparison of the value of the loss function to a threshold value. The loss function of the AI / ML model may determine a metric for training accuracy during the training of the AI / ML model.

[0006] The WTRU may determine a model drift metric. The WTRU may determine a performance of the AI / ML model for CSI estimation based on a model drift by applying a statistical test to the CSI-RS measurements. The WTRU may retrain the AI / ML model based on the model drift metric exceeding a drift threshold value.

[0007] The WTRU may send a report to a network in response to an indication that the model drift metric exceeds the threshold value. The report may comprise details of the model drift or a recommendation for adjusting the AI / ML model.

[0008] The WTRU may send a report to a network based on the CSI-RS measurements and / or a request to be configured with one or more training allocations. The WTRU may receive an indication (e.g., a response) from the network as to whether the request to be configured with one or more training allocations was granted. The indication (e.g., the response) may further comprise seed information used to generate the PRD.

[0009] The WTRU may receive a PRD training allocation configuration. The PRD training allocation may indicate a first type of PRD training allocation, a second type of PRD training allocation, and / or a third type of PRD training allocation. The first type of PRD training allocation may configure the WTRU to use the PRD and / or the CSI-RS measurements for a first set of resource blocks that carry the PRDs and the CSI- RSs. The second type of PRD training allocation may configure the WTRU to use the PRD and / or the CSI- RS measurements for a second set of resource blocks that carry the PRD, the CSI-RSs, and / or resource elements comprising data that do not carry PRD. The PRD and the resource elements comprising data that do not carry PRD may occupy the same resource blocks. The third type of PRD training allocation may configure the WTRU to use the PRD and / or CSI-RS measurements for a third set of resource blocks that carry the PRD, the CSI-RSs, and / or resource elements comprising data that do not carry the PRD. The PRD and the resource elements comprising data that do not carry PRD may occupy different but overlapping resource blocks.

[0010] The first, second, and / or third types of PRD resource allocations may be signaled in downlink control information (DCI), semi-statically via a system information block, and / or via a broadcast message.

[0011] The WTRU may send a first message that comprises an online training request. The online training request may indicate a number of resource blocks carrying PRD for training, channel statistic measurements, a channel diversification request, and / or a priority of request. The WTRU may receive a second message that comprises an online training response. The online training response may indicate the location of the resource blocks that comprise the PRDs, an allocated online training time duration, an allocated online training starting time, and / or an allocated online training learning reference signals (L-RSs).BRIEF DESCRIPTION OF THE DRAWINGS

[0012] FIG. 1 A is a system diagram illustrating an example communications system in which one or more disclosed embodiments may be implemented.

[0013] FIG. 1 B is a system diagram illustrating an example wireless transmit / receive unit (WTRU) that may be used within the communications system illustrated in FIG. 1A according to an embodiment.

[0014] FIG. 1C is a system diagram illustrating an example radio access network (RAN) and an example core network (CN) that may be used within the communications system illustrated in FIG. 1 A according to an embodiment.

[0015] FIG. 1 D is a system diagram illustrating a further example RAN and a further example CN that may be used within the communications system illustrated in FIG. 1 A according to an embodiment.

[0016] FIG. 2 is an example of Type 1 training allocation with pseudo-random data (PRD) and channel state information reference symbol (CSI-RS).

[0017] FIG. 3 is an example of Type 2 training allocation with PRD, CSI-RS, and / or coincident data scheduling.

[0018] FIG. 4 is an example diagram of online data labeling and / or training of artificial intelligence / machine learning (AI / ML) based CSI estimator.

[0019] FIG. 5 is an example diagram of performance monitoring of AI / ML-based CSI estimator using PRD.

[0020] FIG. 6 is an example diagram of performance monitoring of AI / ML-based CSI estimator using learning reference signals (L-RS).

[0021] FIG. 7 depicts an example AI / ML CSI estimation module from the WTRU perspective.DETAILED DESCRIPTION

[0022] FIG. 1A is a diagram illustrating an example communications system 100 in which one or more disclosed embodiments may be implemented. The communications system 100 may be a multiple accesssystem that provides content, such as voice, data, video, messaging, broadcast, etc., to multiple wireless users. The communications system 100 may enable multiple wireless users to access such content through the sharing of system resources, including wireless bandwidth. For example, the communications systems 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 DFT-Spread OFDM (ZT UW DTS-s OFDM), unique word OFDM (UW-OFDM), resource block-filtered OFDM, filter bank multicarrier (FBMC), and the like.

[0023] As shown in FIG. 1 A, the communications system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, a RAN 104 / 113, a CN 106 / 115, a public switched telephone network (PSTN) 108, the Internet 110, and other networks 112, though it will be appreciated 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” and / or a “ST A”, may be configured to transmit and / or receive wireless signals and may include a user equipment (UE), a mobile station, a fixed or mobile subscriber unit, a subscription-based unit, a pager, a cellular telephone, a personal digital assistant (PDA), a smartphone, a laptop, a netbook, a personal computer, a wireless sensor, a hotspot or Mi-Fi device, an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts), a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like. Any of the WTRUs 102a, 102b, 102c and 102d may be interchangeably referred to as a WTRU.

[0024] The communications systems 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 / 115, the I nternet 110, and / or the other networks 112. By way of example, the base stations 114a, 114b may be a base transceiver station (BTS), a Node-B, an eNode B, a Home Node B, a Home eNode B, a gNB, a NR NodeB, a site controller, an access point (AP), a wireless router, and the like. While the base stations 114a, 114b are each depicted as a single element, it will beappreciated that the base stations 114a, 114b may include any number of interconnected base stations and / or network elements.

[0025] The base station 114a may be part of the RAN 104 / 113, which may also include other base stations and / or network elements (not shown), such as a base station controller (BSC), a radio network controller (RNC), relay nodes, etc. The base station 114a and / or the base station 114b may be configured to transmit and / or receive wireless signals on one or more carrier frequencies, which may be referred to as a cell (not shown). These frequencies may be in licensed spectrum, unlicensed spectrum, or a combination of licensed and unlicensed spectrum. A cell may provide coverage for a wireless service to a specific geographical area that may be relatively fixed or that may change over time. The cell may further be divided into cell sectors. For example, the cell associated with the base station 114a may be divided into three sectors. Thus, in one embodiment, the base station 114a may include three transceivers, i.e., one for each sector of the cell. In an embodiment, the 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 transmit and / or receive signals in desired spatial directions.

[0026] 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).

[0027] More specifically, as noted above, the communications system 100 may be a multiple access system and may employ one or more channel access schemes, such as CDMA, TDMA, FDMA, OFDMA, SC-FDMA, and the like. For example, the base station 114a in the RAN 104 / 113 and the WTRUs 102a, 102b, 102c may implement a radio technology such as Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access (UTRA), which may establish the air interface 115 / 116 / 117 using wideband CDMA (WCDMA). 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 UL Packet Access (HSUPA).

[0028] I n 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), which may establish the air interface 116 using Long Term Evolution (LTE) and / or LTE-Advanced (LTE-A) and / or LTE-Advanced Pro (LTE-A Pro).

[0029] I n an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology such as NR Radio Access , which may establish the air interface 116 using New Radio (NR).

[0030] 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 implement LTE radio access and NR radio access together, for instance using dual connectivity (DC) principles. Thus, the air interface utilized by WTRUs 102a, 102b, 102c may be characterized by multiple types of radio access technologies and / or transmissions sent to / from multiple types of base stations (e.g., a eNB and a gNB).

[0031] 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), GSM EDGE (GERAN), and the like.

[0032] The base station 114b in FIG. 1 A may be a wireless router, Home Node B, Home eNode B, or access point, for example, and may utilize any suitable RAT for facilitating wireless connectivity in a localized area, such as a place of business, a home, a vehicle, a campus, an industrial facility, an air corridor (e.g., for use by drones), a roadway, and the like. 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 femtocell. As shown in FIG. 1A, the base station 114b may have a direct connection to the Internet 110. Thus, the base station 114b may not be required to access the Internet 110 via the CN 106 / 115.

[0033] The RAN 104 / 113 may be in communication with the CN 106 / 115, 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 varying quality of service (QoS) requirements, such as differing throughput requirements, latency requirements, error tolerance requirements, reliability requirements, data throughput requirements, mobility requirements, and the like.The CN 106 / 115 may provide call control, billing services, mobile location-based services, pre-paid calling, Internet connectivity, video distribution, etc., and / or perform high-level security functions, such as user authentication. Although not shown in FIG. 1A, it will be appreciated that the RAN 104 / 113 and / or the CN 106 / 115 may be in direct or indirect communication with other RANs that employ the same RAT as the RAN 104 / 113 or a different RAT. For example, in addition to being connected to the RAN 104 / 113, which may be utilizing a NR radio technology, the CN 106 / 115 may also be in communication with another RAN (not shown) employing a GSM, UMTS, CDMA 2000, WiMAX, E-UTRA, or WiFi radio technology.

[0034] The CN 106 / 115 may also serve as a gateway for the WTRUs 102a, 102b, 102c, 102d to access the PSTN 108, the Internet 110, and / or the other networks 112. The PSTN 108 may include circuit- switched telephone networks that provide 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), user datagram protocol (UDP) and / or the internet protocol (IP) in the TCP / IP internet protocol suite. The networks 112 may include wired and / or wireless communications 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 / 113 or a different RAT.

[0035] Some or all of the WTRUs 102a, 102b, 102c, 102d in the communications 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 over different wireless links). For example, the WTRU 102c shown in FIG. 1A 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.

[0036] FIG. 1 B is a system diagram illustrating an example WTRU 102. As shown in FIG. 1 B, the WTRU 102 may include 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, among others. It will be appreciated that the WTRU 102 may include any sub-combination of the foregoing elements while remaining consistent with an embodiment.

[0037] 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 in association with a DSP core, a controller, a microcontroller, Application SpecificIntegrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) circuits, any other type of integrated circuit (IC), a state machine, and the like. The processor 118 may perform signal coding, 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. While FIG. 1B depicts the processor 118 and the transceiver 120 as separate components, it will be appreciated that the processor 118 and the transceiver 120 may be integrated together in an electronic package or chip.

[0038] The transmit / receive element 122 may be configured to transmit signals to, or receive signals from, a base station (e.g. , the base station 114a) over the air interface 116. For example, in one embodiment, the transmit / receive element 122 may be an antenna configured to transmit and / or receive RF signals. In an embodiment, the transmit / receive element 122 may be an emitter / detector configured to transmit and / or receive IR, UV, or visible light signals, for example. In yet another embodiment, the transmit / receive element 122 may be configured to transmit and / or receive both RF and light signals. It will be appreciated that the transmit / receive element 122 may be configured to transmit and / or receive any combination of wireless signals.

[0039] Although the transmit / receive element 122 is depicted in FIG. 1 B as a single element, 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.

[0040] The transceiver 120 may be configured to modulate the signals that are to be transmitted by the transmit / receive element 122 and to demodulate the signals that are received by the transmit / receive element 122. As noted above, the WTRU 102 may have multi-mode capabilities. Thus, the transceiver 120 may include multiple transceivers for enabling the WTRU 102 to communicate via multiple RATs, such as NR and IEEE 802.11 , for example.

[0041] The processor 118 of the WTRU 102 may be coupled to, and may receive user input data from, the speaker / microphone 124, the keypad 126, and / or the display / touchpad 128 (e.g., a liquid crystal display (LCD) display unit or 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 the non-removable memory 130 and / or the removable memory 132. The non-removable memory 130 mayinclude random-access memory (RAM), 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, a secure digital (SD) memory card, and the like. In other embodiments, the processor 118 may access information from, and store data in, memory that is not physically located on the WTRU 102, such as on a server or a home computer (not shown).

[0042] The processor 118 may receive power from the power source 134, and may be configured to distribute and / or control the 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, and the like.

[0043] 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 over the air interface 116 from a base station (e.g., base stations 114a, 114b) and / or determine its location based on the timing of the signals being received from two or more nearby base stations. It will be appreciated that the WTRU 102 may acquire location information by way of any suitable locationdetermination method while remaining consistent with an embodiment.

[0044] The processor 118 may further be coupled to other peripherals 138, which may include one or more software 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 e-compass, a satellite transceiver, a digital camera (for photographs and / or video), a universal serial bus (USB) port, a vibration device, a television transceiver, a hands free headset, a Bluetooth® module, a frequency modulated (FM) radio unit, a digital music player, a media player, a video game player module, an Internet browser, a Virtual Reality and / or Augmented Reality (VR / AR) device, an activity tracker, and the like. The peripherals 138 may include one or more sensors, the sensors may be one or more of a gyroscope, an accelerometer, a hall effect sensor, a magnetometer, an orientation sensor, a proximity sensor, a temperature sensor, a time sensor; a geolocation sensor; an altimeter, a light sensor, a touch sensor, a magnetometer, a barometer, a gesture sensor, a biometric sensor, and / or a humidity sensor.

[0045] 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 particular subframes for both the UL (e.g., for transmission) and downlink (e.g., for reception) may be concurrent and / or simultaneous. The full duplex radio may include aninterference management unit 139 to reduce and or substantially eliminate self-interference via either hardware (e.g., a choke) or signal processing via a processor (e.g., a separate processor (not shown) or via processor 118). In an embodiment, the WRTU 102 may include a half-duplex radio for which transmission and reception of some or all of the signals (e.g., associated with particular subframes for either the UL (e.g, for transmission) or the downlink (e.g, for reception)).

[0046] FIG. 1C is a system diagram illustrating the RAN 104 and the CN 106 according to 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.

[0047] The RAN 104 may include eNode-Bs 160a, 160b, 160c, though it will be appreciated that the RAN 104 may include any number of eNode-Bs while remaining consistent with an embodiment. The eNode-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 eNode-Bs 160a, 160b, 160c may implement MIMO technology. Thus, the eNode-B 160a, for example, may use multiple antennas to transmit wireless signals to, and / or receive wireless signals from, the WTRU 102a.

[0048] Each of the eNode-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, scheduling of users in the UL and / or DL, and the like. As shown in FIG. 1C, the eNode-Bs 160a, 160b, 160c may communicate with one another over an X2 interface.

[0049] The CN 106 shown in FIG. 1 C may include a mobility management entity (MME) 162, a serving gateway (SGW) 164, and a packet data network (PDN) gateway (or PGW) 166. While each of the foregoing elements are depicted as part of the CN 106, it will be appreciated that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0050] The MME 162 may be connected to each of the eNode-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, selecting a particular serving gateway during an initial attach of the WTRUs 102a, 102b, 102c, and the like. 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.

[0051] The SGW 164 may be connected to each of the eNode Bs 160a, 160b, 160c in the RAN 104 via the S1 interface. The SGW 164 may generally route and forward user data packets to / from the WTRUs102a, 102b, 102c. The SGW 164 may perform other functions, such as anchoring user planes during inter- eNode B handovers, triggering paging when DL data is available for the WTRUs 102a, 102b, 102c, managing and storing contexts of the WTRUs 102a, 102b, 102c, and the like.

[0052] 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.

[0053] 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 communicate with, an IP gateway (e.g., an IP multimedia subsystem (IMS) server) that serves 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 the other networks 112, which may include other wired and / or wireless networks that are owned and / or operated by other service providers.

[0054] Although the WTRU is described in FIGS. 1 A-1 D as a wireless terminal, it is contemplated that in certain representative embodiments that such a terminal may use (e.g., temporarily or permanently) wired communication interfaces with the communication network.

[0055] In representative embodiments, the other network 112 may be a WLAN.

[0056] A WLAN in Infrastructure Basic Service Set (BSS) mode may have an Access Point (AP) for the BSS and one or more stations (STAs) associated with the AP. The AP may have an access or an interface to a Distribution System (DS) or another type of wired / wireless network that carries traffic in to and / or out of the BSS. Traffic to STAs that originates from outside the BSS may arrive through the AP and may be delivered to the STAs. Traffic originating from STAs to destinations outside the BSS may be sent to the AP to be delivered to respective destinations. Traffic between STAs within the BSS may be sent through the AP, for example, where the source STA may send traffic to the AP and the AP may deliver the traffic to the destination STA. The traffic between STAs within a BSS may be considered and / or referred to as peer-to- peer traffic. The peer-to-peer traffic may be sent between (e.g., directly between) the source and destination STAs with a direct link setup (DLS). In certain representative embodiments, the DLS may use an 802.11 e DLS or an 802.11 z tunneled DLS (TDLS). A WLAN using an Independent BSS (I BSS) mode may not have an AP, and the STAs (e.g., all of the STAs) within or using the IBSS may communicatedirectly with each other. The IBSS mode of communication may sometimes be referred to herein as an “ad- hoc” mode of communication.

[0057] When using the 802.11 ac infrastructure mode of operation or a similar mode of operations, the AP may transmit a beacon on a fixed channel, such as a primary channel. The primary channel may be a fixed width (e.g., 20 MHz wide bandwidth) or a dynamically set width via signaling. The primary channel may be the operating channel of the BSS and may be used by the STAs to establish a connection with the AP. In certain representative embodiments, Carrier Sense Multiple Access with Collision Avoidance (CSMA / CA) may be implemented, for example in in 802.11 systems. For CSMA / CA, the STAs (e.g., every ST A), 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.

[0058] High Throughput (HT) STAs may use a 40 MHz wide channel for communication, for example, via a combination of the primary 20 MHz channel with an adjacent or nonadjacent 20 MHz channel to form a 40 MHz wide channel.

[0059] Very High Throughput (VHT) STAs may support 20MHz, 40 MHz, 80 MHz, and / or 160 MHz wide channels. The 40 MHz, and / or 80 MHz, channels may be formed by combining contiguous 20 MHz channels. A 160 MHz channel may be formed by combining 8 contiguous 20 MHz channels, or by combining two non-contiguous 80 MHz channels, which may be referred to as an 80+80 configuration. For the 80+80 configuration, the data, after channel encoding, may be passed through a segment parser that may divide the data into two streams. Inverse Fast Fourier Transform (IFFT) processing, and time domain processing, may be done on each stream separately. The streams may be mapped on to the two 80 MHz channels, and the data may be transmitted by a transmitting STA. At the receiver of the receiving STA, the above described operation for the 80+80 configuration may be reversed, and the combined data may be sent to the Medium Access Control (MAC).

[0060] Sub 1 GHz modes of operation are supported by 802.11 af and 802.11 ah. The channel operating bandwidths, and carriers, are reduced in 802.11 af and 802.11 ah relative to those used in 802.11 n, and 802.11ac. 802.11 af supports 5 MHz, 10 MHz and 20 MHz bandwidths in the TV White Space (TVWS) spectrum, and 802.11 ah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non-TVWS spectrum. According to a representative embodiment, 802.11 ah may support Meter Type Control / Machine- Type Communications, such as MTC devices in a macro coverage area. MTC devices may have certain capabilities, for example, limited capabilities including support for (e.g., only support for) certain and / orlimited bandwidths. The MTC devices may include a battery with a battery life above a threshold (e.g., to maintain a very long battery life).

[0061] WLAN systems, which may support multiple channels, and channel bandwidths, such as 802.11 n, 802.11 ac, 802.11 af, and 802.11 ah, include a channel which may be designated as the primary channel. The primary channel may have a bandwidth equal to the largest 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, from among all STAs in operating in a BSS, which supports the smallest bandwidth operating mode. In the example of 802.11 ah, the primary channel may be 1 MHz wide for STAs (e.g., MTC type devices) that support (e.g., only support) a 1 MHz mode, even if the AP, and other STAs in the BSS support 2 MHz, 4 MHz, 8 MHz, 16 MHz, and / or other channel bandwidth operating modes. Carrier sensing and / or Network Allocation Vector (NAV) settings may depend on the status of the primary channel. If the primary channel is busy, for example, due to a STA (which supports only a 1 MHz operating mode), transmitting to the AP, the entire available frequency bands may be considered busy even though a majority of the frequency bands remains idle and may be available.

[0062] In the United States, the available frequency bands, which may be used by 802.11 ah, are from 902 MHz to 928 MHz. In Korea, the available frequency bands are from 917.5 MHz to 923.5 MHz. In Japan, the available frequency bands are from 916.5 MHz to 927.5 MHz. The total bandwidth available for 802.11 ah is 6 MHz to 26 MHz depending on the country code.

[0063] FIG. 1 D is a system diagram illustrating the RAN 113 and the CN 115 according to an embodiment. As noted above, the RAN 113 may employ an NR radio technology to communicate with the WTRUs 102a, 102b, 102c over the air interface 116. The RAN 113 may also be in communication with the CN 115.

[0064] The RAN 113 may include gNBs 180a, 180b, 180c, though it will be appreciated that the RAN 113 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, gNBs 180a, 108b may utilize beamforming to transmit signals to and / or receive signals from the gNBs 180a, 180b, 180c. Thus, the gNB 180a, for example, may use multiple antennas to transmit 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, the gNB 180a may transmit multiple component carriers to the WTRU 102a (not shown). A subset of these component carriersmay be on unlicensed spectrum while the remaining component carriers may be on licensed spectrum. In an embodiment, the 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).

[0065] The WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using transmissions associated with a scalable numerology. For example, the 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 gNBs 180a, 180b, 180c using subframe or transmission time intervals (TTIs) of various or scalable lengths (e.g., containing varying number of OFDM symbols and / or lasting varying lengths of absolute time).

[0066] 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 the standalone configuration, WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c without also accessing other RANs (e.g., such as eNode-Bs 160a, 160b, 160c). In the standalone configuration, WTRUs 102a, 102b, 102c may utilize one or more of gNBs 180a, 180b, 180c as a mobility anchor point. In the standalone configuration, WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using signals in an unlicensed band. In a non-standalone configuration WTRUs 102a, 102b, 102c may communicate with / connect to gNBs 180a, 180b, 180c while also communicating with / connecting to another RAN such as eNode-Bs 160a, 160b, 160c. For example, WTRUs 102a, 102b, 102c may implement DC principles to communicate with one or more gNBs 180a, 180b, 180c and one or more eNode-Bs 160a, 160b, 160c substantially simultaneously. In the non-standalone configuration, eNode-Bs 160a, 160b, 160c may serve as a mobility anchor for WTRUs 102a, 102b, 102c and gNBs 180a, 180b, 180c may provide additional coverage and / or throughput for servicing WTRUs 102a, 102b, 102c.

[0067] Each of the gNBs 180a, 180b, 180c may be associated with a particular cell (not shown) and may be configured to handle radio resource management decisions, handover decisions, scheduling of users in the UL and / or DL, support of network slicing, dual connectivity, interworking between NR and E-UTRA, routing of user plane data towards User Plane Function (UPF) 184a, 184b, routing of control plane information towards Access and Mobility Management Function (AMF) 182a, 182b and the like. As shown in FIG. 1D, the gNBs 180a, 180b, 180c may communicate with one another over an Xn interface.

[0068] The CN 115 shown in FIG. 1 D may include at least one AMF 182a, 182b, at least one UPF 184a, 184b, at least one Session Management Function (SMF) 183a, 183b, and possibly a Data Network(DN) 185a, 185b. While each of the foregoing elements are depicted as part of the CN 115, it will be appreciated that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0069] The AMF 182a, 182b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 113 via an N2 interface and may serve as a control node. For example, the AMF 182a, 182b may be responsible for authenticating users of the WTRUs 102a, 102b, 102c, support for network slicing (e.g., handling of different PDU sessions with different requirements), selecting a particular SMF 183a, 183b, management of the registration area, termination of NAS signaling, mobility management, and the like. Network slicing may be used by the AMF 182a, 182b in order to customize CN support for WTRUs 102a, 102b, 102c based on the types of services being utilized WTRUs 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, services for machine type communication (MTC) access, and / or the like. The AMF 162 may provide a control plane function for switching between the RAN 113 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.

[0070] The SMF 183a, 183b may be connected to an AMF 182a, 182b in the CN 115 via an N11 interface. The SMF 183a, 183b may also be connected to a UPF 184a, 184b in the CN 115 via an N4 interface. The SMF 183a, 183b may select and control the UPF 184a, 184b and configure the routing of traffic through the UPF 184a, 184b. The SMF 183a, 183b may perform other functions, such as managing and allocating WTRU IP address, managing PDU sessions, controlling policy enforcement and QoS, providing downlink data notifications, and the like. A PDU session type may be IP-based, non-IP based, Ethernet-based, and the like.

[0071] The UPF 184a, 184b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 113 via an 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 downlink packets, providing mobility anchoring, and the like.

[0072] The CN 115 may facilitate communications with other networks. For example, the CN 115 may include, or may communicate with, an IP gateway (e.g., an IP multimedia subsystem (IMS) server) that serves as an interface between the CN 115 and the PSTN 108. In addition, the CN 115 may provide theWTRUs 102a, 102b, 102c with access to the other networks 112, which may include other wired and / or wireless networks that are owned and / or operated by other service providers. In one embodiment, the WTRUs 102a, 102b, 102c may be connected to a local Data Network (DN) 185a, 185b through the UPF 184a, 184b via the N3 interface to the UPF 184a, 184b and an N6 interface between the UPF 184a, 184b and the DN 185a, 185b.

[0073] In view of Figures 1A-1 D, and the corresponding description of Figures 1A-1 D, one or more, or all, of the functions described herein with regard to one or more of: WTRU 102a-d, Base Station 114a-b, eNode-B 160a-c, MME 162, SGW 164, PGW 166, gNB 180a-c, AMF 182a-ab, UPF 184a-b, SMF 183a-b, DN 185a-b, and / or any other device(s) described herein, may be performed by one or more emulation devices (not shown). The emulation devices may be one or more devices configured to emulate one or more, or all, of the functions described herein. For example, the emulation devices may be used to test other devices and / or to simulate network and / or WTRU functions.

[0074] The emulation devices may be designed to implement one or more tests of other devices in a lab environment and / or in an operator network environment. For example, the one or more emulation devices may perform the one or more, 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. The one or more emulation devices may perform the one or more, or all, functions while being temporarily implemented / deployed as part of a wired and / or wireless communication network. The emulation device may be directly coupled to another device for purposes of testing and / or may performing testing using over-the-air wireless communications.

[0075] The one or more emulation devices may perform the one or more, including all, functions while not being implemented / deployed as part of a wired and / or wireless communication network. For example, the emulation devices may be utilized in a testing scenario in a testing laboratory and / or a non-deployed (e.g, testing) wired and / or wireless communication network in order to implement testing of one or more components. The one or more emulation devices may be test equipment. Direct RF coupling and / or wireless communications via RF circuitry (e.g., which may include one or more antennas) may be used by the emulation devices to transmit and / or receive data.

[0076] Described herein are methods for the WTRU to enable online learning for channel state feedback (CSF) functions for systems using artificial intelligence / machine learning (AI / ML) for CSF functions.

[0077] Wireless devices need to adapt to their local environmental conditions, such as the radio propagation environment. To better adapt, the channel state information (CSI) may be known at bothtransmitting and receiving nodes. With the CSI known at both transmitting and receiving nodes, the transmission may be properly configured and / or received, respectively. CSI may be used at the transmitter to select modulation and coding scheme (MCS), layers, precoder, power, and / or resource block (RB) allocation, etc., CSI may be used at the receiver (either explicitly and / or implicitly) for demodulation. A non- AI / ML algorithm (e.g., channel estimator and / or predictor) may typically estimate CSI. The non-AI / ML algorithm that itself uses the output of other non-AI / ML algorithms (e.g., Doppler estimator) to estimate, predict, or provide other channel related statistics. AI / ML may be used to learn CSI feedback functions (e.g., an estimator, a predictor, a compressor, and / or a combination of these, etc.).

[0078] When these CSI feedback functions are learned offline, error-free ground truth channels data may be typically available as a dataset utilized for training. The offline training data may be obtained from a handful of available and / or agreed channel models. These channel models may have limited support of true mobility, do not include radio impairments, and / or may not be well matched to the local environment where the WTRU resides. For a device to adapt to its environment, AI / ML may learn CSF functions optimized for the local environment if training data specific to the WTRU and its environment may be measured online.

[0079] The input to the AI / ML model may be the received and / or processed pilot (e.g., reference) signals after application of the fast Fourier Transform (FFT). The desired output of the model, also referred to “labels” and / or “targets”, may not be directly observable. Instead, a method to create high quality estimates of the channel used as labels may be needed. Obtaining high quality training labels is challenging in an operational system as the overhead associated with pilot (e.g., reference) signals may need to be minimized. Minimizing pilot (e.g., reference) signals may optimize throughput and / or other key performance indicators (KPIs). Training an AI / ML based channel estimator with conventional estimators and / or conventional pilots may not produce better estimates than the conventional estimator.

[0080] In examples, it is possible to apply machine learning (ML) algorithms to a combination and / or interpolation of the above-mentioned approaches. For example, a semi-supervised learning approach may use a combination of a small amount of labeled data with a large amount of unlabeled data during training. In this setup, a semi-supervised learning falls between unsupervised learning (e.g., with no labeled training data) and supervised learning (e.g., with only labeled training data).

[0081] Deep learning (DL) refers to a class of ML algorithms that may employ artificial neural networks, specifically Deep Neural Networks (DNNs). DNNs are loosely inspired from biological systems. The DNNs are a special class of ML models inspired by the human brain wherein the input may be linearlytransformed and pass through non-linear activation functions multiple times. DNNs may typically consist of multiple layers. Each layer may consist of a linear transformation and / or given non-linear activation functions. Examples of the linear transformation may include: a multiplication of the input of the layer and the learnable weights of the layer, wherein a learnable bias is added to the resulting multiplication; a convolution operation between the input of the layer and the learnable weights of the layer, wherein a learnable bias is added to the resulting convolution.

[0082] The DNNs may be trained using the training data via a backpropagation algorithm. Backpropagation is the operation of computing the loss of the AI / ML model (e.g., a measured difference between the output of the AI / ML model and the target labels (e.g., the ground truth) for the case of supervised learning). Backpropagation may involve computing the derivatives of the computed loss with respect to the learnable parameters (e.g., weights and / or biases, etc.) of the AI / ML model, and then backpropagating (e.g., passing the derivatives in a backward manner) these derivatives over the layers of the AI / ML model (e.g., from the output layer (the last layer) up to the input layer (the first layer)). Computation may require use of the differentiation chain rule to compute the gradients of each layer with respect to its parameters and / or to update the parameters of each layer using the computed gradients.

[0083] CSI feedback functions may define a series of functions implemented at the WTRU side to enable the estimation of the channel state information (CSI) and / or its transmission to the network. The network may exploit the received CSI feedback to apply link adaptation functions to the WTRU (e.g., appropriate MCS and / or precoding, beam management, power allocation, and / or resource block (RB) allocation, etc.). The CSI feedback functions may primarily include the CSI estimation. From the CSI information, the WTRU may generate a CSI report containing measurements indicators that detail the channel’s state (e.g., channel quality index (CQI), pre-coding matrix indicator (PMI), rank indicator (Rl), and / or layer indicator (LI), etc.), for 5G new radio (NR), as detailed in Third Generation Partnership Project (3GPP) Release 15.

[0084] Mechanisms and / or frameworks for using artificial intelligence / machine learning (AI / ML) based approaches at the air interface level may be enabled, for example, for CSI compression and CSI prediction.

[0085] CSI compression may define the operation of compressing the CSI estimates by the WTRU within a predefined feedback size (in bits) and / or transmitting the compressed CSI estimates to the network. The network, in turn, may reconstruct the CSI estimates by decompressing the received CSI feedback from the WTRU. Additionally, or alternatively, CSI prediction may define the operation of prediction posterior CSI, either by the WTRU or by the network, from historical CSI estimates. These two sub-use cases may be within the CSI feedback functions and / or are implemented after the CSI estimation operation.

[0086] AI / ML operations for CSI feedback functions may be based on models (e.g., neural networks) trained using a substantial amount of data under different scenarios and / or conditions. For a given function, there may be several models (e.g., each trained and / or suitable for different network and / or WTRU conditions and / or characteristics). Once an AI / ML model is trained, the model may be deployed for online inference. After deployment, AI / ML model monitoring may be performed in a real network as the current network and / or WTRU conditions may become different from the scenarios and / or conditions in which the model was trained and / or tested. AI / ML model training and / or monitoring for CSI feedback functions may be performed at the WTRU, at the network, and / or in collaboration between the two.

[0087] After deploying an AI / ML model for online inference, the model’s performance may degrade over time due to a concept called “model drift.” The deployed AI / ML model may constantly receive new data to make predictions upon. However, this data may have a different probability distribution than the one used for training. Hence, using the original AI / ML model with the new data distribution may cause a drop in the model performance. AI / ML model drift may be associated with the model’s performance declining over time. This degradation in performance may cause the model to give poor predictions. AI / ML model drift may be categorized as concept drift or data drift.

[0088] Concept drift may occur when the functional relationship between the inputs and outputs of an AI / ML models changes. The functional relationship between the inputs and outputs of the AI / ML models may have changed, but the model does not know about the change. Accordingly, the model may no longer retain learned parameters. Data drift may occur when the AI / ML model’s input distribution changes. The AI / ML model’s performance may weaken because the model has not received sufficient data and / or been trained enough. Data drift may also occur if the data received by the AI / ML model at the online inference time contains features that were not present during the offline training.

[0089] During the online inference of AI / ML models, the WTRU and / or the network AI / ML endpoint e.g., a gNB, a cloud server, and / or an edge computing server, etc.) may detect that the performance of the AI / ML model has degraded. When a WTRU first enters a new geographic region, the AI / ML model may not be well trained for this new local environment. In this situation, the WTRU may request support for updating the AI / ML model. One solution is that the WTRU may retrain the AI / ML model online with the assistance of another node in the network (e.g., gNB, another WTRU using sidelink (SL) and / or WTRU-to-WTRU direct communication, etc.). Although the proposed solution may apply for any CSI feedback function (e.g., CSI estimation, CSI prediction, CSI compression, and / or the combination of them, etc.), the rest of thedisclosure herein focuses on CSI estimation as an example. In the context of channel estimation, this solution focuses on providing the steps that allow a WTRU to perform online training of AI / ML models.

[0090] The WTRU and / or the network may detect a need to initiate online training of channel estimation algorithm. In examples, the AI / ML model drift detection mechanism in WTRU may indicate that the AI / ML model has drifted and / or is drifting. The WTRU may enter a geographic region and / or a cell for which the channel estimation algorithm has not been previously trained (e.g., WTRU compared global positioning system (GPS) coordinates to measure distance to previously trained regions). If the distance to previously trained regions exceeds a threshold provided by the network, the WTRU may signal the network indicating the threshold is exceeded. The WTRU may further possibly signal its distance and / or current location.

[0091] If the WTRU supports an AI / ML-based CSI estimator that uses data symbols (in addition to channel state information reference signals (CSI-RS)), then the WTRU capabilities message may indicate the support. If the WTRU supports an AI / ML-based CSI estimator that uses data symbols (in addition to CSI- RS) then the WTRU may request radio resource allocations for training that use pseudo-random data (PRD) training allocation in the place of data resource elements (REs).

[0092] The WTRU may also request and / or use learning reference signals (L-RSs) training allocations in addition to PRD training allocations. The WTRU may generate CSI reports based on CSI-RS. The network may use previous CSI report information to choose precoders for the PRD training allocation. The network may analyze the CSI reports. The network may determine whether to make PRD training allocations for the WTRU. The network may acknowledge the request for PRD training allocations, indicating denied or accepted. If accepted, the network may send to the WTRU a key, seed, and / or other information required to locally generate pseudo random bits and / or symbols. After receiving this information, the WTRU may now be ready to receive PRD training allocations. The network may indicate to the WTRU that the WTRU may assume that the PRD is precoded according to the last reported CSI report from the WTRU.

[0093] The WTRU may assume that demodulation reference signal (DMRS) for the PRD training allocation uses a precoder consistent with its recent CSI reports. If the WTRU supports blind detection of PRD training allocations, then the WTRU capabilities message may indicate the support. The network may make PRD training allocations for the WTRU without indicating that the allocation is a PRD training allocation. Aperiodic PRD training allocations may include a PRD training indication in the downlink control information (DCI) and / or other downlink (DL) control channel. Semi-persistent PRD training allocations may be configured without WTRU specific (and / or without WTRU group specific) signaling (e.g., radio resource control (RRC) and / or MAC-CE). Periodic and / or semi-persistent PRD allocations may be indicatedas opportunistic. If a PRD allocation is opportunistic, the WTRU may determine if an allocation contains PRD before using the allocation for training.

[0094] The PRD Training request may indicate the training bandwidth (BW) and / or number of RBs. The network may decide the size e.g., how large) of the training allocations. The request may indicate channel statistics measurements such as Delay Spread and / or Doppler. The request may indicate channel diversification (e.g., increased Delay Spread or Doppler). Therein, the network may decide the size (e.g., how much) of the variation (if any) to add to the channel, (e.g., a time-varying finite impulse response (FIR) filter may augment the channel).

[0095] The request may indicate priority of request. Herein, the WTRU may indicate a high priority if the performance of the channel estimator is below a certain threshold. The WTRU may receive the performance threshold from the network (e.g., gNB). The network may indicate in a system information block (SIB) and / or RRC that an estimated model drift and / or normalized mean square error (NMSE) of channel estimate above a threshold may be required for WTRU to make priority request.

[0096] The network may decide to override PRD training allocations with data REs. The network may also decide to decrease or not increase the training allocations if the priority is not high. The WTRU may indicate a low priority if the request is for predictive maintenance, (e.g., to compare the quality of an AI / ML- based CSI estimator to the CSI estimated by a reference CSI estimator using L-RS). A WTRU that supports PRD training may also request and / or use already present L-RS training allocations to measure performance of the AI / ML estimator. If the comparison indicates the AI / ML performance has degraded compared to a threshold, the WTRU may signal the network with a new message to increase the priority of training allocation.

[0097] The WTRU may use non-specific training allocations. These allocations may be immediate (signaled in DCI and / or similar control channel), and / or semi-static with training allocation information indicated in SIB and / or other broadcast messages. Non-specific training allocations may be non-specific to any WTRU. Accordingly, multiple WTRUs may simultaneously use the same training allocations. Nonspecific training allocations may include PRD training allocations and / or L-RS training allocations. In the PRD training allocation, data REs may be populated with random looking data that may be used as input to the AI / ML-based CSI estimator (e.g., a trainable channel estimator). The same random looking data may be known sequences. The same random looking data may make high-quality channel estimates as labels for trainable (e.g., AI / ML) channel estimators that use normal CSI-RS as inputs.

[0098] Non-specific training allocations may be configured as semi-persistent, periodic, semi-periodic, and / or periodic. FIG. 2 is an example of Type 1 training allocation 200 with PRD and CSI-RS. Type 1 PRD training allocations may include PRD 204 (e.g., symbols) in the REs normally carrying data. In addition, REs used for DMRS may be used for PRD DRMS 208. REs used for DMRS may be precoded in the same way as the PRD 204. In this method, as depicted in FIG. 2, the allocated RBs, (e.g., or resource block groups (RBGs)), may not carry any user data. The data REs may instead be populated with PRD REs 204. CSI-RS 212, control channels such as PDCCH 216, and / or other physical (PHY) channels may be present. The PRD REs 204 may be derived from keyed and / or seeded random number source. The PRD REs 204 may therefore be treated as pilot sequences after preprocessing to make high-quality channel estimates as labels for trainable channel estimators that use normal CSI-RS 212 as inputs.

[0099] The same PRD and / or, e.g., PRD DMRS REs, but without preprocessing based on the known PRD sequence, may then be used as inputs to the AI / ML model during training. Any present CSI-RS may also be used as input to the AI / ML model during training. This way, the model may learn to use data REs as inputs in addition to CSI-RS. The model may learn that high-quality labels may be generated simultaneously using the same PRD data and / or signal. Since the network is not required to use precoders indicated by the WTRU CSI report, the WTRU may use only CSI-RS as inputs to a second trainable channel estimator. This second trainable channel estimator may not use data REs as inputs during interface and / or may not use PRD as inputs during training.

[0100] FIG. 3 is an example of Type 2 training allocation 300 with PRD, CSI-RS, and / or coincident data scheduling. Type 2 PRD training allocations include PRD 304 (e.g., symbols) and / or data REs. Type 2 PRD training allocations may leave a portion of REs available for user data 308. A common DRMS 312 is used for both the data 308 and / or the PRD 304. The same antenna ports and / or the same precoder are used for the data 308, PRD 304, and / or the DMRS 312. In this method, as depicted in FIG. 3, some of the REs in a training allocation may be populated with PRD REs 304 so that high-quality channel estimates may be created as labels for trainable channel estimators that use normal CSI-RS 316 and / or data REs 308 as inputs. CSI-RS 316, control channels such as PDCCH 320, and / or other physical (PHY) channels may be present. Note that the input to the AI / ML model may include both the data 308 and / or the PRD 304 if the PRD 304 uses the same modulation as the data. Note also that the high-quality CSI estimate used for CSI label may also be used as the channel estimate for either CSI reporting and / or demodulation.

[0101] Type 2a PRD training allocations may refer to a training allocation and WTRU data allocation occupying the same (e.g., exactly the same) slot-RBs. WTRUs that have capability to use Type 2a PRDtraining allocations for data may be scheduled for coincident training and / or data. These WTRUs may learn the pattern of PRD and / or WTRU specific REs from SIB, RRC, and / or other broadcast information. WTRUs that have capability to use Type 2a PRD training allocations for training (such WTRUs must be part of a WTRU group for which another WTRU in the group is getting data) may learn the PRD pattern from SIB, RRC, and / or other broadcast information. The Type 2a allocation information in a SIB may indicate that when no data is available in a Type 2a allocation, the WTRU specific REs may be filled with PRD. If indicated in the SIB, the WTRU may detect if the PRD is present in the WTRU specific REs. The WTRU may use the PRD for improved CSI labels and / or additional inputs to the AI / ML estimator.

[0102] Type 2b non-specific training allocations (the training allocation and / or WTRU data allocation may use different but overlapping slot-RBs) may utilize the same procedure as Type 2a. However, from a training perspective, the WTRU should know which RBs of the training allocation contain WTRU specific REs. A WTRU capable of training with a Type 2b training allocation may indicate this capability to the network. The WTRU accordingly indicates to the network that putting this WTRU in a group that uses Type 2b training allocations may not incorrectly incorporate user data REs into AI / ML training data.

[0103] The WTRU may further be configured for blind detection. The WTRU using a Type 2 training allocation may use the REs in each RB and / or RBG to separately determine if the RB and / or RBG may carry user data and / or exclude REs carrying user data from AI / ML training.

[0104] The WTRU may further be configured for control channel indication (e.g., PDCCH). Carrying the allocation for user data may also carry information detectable (at least) by WTRUs with Type 2b capability that may indicate the RBs used in a data allocation. The WTRUs using the training allocation may exclude REs carrying user data from AI / ML training.

[0105] From a data reception perspective, the WTRU should know which RBs of the data allocation contain PRD. A WTRU capable of data reception in a Type 2b PRD training allocation may indicate to the network that the WTRU has identified semi-persistent training allocations. In other words, the WTRU may indicate that the WTRU knows the allocation information for the indicated semi-static training allocations. The network may then permit coincident data and / or training allocations to occur for the WTRU that has made such indication. When the WTRU receives a data allocation, the WTRU may identify the RBs of the data allocation that intersect with the training allocation. Based on the training allocation information, the WTRU may determine which REs from the intersecting RBs may actually be carrying user data. The WTRU may select those REs for data reception accordingly. If the training allocation is opportunistic (e.g., determined also from the training allocation information), the WTRU may also check the REs that similarlycarry PRD. Opportunistic training allocation may provide that a WTRU is capable of detecting the PRD training allocations. The WTRU may perform this check to determine if the REs carry user data and / or PRD and / or selects those REs for data reception accordingly.

[0106] The WTRU may receive broadcast information to learn the current PRD training allocation schedules for the WTRU and / or WTRU group(s). The WTRU may read broadcast information (e.g., SIB and / or RRC) to learn the semi-static schedule of PRD training allocations of Type 1 , Type 2a, and / or Type 2b. The allocation information may be similar to other semi-static allocations (e.g., starting SFN and slot, RBs and / or RBGs, periodicity, and / or frequency hopping, etc.) and / or PRD specific information. The PRD training allocation may include resource allocation information and / or PRD power offset information, (e.g., power offset to CSI-RS).

[0107] The PRD training allocation may include a PRD number generator state and / or seed and / or corresponding system frame number (SFN). The PRD training allocation may include these so that the WTRU may use them to learn the random sequence of symbols used in the PRD REs. The PRD training allocation may include a modulation type indication. The indication may indicate a sequence of different modulation orders. For example, if WTRU reads a modulation indicator of {0,1 , 0,2, 3, 0,0}, then the 1st, 3rd, 6th, and 7thRBs are quadrature phase shift keying (QPSK), the 2ndRB is 16 quadrature amplitude modulation (QAM), the 4thRB is 64QAM, and the 5thRB is 256QAM. The semi-periodic PRD training allocation may include ‘opportunistic’ indication. Opportunistic indication may provide that a WTRU is capable of detecting the PRD training allocations. The opportunistic indication may be present as information bits in the SIB. If indicated, the WTRU may not assume the allocation is a PRD training allocation. Instead, the WTRU may detect if the allocation is a PRD training allocation and / or use the allocation for training. Detection may consist of correlation to the known PRD symbol sequence. Opportunistic indication may apply to Type 1 and / or Type 2 semi-static training allocations.

[0108] The WTRU may be assigned to a WTRU PRD group if multiple PRD training allocations are used. Directional systems may need such grouping. Different ports may be used for different directions (e.g., groups of WTRUs). Moreover, non-directional systems may need such grouping. For example, the WTRU may be grouped according to the likely precoder to be used. The network may learn the likeliest precoder to use from the WTRU CSI reports. The correlation between CSI reports and / or precoder selection helps the trainable channel estimator make use of DMRS and / or data REs. The WTRU may read the schedule for the WTRU PRD group it belongs to.

[0109] FIG. 4 is an example diagram of online data labeling and / or training of artificial intelligence / machine learning AI / ML based CSI estimator. The WTRU 404 may receive the PRD data 408 and / or apply preprocessing 412. Preprocessing 412 in this case may include, but not be limited to extracting the PRD REs from the received PRD signals using the PRD resource allocation. Preprocessing 412 may include division by, or multiplication by conjugate of, the known sequence of PRD symbols. Preprocessing 412 may include a 2D finite impulse response (FIR) filtering option of the PRD carrying REs. The WTRU 404 may use the preprocessed data 412 as input to a channel estimator 416 (e.g, a high- quality channel estimator). The WTRU 404 may use the preprocessed PRD 412 to generate, via the channel estimator 416 (e.g., a high-quality channel estimator), a good quality label 420 for the CSI (e.g., a CSI label) as illustrated in FIG. 4.

[0110] The WTRU 404 may use the PRD 408 as input to a trainable channel estimator 424. Preprocessing 428 in this case may include, but not be limited to extracting the PRD REs from the received PRD signals 408 using the PRD resource allocation. The WTRU 404 may use the preprocessed PRD 428 as input to a trainable channel estimator 424 (e.g., AI / ML model, AI / ML channel estimator, etc.) and / or generate an output CSI 432 as illustrated in FIG. 4.

[0111] The WTRU 404 may train the trainable channel estimator (e.g., AI / ML model, AI / ML channel estimator, etc.) based on the generated CSI label 420 and / or the output CSI 432 as illustrated in FIG. 4. The WTRU 404 may monitor training progress by, e.g., measuring the rate of change of the loss function 436 and / or comparing to threshold. For example, the training of the trainable channel estimator 424 (e.g., AI / ML model, AI / ML channel estimator, etc.) may be based on the CSI label 420 and the output CSI 432 by comparing them with (e.g., inputting them into) the loss function (e.g., the measured difference between the CSI label 420 and the output CSI 432) and computing the loss 436 and using backpropagation to update the parameters (e.g, the weights and / or biases of the layers) of the model. The WTRU 404 may then report to the network 440 (e.g, gNB) that training is sufficient. The WTRU 404 may use the trained channel estimator 424 (e.g, AI / ML model, AI / ML channel estimator, etc.) for online inference. The input to the trained channel estimator 412 may be normal CSI-RS 420 received from network 412 (e.g, gNB). The output may be the output CSI 432 associated to the received normal CSI-RS 444.

[0112] During the online inference of AI / ML models, the WTRU and / or the network AI / ML endpoint (e.g, a gNB, a cloud server, and / or an edge computing server) may need to monitor the performance of the AI / ML model. In the context of channel estimation for sounding, the following procedures focus on providing the steps that allow a WTRU to perform performance monitoring of the AI / ML model using the PRD.

[0113] FIG. 5 is an example diagram of performance monitoring of AI / ML-based CSI estimator using PRD. The WTRU 504 may request and / or monitor the performance of the channel estimation algorithm as illustrated in FIG. 5. The WTRU 504 may use PRD 508 for performance monitoring. The WTRU 504 may use Type 1 , Type 2a, and / or Type 2b PRD scheduling for performance monitoring. The monitoring request may indicate the test bandwidth (BW), bandwidth part (BWP) and / or number of RBs. The network may use the request to decide how large the test allocations should be. The WTRU 504 may receive the performance threshold from the network 540 e.g., gNB). The network (e.g., gNB) may indicate a threshold in SIB and / or RRC. To continue using the same AI / ML model, the estimated NMSE of channel estimate may not exceed this threshold. The WTRU 504 may use non-specific test allocations. These allocations may be immediate (e.g., signaled in DCI and / or similar control channel), and / or semi-static with test allocation information indicated in SIB and / or other broadcast messages.

[0114] The WTRU 504 may receive the PRD 508 and / or apply preprocessing 512. The WTRU 504 may use the preprocessed data 512 as input to a channel estimator 516 (e.g., a high-quality channel estimator). When the WTRU 504 uses PRD 508 to generate CSI labels 520, preprocessing 512 may include extracting the normal PRD REs from the received PRD signals using the PRD resource allocation. Preprocessing 512 may include division by, or multiplication by conjugate of, the known sequence of PRD symbols. Preprocessing 512 may include a 2D FIR filtering option of the PRD carrying REs. The WTRU 504 may use the preprocessed PRD 512 to generate, via the channel estimator 516 (e.g., a high-quality channel estimator), a good quality label 520 for the CSI (e.g., CSI label) as illustrated in FIG. 5.

[0115] The WTRU 504 may use the PRD 508 as input to a trainable channel estimator 524 (e.g., AI / ML model, AI / ML channel estimator, etc.). Preprocessing 528 in this case may include, but not be limited to extracting the PRD REs from the received PRD signals 508 using the PRD resource allocation. The WTRU 504 may use the preprocessed PRD 528 as input to trainable channel estimator 524 (e.g., AI / ML model, AI / ML channel estimator, etc.) and / or generate an output CSI 532 as illustrated in FIG. 5.

[0116] The WTRU 504 may compute the error 536 between CSI labels 520 and the output CSI 532. If the error 536 is below a threshold, the WTRU 504 may report the result to the network 540 (e.g., gNB) and / or may continue to use the same trainable channel estimator 524 (e.g., AI / ML model, AI / ML channel estimator, etc.). If the error 536 exceeds a threshold, the WTRU 504 may report the result to the network 540 (e.g., gNB), may fallback to a legacy non-AI / ML channel estimator, and / or may request retraining the AI / ML model online by requesting training allocations.

[0117] During the online inference of AI / ML models, the WTRU or the network AI / ML endpoint (e.g., a gNB, a cloud server, and / or an edge computing server) may monitor the performance of the AI / ML model. In the context of channel estimation for sounding, the following procedures focus on providing the steps that allow a WTRU to perform performance monitoring of the AI / ML model using L-RSs.

[0118] FIG. 6 is an example diagram of performance monitoring of AI / ML-based CSI estimator using learning reference signals (L-RS). The WTRU 604 may request and / or monitor the performance of the channel estimation algorithm as illustrated in FIG. 6. The WTRU 604 may use L-RSs 608 for performance monitoring. An L-RS 608 is an online learning RS that supports (e.g., in the case of normal CSI-RS 610) semi-persistent, periodic, and / or aperiodic scheduling. The WTRU 604 may use Type 1 , Type 2a, and / or Type 2b L-RS scheduling for performance monitoring. The monitoring request may indicate the test BW, BWP and / or number of RBs. The network may use the request to decide how large and / or how frequent the test allocations should be. The WTRU 604 may receive the performance threshold from the network 640 (e.g., gNB). The network 640 (e.g., gNB) may indicate in SIB and / or RRC that the estimated NMSE of channel estimate is below a threshold to continue using the same AI / ML model). The WTRU 604 may use non-specific test allocations. These allocations may be immediate (e.g., signaled in DCI and / or similar control channel), and / or semi-static with test allocation information indicated in SIB and / or other broadcast messages.

[0119] The WTRU 604 may receive the normal CSI-RS 610 and / or apply preprocessing 612. The WTRU 604 may use the preprocessed CSI-RS 612 as input to a channel estimator 616 (e.g., a high-quality channel estimator). When the WTRU 604 uses CSI-RS 610 to generate CSI labels 620, preprocessing 612 may include extracting the normal CSI-RS 610 received on the allocated resources. Preprocessing 612 may include division by, or multiplication by conjugate of, the known CSI-RS 610 symbols. Preprocessing 612 may include a 2D FIR filtering option of the CSI-RS 610 carrying REs. The WTRU 604 may apply out- of-distribution (OOD) detection test to the extracted CSI-RS 610.

[0120] The WTRU 604 may receive the L-RS 608 and / or apply preprocessing 612. The WTRU 604 may use the preprocessed data 612 as input to a channel estimator 616 (e.g., a high-quality channel estimator). When the WTRU 604 uses L-RS 608 to generate CSI labels 620, preprocessing 612 may include extracting the L-RS 608 received on the allocated resources. Preprocessing 612 may include division by, or multiplication by conjugate of, the known L-RS 608 symbols. Preprocessing 612 and / or a 2D FIR filtering option of the L-RS 608 carrying REs. The WTRU 604 may apply out OOD detection test to the extracted L- RS 608.

[0121] The WTRU 604 may apply preprocessing 628 to the normal CSI-RS 610. The WTRU 604 may use the preprocessed CSI-RS 628 as input to a trainable channel estimator 624 (e.g., AI / ML model, AI / ML channel estimator, etc.) to generate an output CSI 632. When the WTRU 604 uses CSI-RS 608 to generate output CSI 632, preprocessing 628 may include extracting the normal CSI-RS 610 received on the allocated resources. Preprocessing 628 may include division by, or multiplication by conjugate of, the known CSI-RS 610 symbols. Preprocessing 628 may include a 2D FIR filtering option of the CSI-RS 610 carrying REs. The WTRU 604 may apply OOD detection test to the extracted CSI-RS 610. The WTRU 604 may use the preprocessed normal CSI-RS 628 as input to a trainable channel estimator 624 (e.g., AI / ML model, AI / ML channel estimator, etc.). The WTRU 604 may use the trainable channel estimator 624 to generate output CSI 632, as illustrated in FIG. 6.

[0122] The WTRU 604 may compute the error 636 between the CSI labels 620 and / or the output CSI 632. If the error 636 is below threshold, the WTRU 604 may report the result to the network 640 (e.g., gNB) and / or continue to use the same AI / ML model (e.g., AI / ML channel estimator, etc.). If the error exceeds the threshold, the WTRU may report the result to the network 640 (e.g., gNB), may fallback to legacy non- AI / ML channel estimator, and / or may request retraining the AI / ML model online by requesting training allocations.

[0123] Hereinafter described are methods for online learning of a CSI estimator. The WTRU and / or the network may detect that the performance of the AI / ML model has degraded. When the WTRU first enters a new region, the AI / ML model may not be well trained for this new local environment. In this situation, the WTRU may request support for updating the AI / ML model. In examples, the WTRU may retrain the AI / ML model online with the assistance of another node in the network (e.g., gNB, another WTRU using sidelink (SL), and / or WTRU-to-WTRU direct communication, etc.). Another example provides that the WTRU retrains the AI / ML model online with the assistance of another node in the network (e.g., gNB, another WTRU using SL and / or WTRU-to-WTRU direct communication, etc.). This context focuses on providing the steps and / or the details that allow a WTRU to perform online training of AI / ML models.

[0124] When a WTRU is configured to perform online learning of an AI / ML model for CSI estimation for channel sounding, the WTRU may transmit key parameters related to the AI / ML functions to the network. These parameters may include the WTRU’s use of PRD and / or other training signals depending on its capabilities. The capabilities message may include use Type 1 training allocation in non-opportunistic mode and / or use Type 1 training allocation in opportunistic mode.

[0125] Opportunistic mode may occur when the WTRU detects, but does not assume, that the allocation is a PRD training allocation. Supporting opportunistic mode implies that the WTRU may detect when the PRD training allocation is (or is not) replaced by data. Moreover, supporting opportunistic mode implies the WTRU may not use a PRD allocation for training if the WTRU determines that the PRD was replaced by data.

[0126] Opportunistic mode capabilities may further include using Type 2a PRD training allocations for training without extended PRD. The extended PRD may refer to the possibility that during a Type 2a PRD training allocation, the REs reserved for data may instead be replaced by an extension of the PRD, (e.g., if there is no data available to transmit in those REs, then they may contain PRD). If the WTRU supports Type 2a only without extended L-RS, then the WTRU may simply ignore the extended PRD (and does not need to test for extended PRD).

[0127] Opportunistic mode capabilities may further include using Type 2a training allocations for training with extended PRD. Supporting Type 2a PRD training allocation with extended PRD may mean that the WTRU detects when PRD extension occurs and / or will include the extended PRD in the high-quality label generation.

[0128] Opportunistic mode capabilities may further include using Type 2a training allocations for data reception. Supporting Type 2a PRD training allocations for data reception may indicate to the network that it may receive data using Type 2a PRD training allocation.

[0129] Opportunistic mode capabilities may further include using Type 2b training allocations for training without extended PRD; use Type 2b training allocations for training with extended PRD; use Type 2b training allocations for data reception; use Type 2a training allocation in opportunistic mode; and / or use Type 2b training allocation in opportunistic mode.

[0130] Additionally, or alternatively, the network may configure other key parameters related to the AI / ML functions via higher layer signaling (e.g., RRC signaling and / or SIB, etc.). The configuration may include information elements and / or parameters for PRD training allocations scheduling. For example, these elements and parameters may include the PRD being an online learning training signal that supports (like in the case of normal CSI-RS) semi-persistent, periodic, and / or aperiodic scheduling. PRD training allocations may be non-specific to any WTRU (or non-specific to any WTRU within a WTRU-L-RS group(s) indicated for the training allocation). PRD may be configured with higher power and / or higher timefrequency density than normal CSI-RS and / or may be indicated in the training allocation as an offset from CSI-RS power.

[0131] Additional configurations of the WTRLI to perform online learning may include WTRU specific REs multiplexing with PRD in type 2 training allocations. In examples, PRDs may be multiplexed concurrently with data and DRMS REs transmitted in new radio (NR) physical downlink shared channel (PDSCH) as well as other RS (e.g, CSI-RS, etc.). A parameter included in the training allocation information, RPRD, may determine what ratio of the REs in the training allocation are used for PRD. RPRD may be selected from a finite set (e.g, list) of numbers and / or may be referenced as an index into that list. An additional parameter(s) OPRD and / or NPRD may determine the starting RE offset and / or number of PRDs to be inserted in the Type 2 allocation. For example, if the REs are listed (e.g., made ordinal), then OPRD may indicate the first RE to carry PRD; RPRD may determine how many REs to skip until next PRD is inserted; and / or NPRD may determine how many PRD are inserted (and / or the RE number where PRD are no longer inserted). The ordinal list may omit a predetermined set of REs such as DMRS. The predetermined list may be part of the training allocation information.

[0132] The WTRU may be configured with additional elements and / or parameters for online learning, including but not limited to: PRD resource mapping (e.g., number of antenna ports, PRD patterns, etc.); high quality channel estimators: L-RS modulation pattern; grouping of WTRUs for different PRD opportunities; size of training allocations; channel resource configuration; the set of usable RBs; thresholds of ML performance monitoring and associated priority levels; and / or AI / ML model training parameters (e.g., loss function, training performance threshold, etc.).

[0133] For WTRU initiated online training, the WTRU may transmit a message of an online training request transmitted to the network AI / ML end point node. The training request may indicate training BW and / or number of RBs. The network may use the request to decide how large training allocations should be. The training request may indicate channel statistics measurements such as delay spread and / or Doppler; a channel diversification request (e.g, increased delay spread and / or Doppler). The network may use the request to decide how much (if any) variation to add to the channel (e.g, a time-varying FIR filter may be used to augment the channel).

[0134] The training request may include the priority of the request. Therein, the WTRU may indicate a high priority if the performance of the channel estimator is below a certain threshold. The WTRU may receive the performance threshold from the network, (e.g, the network may indicate in SIB and / or RRC that an estimated NMSE of channel estimate exceeds a threshold is required for WTRU to make priority request). The network may decide to not override data REs with training allocations or increase the training allocations if the priority is not high. The WTRU may indicate a low priority if the request is for predictivemaintenance, (e.g., to compare the quality of an AIML CSI estimator to the CSI estimated of a reference CSI estimator using the PRD). If the comparison indicates the AIML performance has degraded compared to a threshold, the WTRU may signal the network with a new message to increase the priority of training allocation.

[0135] The training request may include the message of online training response transmitted by the network AI / ML end point node as the response to the online training request transmitted by the WTRU. The message may include but not be limited to: the location of RBs and / or PRBs where the PRD is transmitted on; allocated online training time duration and the starting time; and / or allocated online training L-RS pattern and / or density, etc. The message of online training request from the WTRU may be transmitted on physical uplink control channel (PUCCH), contention-free physical random access channel (PRACH), PUSCH, medium access control control element (MAC-CE) RRC message, and / or non-access stratum (NAS) message.

[0136] Disclosed herein are procedures for online training, wherein a WTRU may train an AI / ML model for CSI estimation for sounding. The stages of the AI / ML process and / or the steps include application of the AI / ML model (e.g., CSI estimation) and / or input data (e.g. CSI-RS with preprocessing, and / or PRD with preprocessing, etc.).

[0137] When the WTRU receives the PRD, the WTRU may apply preprocessing. Preprocessing may include extracting the PRD REs from the received PRD signals using the PRD resource allocation.

[0138] Regarding the AI / ML model, for CSI estimation, various AI / ML models may be proposed in the literature, wherein Residual channel Estimation Network (ReEsNet) may be a good model in terms of performance-complexity tradeoff. ReEsNet may be based on convolution layers. The next step may regard output data, wherein full channel estimates (e.g., channel impulse response) associated to the input (e.g., extracted normal CSI-RS).

[0139] The training may be performed online in a supervised manner. Features of the training may be the extracted PRD REs. The labels may be the target (e.g., true) full channel estimates obtained using the same PRD used as input data, but after applying additional preprocessing. Preprocessing in this case may include extracting the PRD REs from the received PRD signals using the PRD resource allocation; division by (or multiplication by conjugate of) the known sequence of PRD symbols; and / or a 2D FIR filtering option of the PRD carrying REs.

[0140] The WTRU may use the preprocessed PRD with a channel estimator that uses this high-quality and / or quantity of PRD to generate high-quality labels for the AI / ML CSI estimator (e.g., CSI labels). TheWTRU may train the trainable (e.g., AI / ML) channel estimator based on the generated CSI label and / or the output CSI. The WTRU may then monitor the training progress, (e.g., by measuring the rate of change of the loss function, comparing to threshold, and / or reporting to the network (e.g., gNB) that training is sufficient).

[0141] Next, the WTRU may use the trained channel estimator for online inference. The input to the trained channel estimator may include normal CSI-RS received from network (e.g., gNB). The input to the trained channel estimator may include user data and DMRS received from network (e.g., gNB). The output may be the output CSI estimate.

[0142] Hereinafter described are methods for AI / ML model performance monitoring using learning reference signals. When a WTRU performs drift detection of the AI / ML model used for CSI estimation for channel sounding, the network may configure key parameters via higher layer signaling (e.g., RRC signaling, and / or SIB, etc.). The configuration may include the error threshold and / or PRD test allocations scheduling.

[0143] The steps and / or procedures for the WTRU to perform AI / ML model performance monitoring using PRD for CSI estimation are described herein. The WTRU may request and / or monitor the performance of the channel estimation algorithm, as illustrated in FIG. 5. The WTRU may be configured to use PRD for performance monitoring. The WTRU may use Type 1 , Type 2a, and / or Type 2b PRD scheduling for performance monitoring.

[0144] The monitoring request may indicate test BW, BWP and / or number of RBs. The network may use the request to decide how large the test allocations should be. The WTRU may receive the performance threshold from the network (e.g., gNB). The WTRU may indicate in SIB and / or RRC that the estimated NMSE of channel estimate is below a threshold to continue using the same AI / ML model. The WTRU may be configured to use non-specific test allocations. These allocations may be immediate (e.g., signaled in DCI and / or similar control channel), semi-static with test allocation information indicated in SIB, and / or other broadcast messages.

[0145] The WTRU may receive the normal CSI-RS and / or apply preprocessing. The WTRU may use the PRD to generate CSI labels. Preprocessing may include extracting the PRD REs received on the allocated resources. Preprocessing may include division by, or multiplication by conjugate of, the known sequence of PRD symbols. Preprocessing may include 2D FIR filtering option of the PRD carrying REs.

[0146] The WTRU may use the preprocessed PRD to generate a good quality label for the CSI (e.g., CSI label), as illustrated in FIGs. 4 and 5.

[0147] The WTRU may use the PRD as input to a channel estimator. Preprocessing in this case includes extracting the PRD REs from the received PRD signals using the PRD resource allocation. The WTRU may use the preprocessed PRD as input to a trainable channel estimator and / or generate an output CSI, as illustrated in FIGs. 4 and 5.

[0148] The WTRU may compute the error between the CSI labels and / or the output CSI. If the error is below threshold, then the WTRU may report the result to the network (e.g., gNB) and / or continue to use the same AI / ML model (e.g., AI / ML channel estimator, etc.). If the error is higher than threshold, then the WTRU may report the result to the network (e.g., gNB) and / or may request retraining the AI / ML model online (e.g., AI / ML channel estimator, etc.).

[0149] When a WTRU may perform drift detection of the AI / ML model used for CSI estimation for channel sounding using L-RSs, the network may configure key parameters via higher layer signaling (e.g., RRC signaling, and / or SIB, etc.). The configuration may include the error threshold and / or L-RS test allocations scheduling.

[0150] The steps and / or procedures for the WTRU to perform AI / ML model performance monitoring for CSI estimation may begin when the WTRU requests and / or monitors the performance of the channel estimation algorithm. The WTRU may use L-RS for performance monitoring, as illustrated in FIGs. 6. L- RS is an online learning RS that may support (e.g., normal CSI-RS) semi-persistent, periodic, and / or aperiodic scheduling. Then, the WTRU may use Type 1, Type 2a, and / or Type 2b L-RS scheduling for performance monitoring.

[0151] The monitoring request may indicate test BW, BWP and / or number of RBs. The network may use the request to decide how large the test allocations should be. The WTRU may receive the performance threshold from the network (e.g., gNB). The WTRU may indicate in SIB and / or RRC that the estimated NMSE of channel estimate is below a threshold to continue using the same AI / ML model. The WTRU may use non-specific test allocations. These allocations may be immediate (e.g., signaled in DCI and / or similar control channel), semi-static with test allocation information indicated in SIB, and / or other broadcast messages.

[0152] The WTRU may receive the normal CSI-RS and / or apply some preprocessing, as illustrated in FIG.6. Preprocessing may include extracting the normal CSI-RS received on the allocated resources. The WTRU may apply OOD detection test to the extracted CSI-RS.

[0153] The WTRU may receive the L-RS and / or apply some preprocessing, as illustrated in FIG. 6. Preprocessing may include extracting the L-RS received on the allocated resources. Preprocessing mayinclude division by, or multiplication by conjugate of, the known L-RS symbols. Preprocessing may include a 2D FIR filtering option of the L-RS carrying REs. The WTRU may apply OOD detection test to the extracted L-RS.

[0154] The WTRU may use the L-RS and / or the normal CSI-RS with a channel estimator that uses this high-quality and / or quantity of input data to generate high-quality labels for the AI / ML CSI estimator (e.g., CSI labels), as illustrated in FIG. 6. The normal CSI-RS and / or the CSI labels may be saved in a replay buffer for continued training.

[0155] The WTRU may use the normal CSI-RS as input to a trainable channel estimator (AI / ML-based channel estimator). The WTRU may generate an output CSI, as illustrated in FIG. 6. The normal CSI-RS may be saved in a replay buffer.

[0156] The WTRU may compute the error between the CSI labels and / or the output CSI. If the error is below threshold, then the WTRU may report the result to the gNB and / or continue to use the same AI / ML model. If the error is higher than threshold, then the WTRU may report the result to the gNB and / or may request retraining the AI / ML model online.FIG. 7 depicts an example AI / ML CSI estimation module 700, from the WTRU perspective. The AI / ML channel estimator 704 receives, as an input, CSI-RS 708. After performing the operations described herein, the AI / ML channel estimator 704 generates an output which is estimated channel 712.

Claims

CLAIMSWhat is claimed is:1 . A method implemented by a wireless transmit / receive unit (WTRU), the method comprising: receiving pseudo-random data (PRD); receiving one or more channel state information reference signals (CSI-RSs); receiving one or more learning reference signals (L-RSs); determining CSI measurements based on the received CSI-RSs; determining L-RS measurements based on the received L-RSs; generating one or more CSI labels based on the PRD, the CSI-RS measurements, and the L-RS measurements; generating one or more output CSI using an artificial intelligence / machine learning (AI / ML) model and based on the PRD and the CSI-RS measurements; and training the AI / ML model with the one or more CSI labels and the one or more output CSI.

2. The method of claim 1 , further comprising: determining a change of a loss function of the AI / ML model based on the one or more CSI labels and the one or more output CSI; and determining whether the training is complete based on a comparison of the value of the loss function to a threshold value.

3. The method of claim 2, wherein the loss function of the AI / ML model determines a metric for training accuracy during the training of the AI / ML model.

4. The method of claim 1 , further comprising: determining a model drift metric; determining a performance of the AI / ML model for CSI estimation based on a model drift by applying a statistical test to the CSI-RS measurements; and retraining the AI / ML model based on the model drift metric exceeding a drift threshold value.

5. The method of claim 4, further comprising: sending a report to a network in response to an indication that the model drift metric exceeds the drift threshold value, wherein the report comprises details of the model drift or a recommendation for adjusting the AI / ML model.

6. The method of claim 1 , further comprising: sending a report to a network based on the CSI-RS measurements and a request to be configured with one or more training allocations; receiving an indication from the network as to whether the request to be configured with one or more training allocations was granted.

7. The method of claim 6, wherein the indication further comprises seed information used to generate the PRD.

8. The method of claim 6, further comprising: receiving a PRD training allocation configuration, wherein the PRD training allocation indicates a first type of PRD training allocation, a second type of PRD training allocation, or a third type of PRD training allocation; wherein the first type of PRD training allocation configures the WTRU to use the PRD and the CSI- RS measurements for a first set of resource blocks that carry the PRD and the CSI-RS; wherein the second type of PRD training allocation configures the WTRU to use the PRD and the CSI-RS measurements for a second set of resource blocks that carry the PRD, the CSI-RS, and the resource elements comprising data that do not carry PRD, wherein the PRD and the resource elements comprising data that do not carry PRD occupy the same resource blocks within the second set of resource blocks; and wherein the third type of PRD training allocation configures the WTRU to use the PRD and CSI-RS measurements for a third set of resource blocks that carry the PRD, the CSI-RS, and the resource elements comprising data that do not carry PRD, where the PRD and the resource elements comprising data that do not carry PRD occupy different but overlapping resource blocks within the third set of resource blocks.

9. The method of claim 8, wherein the first, second, and third types of PRD resource allocations are signaled in downlink control information (DCI), semi-statically via a system information block, or via a broadcast message.

10. The method of claim 1 , further comprising, sending a first message, the first message comprising an online training request, wherein the online training request indicates a number of resource blocks carrying PRD for training, channel statistic measurements, a channel diversification request, and a priority of request; and receiving a second message, the second message comprising an online training response, wherein the online training response indicates the location of the resource blocks that comprise the PRDs, an allocated online training time duration, an allocated online training starting time, and an allocated online training L-RSs.

11. A wireless transmit / receive unit (WTRU) comprising a processor and a memory, the processor and memory configured to: receive pseudo-random data (PRD); receive one or more channel state information reference signals (CSI-RSs); receive one or more learning reference signals (L-RSs); determine CSI measurements based on the received CSI-RSs; determine L-RS measurements based on the received L-RSs; generate one or more CSI labels based on the PRD, the CSI-RS measurements, and the L-RS measurements; generate one or more output CSI using an artificial intelligence / machine learning (AI / ML) model and based on the PRD and the CSI-RS measurements; and train the AI / ML model with the one or more CSI labels and the one or more output CSI.

12. The WTRU of claim 11 , the processor further configured to: determine a change of a loss function of the AI / ML model based on the one or more CSI labels and the one or more output CSI; and determine whether the training is complete based on a comparison of the value of the loss function to a threshold value.

13. The WTRU of claim 12, wherein the loss function of the AI / ML model determines a metric for training accuracy during the training of the AI / ML model.

14. The WTRU of claim 11 , the processor further configured to: determining a model drift metric; determine a performance of the AI / ML model for CSI estimation based on a model drift by applying a statistical test to the CSI-RS measurements; and retrain the AI / ML model based on the model drift metric exceeding a drift threshold value.

15. The WTRU of claim 14, the processor further configured to: send a report to a network in response to an indication that the model drift metric exceeds the drift threshold value, wherein the report comprises details of the model drift or a recommendation for adjusting the AI / ML model.

16. The WTRU of claim 1 , the processor further configured to: send a report to a network based on the CSI-RS measurements and a request to be configured with one or more training allocations; receive an indication from the network as to whether the request to be configured with one or more training allocations was granted.

17. The WTRU of claim 16, wherein the indication further comprises seed information used to generate the PRD.

18. The WTRU of claim 16, the processor further configured to: receive a PRD training allocation configuration, wherein the PRD training allocation indicates a first type of PRD training allocation, a second type of PRD training allocation, or a third type of PRD training allocation;wherein the first type of PRD training allocation configures the WTRU to use the PRD and the CSI- RS measurements for a first set of resource blocks that carry the PRD and the CSI-RS; wherein the second type of PRD training allocation configures the WTRU to use the PRD and the CSI-RS measurements for a second set of resource blocks that carry the PRD, the CSI-RS, and the resource elements comprising data that do not carry PRD, wherein the PRD and the resource elements comprising data that do not carry PRD occupy the same resource blocks within the second set of resource blocks; and wherein the third type of PRD training allocation configures the WTRU to use the PRD and CSI-RS measurements for a third set of resource blocks that carry the PRD and the CSI-RS, and the resource elements comprising data that do not carry PRD, where the PRD and the resource elements comprising data that do not carry PRD occupy different but overlapping resource blocks within the third set of resource blocks.

19. The WTRU of claim 18, wherein the first, second, and third types of PRD resource allocations are signaled in downlink control information (DCI), semi-statically via a system information block, or via a broadcast message.

20. The WTRU of claim 11 , the processor further configured to, send a first message, the first message comprising an online training request, wherein the online training request indicates a number of resource blocks carrying PRD for training, channel statistic measurements, a channel diversification request, and a priority of request; and receive a second message, the second message comprising an online training response, wherein the online training response indicates the location of the resource blocks that comprise the PRDs, an allocated online training time duration, an allocated online training starting time, and an allocated online training L-RSs.

Images